Study of Psychological Effects of Administrative Segregation, Colorado DOC, 2010
Download original document:
Document text
Document text
This text is machine-read, and may contain errors. Check the original document to verify accuracy.
The author(s) shown below used Federal funds provided by the U.S.
Department of Justice and prepared the following final report:
Document Title:
One Year Longitudinal Study of the
Psychological Effects of Administrative
Segregation
Author:
Maureen L. O’Keefe, M.A., Kelli J. Klebe, Ph.D.,
Alysha Stucker, B.A., Kristin Sturm, B.A.,
William Leggett, B.A.
Document No.:
232973
Date Received:
January 2011
Award Number:
2006-IJ-CX-0015
This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federallyfunded grant final report available electronically in addition to
traditional paper copies.
Opinions or points of view expressed are those
of the author(s) and do not necessarily reflect
the official position or policies of the U.S.
Department of Justice.
2010
IIIIMMININE
i
One Year Longitudinal
Study of the
Psychological Effects of
Administrative Segregation
By:
Maureen L. O’Keefe, M.A.
Colorado Department of Corrections
2862 South Circle Drive
I
Colorado Springs, CO 80906
(719) 226‐4364
Kelli J. Klebe, Ph.D.
Alysha Stucker, B.A.
Kristin Sturm, B.A.
William Leggett, B.A.
w.
w.,
,
w.,
University of Colorado – Colorado Springs
w.
w.,
,
Department of Psychology
w.
w.,
,
w.
,
w.
,
1420 Austin Bluffs Parkway
w.
w.,
,
w.,
Colorado Springs, CO 80918
(719) 255‐4175
Submission Date to NIJ: October 31, 2010
Aristedes W. Zavaras
Executive Director
Department of Corrections
Pamela Shockley‐Zalabak
Chancellor
University of Colorado
Colorado Springs
This project was supported by Grant No. 2006‐IJ‐CS‐0015 awarded by the Na‐
tional Institute of Justice, Office of Justice Programs, U.S. Department of Justice.
Points of view in this document are those of the author and do not necessarily
i
represent the official position or policies of the US Department of Justice.
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
ACKNOWLEDGEMENTS
The research team would like to gratefully acknowledge the following individuals who served on the
project’s advisory board for their significant time commitment and invaluable insights into this project:
Jamie Fellner, Esq.
Human Rights Watch
Jeffrey Metzner, M.D.
University of Colorado – Denver
Joel Dvoskin, Ph.D.
University of Arizona
Larry Reid, M.A.
Susan Jones, M.C.J.
Joanie Shoemaker, M.B.A.
Jim Michaud, Ph.D.
Elizabeth Hogan, M.D.
Colorado Department of Corrections
John Stoner, Ph.D.
Colorado Department of Corrections, Retired
The researchers would also like to acknowledge the support of Mr. Aristedes Zavaras and the staff at the
Department of Corrections. This project would never have been possible without the commitment and con‐
tributions of management, mental health, case management, offender services, and security staff.
i
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
ABSTRACT
One of the most widely debated topics in the field of corrections – the use of long‐term administrative se‐
gregation (AS) – has suffered from a lack of empirical research. Critics have argued that the conditions of AS
confinement exacerbate symptoms of mental illness and create mental illness where none previously ex‐
isted. Empirical research has had little to offer this debate; the scant empirical research conducted to date
suffers from research bias and serious methodological flaws. This study seeks to advance the literature in
this regard.
This study tested three hypotheses: (1) offenders in AS would develop an array of psychological symptoms
consistent with the security housing unit (SHU) syndrome, (2) offenders with and without mental illness
would deteriorate over time in AS, but at a rate more rapid and extreme for the mentally ill, and (3) inmates
in AS would experience greater psychological deterioration over time than the comparison groups.
Study participants included male inmates who were placed in AS and comparison inmates in the general
population (GP). Placement into AS or GP conditions occurred as a function of routine prison operations. GP
comparison participants included those at risk of AS placement due to their institutional behavior. Inmates
in both of these study conditions (AS, GP) were divided into two groups – inmates with mental illness (MI)
and with no mental illness (NMI). A third comparison group of inmates with severe mental health problems
placed in San Carlos Correctional Facility, a psychiatric care prison facility, was also included. A total of 302
inmates were approached to participate in the study, and 55 refused to participate or later withdrew their
consent. Participants were tested at 3‐month intervals over a yearlong period.
Standardized test data were collected through self‐report, correctional staff and clinical staff measures. Tests
with demonstrated reliability and validity were selected to assess the eight primary constructs of interest: (1)
anxiety, (2) cognitive impairment, (3) depression‐hopelessness, (4) hostility‐anger control, (5) hypersensitivity,
(6) psychosis, (7) somatization, and (8) withdrawal‐alienation. Extensive analyses of psychometric properties
revealed that inmates self‐reported psychological and cognitive symptoms in remarkably reliable and valid
ways.
The results of this study were largely inconsistent with our hypotheses and the bulk of literature that indi‐
cates AS is extremely detrimental to inmates with and without mental illness. Similar to other research, our
study found that segregated offenders were elevated on multiple psychological and cognitive measures
when compared to normative adult samples. However, elevations were present among the comparison
groups too, suggesting that high degrees of psychological disturbances are not unique to the AS environ‐
ment. In examining change over time patterns, there was initial improvement in psychological well‐being
across all study groups, with the bulk of the improvements occurring between the first and second testing
periods, followed by relative stability for the remainder of the study. Patterns indicated that the MI groups
tended to be similar to one another but were significantly elevated compared to the NMI groups, regardless
of their setting. Contrary to our hypothesis, offenders with mental illness did not deteriorate over time in AS
at a rate more rapid and more extreme than for those without mental illness. Finally, although AS inmates in
this study were found to possess traits believed to be associated with long‐term segregation, these features
cannot be attributed to AS confinement because they were present at the time of placement and also oc‐
curred in the comparison study groups. Implications for policy and future research are discussed.
ii
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
CONTENTS
Executive Summary ................................................. iv
Official Record Data .......................................... 40
Introduction ........................................................ iv
Group Comparisons .......................................... 45
Purpose of Present Study .................................... iv
Normative Comparisons ................................... 46
Method ................................................................ v
Change over Time ............................................. 52
Findings ............................................................. viii
Predictor Analyses............................................. 75
Policy Implications .............................................. ix
Discussion .............................................................. 78
Introduction ............................................................. 1
Limitations ......................................................... 79
Characteristics of Long‐Term Segregation ........... 1
Future Research ................................................ 80
Criticisms of the AS Model ................................... 2
Policy Implications ............................................ 81
Case Law Review .................................................. 3
References ............................................................ 84
Research Review .................................................. 5
Appendix A ............................................................ 91
The Colorado System ........................................... 7
Appendix B ............................................................ 95
Purpose of Present Study ...................................15
Summary of Study Measures ............................ 95
Method ..................................................................17
Description of Individual Measures .................. 96
Group Assignment .............................................17
Normative Comparisons ................................. 125
Participants ........................................................19
Composite Scores ............................................ 126
Materials ............................................................19
Summary ......................................................... 138
Procedures .........................................................28
Appendix B References ................................... 139
Results ....................................................................31
Appendix C .......................................................... 145
Data Analysis Plan ..............................................31
Prison Symptom Inventory Analyses .............. 145
Sampling .............................................................32
Summary ......................................................... 150
Validity of Responses .........................................36
iii
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
EXECUTIVE SUMMARY
INTRODUCTION
One of the most widely debated topics in the field of corrections – the use of long‐term administrative se‐
gregation (AS) – has suffered from a lack of empirical research. The placement of offenders in AS environ‐
ments, particularly those with serious mental illness, has been a point of contention. Critics have argued
that the conditions of AS confinement exacerbate symptoms of mental illness and create mental illness
where none previously existed. The use of AS across the country has persisted as a corrections management
tool despite litigation, although in many states, the placement of mentally ill into AS is no longer permitted.
Empirical research has had little to offer this debate; the scant empirical research conducted to date suffers
from research bias and serious methodological flaws.
Now decades after the deinstitutionalization of states’ mental health hospitals, corrections agencies have
seen a surge of offenders with serious mental illness in their prisons. The rate of serious mental illness in the
community is 6% (National Institute of Mental Health, 2010). Among the incarcerated, the rate of serious
mental illness is tripled at about 18% (Ditton, 1999; O’Keefe & Schnell, 2008). A similar phenomenon is oc‐
curring within prisons, whereby a disproportionate rate of mentally ill are found within AS, estimated to be
50% higher than the rate within the general prison population (O’Keefe, 2008a). It is not known the extent
to which this difference is caused by the AS environment. Researchers have been unable to settle the ques‐
tion of whether these high rates of mental illness are caused by AS relative to the general prison population
or whether there is a selection bias such that offenders with mental illness, unable to adapt to general pris‐
on settings, are placed in AS at higher rates. This study seeks to advance the literature in this regard.
PURPOSE OF PRESENT STUDY
The broad purpose of the project was to evaluate the psychological effects of long‐term segregation on of‐
fenders, particularly those with mental illness. This study examined conditions as they existed in the Colora‐
do prison system with respect to AS, using the Colorado State Penitentiary (CSP) as the AS study facility. On‐
ly males were included because females represent 2% of Colorado’s AS population. We did not assign in‐
mates to segregation, but studied those conditions as they naturally occurred. The following are the primary
goals and hypotheses.
Goal 1: To determine which, if any, psychological domains are affected, and in which direction, by the differ‐
ent prison environments. A multitude of psychological dimensions were examined, drawing from those most
often cited in the literature. The broad constructs of interest were depression/hopelessness, anxiety, psy‐
chosis, withdrawal and alienation, hostility and anger control, somatization, hypersensitivity, and cognitive
impairment. We hypothesized that offenders in segregation would develop an array of psychological symp‐
toms consistent with the security housing unit (SHU) syndrome, with elevations across the eight constructs.
Goal 2: To assess whether offenders with mental illness decompensate differentially from those without
mental illness. We were particularly interested in whether long‐term segregation had a differential impact
based on the presence of mental illness in offenders. We sought answers to the following questions: Does
AS exacerbate symptoms in offenders with mental illness? Does AS create symptoms of mental illness in
those who did not exhibit any at placement? It was hypothesized that offenders with and without mental
illness would deteriorate over time, but the rate at which it occurred would be more rapid and more ex‐
treme for the mentally ill.
iv
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Goal 3: To compare the impact of long‐term segregation against the general prison setting and a psychiatric
care prison. In this study, the psychological and behavioral symptoms of offenders in AS were compared to
similar offenders who were sent to San Carlos Correctional Facility (SCCF) or returned to the general prison
population pursuant an AS hearing. This study used a repeated measures design over the course of a year to
explore whether psychological distress was attributable to the various prison environments. It was hypothe‐
sized that inmates in segregation would experience greater psychological deterioration over time than the
comparison groups.
This study also included an examination of individual characteristics such as mental health status, personali‐
ty, and trauma history to determine if certain factors could predict patterns of change. Prediction analyses
were exploratory in nature and we did not formulate a hypothesis about the variables that might predict
differential rates of psychological decompensation.
METHOD
Group Assignment
Study participants included male inmates who were placed in AS and comparison inmates in the general
population (GP). Placement into AS or GP conditions occurred as a function of routine prison operations,
pending the outcome of their AS hearing, without involvement of the researchers. All study participants
classified to AS were waitlisted for and placed in CSP. Inmates who returned to GP following an AS hearing
were assumed to be as similar as possible to AS inmates and, therefore, comprised the comparison groups.
Comparison participants also included inmates targeted for a diversionary program that identified inmates
at high risk of AS placement due to their disruptive behavior. This program discontinued shortly after the
study commenced, hence few participants were identified through this method. Inmates in both of these
study conditions (AS, GP) were divided into two groups – inmates with mental illness (MI) and with no men‐
tal illness (NMI). There are fewer inmates with mental illness than without, but because both subgroups were of
equal interest to this study, separate groups enabled over‐selection of inmates with mental illness.
A third comparison group was included. This group included inmates with severe mental health problems
placed in SCCF. Of the inmates placed in SCCF, only those with patterns of prison misbehavior, as measured
by disciplinary violations, were included in the study. The purpose of the SCCF comparison group was to
study inmates with serious mental illness and behavioral problems who were managed in a psychiatric pris‐
on setting.
Participants
A total of 302 male inmates were approached to participate in the study. Thirty refused to participate. Two
more offenders were considered a passive refusal and were removed for inappropriate sexual behavior to‐
wards the researcher during the first testing session. An additional 23 offenders later withdrew their con‐
sent, although the data collected to the point of their withdrawal was used. In addition to refusals and with‐
drawals, 10 inmates released prior to the end of the study due to discretionary releases by the Parole Board
and one participant death.
Five testing sessions were initially established at 3‐month intervals, beginning with the date of consent and
initial administration. Therefore, tests were scheduled at 3 months, 6 months, 9 months and 12 months af‐
ter the baseline assessment. However, this schedule was problematic for the AS groups. When the study
v
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
began, there was a 3‐month average wait for inmates to be transferred to CSP due to a shortage of AS beds.
While on the waitlist, AS inmates were held in a punitive segregation bed at their originating facility. It was
determined that the primary goal was to study inmates in a single long‐term segregation facility (CSP) to lim‐
it confounding variables and that therefore the baseline measure should be collected upon placement into
CSP. However, it was also recognized that significant changes could occur while inmates were held in segre‐
gation at their originating facility. Therefore, a “pre‐baseline” measure was collected as close to the AS hear‐
ing as possible, which meant that the CSP groups completed six test intervals rather than five. The time be‐
tween the pre‐baseline and baseline measure varied according to how long the inmate was on the waitlist.
The median time between pre and baseline tests was 99 days, although eight offenders were moved into
CSP so quickly that they did not have a pre‐baseline measure. In the analyses, tests were aligned across
groups according to the test number, such that the CSP groups had an additional test at the end rather than
at the beginning.
Participants’ ages ranged from 17 to 59 at the time of consent, with a mean age of 31.8 (SD = 9.1). The ra‐
cial/ethnic breakdown of participants was 40% white, 36% Hispanic, 19% African American, 4% Native Amer‐
ican, and 1% Asian. Of the inmates with mental illness who were included in this study, 56% were identified with
a serious and pervasive disorder.
Materials
Assessment tools were selected to comprehensively cover the variety of psychological constructs associated
with AS (e.g., Arrigo & Bullock, 2008; Grassian, 1983; Haney, 2003). The primary constructs assessed in this
study were as follows: (1) anxiety, (2) cognitive impairment, (3) depression/hopelessness, (4) hostility/anger
control, (5) hypersensitivity, (6) psychosis, (7) somatization, and (8) withdrawal/alienation. Additionally, malin‐
gering, self‐harm, trauma, and personality disorders were assessed.
Research materials were selected to meet the following criteria: (1) use of assessments with demonstrated
reliability and validity, (2) use of multiple sources for providing information (e.g., self‐report, clinician rat‐
ings, files), (3) use of multiple assessments of each construct of interest, (4) ability to use within the prison
setting, and (5) ease of administration, including no specialized equipment, no physical contact, length of
time, and appropriate reading level.
The 12 self‐report instruments used in this study were: (1) Beck Hopelessness Scale, (2) Brief Symptom In‐
ventory, (3) Coolidge Correctional Inventory, (4) Deliberate Self‐Harm Inventory, (5) Personality Assessment
Screener, (6) Prison Symptom Inventory, (7) Profile of Mood States, (8) Saint Louis University Mental Status,
(9) State‐Trait Anxiety Inventory, (10) Structured Inventory of Malingered Symptomatology, (11) Trail Mak‐
ing Test, and (12) Trauma Symptom Inventory.
In addition to self‐report assessments, ratings of psychological functioning were obtained from clinical staff
and ratings of behavior in the housing unit were obtained from correctional staff. The Brief Psychiatric Rat‐
ing Scale (BPRS) was completed by clinical staff and the Prison Behavior Rating Scale (PBRS) was completed
by correctional staff.
Most assessments were collected at each testing period, although personality disorders, self‐harm, and
trauma history were not. It was determined that personality and trauma history were relatively stable con‐
structs that needed to be assessed only once to limit the testing burden on study participants. Also, due to
vi
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
the burden on already limited mental health resources, the BPRS was only administered at the first, third,
and fifth testing intervals.
Data from official records were collected primarily from the Department of Corrections Information System,
which is an administrative database of offender data. Offender characteristics to include demographic histo‐
ry, criminal history and offense data, institutional behavior, and needs levels were electronically down‐
loaded.
Certain data elements were collected only for study participants during the course of their participation in
the study. The following were collected and coded for the period of time between each testing interval for
each participant: the amount of time spent in various settings (e.g., segregation, GP, hospital), phone
records, and mental health crisis data. Additionally, activity logs from paper files for the CSP participants
were collected and coded.
Procedure
Study enrollment began July 2007 and ended March 2009, with final testing of all participants completed in
March 2010. The project operated under the approval of the institutional review board at the University of
Colorado at Colorado Springs.
The research team was notified of AS hearings by the case management supervisor at each facility and of
SCCF placements by the clinician who scheduled the facility transfers. Notification typically occurred before
the hearings or SCCF placement to give the field researcher maximal lead time. Researchers reviewed elec‐
tronic records to screen inmates for study eligibility.
The field researcher was a female university employee who completed the full training academy and had a
badge that permitted her unescorted access to the facilities. In advance of each visit, the researcher con‐
tacted prison security to arrange visits with specific inmates. All inmates were escorted by security staff to
the visiting room, which contained a noncontact booth for inmates in AS or punitive segregation conditions.
The researcher met individually with each inmate to review the consent form, which included the general
purpose of the study, voluntary nature of participation, risks and benefits, and remuneration. Inmates were
advised that the purpose of the study was to learn about adjustment to prison and offenders in prisons
across the state were participating in this study.
At the time of consent, the initial test battery was administered. The field researcher instructed participants
to read the directions for each test. Instructions were highlighted by researchers when there was an indica‐
tion on the test to respond with respect to a certain timeframe (e.g., in the past week). The researcher ad‐
ministered the timed tests, and she assisted if they had questions, most frequently with the definition of a
word. The researcher collected the test packet immediately following its completion, so it was not ever han‐
dled by security staff.
The field researcher distributed the PBRS to housing staff at each testing interval and collected the com‐
pleted forms upon return visits to the facility. Mental health clinicians were generally notified that a BPRS
was needed a couple weeks prior to the researcher testing to give them time to complete the assessment.
Participants’ data were kept in two separate databases. The eligibility database tracked the eligible pool of
offenders, such as identifying information, current location, date of AS hearing or SCCF placement, expected
vii
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
release date, mental health status and clinician approval, selection into study or reason for exclusion, and
date of consent or refusal. A testing schedule for study participants was incorporated into the database,
which also had reporting capabilities in order to manage the project. A separate database tracked partici‐
pants’ responses to the standardized tests; no identifying information was included in this database other
than a secure researcher‐assigned identification number. Both databases were stored on a secured server
with access restricted to project researchers.
FINDINGS
The results of this study were largely inconsistent with our hypotheses and the bulk of literature that indi‐
cates AS is extremely detrimental to inmates with and without mental illness. We hypothesized that inmates
in segregation would experience greater psychological deterioration over time than comparison inmates,
who were comprised of similar offenders confined in non‐segregation prisons. Consistent with other re‐
search, our study found that segregated offenders were elevated on multiple psychological and cognitive
measures when compared to normative adult samples (Haney, 2003; Suedfeld, Ramirez, Deaton, & Baker‐
Brown, 1982). However, there were elevations among the comparison groups too, suggesting that high de‐
grees of psychological disturbances are not unique to the AS environment. The GP NMI group was the only
one that was similar to the normative group on a number of scales.
In examining change over time patterns, there was initial improvement in psychological well‐being across all
study groups, with the bulk of the improvements occurring between the first and second testing periods,
followed by relative stability for the remainder of the study. On only one measure – withdrawal – did of‐
fenders worsen over time, but this finding was only true for the two NMI groups, so it is not attributable to
AS. Even given the improvements that occurred within the study timeframe, the elevations in psychological
and cognitive functioning that were evident at the start of the study remained present at the end of the
study.
Another hypothesis was that offenders with mental illness would deteriorate over time in AS at a rate more
rapid and more extreme than for those without mental illness. Patterns indicated that the MI groups (CSP
MI, GP MI, SCCF) tended to look similar to one another but were significantly elevated compared to the NMI
groups (CSP NMI, GP NMI), regardless of their setting. For the AS offenders, the MI group scored worse than
the NMI group on all self‐report measures except the Trails test and all staff measures except the PBRS Anti‐
Authority scale. In addition to the changes over time described above, PBRS scores decreased significantly
for segregated inmates regardless of their mental health status, which would be an indicator that staff may
be perceiving improvements, but the significant differences were from the first to the second assessment
periods when the majority of participants changed facilities, which suggests this is perhaps a measurement
error rather than a true improvement. As hypothesized there was a differential time effect for the MI and
NMI groups on several composite measures (i.e., anxiety, hostility‐anger control, hypersensitivity, somatiza‐
tion), but the interactions were in the opposite direction of our hypothesis; on average, the CSP NMI group
did not change while the CSP MI group improved.
We stated that offenders in segregation would develop an array of psychological symptoms consistent with
the SHU syndrome. As already discussed, all of the study groups, with the exception of the GP NMI group,
showed symptoms that were associated with the SHU syndrome. These elevations were present from the
start and were more serious for the mentally ill than non‐mentally ill. In classifying people as improving, de‐
viii
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
clining, or staying the same over time, the majority remained the same. There was a small percentage (7%)
who worsened and a larger proportion (20%) who improved. Therefore, this study cannot attribute the
presence of SHU symptoms to confinement in AS. The features of the SHU syndrome appear to describe the
most disturbed offenders in prison, regardless of where they are housed. In fact, the group of offenders who
were placed in a psychiatric care facility (SCCF) had the greatest degree of psychological disturbances and
the greatest amount of negative change.
Finally, in this study, we conducted some exploratory predictive analyses to determine if there were individ‐
ual characteristics that could identify who may be at greater risk of psychological harm from segregation.
There were no individual predictors that showed strong effects for predicting change. This could indicate
that we did not have the correct predictors or that patterns of decompensation are individualized (i.e., not
predictable), but it is more likely that the relative stability over time makes it difficult to predict change.
A review of the findings warrants a discussion of plausible alternative explanations that might account for
our results. The use of a repeated measures design enabled us to determine that change was occurring and
in which direction. Even given the debate about whether or not harmful effects resulted from AS, it was
never suggested that inmates might improve as this study found. The presence of comparison groups avoids
an attribution error; the changes, improvements in this case (i.e., 20%), are not due to segregation. These
conclusions replicate those drawn by Zinger and colleagues (2001) where there was a similar lack of evi‐
dence of harm. These studies suffered criticism for high refusal rates, high attrition rates, small sample sizes,
and short durations – limitations that were corrected in the present study (note, however, that no generali‐
zations should be made beyond the 1 year follow‐up period in this study). Furthermore, the use of reliable
and valid standardized measures enabled the present research study to assess psychological functioning in
an objective manner. Although the majority of these tests were not normed for prisoner populations, the
current reliability and validity findings increased our confidence in these measures.
POLICY IMPLICATIONS
Does this study legitimize the use of segregation with offenders, including those with serious and persistent
mental illness? Because this study may not generalize to other prison systems, especially those that have
conditions of confinement dissimilar to CSP, it is not possible to conclude that AS is not detrimental for all
offenders. Systems that are more restrictive and have fewer treatment and programming resources should
not generalize these findings to their prisons. Replication is needed to understand how increased services,
privileges, and out of cell time ameliorate the unintended consequences of AS, and research needs to inform
prison systems about the standards and practices necessary to protect inmates in segregation from harmful
psychological effects.
It is also important to note that there may be other negative consequences of AS that we did not study. For
example, Lovell, Johnson, and Cain (2007) found that inmates released directly from segregation to the
streets had dramatically higher rates and severity of detected recidivism than AS inmates who first released
to GP (but see Mears & Bales, 2009). We also did not study the degree to which AS met its purported goal of
changing inmate behavior for the better over time. The only questions addressed by this study were related
to psychological changes over time in segregation. Thus, we make no empirical or value judgments about
whether and to what degree the use of AS balances the benefits (e.g., a safer prison system) with costs (e.g.,
significant reductions in freedom).
ix
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
It is impossible to ignore the extremely disproportionate rate at which inmates with serious mental illness
are assigned to AS (Lovell, 2008; Metzner & Fellner, 2010; O’Keefe, 2008a), which has to some degree
“shocked the conscience” of the courts (see Jones ‘El v. Berge, 2001; Madrid v. Gomez, 1995; Ruiz v. John‐
son, 1999). In an era when prisons are expected to implement evidence‐based practices and to rehabilitate
offenders who will be releasing back to the community, is it enough to avoid harm? Must we ask ourselves
another question: what are the conditions required to improve inmates’ mental well‐being while in segrega‐
tion? Prison systems are held to a standard of treatment that is at least equivalent to community standards.
It is likely that this most difficult segment of society has failed at all levels of community treatment and ear‐
lier criminal justice interventions, but the quest to treat and improve services for the most needy is an im‐
portant reality facing corrections agencies.
Regarding their psychological functioning and levels of distress, these data suggest, although the differences
were small, that inmates with serious mental illness are less likely to improve in segregation and are less
likely to get worse compared to mentally ill inmates in GP. We do not assume that the reasons for these ap‐
parently contradictory findings are the same. For example, it is possible that fewer inmates with mental ill‐
ness get worse because segregation is a safer and more structured environment. On the other hand, hypo‐
theses regarding their unlikeliness to improve include the significant limitations that segregation places on
various types of therapeutic activities and services such as group therapy. Further, the data do not tell us
which aspects of AS prevent psychological improvement and deterioration, respectively, among inmates
with mental illness. However, since prisons have a constitutional duty to respond to serious medical (includ‐
ing psychiatric) needs, the possibility that segregation may prevent improvement is cause for concern and
further study.
There remain significant implications for mental health staff who work in prison systems that permit the
placement of mentally ill in long‐term segregation. It is critical for mental health staff to screen and assess
offenders prior to AS placement to determine their vulnerability to harm that might occur as a result of their
segregation. While in segregation, it is important that the mental status of all offenders be assessed on a
frequent, regular basis through rounds and individual sessions. Prison systems need to have a range of con‐
finement options, such that offenders who are at risk of or are showing signs of decompensation can be re‐
moved from segregation and placed in an alternative high security environment that permits greater out of
cell time and interaction with others.
Other systems have rejected confinement models that isolated offenders and held them in extremely re‐
strictive spaces. Even if the segregation models of the early 1900’s and the state psychiatric hospitals of the
mid‐19th century are viewed as “primitive” compared to modern‐day AS facilities, it is important to examine
and understand why these models failed and were ultimately dismantled. Although there are a number of
researchers who predict that there is no end in sight to the supermax model (King, 1999; Mears, 2008; Pizar‐
ro & Narag, 2008; Pizarro & Stenius, 2004), they have also raised empirical questions regarding their effica‐
cy. Questions about the efficacy of AS will be asked until more is known about whether the use of AS in pris‐
on systems improves conditions for the rest of the system, whether and how they improve inmate behavior
within and beyond the prison walls, whether they are cost‐effective, whether they increase risks to public
safety, and whether there are settings or individuals that are prone to psychological deterioration.
x
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
INTRODUCTION
One of the most widely debated topics in the field of corrections – the use of long‐term administrative se‐
gregation (AS) – has suffered from a lack of empirical research. The placement of offenders in AS environ‐
ments, particularly those with serious mental illness, has been a point of contention. Critics have argued
that the conditions of AS confinement exacerbate symptoms of mental illness and create mental illness
where none previously existed. The use of AS across the country has persisted as a corrections management
tool despite litigation, although in many states, the placement of mentally ill into AS is no longer permitted.
Empirical research has had little to offer this debate; the scant empirical research conducted to date suffers
from research bias and serious methodological flaws.
Now decades after the deinstitutionalization of states’ mental health hospitals, corrections agencies have
seen a surge of offenders with serious mental illness in their prisons. The rate of serious mental illness in the
community is 6% (National Institute of Mental Health, 2010). Among the incarcerated, the rate of serious
mental illness is tripled at about 18% (Ditton, 1999; O’Keefe & Schnell, 2008). A similar phenomenon is oc‐
curring within prisons, whereby a disproportionate rate of mentally ill are found within AS, estimated to be
50% higher than the rate within the general prison population (O’Keefe, 2008a). It is not known the extent
to which this difference is caused by the AS environment. Researchers have been unable to settle the ques‐
tion of whether these high rates of mental illness are caused by AS relative to the general prison population
or whether there is a selection bias such that offenders with mental illness, unable to adapt to general pris‐
on settings, are placed in AS at higher rates. This study seeks to advance the literature in this regard.
CHARACTERISTICS OF LONG‐TERM SEGREGATION
“Supermax” is the popular term used to describe the technologically advanced, supermaximum security
prisons designed for single‐cell occupancy that were rapidly being constructed across the nation during the
1990’s. Even when new construction was not possible, existing prisons were retrofitted to conform to this
new model. Therefore, a supermax facility may refer to an entire facility or a distinct unit within a facility
(National Institute of Corrections, 1997). Although there was a virtual explosion of supermax facilities over
the past two decades, similar units have operated on a smaller scale for decades (Zinger, Wichman, & An‐
drews, 2001).
The modern‐day supermax model is traced back to the U.S. Penitentiary in Marion, Illinois, that went into
permanent lockdown status in 1983. Prior to Marion, the Federal Bureau of Prisons operated solitary con‐
finement at the Alcatraz Island Prison until it closed in 1963. History points to even earlier uses of solitary
confinement including Pennsylvania’s Eastern State Penitentiary, which opened in 1829 and was later mod‐
eled in European prisons (Smith, 2008). However, these early models featured such extreme social isolation
and sensory deprivation (Cohen, 2008) and were so primitive that there is little comparison between them
and today’s modern supermaxes (National Institute of Corrections, 1999).
Across prison systems, different terms are used to describe the same concept: administrative segregation or
AS, control units, security housing units or SHUs, and security controls unit (Haney, 2003; NIC, 1999). In Col‐
orado, it is known as AS. Just as the names vary, so do the conditions. However, the defining feature that is
frequently associated with this model is single‐cell confinement for 23 hr per day, with 1 hr allowed out of
cell for showers and exercise. AS is differentiated from punitive or disciplinary segregation, which is a time‐
INTRODUCTION
1
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
limited punishment enforced for a prison violation pursuant a full due process hearing; placement in AS is an
administrative decision that often extends for an indefinite time period.
AS prisons are costly to build and operate due to costs associated with high security that include single‐
occupancy cells, high staff to inmate ratios, and technology (Mears & Bales, 2009). Because inmate move‐
ment requires multiple restraints and staff, many services are provided at the cell door, including meals, li‐
brary, mental health services, and programs. Newer AS facilities are equipped with advanced technological
equipment, which enables delivery of even more services to inmates in their cells (e.g., visitation through
videoconferencing) or within the facility (e.g., medical and dental procedures). Although technological
equipment is designed and used to reduce security breaches, it also increases the degree of isolation expe‐
rienced by inmates.
It is difficult to establish the number of inmates held in AS nationally. In 1999, King estimated that 1.8% of all
state prisoners were housed in AS. Although prevalence estimates are higher now than in 1999, prison sys‐
tems under‐report the actual use of AS, likely due to the negative connotation associated with the supermax
label used in national reporting (Naday, Freilich, & Mellow, 2008). For example, the Federal Bureau of Pris‐
ons reported no inmates in AS, protective custody, or supermax beds in 2008 (American Correctional Asso‐
ciation, 2009), which is inaccurate. Additionally, states reported drastically different numbers of offenders in
AS from year to year (see Naday et al., 2008). Given these limitations, it is estimated that at least 3.2% of all
state prisoners in 2008 were housed in AS or protective custody (American Correctional Association, 2009),
although this appears to be an under‐estimation of the true prevalence rates.
CRITICISMS OF THE AS MODEL
The use of AS has sparked a controversy resulting in considerable criticism of the prison system and its ad‐
ministrators. The limited number of research studies and the inadequacies of existing research on AS have
only fueled the controversy. Numerous researchers and forensic professionals have called for more research
to examine whether evidence based practices are in place and to examine whether harm is being done by
confining inmates to segregation (Kurki & Morris, 2001; Lovell et al., 2007; Mears, 2008; Metzner & Dvoskin,
2006; Pizarro & Narag, 2008), but the topics and setting are difficult ones in which to conduct research
(Mears & Watson, 2006; Naday et al., 2008).
One criticism has been the lack of evidence that segregation has achieved its intended goal of reducing vi‐
olence in the prison system (Kurki & Morris, 2001; Mears, 2008). There is some literature to suggest that
wardens and prison systems find this model to be effective in reducing violence and increasing order within
the larger prison system (Atherton, 2001; Mears & Watson, 2006; Ward & Werlich, 2003). However, these
studies lack the appropriate statistical controls to assert that the improvements are measurable and attri‐
butable to AS rather than merely perceptions of wardens or the result of other management controls also
put into place at the same time. In an empirical study of institutional violence in three states, Briggs, Sundt,
and Castellano (2003) did not find that AS reduced inmate‐on‐inmate violence. However, in a follow‐up
study, Sundt, Castellano and Briggs (2008) found that permanent reductions in inmate‐on‐staff violence
were attributable to the opening of an AS prison in Illinois.
Corrections departments have been moving towards evidence‐based models and practices to improve the
rehabilitation opportunities for offenders. These practices include standardized assessments, matching of‐
fender needs to services, cognitive‐behavioral programs, re‐entry services, structured decision making
2
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
guidelines, and intensive treatment programs such as therapeutic communities. By endorsing an administra‐
tive action to determine placement of inmates into AS, corrections agencies have moved away from the evi‐
dence‐based risk and classification instruments in favor of more subjective decisions. This criticism applies to
both the decision to place inmates in AS and their continuation in AS (Human Rights Watch, 2000; O’Keefe,
2008b; Pizarro & Narag, 2008).
An emerging concern is the return of offenders from AS to society, which may occur with little or no step‐
down process such that offenders are released directly to the streets from 23/7 confinement. Although the
adjustment required for offenders to adapt to rapid and extreme socialization changes is of concern, the
issue of public safety is perhaps of even greater concern. Research has indicated that AS inmates have high‐
er recidivism rates than non‐AS offenders (Mears & Bales, 2009; Motiuk & Blanchette, 2001; O’Keefe, 2005),
but this is likely due to the selection effects of who is confined to AS. When matching procedures were en‐
gaged, no differences in overall recidivism rates were found between AS and matched non‐AS inmates (Lo‐
vell et al., 2007; Mears & Bales, 2009). Mears and Bales (2009) found a small, but significant difference when
violent recidivism was the outcome measure rather than general recidivism; 24.2% of AS inmates had a vio‐
lent re‐offense compared to 20.5% of matched non‐AS inmates. Lovell et al. (2007) found that inmates who
released directly from AS had a higher recidivism rate than matched offenders who transitioned from AS to
a lower security facility prior to release. In contrast, Mears and Bales (2009) found neither a recency effect
(i.e., amount of time that elapsed between AS confinement and release) nor an exposure effect (i.e., total
amount of time spent in AS confinement) on recidivism rates.
Human rights concerns are tantamount to a discussion of the criticisms of the AS model. The use of AS has
been called a human rights violation, and some have even labeled it torture (Gawande, 2009; Metzner &
Fellner, 2010). Many find the conditions of solitary confinement to be excessively harsh and inhumane (Co‐
hen, 2008; Haney, 2003, 2008; Human Rights Watch, 1997, 1999, 2000; King, 1999; Kupers, 2008; Kurki &
Morris, 2001; Toch, 2001). Specifically, the lack of treatment, programs, and activities to engage the mind;
the restricted personal contact; lack of control over light and sound; lack of windows; and little or no access
to the outdoors are considered to be more extreme than is required for the safe operation of prisons. Addi‐
tionally, when people are held in highly restrictive environments where they have little control over their
life, there is a greater opportunity for staff to inflict abuses upon those confined within (Haney, 2008; Hu‐
man Rights Watch, 2000; Kurki & Morris, 2001).
The most significant issue is the question of whether prisoners are able to psychologically adapt to the con‐
ditions of AS. There is concern that mentally healthy individuals will decompensate in segregation, but re‐
cent discussions have centered on the placement of offenders with mental illness in such environments. Be‐
cause the harmful effects of AS is the central focus of this study, we will examine the evidence as it is availa‐
ble both in case law and in the research literature.
CASE LAW REVIEW
As is the case with many important issues that affect the correctional system, conditions of AS confinement
have been challenged in U.S. courts. In a pivotal First Amendment case heard in the Supreme Court, Turner
v. Safley (1987) set a standard for lower courts to evaluate the claims of prisoners such that deference is
given to prison administrators to set policies to ensure the safe operation of their prisons. Although the
Court’s decision does not prevent inmates from making claims against AS confinement, it limits the scope of
claims that they might successfully litigate to conditions that are needlessly harsh or unreasonable (Pizarro
INTRODUCTION
3
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
& Narag, 2008). Additionally, the Prison Reform Litigation Act of 1996 was enacted to restrict the filing of
prisoners’ cases in federal court. Consequently, most of the case law surrounding AS has been on the
grounds of a Fourteenth or Eighth Amendment violation (Collins, 2004).
Fourteenth Amendment
Under the Fourteenth Amendment of the U.S. Constitution, the state must adhere to certain procedures in
deciding to deprive inmates of their liberty interest (Collins, 2004). In Sandin v. Conner (1995), it was deter‐
mined that a liberty interest is created only when there is an “atypical and significant deprivation in relation
to the normal incidents of prison life.” In such cases, certain due process procedures are required.
Segregation that does not pose an atypical and significant hardship is not subject to due process, including
such confinement that may occur during a period of investigation into inmates’ misconduct (Jones v. Baker,
1998). However, in Wilkinson v. Austin (2005), it was decided that the plaintiffs’ due process and liberty in‐
terest had been violated because the combination of conditions were significantly more restrictive than
other Ohio state correctional facilities (e.g., isolation, lack of control over heating and lighting, no outside
recreation) and because of the length of confinement. The court upheld the Hewitt v. Helms (1983) decision
that these inmates were entitled to minimal procedural requirements, specifically timely notice of an AS
evidentiary hearing, reason for confinement, and sufficient opportunity for response.
Extended confinement in segregation without a review hearing was also determined to be a violation of the
Fourteenth Amendment. A New York court found that periodic review of inmates’ continued need for such
confinement is required (McClary v. Kelly, 1998).
Eighth Amendment
The Eighth Amendment ensures prisoners protection from cruel and unusual punishment. Because this con‐
cept is subjective, the Supreme Court has established the following standards:
(a) shocks the conscience of the Court, (b) violates the evolving standards of decency of a ci‐
vilized society, (c) punishment that is disproportionate to the offense, and (d) involves the
wanton and unnecessary infliction of pain (Collins, 2004, p. 106).
In examining the conditions of confinement, the totality of circumstances must be weighed; although each
individual condition might not be a violation, the combination of conditions might constitute one. Further‐
more, prison officials must demonstrate “deliberate indifference” to a prisoner’s basic human need in order
for there to be an Eighth Amendment violation.
The use of prolonged segregation was tested in three significant cases in California (Madrid v. Gomez, 1995),
Texas (Ruiz v. Johnson, 1999), and Wisconsin (Jones ‘El v. Berge, 2001). Long‐term segregation was not
deemed a violation, except in the case of inmates with serious mental illness where extended stays were
ruled unconstitutional. In Madrid v. Gomez (1995), not only was it ruled cruel and unusual punishment to
place mentally ill inmates in the SHU, those at reasonably high risk of suffering mental illness as a result of
SHU conditions were also restricted. Explicit in these cases is the requirement of correctional mental health
staff to screen, assess, and monitor offenders for mental illness or emerging symptoms resulting from their
placement in segregation.
It is also significant to note that in a number of states, settlement cases have also prevented or mitigated
the placement of inmates with serious mental illness into long‐term segregation. These states include Ohio,
4
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Connecticut, Indiana, New Mexico, New York, and Mississippi. Other cases, in states such as New Jersey and
Florida, have led to agreements to modify the terms under which prisoners with mental illness can be put or
kept in segregation (Jamie Fellner, personal communication, June 10, 2010).
RESEARCH REVIEW
There exists a large body of peer‐reviewed literature surrounding long‐term segregation and solitary con‐
finement. Many of these publications are literature reviews, theoretical articles, and case studies; few meet
the American Psychological Association (2009) standard of empirical study article defined as reporting on
original research or presenting new data analyses not addressed in previous reports, whether qualitative or
quantitative. For example, in the 2008 special edition of The Disturbed Offender in Confinement published by
Criminal Justice and Behavior, many of the nine articles focused on AS or other types of high security set‐
tings but only one (Lovell, 2008) presented an empirical study. Also in 2008, The Prison Journal released a
special issue entitled Supermax Prisons. Only two of the eight articles (Sundt et al., 2008; O’Keefe, 2008b)
meet the American Psychological Association standard for empirical research (2009). The large number of
articles and corresponding lack of empirical research reinforce this as an important area of forensic psychia‐
try in which it is very difficult to conduct viable research.
The entire body of literature has been critical to advancing our understanding of AS confinement and its re‐
lated issues. We relied on this literature to shape our hypotheses and research design in the present study.
The case study research in particular has been useful to illustrate problems that might be attributed to AS
(i.e., serious psychological harm) and highlight the need for research (see Benjamin & Lux, 1975; Human
Rights Watch, 1997, 1999; King, 1999; Kurki & Morris, 2001; Rhodes, 2004). However, there are serious limi‐
tations with case studies. Small sample sizes, as are the norm in case studies, mean findings may not gene‐
ralize to all, or even most, segregated offenders. Particularly concerning is that sampling procedures are of‐
ten not discussed, suggesting that special care was not taken to select a representative sample. Additionally,
these approaches do not provide a relative comparison of the participants’ behavior in other settings; in‐
mates who report serious psychological difficulties in segregation may experience those same problems in
other prison settings or in society. Because we are interested in conducting an empirical study, our review of
the research focuses on other empirical studies of the psychological effects of AS along with several key ar‐
ticles that informed our selection of psychological measures.
The SHU Syndrome
In 1983, Dr. Grassian described the psychopathological features resulting from AS that he believed to form a
clinical syndrome, which later became known as the SHU syndrome in the wake of Madrid v. Gomez (1995)
case. He interviewed 14 plaintiffs in a conditions‐of‐confinement lawsuit and described his clinical observa‐
tions resulting from those interviews. Grassian noted perceptual changes, affective disturbances, cognitive
difficulties, disturbing thought content, and impulse control problems that immediately subsided following
release from such confinement. In more recent research, Haney (2003) found elevated symptoms of psycho‐
logical trauma (e.g., anxiety, headaches, impending nervous breakdown, lethargy) and psychopathological
features (e.g., ruminations, social withdrawal, irrational anger) among 100 SHU prisoners as compared to
national probability samples. This constellation of symptoms composes the primary features of what has
been coined the SHU syndrome.
INTRODUCTION
5
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Quantitative Research
Well‐designed quantitative studies, although often not as rich in detail as case studies or qualitative re‐
search, can provide information about the impact of segregation on psychological well‐being through the
use of randomly sampled participants, representative samples, comparison groups, objective data collection
strategies, standardized procedures, and analytical strategies that account for random error. Research on
the effects of AS have been criticized for lacking these quality components that allow one to rule out plausi‐
ble alternative explanations (Arrigo & Bullock, 2008; Metzner & Dvoskin, 2006; Pizarro & Narag, 2008; Zinger
et al., 2001).
A key component that distinguishes research from demonstrations is the use of control or comparison
groups. Because of the lack of a comparison group, some frequently cited studies are actually demonstra‐
tions of the potential impacts of AS (e.g., Brodsky & Scogin, 1988; Haney, 1993; Grassian, 1983). In the sim‐
plest research design, a study will compare a “treated” group to a control or comparison group to determine
if the groups are different on the variable of interest. In a pure experimental design where participants are
randomly assigned to conditions (e.g., segregation, general prison population), differences between groups
would indicate the impact of segregation on the outcome variable; however in applied studies where ran‐
dom assignment to conditions is not feasible, the differences between the segregation group and a compari‐
son group may be due to segregation or to other uncontrolled factors. The quality of the comparison de‐
pends on the similarity between the control/comparison group and the experimental/treated group.
Several quantitative studies have used comparison groups to explore the impact of segregation on psycho‐
logical outcomes. Several of these studies have been experimental in nature in that inmates who volunteer
to be randomly assigned to either segregation or comparison conditions for a short period of time (e.g., Ec‐
clestone, Gendreau, & Knox, 1974; Gendreau & Bonta, 1984; Gendreau, Freedman, Wilde, & Scott, 1968,
1972; Gendreau, McLean, Parsons, Drake, & Ecclestone, 1970). These studies tend to show little impact of
segregation on mental well‐being but can be criticized for lacking ecological validity by using participants
who volunteered to be placed in segregation, using small samples sizes, and for being short‐term, all of
which do not match the current reality of how AS exists in U.S. prisons today. To demonstrate ecological va‐
lidity, conditions under investigation should reflect real life conditions. Similarly, comparisons to prisoners of
war or use of college students and inmate volunteers, lacks the ecological validity necessary to generalize
the findings to inmates in segregation.
Cross‐Sectional Designs
Non‐experimental research, which may demonstrate more ecological validity, have used a variety of com‐
parison groups including general, non‐inmate populations and norms (e.g., Haney, 2003; Hodgins & Côté,
1991; Suedfeld et al., 1982), general population prisoners (e.g., Hodgins & Côté, 1991), and inmates in dif‐
ferent security levels who report being in segregation or not ever experiencing segregation (Suedfeld et al.,
1982). Most, although not all, of these studies concluded that inmates in AS demonstrate higher levels of
psychological distress. Because the quality of the conclusions depends on the similarity between the com‐
parison group and the AS group, these cross‐sectional studies lack the ability to attribute these differences
to the conditions of confinement. In these studies, it is not possible to rule out alternative explanations due
to selection bias and potential pre‐existing differences, including psychological differences that may have
existed prior to entering AS (i.e., there has been an inability to establish the time precedence between AS
6
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
and psychological well‐being). An improved design strategy is to select a comparison group that has been
matched to the segregation group on important variables (e.g., Lovell et al., 2007; Mears & Bales, 2009).
Longitudinal Designs
In order to truly understand how AS impacts the well‐being of inmates, an improvement over cross‐sectional
design strategies with a comparison group is to study how inmates change over time using a longitudinal (or
repeated measures) design. Studying intra‐individual change allows for better understanding on whether
change occurs as well as explication of how change occurs. In longitudinal designs, individuals serve as their
own control group, and comparisons from baseline allow one to see how change is occurring. Adding a
comparison group in a longitudinal design will allow one to rule out additional alternative explanations
when change is (or is not) occurring.
There have been few longitudinal studies about the effects of segregation. Early studies by Gendreau and
colleagues (Ecclestone et al., 1974; Gendreau & Bonta, 1984; Gendreau et al., 1968, 1970, 1972) used re‐
peated measures experimental designs over periods of up to 10 days to explore the effects of segregation
on psychological and physiological measures. Few negative impacts of segregation were found over these
brief time periods. Although use of a repeated measures experimental paradigm improves over cross‐
sectional studies which may have selection bias issues, the short confinement periods are unrealistic for
providing information on the effects of segregation as it is currently being used in U.S. prisons.
Only two recent studies were found that followed inmates for longer time periods after placement in segre‐
gation (Andersen et al., 2000; Andersen, Sestoft, Lillebaek, Gabrielsen, & Hemmingsen, 2003; Zinger et al.,
2001). Andersen et al. (2000) studied participants over a 4 month period, but the majority of participants
had data for less than a month. Zinger et al. (2001) followed inmates over a 60 day period. Both of these
studies had high attrition rates (usually due to release from segregation), leading to a small percentage of
participants who had complete data. Attrition is a major problem in longitudinal designs both for generali‐
zability issues (i.e., are the participants who remain different from those who drop out) as well as analysis
problems for those methodologies which require complete data from all participants (e.g., analysis of va‐
riance techniques). Newer methodologies developed for studying intra‐individual change are less impacted
by attrition rates. Although conclusions from these studies are limited by methodological weaknesses, both
Andersen et al. (2000) and Zinger et al. (2001) demonstrated that segregated populations have more psy‐
chological disorders at the start than comparison subjects. However, these two studies provide conflicting
evidence on whether conditions get worse over time. Thus, further longitudinal studies are needed to sort
out these discrepancies and understand the long‐term impacts of segregation.
THE COLORADO SYSTEM
In Colorado at the time of this study, there were four designated AS facilities. Colorado State Penitentiary
(CSP) opened in 1993 as a 756‐bed male AS facility in its entirety. At the Sterling Correctional Facility, 192 of
its 2,545 beds were constructed to house male AS inmates in three units that are separate from the rest of
the facility. The San Carlos Correctional Facility (SCCF) is a male acute care psychiatric prison, with nine units
of varying security levels. One 26‐bed unit at SCCF is designated for AS classified inmates. Generally, AS at
SCCF is reserved for inmates already housed at SCCF needing high security or for inmates in AS at CSP or
Sterling Correctional Facility whose psychiatric needs exceed those available at their current facility. The
fourth AS facility is a 24‐bed unit located at the multi‐custody Denver Women’s Correctional Facility. Be‐
INTRODUCTION
7
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
cause it houses the largest number of AS inmates and no other custody levels, CSP was the only AS site in‐
cluded in this study.
The Colorado Department of Corrections (CDOC) had 25 state and 7 private prisons that managed 19,279
inmates as of June 30, 2007, which marked the start of data collection. There are five security levels – mini‐
mum, minimum‐restrictive, medium, close, and AS – to which offenders are assigned. CDOC uses a standar‐
dized, objective classification instrument that was developed specifically for the management of Colorado’s
inmate population (Austin, Alexander, Anuskiewicz, & Chin, 1995). The classification instrument is used to
assign inmates to minimum through close security levels. However, AS is a long‐term segregation placement
for inmates who display violent, dangerous, and disruptive behaviors and placement is determined through
an administrative action that is separate and distinct from both the usual classification system and the dis‐
ciplinary system. Although disciplinary infractions may affect classification at all levels, the disciplinary
process is a punitive response to a finding of guilt for an institutional rule violation and may result in punitive
segregation, which can extend up to 60 days. Therefore, punitive segregation is of short duration used for
punishment and AS is of long duration used for management purposes.
The administrative action to classify an offender to AS begins with a hearing, frequently following either a se‐
rious violation or a series of less serious infractions. Colorado does not house protective custody inmates;
therefore, no AS placements occur at the request of inmates. Also, during the study, newly arrived inmates
were not placed directly into AS upon intake into DOC except in rare cases for violent behavior in county jail or
for an interstate compact case transferred from AS in another prison system. Although the disciplinary system
only allows for punitive segregation following a finding of guilt, pre‐hearing segregation (removal from popula‐
tion) may occur immediately following a serious incident for the safety and security of the facility. Therefore, in
the time leading up to and during their AS hearing, inmates have typically been in segregation.
AS Offenders in Punitive Segregation
All facilities across the state of Colorado have punitive segregation beds with the exception of CSP and min‐
imum security facilities. Minimum custody offenders are transported to a higher security facility to complete
their punitive segregation time. When offenders are placed in punitive segregation, they are removed from
the general population (GP) and taken to an isolated part of the facility to be placed in a single cell. Punitive
segregation offenders remain in their cell for 23 to 24 hours a day, only coming out for recreation and
showers, both of which are located within the living unit. Therefore, most do not leave the unit during their
segregation time. Services including meals, library, laundry, and even medical and mental health appoint‐
ments occur at the cell door. If a situation warrants an offender to be out of cell, the offender is placed in
full‐restraints and escorted to a room within the unit where he or she can meet with staff privately. Many
offenders do not like being taken out of their cell unless absolutely necessary because of the use of full re‐
straints. Additionally, they may not like leaving their cell because officers may take the opportunity to search
the cell for contraband.
Due to the disciplinary nature of punitive segregation, offenders are stripped of most privileges during their
stay. Punitive segregation inmates are neither allowed to work nor are they permitted to participate in pro‐
grams or education. Furthermore, their televisions are removed, and they cannot order canteen beyond es‐
sential hygiene items.
8
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Punitive segregation is a highly restrictive environment, only intended to be used for a short period of time.
Once reclassified to AS, offenders may remain in a punitive segregation bed while waiting for an AS bed to
become available. This can be problematic as GP facilities are not designed to house offenders in long‐term
segregation and the small number of punitive segregation beds at each GP facility can fill up quickly. Fur‐
thermore, while punitive segregation offenders are not afforded privileges, AS offenders are granted limited
privileges such as visiting, which happens outside of the unit. Visitation is labor intensive because it requires
escort by two correctional staff. In addition, while being held at an AS facility, offenders who behave well
and complete their required programming and education are able to progress through a step program whe‐
reby they earn more phone sessions, visiting time, and privileges (e.g., TV, canteen). Only two punitive se‐
gregation facilities offer a step program for privileges, and there are none that provide the opportunity for
programming or education. This means that while AS offenders are held in a punitive segregation bed, they
are unable to begin working their way toward leaving segregation.
CSP Conditions of Confinement
Once an AS offender is moved from a GP facility and assigned to CSP, he is transported to CSP where he
completes his AS time. Offenders are taken into CSP through intake, which is located on the lowest level of
the facility. While in intake, offenders are placed in a holding cell that is similar to their permanent cell. Dur‐
ing this time, the offender watches an orientation video that outlines what he can expect and what is ex‐
pected of him during his time at CSP. He also has a brief visit from mental health, conducted at the cell door.
While the offender is going through orientation, property staff assesses his belongings to ensure that no un‐
allowable items enter with the offender, as they are permitted fewer property items than in GP facilities.
This also prevents dangerous contraband such as drugs or weapons from entering the facility. Once the of‐
fender has completed orientation, usually within the first few hours, he is escorted to his permanent cell in a
different area of the facility.
Physical Environment. CSP has six identical pods, or living units. When the offender enters the pod, he is es‐
corted down a long hallway that opens into a circular area. In the center of the area is a tower with an office
for housing unit staff on the lower level and the pod’s control center on the upper level. Officers manning
the control center operate all doors or sliders into the pod, including those to offenders’ cells. Correctional
staff standing in either the lower or upper levels of the tower can see into all eight of the day halls. Each day
hall contains 15 to 16 offender cells separated onto two tiers with each tier having 7 or 8 cells, a shower,
and a recreation room.
The cells in CSP are 80 square feet with 35 square feet of unencumbered floor space and contain a bunk,
toilet, sink, desk, and stool. Each of these items is made of metal and is mounted to the wall or floor for se‐
curity. Every cell has a 5” x 45” window on the exterior wall above the offender’s bunk through which the
offender can see outside. There is also a window on the cell door that faces the day hall. Depending on the
pod, the window is either 3.5” x 20.5” or 5” x 15”. Neither of these windows opens, which precludes the of‐
fender from receiving outside air while in his cell.
Per CSP policy, offenders wanting to participate in recreation are generally permitted at least one hour five
times per week (as well as to shower for 15 minutes three times per week which generally coincides with an
offer to exercise), assuming that there are no facility occurrences disrupting this schedule. When an offend‐
er is offered recreation and chooses to participate, he is placed in full‐restraints and escorted from his cell to
the recreation room at the end of the tier. The recreation room is a 90‐square foot cell that contains a pull‐
INTRODUCTION
9
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
up bar mounted to the wall. No other exercise equipment is allowed. The only opportunity offenders have
to receive fresh outside air is through two 5” x 60” grated windows on the exterior wall of the recreation
room. On the interior, a glass wall faces the V‐shaped day hall, so the offender in recreation is fully visible.
Though prohibited by the facility, an offender in the recreation room may call out exercises to other offend‐
ers who in turn workout in their cells.
There are light and sound standards for CSP. Standards for CSP require that ambient sound does not exceed
70dBA during the day or 45dBA at night. A sound measurement of offender housing units at CSP, on a single
day, returned an average of 55dBA at 7:50 AM and 42dBA at 10:40 PM. Although staff attempt to regulate
the ambient sound of the facility, it can be difficult to regulate the noise level of 756 offenders; these mea‐
surements do not reflect periods of sound elevations produced by inmates’ yelling and banging. Additional‐
ly, each offender is entitled to at least 20 foot‐candles of light in the desk area of his cell. A light measure‐
ment of offender cells returned an average of 55 foot‐candles of light in offenders’ cells. Offenders have two
32‐watt lights over the desk in each cell that they are able to control. In addition, each cell contains a 7‐watt
security light underneath the desk that stays on 24 hours per day.
Interpersonal Communication. Each cell has an intercom system through which correctional officers can con‐
tact each offender from the unit’s control center. Officers use the intercom system to ask prisoners ques‐
tions such as whether or not they want to attend recreation or take a shower. They also use the intercom to
inform inmates when they will be leaving their cell for such things as a mental health visit, a family or friend
visit, or if the offender will be escorted to another part of the facility or off grounds. Conversely, inmates can
use the intercom system by pushing a button in their cell to contact staff, which they may do to request
items (e.g., razor, toilet paper) or simply to chat. Staff also has the ability to monitor conversations using the
intercom system.
While the intercom system provides a means for correctional staff and offenders to communicate with each
other relatively easily, it does not afford offenders the opportunity to communicate with one another. Many
offenders at CSP have become skilled in sign language. Since each day hall is V‐shaped and cell doors have
windows, offenders are able to communicate with each other using sign language. This aids in keeping the
noise level down in the day hall and gives inmates the opportunity to speak to each other without the risk of
staff overhearing. At times, however, many inmates simply yell through their cell door so that other offend‐
ers can hear. When this happens, the day hall can become very noisy.
Due to the safety concerns of the facility and the fact that moving an AS offender from his cell is staff inten‐
sive, offenders in AS receive many services at their cell door. At CSP, officers make rounds every 30 minutes
to do a visual check into the cell of every offender. Mental health clinicians are required to do monthly
rounds as well. During these rounds, clinicians go to the cell door of every offender in their assigned pod and
check in with the inmate to see how he is doing. If the offender is well, the clinician moves on; however, if
the clinician feels the offender needs follow‐up, he or she will schedule an appointment with the offender
for a later time. This appointment will be conducted in the visiting room, not cell side. In addition to rounds,
offenders receive their library service and educational services at their cell door. Once a week, a librarian
picks up library kites, or requests, and distributes books and magazines to offenders who put in a kite the
previous week. When an offender is participating in programming or education, the teacher or counselor
distributes homework to each inmate through the cell door and also collects completed assignments in the
same manner.
10
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Mental Health Services. In addition to mental health services received cell‐side, offenders who are diag‐
nosed with a mental illness receive more in‐depth mental health services. Offenders with mental illness who
are stable are offered a one‐on‐one session at least once every 90 days. Those with acute mental health
needs are required to be seen at least once every 30 days. Although there are requirements on mental
health staff to schedule appointments, offenders may refuse these appointments. Conversely, if an offender
feels his mental health status has changed since his last monthly round, he may put in a request to see a cli‐
nician sooner than scheduled. If necessary, clinicians will schedule an offender for a mental health session
for 1 to 2 hours per week as they are available; this is infrequent but most likely to occur following a crisis
event. Additionally, if a mental health clinician feels a prisoner requires psychotropic medication, an ap‐
pointment is made for him to meet with a psychiatrist. This visit may happen in a noncontact visiting booth
or via teleconference.
Mental health appointments occur in a noncontact booth in the visiting room, unless the offender has de‐
clared a mental health emergency. If an offender has threatened self‐harm, he is often taken to intake and
placed in a special observation cell where he is stripped of his belongings and can easily be observed by staff
for his safety and the safety of staff. An offender is kept in the observation cell until the clinician can make a
reasonable assumption that the offender no longer plans to self‐harm or for 72 hours, whichever comes
first. If the clinician determines the offender needs to be observed beyond 72 hours, approval is needed
from administrators and a mental health supervisor outside of the facility. Offenders who remain in a men‐
tal health crisis situation beyond the three to five day window are then sent to the infirmary at a different
facility. There are generally four to six mental health clinicians who are responsible for managing the mental
health needs of offenders at CSP. When the facility is fully staffed with six clinicians, each is assigned to a
pod of 126 offenders, but when there are vacant positions, clinicians are required to cover their pod’s men‐
tal health needs and split an additional pod with another clinician.
Quality of Life Program. When an offender arrives at CSP, his length of stay is indeterminate because it is
based upon his behavior and ability to comply with programming requirements. The average length of stay
at CSP is two years (O’Keefe, 2005). CSP provides incentive‐based behavior modification and cognitive pro‐
grams. Every offender must successfully complete three cognitive classes with each lasting three months.
Successful completion of the required programming along with modeling appropriate behavior is the prima‐
ry way for an offender to work his way out of CSP. The goal of these programs is to provide offenders with
tools so they may be successfully reintegrated into lower security prisons.
CSP’s incentive‐based programming consists of three quality of life (QOL) levels. Each level brings with it
more privileges; however, these privileges must be earned by the offender through appropriate behavior
and compliance with CSP rules. Each level has a prescribed minimum number of days: 7 for level one, 90 for
level two, and 90 for level three. Because offenders are required to complete three 90‐day cognitive courses
and there are often program waitlists that may result in an offender staying on levels two or three for longer
than 90 days, the total program length is expected to last a minimum of one year. Additionally, offenders
who misbehave may be regressed through the levels, extending their time in the program.
QOL level one is much like punitive segregation in that offenders are not permitted to have a television or to
participate in programs or work. Furthermore, offenders at this level are only allowed one 20‐minute phone
session and one 2‐hour noncontact visit per month, should they remain at level one for that length of time.
They are able to order items from the canteen at a maximum of $10 per week. Though limited in compari‐
INTRODUCTION
11
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
son to what GP offenders are able to buy, offenders at CSP have a variety of food, hygiene, faith, and per‐
sonal items available for purchase. Additionally, AS inmates are permitted three library books at any given
time. All other property must fit inside a 2 cubic feet duffle bag; however, as long as the property can fit in
the bag, they are allowed two personal books, two magazines, and one newspaper. Other items that of‐
fenders at this level may have are photographs and an address book.
Level one offenders are automatically reviewed on their seventh day by the unit sergeant. Unless they re‐
ceived a negative write‐up or report, offenders automatically progress to level two after seven days. Those
who do not progress are extended another seven days until their next review.
Inmates become eligible for cognitive classes when they have been elevated to QOL level two. Offenders at
this level are allowed a television, and if they are indigent and unable to afford one, the facility will loan one
to them. This is beneficial as CSP provides some of their programs and recreational activities through the
television. Offenders also receive 20 television channels that they are able to control from their cell and
view at their leisure 24 hours per day. Also available through the television is a music channel that plays at
designated times and rotates through music genres. Additionally, offenders have the opportunity to play
bingo on a monthly basis. The bingo numbers are selected and aired over closed circuit TV at the facility.
Offenders who wish to participate in Bingo receive six board games and are awarded a candy bar for each
verified bingo.
In addition to programming received through their television, offenders at this level are permitted art sup‐
plies (colored pencils, art paper, drawing patterns, and coloring pictures), games (solitaire and kings table),
puzzles (crossword, word fill‐ins, word search, and Sudoku), and pamphlets for in‐cell exercises offenders
(push‐ups, stretching, and isometrics). Offenders may request a new supply of colored pencils every six
months and are able to receive four new sheets of art paper and new puzzles on a weekly basis.
At level two, offenders are permitted to increase their weekly canteen order to $20 and have an increase in
both their phone privileges, to two 20‐minute phone sessions per month, and their visiting privileges, to two
2‐hour noncontact visits per month. However, offenders at this level remain unable to work. Once an of‐
fender has completed a minimum of 90 days on level two, has been compliant with programming, has not
had any negative write‐ups for at least 90 days, has had appropriate interaction with staff, and has sustained
suitable cell conditions, he may be progressed to QOL level three. Offenders’ case manager initiates the pa‐
perwork for a level progression, which requires approval by the housing captain.
Arguably one of the most important benefits of QOL level three is an offender’s ability to have more contact
with friends and family. While offenders’ visits remain noncontact, they are increased to four 3‐hour visits
per month and four 20‐minute phone sessions. Offenders are also permitted to order as much as $25 worth
of canteen per week. One additional benefit is that offenders may now be eligible to work as a porter or
barber. There are 54 positions available to offenders at CSP. Benefits to being offered a job position include
the ability to earn money, increased time out of their cell, and two additional phone sessions per month.
However, simply being at QOL level three does not automatically qualify an offender for a job. If intelligence
officers feel the offender is a threat to the facility, he will not be permitted to work as a porter or barber. A
QOL level three inmate may be deemed a continued threat due to an institutional history of assaultive be‐
havior or from intelligence that suggests he may use the opportunity to intimidate or pass gang information
or contraband to other offenders. There are no time limits restricting how long an offender can be in a job
position and there is a waitlist of offenders who have put in a request to work; however, because of the
12
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
progression of offenders out of CSP and offenders who their work privilege, the same 54 offenders are not
usually working for more than a couple of months.
The restrictions inherent in AS diminish staff’s ability to impose traditional sanctions for institutional rule
violations. Offenders in GP who are found guilty of a rule violation can receive a maximum of 60 days in pu‐
nitive segregation or up to 180 days in loss of privileges (e.g., TV, canteen, visiting). Offenders in AS are re‐
quired to follow the same institutional rules as GP offenders; however, because AS offenders are already in
segregation for an indeterminate amount of time, they cannot receive additional segregation time as pu‐
nishment. They can still lose privileges and be regressed through the QOL system. Sanctions are tailored to
the seriousness of the infraction.
Offenders who engage in minor rule violations may initially receive a warning that is documented in a chro‐
nological record report. If the behavior continues, the offender may lose a privilege for a short amount of
time (e.g., three days) without losing a QOL level. For example, this may happen if an offender covers the
security light in his cell to make it darker for sleeping. This may also happen if an offender is caught “rat lin‐
ing” or “fishing,” which are forms of communication or exchange of items between offenders locked in their
cells.
When an offender in AS violates a serious institutional rule, the officer initiates documentation on which he
may recommend that the offender be regressed to a lower QOL level. This recommendation is approved or
denied by a housing lieutenant. In general, if the lieutenant approves the offender’s level regression, he is
dropped one level. This process is kept separate from the disciplinary process, which may or may not result
in a guilty finding, in order to have an immediate response to an offender’s negative behavior. The discipli‐
nary process can be lengthy because of due process requirements, but he may also receive a loss of privileg‐
es sanction through the disciplinary process.
Regardless of the offender’s level, if he engages in behavior that dangerously disrupts the operation of the
facility, he will be placed on special controls in the intake unit where he can be carefully monitored. This of‐
ten happens during what is referred to as a use of force incident, which is any time an officer uses any level
of force against an offender. A use of force incident generally occurs when an offender assaults a staff
member or refuses to comply with a lawful order (e.g., refuses to be restrained for escort). An officer’s re‐
sponse can include the use of simple pressure point tactics, the use of agents such as oleoresin capsicum
(OC), or a forced cell extraction of the offender. During both fiscal years 2008 and 2009, CSP had an average
of seven use of force incidents per month. Upon an offender’s return to his cell, he will automatically begin
at QOL level one again. Though the offender will not be required to retake any of the cognitive classes that
he has already completed, he will be terminated from any classes in which he is currently enrolled and will
be required to begin his process through the QOL levels again. Additional sanctions may be imposed through
the disciplinary process.
Offenders who have difficulty progressing through the QOL level system may require special consideration.
Offenders in segregation can accumulate a high number of sanctions through behaviors such as breaking the
sprinkler head in their cells or overflowing the toilet in their cells, causing flooding on the tier. It is difficult to
manage and change the behavior of offenders who have so many sanctions that there is no tangible incen‐
tive to improve their behavior. When this is the case, case managers and housing staff enact a behavior
management plan. In a behavior management plan, case managers and correctional staff will use one privi‐
INTRODUCTION
13
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
lege (e.g., TV) that is highly valuable to the offender as an incentive. If the offender can behave well for a
short period of time (generally 7 to 10 days), he may receive a television despite his loss of privilege status.
When an offender has been at level three for at least 90 days with good behavior and has successfully com‐
pleted the requirements of the program, he is interviewed for progression out of CSP. A classification com‐
mittee must approve the decision to reclassify him to close custody, and then he is moved to the Centennial
Correctional Facility when a bed becomes available, where he continues to work toward completing his rein‐
tegration programming. It is less common that an offender transitions out of CSP any other way; however,
offenders do sometimes parole from CSP or release when they reach the end of their sentence while in AS.
Additionally, an offender may be released from AS based on a warden’s review. An offender may receive a
warden’s review if he has been in CSP for more than two years but has been unable to progress out of CSP.
If it is felt that the offender no longer needs to be in CSP, he may be released back to GP without transition‐
ing through Centennial Correctional Facility.
Progressive Reintegration Opportunity (PRO) Unit. At the Centennial Correctional Facility, there is a contin‐
ued focus on behavior modification and cognitive programs to transition disruptive offenders to less secure
environments. Most offenders complete QOL levels four through six in the PRO unit. Upon transfer from
CSP, offenders are reclassified from AS to close custody, the next highest custody level. Upon arrival, little is
different for the newly classified close custody offenders; however, as offenders work their way through the
PRO unit levels, they work toward contact visits with friends and family and are eventually allowed
recreation time in the gym with other inmates. Ultimately, offenders who are successful in completing all six
QOL levels are released back to GP.
Offenders with Mental Illness (OMI) Management Program. During the course of the research project, the
OMI management program was opened at Centennial Correctional Facility. In addition to the PRO unit, the
OMI program was designed to be a transitional program from CSP specifically for prisoners with a mental
illness. Offenders are selected for the OMI program by a multi‐disciplinary committee and must be approved
for reclassification as a close custody inmate. In order to be considered for transfer to the program, offend‐
ers must have been in AS for a minimum of six months, enrolled in a cognitive program, have a mental
health disorder, and be actively working with a mental health clinician.
Upon transfer to the OMI program, inmates are automatically placed in the intermediate program level. The
OMI program has three levels: high, intermediate, and low. High is the most restrictive level with low the
least. Depending on the individual’s behavior, he can be moved to high or low levels. The program focuses
on treatment and socialization. Offenders in this program work their way toward earning more privileges
than are available in AS, contact visits with friends and family, and recreation in the gym with other inmates,
much like PRO unit offenders do; however, the OMI program has the added benefit of group therapy. Initial‐
ly, offenders are afforded the opportunity to participate in group therapy by being tethered to a special ta‐
ble. As offenders progress through the program, they are eventually allowed in groups of eight untethered
inmates. The goal is to transition offenders to GP or the community although placements in the program
may be long term.
San Carlos Correctional Facility (SCCF)
SCCF is a 255‐bed special needs prison designed to stabilize and treat offenders with the most acute psy‐
chiatric symptoms or with developmental disabilities who are at risk for self injury as a result of their illness
14
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
and who have shown a substantial impairment in their ability to function at another correctional facility.
SCCF houses inmates at all five custody levels. SCCF is unique in that offenders of all custody levels live and
interact with one another on their living unit, with the exception of AS offenders who are housed in a sepa‐
rate unit.
All offenders who arrive at SCCF are processed through the intake/assessment unit. New arrivals are inter‐
viewed by both the mental health clinician assigned to the unit and a psychiatrist. New offenders are not
permitted interaction with other inmates for the first 72 hours. If after 72 hours, clinicians and correctional
staff feel the offender can reasonably interact with other offenders, he will be allowed in the day hall with as
many as five other inmates. Offenders on the intake unit are permitted out of their cell in the day hall for at
least one hour a day, five days per week. During this time, they have open access to the phones and show‐
ers.
Offenders typically progress through the programming levels as their mental health status improves. Of‐
fenders are continually monitored by a psychiatrist with an appointment every 30 days for the most severe
offenders or every 60 days for those who are progressing well. One‐on‐one sessions occur with a mental
health clinician as needed and are not scheduled on a regular basis; however, there is a clinician assigned to
each unit, with each unit housing fewer than 30 inmates.
As an offender continues to progress through the facility, he will work his way towards open access to the
day hall, phones, and shower. As he progresses through the facility he will then be allowed out with seven
inmates and then fifteen, eventually earning all day open access. Those who have progressed to the lowest
levels are also permitted one hour of recreation five days per week in the yard or gym plus three hours per
week at the library. Additionally, they are able to participate in group therapy sessions, which happen once
or twice a week depending on the topic. Group therapy subjects include anger management, understanding
one’s mental illness, and other related topics. Once mental health, psychiatric, and correctional staff deter‐
mine that the offender has improved enough to function in GP, he is then transferred to a facility at his cus‐
tody level.
PURPOSE OF PRESENT STUDY
The broad purpose of the project was to evaluate the psychological effects of long‐term segregation on of‐
fenders, particularly those with mental illness. This study examined conditions as they existed in the Colora‐
do prison system with respect to AS, using CSP as the AS study facility. Only males were included because
females represent 2% of Colorado’s AS population. We did not assign inmates to segregation, but studied
those conditions as they naturally occurred. The following were the primary goals and hypotheses of the
grant.
Goal 1: To determine which, if any, psychological domains are affected, and in which direction, by the differ‐
ent prison environments. A multitude of psychological dimensions were examined, drawing from those most
often cited in the literature. The broad constructs of interest were depression/hopelessness, anxiety, psy‐
chosis, withdrawal and alienation, hostility and anger control, somatization, hypersensitivity, and cognitive
impairment. We hypothesized that offenders in segregation would develop an array of psychological symp‐
toms consistent with the SHU syndrome, with elevations across the eight constructs.
INTRODUCTION
15
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Goal 2: To assess whether offenders with mental illness decompensate differentially from those without
mental illness. We were particularly interested in whether long‐term segregation had a differential impact
based on the presence of mental illness in offenders. We sought answers to the following questions: Does
AS exacerbate symptoms in offenders with mental illness? Does AS create symptoms of mental illness in
those who did not exhibit any at placement? It was hypothesized that offenders with and without mental
illness would deteriorate over time, but the rate at which it occurred would be more rapid and more ex‐
treme for the mentally ill.
Goal 3: To compare the impact of long‐term segregation against the general prison setting and a psychiatric
care prison. In this study, the psychological and behavioral symptoms of offenders in AS were compared to
similar offenders who were sent to SCCF or returned to the general prison population pursuant an AS hear‐
ing. This study used a repeated measures design over the course of a year to explore whether psychological
distress was attributable to the various prison environments. It was hypothesized that inmates in segrega‐
tion would experience greater psychological deterioration over time than the comparison groups.
This study also included an examination of individual characteristics such as mental health status, personali‐
ty, and trauma history to determine if certain factors could predict patterns of change. The prediction ana‐
lyses were exploratory in nature and we did not formulate a hypothesis about the variables that might pre‐
dict differential rates of psychological decompensation.
16
INTRODUCTION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
METHOD
GROUP ASSIGNMENT
Study participants included male inmates placed in AS and comparison inmates in the GP. Placement into AS or
GP conditions occurred as a function of routine CDOC operations, pending the outcome of their AS hearing, with‐
out involvement of the researchers. Inmates were identified as study candidates at the point the offenders were
notified that they would have an AS hearing. Oftentimes, it was unknown whether a particular inmate would be
placed in AS or returned to GP at the time of his study consent; approximately 10% of hearings do not result in AS
placement. For the purposes of this study, all study participants classified to AS were waitlisted for and placed in
CSP (as opposed to Sterling Correctional Facility). Inmates who returned to GP following an AS hearing were as‐
sumed to be as similar as possible to AS inmates and, therefore, comprised the comparison groups. Comparison
participants also included inmates targeted for a diversionary program that identified inmates at high risk of AS
placement due to their disruptive behavior. This program discontinued shortly after the study commenced,
hence few participants were identified through this method.
Inmates in both of these settings (CSP, GP) were divided into two groups – offenders with mental illness (MI) and
with no mental illness (NMI). There are fewer inmates with mental illness than without, but because both sub‐
groups were of equal interest to this study, separate groups enabled over‐selection of inmates with mental ill‐
ness. All offenders are rated on a psychological needs level by trained clinicians upon intake into CDOC and pe‐
riodically during their incarceration as warranted. The psychological needs level has a 5‐point rating, where high‐
er values indicate the need for more intensive services, and a qualifier code that indicates whether the offender
has a serious and persistent mental disorder. Most inmates rated 3 through 5 have an Axis I diagnosis, although
certain Axis II diagnoses may infrequently warrant this rating (e.g., borderline, schizotypal). Disorders that typical‐
ly qualify as serious and pervasive are mood disorders including major depression, other depressive disorders,
dysthymic, and bipolar disorders; psychotic disorders including schizophrenic, paranoid, delusional, and schizoph‐
reniform disorders; dissociative identity disorder; and posttraumatic stress disorder. In this study, inmates as‐
sessed with a psychological needs level of 3 through 5 were defined as MI and levels 1 or 2 were defined as NMI.
A third comparison group was included. This group included inmates with severe mental health problems placed
in SCCF. Of the inmates placed in SCCF, only those with patterns of prison misbehavior, as measured by discipli‐
nary violations, were included in the study. However, inmates placed into AS at SCCF were excluded because of
the small number and because many had transferred from AS at CSP or Sterling Correctional Facility, where the
effects of the earlier placement would be unknown. The purpose of the SCCF comparison group was to study in‐
mates with serious mental illness and behavioral problems who were managed in a psychiatric prison setting.
Figure 1 illustrates the number of offenders who were eligible for the study and details the selection of offenders
within each of the five study groups. Given that the purpose of this project was to study long‐term segregation,
inmates projected to release from prison before administration of the final testing session were excluded. In‐
mates were also excluded if they could not read English or if their reading level was not high enough (roughly
eighth grade) to complete the battery of tests. SCCF inmates were excluded if they did not have significant discip‐
linary violations in their history. Infrequently, offenders were excluded for other reasons such as being an inter‐
state compact offender, being the suspect in a high‐profile murder investigation (as reason for AS placement), or
a visual impairment prohibiting them from reading. Finally, inmates were sampled as a matter of convenience.
Because this project funded only one field researcher, participants were selected based on their proximity by ei‐
ther timing or location to others who could be included in this study.
METHOD
17
18
1 Other
10 Other
16 Location
9 Other
28 Violation
H istory
78 Sentence
Length
Approached Inmate to
Participate in Study
71
17
15 Other
7 Other
10 Location
51 Sentence
Length
203 MI
203 Placed in
SCCF
Not Selected
125 Location
17 Location
22 Sentence
Length
220 Return to GP or
diversion placement
88
73 Sentence
Length
■ 64 Sentence
Length
664 Decision to Ad
Seg
884 Ad Seg hearing or
diversion placement
# Eligible
Excluded from Participating
Mental Illness Designation
CDOC Decision
At Risk of Ad Seg Placement
Figure 1. Eligibility and Selection of Study Participants
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
PARTICIPANTS
Figure 2 details the flow of participants through the study, including an account of how many offenders
completed the testing at each interval. A total of 302 male inmates were approached to participate in the
study. Thirty refused to participate. Two more offenders were considered a passive refusal and were re‐
moved for inappropriate sexual behavior towards the researcher during the first testing session. An addi‐
tional 23 offenders later withdrew their consent, although the data collected to the point of their withdraw‐
al was used. In addition to refusals and withdrawals, 10 inmates released prior to the end of the study due
to discretionary releases by the Parole Board and one GP participant died of a drug overdose.
Five testing sessions were initially established at 3‐month intervals, beginning with the date of consent and
initial administration. Therefore, tests were scheduled at 3 months, 6 months, 9 months and 12 months af‐
ter the baseline assessment. However, this schedule was problematic for the CSP groups. When the study
began, there was a 3‐month average wait for inmates to be transferred to CSP due to a shortage of AS beds.
While on the waitlist, AS inmates were held in a punitive segregation bed at their originating facility. It was
determined that the primary goal was to study inmates in a single long‐term segregation facility (CSP) to lim‐
it confounding variables and that therefore the baseline measure should be collected upon placement into
CSP. However, it was also recognized that significant changes could occur while inmates were held in segre‐
gation at their originating facility. Therefore, a “pre‐baseline” measure was collected as close to the AS hear‐
ing as possible, which meant that the CSP groups completed six test intervals rather than five. The time be‐
tween the pre‐baseline and baseline measure varied according to how long the inmate was on the waitlist.
The median time between pre and baseline tests was 99 days, although eight offenders were moved into
CSP so quickly that they did not have a pre‐baseline measure. In the analyses, tests were aligned across
groups according to the test number, such that the CSP groups had an additional test at the end rather than
at the beginning.
Participants’ ages ranged from 17 to 59 at the time of consent, with a mean age of 31.8 (SD = 9.1). The ra‐
cial/ethnic breakdown of participants was 40% white, 36% Hispanic, 19% African American, 4% Native Amer‐
ican, and 1% Asian. Of the inmates with mental illness who were included in this study, 56% were identified with
a serious and pervasive disorder. Other participant characteristics are described in greater detail in the results
section, including comparisons of study samples to eligible pools and comparisons of refusers to partici‐
pants.
MATERIALS
Assessment tools were selected to comprehensively cover the variety of psychological constructs associated
with AS (e.g., Arrigo & Bullock, 2008; Grassian, 1983; Haney, 2003). The primary constructs assessed in this
study were as follows: (1) anxiety, (2) cognitive impairment, (3) depression/hopelessness, (4) hostility/anger
control, (5) hypersensitivity, (6) psychosis, (7) somatization, and (8) withdrawal/alienation. Additionally, ma‐
lingering, self‐harm, trauma, and personality disorders were assessed.
Research materials were selected to meet the following criteria: (1) use of assessments with demonstrated
reliability and validity, (2) use of multiple sources for providing information (e.g., self‐report, clinician rat‐
ings, files), (3) use of multiple assessments of each construct of interest, (4) ability to use within the prison
setting, and (5) ease of administration, including no specialized equipment, no physical contact, short length
of time, and appropriate reading level.
METHOD
19
20
Test 6
Test 5
Test 4
Test 3
Test 2
Completed Test 1
Consented &
1 Released
51 Completed
2 Withdrew
2 Missed Test
2 no pre-baseline
54 Completed
56 Completed
2 Withdrew
1 Withdrew
56 Completed
56 Completed
1 Released
1 Withdrew
1 Missed Test
29 Completed
1 Withdrew
2 Released/Died
29 Completed
1 Withdrew
32 Completed
1 Released
38 Completed
1 Released
1 Withdrew
39 Completed
41 Completed
1 Released
2 Withdrew
59 Completed
1 Released
3 Withdrew
1 Missed Test
61 Completed
2 Missed Test
64 Completed
4 Refused
67 Consented
SCCF (n = 71)
57 Completed
2 Withdrew
41 Completed
11 Refused
43 Consented
GP NMI (n = 54)
1 Withdrew
1 Missed Test
32 Completed
0 Refused OF
33 Consented
GP MI (n = 33)
3 Withdrew
1 Missed Test
59 Completed
60 Completed
1 Withdrew
1 Missed Test
60 Completed
2 Withdrew
62 Completed
12 Refused
63 Consented
64 Consented
5 Refused
CSP NMI (n = 75)
(N = 302)
Approached to Participate in Study
CSP MI (n = 69)
6 no pre-baseline
Figure 2. Flow of Participants through Study
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
After selection of the self‐report assessments was complete, there remained several areas of interest (e.g.,
panic disorder, hypersensitivity to external stimuli, physical hygiene) for which there was no established
measure that met our criteria. In conjunction with the study advisory board, the research team developed a
39‐item instrument to assess these areas. This instrument, called the Prison Symptom Inventory (PSI), is
shown in Appendix A.
In addition to self‐report assessments, ratings of psychological functioning were obtained from clinical staff
and ratings of behavior in the housing unit were obtained from correctional staff. Official record data were
also gathered from electronic and paper files. This section summarizes information for self‐report assess‐
ments, staff ratings, and behavioral data. Complete descriptions of the individual measures and their known
psychometric properties from past research and for the current study are provided in Appendix B. Additional
analyses of the psychometric properties of the PSI are presented in Appendix C.
METHOD
C
C
C
C
I
I
I
I
I
I
I
I
I
I
Trauma
Self‐Harm
C
I
Personality
Disorder
Malingering
Somatization
Psychosis
I
Withdrawal –
Alienation
Beck Hopelessness Scale (BHS)
Brief Psychiatric Rating Scale (BPRS)
Activity
Anxious‐Depressed
Hostility/Suspiciousness
Thought Disorder
Withdrawal
Brief Symptom Inventory (BSI)
Anxiety
Depression
Hostility
Interpersonal Sensitivity
Obsessive‐Compulsive
Paranoid Ideation
Phobic Anxiety
Psychoticism
Somatization
Coolidge Correctional Inventory (CCI)
Deliberate Self‐Harm Inventory (DSHI)
Personality Assessment Screener (PAS)
Acting Out
Alienation
Hypersensitivity
Anxiety
Table 1. Assessments and Constructs
Cognitive
Impairment
Depression –
Hopelessness
Hostility – Anger
Control
Data were collected directly from participants on 12 self‐report assessments (ten paper‐and‐pencil tests,
two administered by the researcher) to assess 12 different constructs. Table 1 provides a list of the assess‐
ment tools for each construct. Most assessments were collected at each testing period, although personality
disorders, self‐harm, and trauma history were not collected at all time periods. It was determined that per‐
sonality and trauma history were relatively stable constructs that needed to be assessed only once to limit
the testing burden on study participants. Also, due to the burden on already limited mental health re‐
sources, the BPRS was only administered at the first, third, and fifth testing intervals.
I
I
21
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
I
I
I
I
I
I
O
O
O
I
I
I
I
I
R
R
Trauma
I
Self‐Harm
Withdrawal –
Alienation
Somatization
Psychosis
I
Personality
Disorder
I
O
I
I
I
I
Malingering
Anger Control
Health Problems
Hostile Control
Negative Affect
Psychotic Features
Social Withdrawal
Suicidal Thinking
Prison Behavior Rating Scale (PBRS)
Anti‐Authority
Anxious‐Depressed
Dull‐Confused
Prison Symptom Inventory (PSI)
Panic Disorder
Hypersensitivity/External Stimuli
Physical Symptoms
Profile of Mood States (POMS)
Anger‐Hostility
Depression‐Dejection
Fatigue‐Inertia
Tension‐Anxiety
St Louis Univ Memory Scale (SLUMS)
State‐Trait Anxiety Inventory (STAI)
State Anxiety
Trait Anxiety
Structured Inventory of Malingered
Symptomatology (SIMS)
Trail Making Test (TMT)
Trauma Symptom Inventory (TSI)
Hypersensitivity
Anxiety
Cognitive
Impairment
Depression –
Hopelessness
Hostility – Anger
Control
I
Note. C = Clinician rating; I = Inmate self‐report; O = Officer rating; R = Researcher administered. Shaded tests not administered at
every testing interval.
Self‐Report Assessments
A composite score was developed for seven of the eight primary constructs by standardizing scores from the
scales on the self‐report assessments. Standardized scores were used so that comparisons between con‐
structs could be made more easily and to create a single measure for constructs assessed by multiple self‐
report assessments. Scores were standardized by centering on the mean of the entire sample at the first
assessment and dividing by the standard deviation. A composite score was computed by standardizing each
assessment and averaging the standardized scores across the individual assessments as the composite
score. Reliabilities for these composites are presented in the discussion of each construct.
Anxiety Construct. Anxiety was measured by eight self‐report variables assessed at each time period. The
self‐report measures used to create the anxiety composite score were the State and Trait subscales from the
State‐Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, & Lushene, 1970); the Obsessive‐Compulsive, An‐
xiety, and Phobic Anxiety subscales from the Brief Symptom Inventory (BSI; Derogatis, 1993); the Negative
22
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Affect subscale from the Personality Assessment Screener (PAS; Morey, 1997); the Tension‐Anxiety subscale
from the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1992); and the Panic Disorder subscale
from the PSI. The following PSI items were included on the Panic Disorder subscale: 2, 6, 10, 13, 16, 17, 20,
25, and 30.
Internal consistency reliabilities were computed for each assessment period for the entire sample. The mean
Cronbach’s alpha across individual anxiety measures and time periods was .87 (range = .60 to .95). The
Cronbach’s alphas for the composite ranged between .89 and .91 for the six time periods. Reliabilities were
similar across testing intervals, and they were similar to internal consistency estimates from normative sam‐
ples. Test‐retest correlations between sequential time periods ranged between .49 and .86 (M = .76) indicat‐
ing reasonable stability over 3 month assessment periods. The validity coefficients between self‐
assessments of the anxiety construct indicated evidence for convergent validity, with correlations between
measures ranging from .36 to .85 (M = .65) across all time periods.
Cognitive Impairment Construct. Cognitive impairment was assessed by two individually administered tests.
The Saint Louis University Memory Scale (SLUMS; Tariq, Tumosa, Chibnall, Perry, & Morely, 2006) was used
to assess orientation, memory, attention, and executive function. The SLUMS is an 11‐item scale and yields a
single total score ranging from 0 to 30, where higher scores indicate stronger cognitive abilities. The Trail
Making Test (TMT; Reitan, 1958) was used to assess attention. The time required to complete the A (con‐
nect sequential numbers) and B (connect alternating numbers and letters) tasks were collected, and the ra‐
tio of times (B/A) was used as the attention measure in subsequent analyses.
The SLUMS demonstrated low internal consistency with a mean Cronbach’s alpha of .52 across groups and
time periods (range = .48 to .60). We could not find comparative information on this newly developed
measure. Because this is a multidimensional measure of cognitive function, internal consistency may not be
the correct assessment of quality. The correlations between sequential time periods ranged from .38 to .84
(M = .67), indicating good test‐retest reliability. The Trails B/A ratio and SLUMS total score were negatively
correlated (range = ‐.17 to ‐.31), as would be expected because the tests are scaled in opposite directions;
however, these correlations are fairly small, indicating that these two measures are assessing different cog‐
nitive functions. Because of the weak correlations between the SLUMS and TMT, each of these assessments
was used individually to assess cognitive impairment rather than combining them to yield a composite
score.
Depression‐Hopelessness Construct. The depression‐hopelessness construct was assessed using five self‐
report measures. The scales used to create this construct were the Beck Hopelessness Scale (BHS; Beck &
Steer, 1993), the BSI Depression subscale (Derogatis, 1993), the PAS Negative Affect and Suicidal Thinking
subscales (Morey, 1997), and the POMS Depression‐Dejection subscale (McNair et al., 1992).
Internal consistency reliabilities were computed for each assessment period for the entire sample. The mean
Cronbach’s alpha across depression measures and time periods was .87 (range = .60 to .96). The Cronbach’s
alpha for the composite ranged between .71 and .77 (M = .75) for the six time periods. Internal consistency
estimates for the subscales were similar to reliabilities from normative data. The test‐retest correlations for
the depression‐hopelessness composite were strong (M = .76, range = .57 to .90) indicating good stability
over time. The validity coefficients between self‐assessments of the depression‐hopelessness construct indi‐
cated good convergent validity with estimates ranging from .35 to .89 (M = .60) across all measures and time
periods.
METHOD
23
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Hostility‐Anger Control Construct. The hostility‐anger control composite was assessed using five self‐report
measures: the BSI Hostility subscale (Derogatis, 1993); the Anger Control, Hostile Control, and Acting Out
subscales on the PAS (Morey, 1997); and the POMS Anger‐Hostility subscale (McNair et al., 1992).
Internal consistency reliabilities were computed for each scale at each assessment period for the entire
sample and a mean Cronbach’s alpha of .62 (range = .27 to .94) was obtained. The Cronbach’s alphas for the
composite ranged between .54 and .61 (M = .57) for the six time periods. Although these reliabilities were
lower than expected, the smaller internal consistency estimates were for the scales with a small number of
items (i.e., PAS subscales with two items) and these reliability estimates are similar to other literature. The
correlations between sequential time periods ranged between .56 and .84 (M = .75) and suggest that the
hostility composite is stable over 3 month periods. The validity coefficients between self‐assessments of the
hostility‐anger control construct were quite variable with validity coefficients ranging between .11 and .84
(M =.42) across all measures and time periods; it was the PAS Acting Out and Hostile Control subscales that
tended to have lower correlations for this composite. These lower correlations along with the lower compo‐
site internal consistency estimates suggest a potential multidimensional construct. Because the composite
was stable and the different aspects of hostility‐anger control were relevant to this study, all subscales were
kept together for the composite measure. Analyses were conducted without the PAS subscales and the
overall study results did not change substantially (results available from the authors upon request), thus all
measures were included as part of the composite.
Hypersensitivity Construct. The hypersensitivity construct was measured by two self‐report measures—the
Hypersensitivity to External Stimulus subscale of the PSI and the Interpersonal Sensitivity subscale of the BSI
(Derogatis, 1993). Items 1, 7, 31, 34, and 37 were included on the PSI Hypersensitivity to External Stimulus
subscale. This composite is assessing two different aspects of hypersensitivity—environmental and interper‐
sonal.
Internal consistency reliabilities computed for each scale at each assessment period for the entire sample
indicated highly variable internal consistency estimates (.22 to .86) with a mean estimate of .56. However,
examination of each scale showed that the BSI had strong internal consistency estimates (.71 to .86) whereas
the PSI has low estimates (.22 to .39). The PSI was created by the researchers and its purpose was to capture
variables not measured by existing measures, thus it may not be a homogeneous construct. Although inter‐
nal consistency estimates of the composite were low (.47 to .61 with M = .53), the composite showed good
test‐retest reliability (.21 to .80 with a mean of .63) and the correlations between these two subscales pro‐
vided evidence of convergent validity (range = .31 to .44); thus these scales were analyzed in the rest of the
study as a composite variable. Analyses using each measure separately are available from the researchers
upon request.
Psychosis Construct. The psychosis construct was assessed by three self‐report measures. These included the
Paranoid Ideation and Psychoticism subscales of the BSI (Derogatis, 1993) and the Psychotic Features subs‐
cale of the PAS (Morey, 1997).
Internal consistency estimates for the subscales ranged between .62 and .83 (M = .77) and estimates for the
composite ranged between .73 and .80 (M = .78) indicating adequate internal consistency estimates for this
composite and its components. Internal consistency estimates for the subscales were similar to those found
with normative samples. Test‐retest correlations between sequential time periods indicated strong stability
over time (range .52 to .87 with a mean correlation of .71). The validity coefficients between self‐
24
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
assessments of the psychosis construct provided evidence of convergent validity, ranging between .35 and
.79 (M = .63) across all measures and time periods.
Somatization Construct. The somatization construct was measured by four self‐report assessments, includ‐
ing the Somatization subscale of the BSI (Derogatis, 1993), the Health Problems subscale of the PAS (Morey,
1997), the POMS Fatigue‐Inertia subscale (McNair et al., 1992), and the Physical Symptoms subscale of the
PSI. Items 5, 8, 11, 15, 19, 24, 27, and 28 were included on the PSI Physical Symptoms subscale.
The mean Cronbach’s alpha across somatization measures and time periods was .79 (range = .56 to .94) and
for the composite the mean alpha was .77 (range = .73 to .79). Test‐retest reliability estimates were strong
with correlations ranging between .58 and .86 (M = .76). The correlations between the self‐assessments of
the somatization construct indicated good convergent validity with coefficients ranging between .38 and .67
(M = .54) across all measures and time periods.
Withdrawal‐Alienation Construct. The withdrawal‐alienation construct was assessed using two PAS subs‐
cales—Alienation and Social Withdrawal. Internal consistency reliabilities were computed for each assess‐
ment period for the entire sample and the median Cronbach’s alpha across withdrawal‐alienation measures
and time periods was .75 (range = .69 to .83). The Cronbach’s alphas for the composite ranged between .62
and .71 (M = .67) for the six time periods. Correlations between sequential time periods (range = .49 to .87;
M = .68) indicated stability. Reliabilities were similar across testing intervals and were similar to reliabilities
found in the normative samples. The correlations between the subscales used in the withdrawal‐alienation
construct indicated good convergent validity with coefficients ranging between .45 and .55 (M = .51) across
time periods.
Malingering. The Structured Inventory of Malingered Symptomatology (SIMS; Widows & Smith, 2005) was
used to assess malingering on mental health disorders. Scores on five subscales (Psychosis, Neurologic Im‐
pairment, Amnestic Disorders, Low Intelligence, and Affective Disorders) were obtained at each testing pe‐
riod. The SIMS was used in this study as one of the tools to determine if a participant’s responses may be
truthful.
The subscales of the SIMS tended to be positively correlated (range = .19 to .63; M = .51) with each other.
The median Cronbach’s alpha across malingering subscales and time periods was .76 (range = .50 to .93).
Lower correlations and reliability estimates tended to be with the Affective Disorder and Low Intelligence
subscales.
The SIMS manual provides cut‐off scores to suggest malingering on each of the subscales as well as a total
score. The cut‐offs were scores greater than 1 for the Psychosis subscale and scores greater than 2 for the
Neurological Impairments, Amnestic Disorders, Low Intelligence, and Affective Disorders subscales. The total
SIMS scale cut‐off included scores greater than 14.
Personality Disorders. The Coolidge Correctional Inventory (CCI; Coolidge, 2004) was utilized to identify per‐
sonality disorders among individuals. For this study, the CCI was used to assess 14 personality disorders
identified in the current and past American Psychiatric Association’s (1980, 2000) Diagnostic and Statistical
Manuals (DSM): Antisocial, Avoidant, Borderline, Dependent, Depressive, Histrionic, Narcissistic, Obsessive‐
Compulsive, Paranoid, Passive‐Aggressive, Sadistic, Schizoid, and Schizotypal. The CCI also has other subs‐
cales to assess, among others, DSM Axis I variables, neuropsychological functioning, and response validity.
METHOD
25
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
For this study we also used the CCI measures of Axis I and personality disorders (Axis II). These variables
have been hypothesized as potential predictors of the impact of segregation on psychological distress. Be‐
cause personality disorders are considered relatively stable constructs, this measure was given only at the
baseline assessment period. Therefore, they were not included in the change over time measures. The me‐
dian Cronbach’s alpha across CCI subscales was. 75. The Cronbach’s alphas ranged between .46 and .88 for
different subscales.
Self‐Harm Construct. The Deliberate Self‐Harm Inventory (DSHI; Gratz, 2001) was used to assess the delibe‐
rate self‐harm history at the initial assessment. The data obtained from the DSHI was coded to provide a
quantitative severity rating based on the frequency of the self harming behavior and whether or not the be‐
havior resulted in hospitalization. This variable was considered to be a potential predictor of outcomes. The
baseline assessment was used to assess lifetime history of self‐harm and each harming behavior was coded
as having occurred in lifetime or not occurring. Scores were summed across the 17 items for a total score.
This measure is meant to be given as an interview rather than a paper‐and‐pencil test; we modified to fit the
testing situation. We had hoped to use this assessment as a repeated measure; however, misunderstanding
of instructions did not allow for integrity of the data and only the lifetime assessment of self‐harm was used
as a potential predictor of outcomes. The internal consistency estimate for the total score was .84 indicating
that it is reasonable to sum the 17 indicators into a total score.
Trauma. The Trauma Symptom Inventory (TSI; Briere, 1995) was used to assess the ongoing impact of trau‐
matic history. This measure was selected as a potential predictor of outcomes. This was administered once
at the second assessment period. Participants use a 4‐point rating scale for frequency of occurrence (0 –
never to 3 – often) of 100 events (e.g., flashbacks, wanting to cry, feeling jumpy) experienced within the last
6 months. Scores are obtained on 3 validity scales and 10 clinical subscales. For this study, the total score
was used. The Cronbach’s alpha for the total score was .97.
Staff Ratings
Two measures were completed by prison staff to assess the constructs of interest. The Brief Psychiatric Rat‐
ing Scale (BPRS; Ventura, Lukoff, Nuechterlein, Liberman, Green, & Shaner, 1993) was completed by clinical
staff and the Prison Behavior Rating Scale (PBRS; Cooke, 1998) was completed by correctional officers and
case managers.
Clinician Ratings. The BPRS (Overall & Gorman, 1962) is a 24‐item scale most commonly used to assess pa‐
tients with psychiatric disorders. It is designed to assess rapidly changing symptoms (Lukoff, Nuechterlein, &
Ventura, 1986; Ventura et al., 1993). It measures positive symptoms, general psychopathology, and affective
symptoms. Some items can be rated after observation of the patient; others require clinical interview to ob‐
tain the patient’s self report information. Each of the 24 symptom constructs are rated on a 7‐point scale of
severity ranging from 1 (not present) to 7 (extremely severe).
Research has indicated that there are five factors: Thought Disorder, Withdrawal, Anxious‐Depressed, Hostil‐
ity‐Suspiciousness, and Activity (Burger, Calsyn, Morse, Klinkenberg, & Trusty, 1997; Hedlund & Vieweg,
1980). The BPRS subscales and total scores demonstrated low internal consistency with alpha estimates
ranging between .40 and .66 (M = .55); these estimates are lower than those found with normative samples.
The correlations between sequential time periods ranged from .23 to .58 (M = .40), indicating moderate sta‐
bility over a 6 month period. The BPRS subscales had low correlations with self‐report measures of the same
26
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
underlying construct, with validity correlations ranging between .03 and .49 among the corresponding
measures and the average validity coefficients of the BPRS with all self‐report assessments at .28. The An‐
xious‐Depressed subscale had the strongest correlations with the self‐report measures and Withdrawal had
the lowest. In general, the BPRS scales had low scores, indicating a possible floor effect (see means in results
section) and impacting variability as well as relationships between measures.
Correctional Staff Ratings. The PBRS was designed to assess psychological features common to prison life
(Cooke, 1998). The instrument was developed for a British prison population. Therefore, some words that
are not common in the U.S. were changed to be culturally appropriate (see Appendix A). Correctional staff
rated 36 behaviors using a 4 point rating scale (0 – never/rarely; 1 – sometimes, 2 – often, 3 – most of the
time) at each of the six time periods. There are three scales: Anti‐Authority, Anxious‐Depressed, and Dull‐
Confused. All items were summed to provide a total score. Internal consistencies were good for the PBRS
scales with a mean Cronbach’s alpha across groups and times of .93 (range = .90 to .95) for the Anti‐
Authority scale, .91 (range .90 to .95) for the Anxious‐Depressed scale, and .83 (range = .78 to .87) for the
Dull‐Confused scale. Total score internal consistency estimates ranged from .94 to .95. Test–retest reliabili‐
ties were highly variable with correlations ranging between .08 and .50. Correlations between testing pe‐
riods were lowest from first to second assessments and tended to increase over time, which might be a
function of familiarity. Correlations with self‐report measures tended to be highly variable and mostly small
(‐.06 to .46), as they were with clinician ratings (.08 to .16).
Official Records Data
Data from official records were collected primarily from the Department of Corrections Information System
(DCIS), which is an administrative database of offender data. Offender characteristics to include demograph‐
ic history, criminal history and offense data, institutional behavior, and needs levels were electronically
downloaded. Inmates are routinely processed through the diagnostic unit upon intake into prison, and data
are gathered through various sources including arrest and pre‐sentence investigation records, diagnostic
interview, and pencil‐and‐paper tests.
Two standardized tests administered to all inmates at the diagnostic unit were included in this study to de‐
scribe the population. These were the Level of Service Inventory – Revised (LSI‐R; Andrews & Bonta, 1995,
2003) and the Tests of Adult Basic Education (TABE; CTB/McGraw‐Hill, 1994). The LSI‐R is a semi‐structured
interview tool that assesses criminal risk, with information verified through official records. The LSI‐R total
score ranges from 0 to 54 and is used to assign the level of supervision for community‐based offenders and
to determine allocation of services (Motiuk, Motiuk, & Bonta, 1992). The LSI‐R showed moderately strong
predictive validity (r = .31) for 1‐year recidivism rates with Colorado parolees (O’Keefe, Klebe, & Hromas,
1998). The TABE is designed to measure adult proficiency in reading, mathematics, language, and spelling. It
gives the information needed to place learners in the appropriate lessons for their particular skill deficien‐
cies. Final scoring of the tests can yield grade equivalent scores. The correlation between the TABE total bat‐
tery score and the GED average score was .63 (CTB/McGraw‐Hill, 2004).
Resulting from the diagnostic assessment process are ratings across different needs levels, including aca‐
demic, vocational, sex offender, substance abuse, medical, psychological, intellectual disabilities, assaultive‐
ness, and self‐destruction. Each level is rated on a 5‐point scale, where scores of 3 through 5 indicate prob‐
lem areas. Similar to the other scales, a psychological rating of 3 or greater indicates the need for mental
health services. Levels may be reevaluated during an offender’s incarceration.
METHOD
27
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Institutional behavior, such as disciplinary violations and involvement in gangs, are recorded electronically
over the course of an offender’s incarceration. Disciplinary violations are grouped into three categories ac‐
cording to their seriousness. Because patterns were similar when analyzing either violation type or total vi‐
olations, only totals are reported. There are three levels of gang involvement: member, associate, and sus‐
pect. Levels are ascertained by field intelligence officers who rate offenders’ involvement across 11 items
(e.g., self admission, moniker, gang tattoos, identification by law enforcement). Each item carries a weight
ranging from 5 to 20 points, and summative scores determine the degree of gang membership or involve‐
ment. To clearly delineate offenders actively involved in gangs, only those scored as gang member were
considered to have gang involvement in the following analyses.
Certain data elements were collected only for study participants during the course of their participation in
the study. The following were collected and coded for the period of time between each testing interval for
each participant: the amount of time spent in various settings (e.g., segregation, GP, hospital), phone
records, and mental health crisis data. Additionally, activity logs from paper files for the CSP participants
were collected and coded.
Phone records were received electronically from the Colorado Inmate Phone System (CIPS). From these
records, researchers coded the number of phone calls attempted, the number of calls completed, and the
time spent on the phone across all calls.
Mental health staff is required to make a written report in DCIS following any unscheduled mental health
visit or crisis. All reports completed for participants during their participation in the study were reviewed
and coded by researchers on a 3‐point self‐harm scale (1 – ideation, 2 – self‐harm behavior, 3 – attempted
suicide) and whether or not there was a report of a psychotic symptom during the crisis.
Pod activity records are kept by CSP correctional staff and are updated on a daily basis to provide informa‐
tion on an offender’s time outside of his cell for shower, exercise, and porter duties. These forms also track
the number of times each offender refused or was not offered these activities. Data were coded to reflect a
refusal or an activity not offered on a specific day as well as the actual amount of time the offender spent
participating in the activity. When the records were unclear or no information was recorded on a specific
day, it was coded as unknown. Researchers coded the pod activity sheets for each offender between each
testing interval and summed for the number of refusals, days an activity was not offered, unknowns, and
average time spent for shower and exercise activities.
PROCEDURES
Study enrollment began July 2007 and ended March 2009, with final testing of all participants completed in
March 2010. The project operated under the approval of the institutional review board at the University of
Colorado at Colorado Springs (UCCS).
The research team was notified of AS hearings by the case management supervisor at each facility and of
SCCF placements by the clinician who scheduled the facility transfers. Notification typically occurred before
the hearings or SCCF placement to give the field researcher maximal lead time. Researchers reviewed elec‐
tronic records to screen inmates for study eligibility.
Per the UCCS institutional review board, a stipulation was added to provide greater protection to inmates
with mental illness. Before consenting them, researchers were required to contact mental health staff, who
28
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
in turn were asked to assess whether the offender would be able to understand the consent form and to
weigh the study risks against the benefits. Other than the SCCF group, there were rarely issues. However, it
was not uncommon for the SCCF clinician to wait several days or even occasionally weeks for a new arrival
to stabilize prior to giving researchers approval to consent participants; these inmates were then included in
the study. Two inmates were excluded from the study because clinicians did not believe they had the capaci‐
ty to fully understand the consent process; both were SCCF inmates.
The field researcher was a female university employee who completed the full CDOC training academy and
had a CDOC badge that permitted her unescorted access to the facilities. In advance of each visit, the field
researcher contacted prison security to arrange visits with specific inmates. All inmates were escorted by
security staff to the visiting room, which contained a noncontact booth for inmates in AS or punitive segre‐
gation conditions. The field researcher met individually with each inmate to review the consent form, which
included the general purpose of the study, voluntary nature of participation, risks and benefits, and remune‐
ration. Inmates were advised that the purpose of the study was to learn about their adjustment to prison
and offenders in prisons across the state were being included in this study. Inmates who agreed to partici‐
pate were given $10 for each testing interval. Although this amount may at first seem high for AS inmates
who do not have an opportunity to earn income, it was important that AS inmates were compensated at the
same rate as GP inmates since the activities were exactly the same. Additionally, all deposits into inmate
bank accounts were subject to a 30% restitution recovery fee and deposits to inmates with negative bal‐
ances (common among AS inmates) were subject to a 50% reduction of the deposited amount. Therefore,
actual payments ranged from $2 to $7.
Inmates were screened for their native language and reading abilities. Although this was done when deter‐
mining study eligibility, the field researcher further assessed them at the time of consent. The testing bat‐
tery was not available in alternative languages and it was determined that using interpreters could negative‐
ly impact the validity of the tests. However, the field researcher attempted to include inmates with border‐
line English or reading skills by helping them to understand difficult words. Eighteen inmates across the five
study groups were specifically excluded from the study for lacking adequate language or reading abilities.
At the time of consent, the initial test battery was administered. The field researcher instructed participants
to read the directions for each test. Instructions were highlighted by researchers when there was an indica‐
tion on the test to respond with respect to a certain timeframe (e.g., in the past week). The field researcher
administered the timed TMT and the SLUMS tests, and she assisted if they had questions, most frequently
with the definition of a word. The researcher collected the test packet immediately following its completion,
so it was not ever handled by security staff. At the same time, she visually scanned the packet before the
inmate was returned to his cell to ensure that he had not inadvertently skipped a test or section of items.
Prior to leaving the facility, the researcher conducted a further review of inmates’ responses for indications
of intent to harm self or others. There were no items that assessed intent to harm others, but numerous
items were identified as potential indicators of suicide ideation. Participants were notified at the time of
consent that confidentiality would be broken if they responded affirmatively to any of these items. When
participants endorsed a suicidal item, the field researcher notified mental health staff and the principal in‐
vestigator immediately. Mental health staff then followed up with each case following notification to assess
the seriousness or intent to self‐harm. There were no participant suicides during the course of the study.
METHOD
29
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
The field researcher distributed the PBRS to housing staff at each testing interval and collected the com‐
pleted forms upon return visits to the facility. Mental health clinicians were generally notified that a BPRS
was needed a couple weeks prior to the researcher testing to give them time to complete the assessment.
In the CSP groups, 18 out of 127 participants were consented and tested prior to their AS hearing. On aver‐
age, CSP participants completed their initial test 7 days (SD = 7.3) after their AS hearing. Thirteen partici‐
pants in the GP groups were selected from the diversion program (for being at risk of AS placement) and
seven were tested prior to an AS hearing. On average, however, GP participants were tested 16 days (SD =
18.9) after their hearing or placement into the diversion program. At the time of consent and the initial test‐
ing, 43% of inmates had been confined in segregation (40% in AS groups and 3% in GP groups) for an aver‐
age of 18.2 days (SD = 18.1). SCCF participants were tested within 13 days of placement on average (SD =
8.9).
Participants’ data were kept in two separate databases. The eligibility database tracked the eligible pool of
offenders, such as identifying information, current location, date of AS hearing or SCCF placement, expected
release date, psychiatric status and clinician approval, selection into study or reason for exclusion, and date
of consent or refusal. A testing schedule for study participants was incorporated into the database, which
also had reporting capabilities in order to manage the project. A separate database tracked participants’
responses to the standardized tests; no identifying information was included in this database other than a
secure researcher‐assigned identification number. Both databases were stored on a secured server with
access restricted to the project researchers.
30
METHOD
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
RESULTS
DATA ANALYSIS PLAN
We first present results that speak to the quality of the research design, addressing issues concerning sam‐
pling and group representativeness; comparing those who participated fully, partially, or refused to partici‐
pate; evaluating the fidelity to confinement conditions; and examining the validity of self‐report responses.
Following these analyses, we present results addressing the hypotheses of interest. This study had three
goals and related hypotheses:
•
To determine which, if any, psychological domains are affected, and in which direction, by the different
prison environments; it was hypothesized that offenders in segregation would develop an array of psy‐
chological symptoms consistent with the SHU syndrome, characterized by elevations across the eight
constructs.
•
To assess whether offenders with mental illness decompensate differentially from those without mental
illness in AS by testing the hypothesis that both groups will get worse over time but that the rate of de‐
terioration would be greater for the mentally ill.
•
To compare the impact of AS against other prison conditions by testing the hypothesis that inmates in
segregation experience greater psychological deterioration over time compared to inmates in other con‐
finement conditions.
To test the first hypothesis, one sample t‐tests are completed to see if study groups are significantly differ‐
ent from normative data on the study measures at each time period.
To test the second hypothesis, analysis of variance statistical techniques are used to assess if the AS groups
have differential change over time. Comparisons are made on mean change over time for each construct for
the mentally ill and non‐mentally ill groups in AS confinement conditions.
To test the third hypothesis, analysis of variance statistical methods are used to assess mean change over
time and groups for each construct of interest. In particular, it is of interest to determine whether there is a
significant interaction between time and group to indicate that there is differential change over time de‐
pending on condition of confinement. An analysis is completed for those with different mental illness status.
That is, the mentally ill group in AS is compared to the mentally ill groups in the general prison and in the
psychiatric prison, whereas the non‐mentally ill in AS are compared to the non‐mentally ill in the general
prison population.
Mean difference statistical results for all three analyses are supplemented with effect size measures assess‐
ing proportion of variance accounted for by the time and group variables. Significant main effects will be
further investigated using pairwise comparisons to explore group differences and comparisons between
means at consecutive time periods to explore time effects. If there is a statistically significant interaction,
simple main effects exploring change over time for each group will be completed. All statistical tests are
completed at the .05 significance level.
In addition to the analysis of variance methods to explore mean change over time, regression analysis is
used to predict change over time using individual variables as potential predictors. Change over time is as‐
sessed by computing an individual slope estimate for each person on each construct of interest. Predictors
RESULTS
31
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
include demographic variables, criminal history variables, personality variables, and confinement conditions.
Within each section we describe the data analytical tools used to complete the analyses.
SAMPLING
Group Representativeness
Because random assignment procedures were not engaged, comparisons of offender characteristics across
variables routinely collected in DCIS were conducted between eligible inmates and study participants to de‐
termine the study sample’s representativeness (see Table 2). Some data are dynamic and, therefore, these
data represent those that were current for each offender at the point of his eligibility for the study. The in‐
stitutional behavior measures of disciplinary violations and prior AS placement were collected over their en‐
tire incarceration up to study eligibility.
Table 2. Representativeness of CSP Study Groups to Eligible Pool
CSP MI
Sample
Eligible Pool
p
Demographics
(n = 64)
(n = 232)
Mean age (SD)
31.2 (9.7)
32.1 (9.2)
n.s.
Ethnicity/Race
n.s.
White
41%
45%
Hispanic
33%
32%
African American
19%
17%
Other
8%
6%
High school achievement
n.s.
HS diploma
12%
12%
HS equivalency
51%
45%
Neither
37%
43%
Test of Adult Basic Education
Mean reading score (SD)
8.7 (3.6)
8.0 (3.6)
n.s.
Mean math score (SD)
6.7 (2.5)
6.3 (2.8)
n.s.
Mean language score (SD)
7.7 (4.0)
7.1 (4.1)
n.s.
Mean total score (SD)
7.7 (3.5)
7.1 (3.5)
n.s.
Sentence and Criminal History
Mean prior incarcerations (SD) .5 (0.9)
.5 (0.9)
n.s.
Mean felony class 1 – 6 (SD)
3.4 (1.1)
3.7 (1.1)
.02
Mean LSI‐R (SD)
35.3 (7.4)
34.8 (6.9)
n.s.
% Sentenced for violent crime
67%
54%
.03
Institutional Behavior
Mean # disc. violations (SD)
22.0 (27.5)
20.7 (20.1)
n.s.
% Prior AS placement
38%
38%
n.s.
% Gang member
30%
33%
n.s.
Need Levels (% scored 3‐5)
% Academic
42%
45%
n.s.
% Vocational
83%
82%
n.s.
% Medical
23%
17%
n.s.
% Substance abuse
83%
80%
n.s.
% Sex offender
44%
33%
n.s.
% Intellectual disability
11%
10%
n.s.
% Anger
69%
61%
n.s.
% Self‐destruction
34%
25%
n.s.
32
CSP NMI
Sample
Eligible Pool
(n = 63)
(n = 432)
30.0 (9.9)
30.4 (8.5)
19%
54%
22%
5%
p
n.s.
n.s.
(3.3)
(2.5)
(3.8)
(3.4)
27%
55%
16%
2%
13%
58%
30%
8.6 (3.6)
7.2 (3.0)
7.6 (3.9)
7.8 (3.5)
.4 (0.8)
3.2 (1.1)
33.1 (5.8)
70%
.4 (0.7)
3.5 (1.1)
33.0 (6.6)
59%
n.s.
n.s.
n.s.
n.s.
13.2(10.8)
32%
43%
13.9 (14.1)
29%
45%
n.s.
n.s.
n.s.
41%
87%
10%
71%
30%
3%
70%
10%
37%
83%
9%
81%
22%
3%
64%
9%
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
10%
54%
36%
7.8
6.7
7.2
7.4
n.s.
n.s.
n.s.
n.s.
n.s.
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Nonparametric chi‐square and t test analyses were conducted for both sets of group comparisons. There were
no differences between the CSP NMI study sample and eligible pool. The only difference for the CSP MI group
was that study participants had a more serious felony offense, as measured by felony class (class 1 is most se‐
rious and class 6 is least) and percent with a violent crime, than individuals in the eligible pool.
Refusals
The field researcher asked inmates who refused to participate or who withdrew their consent for their reasons.
Half of them gave no reason for doing so. Of those who listed their reasons, 10 inmates stated general disinter‐
est, 6 were skeptical of the research, 4 feared retaliation from their gang or other inmates for participating, 3
listed monetary reasons, and 3 expected imminent release due to an appeal. Chi‐square analyses and t‐tests
were conducted between study participants and inmates who refused to participate in the study or withdrew
their consent to determine if significant differences existed (see Table 3). The only measured difference between
the two groups was that participants had higher LSI‐R scores, which indicates higher recidivism risk.
Table 3. Comparison of Refusals to Study Participants
Refusals
Participants
Demographics
(n = 55)
(n = 247)
Mean age (SD)
33.2 (10.5)
31.7 (8.9)
Ethnicity/Race
White
44%
40%
Hispanic
33%
37%
African American
20%
18%
Other
4%
5%
High school achievement
HS diploma
23%
13%
HS equivalency
42%
56%
Neither
36%
32%
Test of Adult Basic Education
Mean reading score (SD)
8.2 (3.6)
8.5 (3.5)
Mean math score (SD)
7.1 (3.1)
6.6 (2.8)
Mean language score (SD)
7.9 (4.0)
7.3 (3.9)
Mean total score (SD)
7.8 (3.4)
7.5 (3.5)
Sentence and Criminal History
Mean prior incarcerations (SD) .3 (0.6)
.5 (0.8)
Mean felony class 1 – 6 (SD)
3.2 (1.1)
3.4 (1.1)
Mean LSI‐R (SD)
31.4 (7.4)
34.3 (7.3)
% Sentenced for violent crime
70%
60%
Institutional Behavior
# Disciplinary violations
16.9 (24.0)
15.8 (16.8)
% Prior AS placement
26%
23%
% Gang member
22%
28%
Need Levels (% scored 3‐5)
% Academic
42%
39%
% Vocational
87%
83%
% Medical
13%
19%
% Substance abuse
70%
79%
% Sex offender
35%
31%
% Intellectual disability
15%
9%
% Anger
72%
60%
% Self‐destruction
24%
23%
RESULTS
p
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
.02
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
33
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Complete and Incomplete Testers
There were 222 participants who completed all testing sessions of the study; thus only 18% of participants did
not have all self‐report assessments for every time period. There were a number of reasons why participants did
not complete all testing periods: some withdrew their consent, others were paroled before the end of the
study, and some were not available for a specific testing interval (e.g., out to court for extended trial). There was
not a significant differential incompletion rate across the five study groups, χ2(4, N = 270) = 3.71, p = .45, with
incompletion rates of 25% for CSP MI, 16% for CSP NMI, 18% for GP NMI, 15% for GP MI, and 18% for SCCF.
Comparisons were made between those who did and did not complete all assessments on demographic, back‐
ground variables, and the dependent variables. To compute a score on the dependent variables, the mean
across scores for available time periods was computed for the self‐report composites and cognitive variables.
Table 4 provides information on these comparisons. Participants who did not complete the entire study had
significantly higher self‐destruction needs, higher mean hostility composite scores, and lower cognitive function
as demonstrated by significantly lower scores on both measures of cognitive performance (SLUMS, Trails B/A).
Table 4. Comparison of Incomplete to Complete Testers
Incomplete
Demographics
(n =48)
Mean age (SD)
31.4 (8.7)
Ethnicity/Race
White
49%
Hispanic
34%
African American
13%
Other
4%
High school achievement
HS diploma
12%
HS equivalency
57%
Neither
31%
Test of Adult Basic Education
Mean reading score (SD)
8.6 (3.5)
Mean math score (SD)
6.6 (2.8)
Mean language score (SD)
7.4 (3.9)
Mean total score (SD)
7.6 (3.5)
Sentence and Criminal History
Mean prior incarcerations (SD)
0.5 (0.8)
Mean felony class 1 – 6 (SD)
3.5 (1.1)
Mean LSI‐R (SD)
34.0 (7.5)
% Sentenced for violent crime
65%
Institutional Behavior
# Disciplinary violations
15.1 (15.4)
% Prior AS placement
22%
% Gang member
27%
Need Levels (% scored 3‐5)
% Academic
37%
% Vocational
84%
% Medical
18%
% Substance abuse
78%
% Sex offender
32%
% Intellectual disability
9%
% Anger
60%
% Self‐destruction
21%
34
Complete
(n =222)
34.1 (10.6)
p
n.s.
n.s.
39%
36%
20%
5%
n.s.
19%
44%
37%
7.7
7.1
7.5
7.4
(3.7)
(3.3)
(4.1)
(3.6)
n.s.
n.s.
n.s.
n.s.
0.4 (0.8)
3.2 (1.2)
33.3 (7.5)
61%
n.s.
n.s.
n.s.
n.s.
22.5 (28.8)
27%
29%
n.s.
n.s.
n.s.
43%
85%
15%
72%
30%
13%
64%
36%
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
.04
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Composites and Cognitive Measures
Mean Anxiety (SD)
Mean Depression‐Hopelessness (SD)
Mean Hostility‐Anger Control (SD)
Mean Hypersensitivity (SD)
Mean Psychosis (SD)
Mean Somatization (SD)
Mean Withdrawal‐Alienation (SD)
Mean SLUMS (SD)
Mean Trails (SD)
Incomplete
‐.18 (.72)
‐.16 (.73)
‐.16 (.57)
‐.13 (.68)
‐.18 (.75)
‐.15 (.69)
‐.22 (.63)
23.18 (3.36)
2.82 (0.75)
Complete
p
‐.08 (.80)
‐.01 (.76)
.02 (.65)
‐.12 (.77)
‐.06 (.73)
.06 (.82)
‐.03 (.70)
22.02 (3.76)
3.18 (1.12)
n.s.
n.s.
.05
n.s.
n.s.
n.s.
n.s.
.03
.01
The amount of complete data was higher on self‐report assessments (85%) than clinician ratings (76%) and
correctional staff ratings (57%). There were significant differences in groups’ completion rates for clinician
ratings with the MI groups having more complete data (CSP MI 70%, GP MI 76%, SCCF 75%) than the NMI
groups (CSP NMI 56%, GP NMI 58%). There were also significant differences between completion rates for
the correctional staff ratings with the GP groups having less complete data (GP NMI 47%, GP MI 49%) than
the other three groups (CSP MI 64%, CSP NMI 75%, SCCF 61%).
Group Fidelity to Conditions of Confinement
Participants remained in their assigned group regardless of later placements throughout the prison system.
Table 5 summarizes the locations of study participants by group and by testing interval (each interval is the
period of time between two assessment periods).
One of the challenges of applied research is the researchers’ lack of control over the independent variable,
which in this case is the condition of confinement. Therefore, all offenders in AS were not confined in segre‐
gation for their entire period of participation in the study. Over the course of the study, 15 offenders in the
CSP MI group were placed in the specialized OMI program; most completed at least three tests prior to the
transfer. Some of the inmates placed in CSP were taken to county jail for a court appearance. Conversely,
inmates in the GP groups may have at some time during their study participation been placed in punitive
segregation or even AS. There were five GP MI and four GP NMI participants who were placed in AS during
the course of their segregation; the remainder of GP inmates who had time in segregation were in punitive
segregation. Seven of the nine GP inmates were reclassified to AS primarily between the third and fourth
assessment periods, one was reclassified after only two tests, and one was reclassified two days prior to his
final test.
Due to the contamination across groups, a separate set of analyses were conducted using only the “pure”
cases, which included those who only experienced a single condition of confinement during their entire
study participation. There were 26 pure cases in the CSP MI group, 39 in the CSP NMI group, 13 in the GP
MI, and 11 in the GP NMI. The p values and partial eta‐squares for the self‐report composites were com‐
pared for these pure cases and the original study groups. The SCCF group was not included because those
participants were expected to transfer from SCCF once stabilization occurred. A result would be considered
different if both the p value changed significance and the effect size was not of the same magnitude. Be‐
cause of the smaller sample size in the pure group analysis, it might be possible for an effect of the same
magnitude to no longer be statistically significant, thus we did not count this as a different result. The same
pattern of results was found for both samples (total vs. pure) except on the hypersensitivity composite. For
RESULTS
35
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
this variable, there was a significant time effect for the entire sample (p = .001, η2 = .026) demonstrating
higher scores at the first assessment compared to all other periods but no significant time effect for the pure
sample (p = .56, η2 = .009). As it does not appear that changing locations was a major explanation for the
results, subsequent analyses included all offenders who participated in the study. (Complete statistical re‐
sults are available upon request from the authors.)
Table 5. Number of Days by Location for each Group at each Testing Interval
Interval 1
Interval 2
Interval 3
Interval 4
Group
Location
n M (days)
n M (days)
n M (days)
n M (days)
CSP
62
19.8
60
78.3
57
91.2
50
88.2
Other seg
56
88.9
2
4.5
2
9.5
0
‐‐
CSP MI
SCCF
1
89.0
0
‐‐
0
‐‐
1
69.0
(n = 64)
GP
4
43.8
3
31.0
7
63.0
8
74.6
Othera
6
32.0
5
12.4
9
8.9
5
20.2
CSP
59
14.2
57
82.7
56
92.2
56
92.7
Other seg
57
90.3
2
3.5
2
43.0
0
‐‐
CSP NMI
SCCF
0
‐‐
0
‐‐
0
‐‐
0
‐‐
(n = 63)
GP
4
5.0
0
‐‐
0
‐‐
0
‐‐
Othera
5
39.0
5
8.4
2
15.5
4
1.8
CSP
0
‐‐
0
‐‐
4
24.3
4
80.5
Other seg
9
12.4
7
23.0
9
34.2
6
30.7
GP MI
SCCF
0
‐‐
0
‐‐
0
‐‐
0
‐‐
(n = 33)
GP
32
87.2
32
89.9
27
84.8
25
88.8
Othera
0
‐‐
3
1.0
1
1.0
1
11.0
CSP
0
‐‐
0
‐‐
0
‐‐
0
‐‐
Other seg
10
39.8
9
13.3
5
39.6
15
37.9
GP NMI
SCCF
0
‐‐
0
‐‐
0
‐‐
1
46.0
(n = 43)
GP
41
79.2
41
89.1
39
85.3
35
83.7
Othera
2
23.5
4
11.5
2
12.5
2
2.5
CSP
0
‐‐
0
‐‐
0
‐‐
1
53.0
Other seg
1
37.0
4
7.5
4
10.3
8
24.0
SCCF
SCCF
64
77.3
54
79.1
40
92.4
34
80.6
(n = 67)
GP
7
26.6
21
66.2
24
77.3
29
75.6
a
Other
5
12.0
6
32.5
3
39.7
6
9.7
Interval 5
n M (days)
44 83.0
2 10.5
1 71.0
15 75.7
5 14.2
54 91.4
0
‐‐
0
‐‐
3 62.3
1
1.0
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
‐‐
Note. Individuals may have multiple locations within a study period, so the n’s within a group and interval can be larger than the
group sample size.
a
Other included out to court (county jail), in custody of US Marshall, hospital or external medical, community placement, and time
in transport.
VALIDITY OF RESPONSES
Most of the assessments used in this study were self‐report measures, which always carry the risk of not
being completed accurately by the participant. Because of this risk, several measures were collected to as‐
sess the validity of individual responses.
During data collection and data entry, responses were scanned for abnormal pattern of responses (e.g., the
same response selected for all items). Each person’s pattern of response was coded as potentially question‐
able or not. If the pattern was noticed during data collection, then the participant was questioned about his
response pattern and asked to redo the test if he admitted to not being truthful. If the participant said he
was being honest and the researcher still did not believe him, she marked the test as questionable. Overall,
12% of participants had a questionable response pattern on any measure at any time period (see Table 6);
36
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
however, there were no differences between the number of questionable response patterns across groups,
χ2(4, N = 270) = 3.87, p = .42.
Table 6. Percentage of Participants with Questionable Response Patterns
CSP MI CSP NMI GP MI
GP NMI
SCCF
All
Time 1
5%
0%
6%
0%
2%
2%
Time 2
6%
7%
0%
0%
3%
4%
Time 3
3%
5%
3%
5%
0%
3%
Time 4
5%
5%
0%
3%
0%
3%
Time 5
4%
7%
0%
10%
2%
5%
Time 6
8%
7%
NA
NA
NA
8%
Potential malingering was assessed using the SIMS, a 75‐item screening measure for detecting feigned
symptoms of psychopathology and cognitive functioning in clinical and forensic settings. A total score and
scores on five subscales (i.e., Psychosis, Neurological Impairment, Amnestic Disorders, Low Intelligence, Af‐
fective Disorders) were obtained. The SIMS assesses whether respondents endorse atypical, improbable,
inconsistent, or illogical symptoms. Scores above the cutoff suggest malingering but may also suggest ge‐
nuine psychopathology. Eighty‐five percent of participants had at least one elevated score across the differ‐
ent subscales of the SIMS (see Table 7). The percentage of participants with elevated scores was significantly
different across groups, χ2(4, N = 270) = 56.82, p < .001, with the MI groups (CSP MI 92%, SCCF 96%, GP MI
97%) demonstrating more elevated scores than the NMI groups and with the CSP NMI group (86%) showing
more elevations than the GP NMI group (49%). We also considered using a rule of removing participants if
they were elevated on multiple scales; however, multiple elevations within a time period were still high
among the mentally ill groups (47% to 62%). Because elevated scores may actually reflect psychopathology,
we did not eliminate anyone from the study based on this measure. The SIMS was administered to detect
potential malingering and was not intended as an outcome measure, thus there are no further analyses with
this measure (Appendix B provides summary statistics for the sample).
We further examined response patterns within the main constructs of interest. Because multiple measures
were used for each construct, we computed variability across standardized measures of the same construct
in order to see if a person was responding in an inconsistent fashion (see Table 8). For example, inconsistent
responses within the depression construct might entail a high score on the BHS but a low score on the BSI
Depression scale, where one might expect the pattern of scores to be similar. If the variability score for a
participant was greater than two standard deviations from the mean on any composite, responses were ex‐
amined. Approximately 17% of participants had a value greater than this cutoff. Different rates of inconsis‐
tent responses were found across the groups, χ2(4, N = 270) = 10.09, p = .04, with the lowest incidence of
inconsistency for GP MI (9%), CSP NMI (10%), and GP NMI (12%) groups, and higher incidences for CSP MI
(20%) and SCCF (27%) groups.
To explore if results were influenced by participants with inconsistent or questionable responses, three sets
of analyses comparing group differences on composite variables were completed using (1) all participants,
(2) those who did not have a questionable response, and (3) those who did not have inconsistent responses.
Removal of persons with questionable or inconsistent responses did not change the overall effects and re‐
sults, so all participants are used in the analyses for this report (full statistical results are available from the
authors upon request).
RESULTS
37
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 7. Elevation on SIMS Scales
CSP MI CSP NMI
Affective Disorders
Time 1
59%
35%
Time 2
60%
28%
Time 3
68%
36%
Time 4
58%
39%
Time 5
41%
29%
Time 6
61%
37%
Neurological Impairment
Time 1
52%
32%
Time 2
42%
30%
Time 3
42%
33%
Time 4
48%
41%
Time 5
43%
25%
Time 6
41%
24%
Psychosis
Time 1
47%
29%
Time 2
50%
22%
Time 3
53%
24%
Time 4
52%
23%
Time 5
39%
18%
Time 6
39%
18%
Low Intelligence
Time 1
11%
5%
Time 2
10%
7%
Time 3
17%
9%
Time 4
13%
9%
Time 5
12%
11%
Time 6
16%
11%
Amnestic Disorders
Time 1
20%
2%
Time 2
26%
5%
Time 3
23%
5%
Time 4
22%
7%
Time 5
16%
5%
Time 6
16%
4%
Total Score
Time 1
50%
29%
Time 2
50%
22%
Time 3
57%
26%
Time 4
52%
27%
Time 5
41%
27%
Time 6
45%
24%
GP MI
GP NMI
SCCF
All
55%
58%
62%
59%
55%
NA
16%
12%
17%
13%
13%
NA
72%
69%
69%
59%
65%
NA
49%
47%
52%
47%
41%
48%
48%
42%
56%
45%
45%
NA
21%
17%
12%
10%
16%
NA
67%
58%
56%
51%
46%
NA
46%
40%
40%
41%
35%
32%
46%
36%
38%
34%
28%
NA
7%
12%
15%
10%
10%
NA
76%
69%
64%
61%
60%
NA
43%
41%
41%
39%
33%
29%
3%
12%
3%
10%
3%
NA
2%
2%
2%
8%
0%
NA
12%
11%
13%
10%
9%
NA
7%
8%
10%
10%
8%
12%
21%
15%
12%
24%
17%
NA
2%
2%
0%
3%
0%
NA
33%
40%
31%
30%
26%
NA
16%
20%
16%
18%
14%
9%
46%
48%
53%
48%
41%
NA
7%
10%
7%
10%
8%
NA
78%
72%
60%
61%
53%
NA
44%
42%
42%
41%
35%
34%
38
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 8. Potentially Inconsistent Responses within a Composite Score
CSP MI CSP NMI GP MI
GP NMI
SCCF
All
Anxiety
Time 1
0%
0%
0%
0%
0%
0%
Time 2
0%
2%
0%
0%
2%
1%
Time 3
0%
0%
0%
0%
0%
0%
Time 4
0%
0%
0%
0%
0%
0%
Time 5
0%
0%
0%
0%
2%
0%
Time 6
0%
0%
NA
NA
NA
0%
Depression‐Hopelessness
Time 1
0%
0%
0%
0%
0%
0%
Time 2
0%
0%
0%
0%
2%
0%
Time 3
0%
0%
0%
0%
0%
0%
Time 4
2%
0%
0%
0%
0%
0%
Time 5
2%
0%
0%
0%
0%
0%
Time 6
0%
0%
NA
NA
NA
0%
Hostility‐Anger Control
Time 1
5%
0%
0%
0%
2%
2%
Time 2
6%
2%
0%
0%
3%
3%
Time 3
3%
4%
0%
0%
3%
2%
Time 4
5%
2%
0%
0%
5%
3%
Time 5
5%
4%
0%
0%
2%
2%
Time 6
0%
2%
NA
NA
NA
1%
Hypersensitivity
Time 1
8%
3%
9%
5%
12%
7%
Time 2
6%
2%
9%
0%
8%
5%
Time 3
5%
4%
9%
2%
12%
6%
Time 4
9%
7%
7%
3%
12%
8%
Time 5
4%
2%
10%
5%
9%
6%
Time 6
12%
2%
NA
NA
NA
7%
Psychosis
Time 1
5%
2%
3%
0%
6%
3%
Time 2
2%
0%
0%
0%
0%
0%
Time 3
0%
0%
0%
2%
0%
0%
Time 4
2%
0%
0%
0%
0%
0%
Time 5
2%
4%
0%
0%
2%
2%
Time 6
4%
2%
NA
NA
NA
3%
Somatization
Time 1
2%
0%
0%
2%
2%
1%
Time 2
0%
2%
0%
0%
0%
<1%
Time 3
7%
2%
0%
0%
0%
2%
Time 4
3%
2%
0%
0%
0%
1%
Time 5
2%
2%
3%
0%
0%
1%
Time 6
0%
0%
NA
NA
NA
0%
Withdrawal‐Alienation
Time 1
2%
2%
0%
5%
4%
3%
Time 2
2%
3%
0%
5%
3%
12%
Time 3
2%
4%
6%
2%
2%
3%
Time 4
0%
2%
0%
3%
2%
1%
Time 5
4%
2%
0%
3%
2%
2%
Time 6
6%
4%
NA
NA
NA
5%
RESULTS
39
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
OFFICIAL RECORD DATA
Several sets of official record data were gathered and coded to use as potential outcomes or predictors of
change. It was expected that CSP inmates might experience varying levels of isolation based on the amount
of time spent at the different QOL levels, the amount of visits and phone contacts, and out of cell time for
showers and recreation. In gathering QOL levels, however, data were obtained from two different sources
that had conflicting information. Because of the quality of this data, it was not possible to code or use in this
study. Following is a discussion of the other official record data gathered and coded.
CIPS Data
CIPS data were collected on the five study groups by testing interval in order to examine amounts of phone
contact. A testing interval consisted of the day the offender tested on their self‐report measures through
the day before the next battery, generally three months. The following data were collected on each offender
for each interval: 1) total number of calls attempted, 2) total calls completed (i.e., offender was able to
reach another person), and 3) the average duration in minutes per week of all completed calls. A total of 75
offenders did not have any calls during at least one time period. Though the CSP groups had one more test‐
ing interval than the other groups, summary statistics are presented only for the four common intervals for
each group. Table 21 provides the mean number (and standard deviation) of total calls attempted, total calls
completed, and average duration (minutes/week).
Table 21. Mean (and SD) for Phone Call Data for each Time Interval by Study Group
Group
Measure
Time 1 to 2
Time 2 to 3
Time 3 to 4
Attempted
30.80 (57.72)
9.13 (10.11)
12.06 (14.53)
CSP MI
Completed
6.24 (13.11)
2.56 (2.65)
3.58
(4.40)
(n = 55)
Avg mins/wk
6.86 (15.17)
3.77 (4.17)
4.61
(5.89)
Attempted
41.52 (80.14)
14.55 (15.94)
23.46 (76.43)
CSP NMI
Completed
11.11 (27.52)
3.68 (3.93)
6.86
(7.66)
(n = 56)
Avg mins/wk
12.54 (32.11)
5.26 (5.46)
9.22 (10.60)
Attempted
86.03 (95.49)
129.45 (312.03)
102.31 (202.69)
GP MI
Completed
22.59 (42.72)
26.62 (59.33)
18.17 (28.94)
(n = 29)
Avg mins/wk
23.40 (38.42)
31.01 (75.21)
18.82 (32.40)
Attempted
122.40 (131.74)
118.87 (137.45)
105.34 (136.08)
GP NMI
Completed
27.08 (40.75)
24.03 (27.15)
18.08 (33.38)
(n = 38)
Avg mins/wk
29.78 (47.09)
26.38 (30.26)
21.36 (39.62)
Attempted
59.11 (80.98)
59.29 (76.43)
56.71 (88.29)
SCCF
Completed
14.41 (27.22)
15.36 (30.63)
13.07 (30.22)
(n = 56)
Avg mins/wk
15.85 (27.93)
13.94 (26.11)
11.39 (27.67)
Time 4 to 5
17.62 (20.94)
5.47 (7.89)
6.92 (3.20)
20.75 (22.25)
6.00 (6.38)
7.96 (8.61)
432.17 (641.45)
97.69 (143.62)
104.00 (159.71)
496.63 (495.10)
98.71 (119.22)
105.40 (131.08)
241.13 (325.30)
58.12 (125.70)
52.87 (111.57)
During the course of the project, important changes were made to the CIPS program. On July 1, 2008, the
CIPS pricing was changed so that all offender calls generated from any CDOC facility dialing someone within
the continental United States cost the same price. Previously, it was more costly for an offender to make a
phone call to someone located outside of Colorado. Additionally, in July of 2007, one trial pod at CSP
changed how offenders were able to access the phone system by providing cordless phones that inmates
were able to use in their cells. After the trial period ended, the remainder of the facility transitioned to the
cordless phone system in July 2009. This change allowed prisoners at CSP to access phones more frequently.
Prior to the introduction of the cordless phones, inmates were required to be escorted by two staff mem‐
bers from their cells to the day hall where they would be tethered near the phone. This method is highly
40
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
staff intensive and because of other required staff duties, staff were not always able to escort offenders to
the phone when requested.
The changes in the CIPS program makes interpretation of change over time difficult; it also impacts group
comparisons because group assessments did not occur evenly over the study period (e.g., because there
were a smaller number of potential CSP MI participants, we started data collection activities earlier for that
group). Thus, further statistical analyses were not completed on these data because it is difficult to know to
what causes any potential findings could be attributed.
Pod Activity Data
Pod activity data were collected on the two CSP groups. Pod activity data were gathered from records that
housing staff at CSP keep on every offender at the facility to track offenders’ exercise and shower habits.
Researchers were provided access to this data for use in the study. Data were coded by testing interval (i.e.,
activities that occurred from one testing period until the next testing period). The following data were col‐
lected on each offender: 1) number of days each offender refused an offer to exercise and/or shower, 2)
number of days each offender was not offered exercise and/or shower, 3) average number of hours per
week an offender participated in exercise and/or shower, 4) the number of days the prisoner participated in
exercise and/or shower, and 5) the number of unknowns for recreation and/or shower in that time period.
Pod activity data also track an inmate’s work record, but because so few participants held jobs during the
study period, this data was not included. Table 22 provides the mean number (and standard deviation) for
each variable by group and testing interval.
Table 22. Summary Statistics for Pod Activities (Exercise and Showers)
Activity
Interval:
Time 2 to 3
Time 3 to 4
Time 4 to 5
Exercise Activity
M
SD
M
SD
M
SD
Days refused
26.02 12.43
29.52
16.38
30.18
17.47
Days not offered
1.85
2.06
2.39
1.92
2.92
2.50
CSP MI
Hours per week
0.71
1.02
0.98
1.17
0.98
1.26
(n = 61)
Days of activity
9.23 12.33
13.41
15.61
14.10
16.99
Unknown
43.95 12.08
47.69
12.59
46.30
15.34
Days refused
22.10 13.35
23.28
14.98
28.25
15.73
Days not offered
2.16
2.05
2.80
1.78
2.69
2.16
CSP NMI
Hours per week
1.41
1.05
1.41
1.22
1.48
1.16
(n = 61)
Days of activity
18.00 13.23
20.69
16.46
22.07
16.12
Unknown
41.49
9.80
44.20
12.29
39.82
6.99
Shower
Activity
M
SD
M
SD
M
SD
Days refused
9.90 10.36
12.74
12.76
12.61
12.55
Days not offered
1.30
1.54
1.61
1.57
1.90
2.01
CSP MI
Hours per week
1.41
0.59
1.43
0.65
1.37
0.68
(n = 61)
Days of activity
36.07 14.61
39.69
17.01
39.49
19.06
Unknown
33.79 14.02
38.98
15.78
39.49
17.38
Days refused
6.82
9.74
7.02
9.35
10.00
12.20
Days not offered
1.44
1.36
1.77
1.43
1.93
1.97
CSP NMI
Hours per week
1.75
0.42
1.75
0.51
1.70
0.53
(n = 61)
Days of activity
44.07
9.77
48.95
12.58
48.70
12.80
Unknown
31.41
6.13
33.23
8.46
32.18
5.02
Time 5 to 6
M
SD
29.74
19.38
3.28
3.54
1.00
1.19
14.54
17.17
46.1
23.34
29.48
16.08
4.59
5.27
1.68
1.24
23.66
17.39
37.39
9.47
M
SD
12.28
12.97
2.34
3.14
1.22
0.64
36.44
18.91
42.59
26.65
10.36
10.69
3.52
5.09
1.73
0.58
50.13
14.92
31.1
7.19
RESULTS
41
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Due to problems with coding this data, it was not felt that further analyses could be run. As was previously
noted, the data were collected from records that CSP staff were already keeping on offenders and it was
often difficult to decipher and/or interpret the records. For example, if a variable was left blank it was not
known if that meant the offender was not offered the activity or if he refused to participate resulting in
much of the data being coded as unknown.
Mental Health Crisis Data
Any situation that is not a scheduled appointment and requires immediate psychological intervention is consi‐
dered a crisis event; crisis events are documented by clinicians in DCIS. For this study, these events were re‐
viewed and coded for whether there was self‐harming ideation or behavior and whether there was a report of an
inmate experiencing at least one symptom commonly associated with psychosis. A total of 36 participants had a
self‐harming ideation or behavior or a report of altered thought patterns commonly associated with psychosis
(see Figures 29 and 30). The self‐harm data were coded into three categories to indicate a range from self‐harm
ideation to suicide attempt. Psychotic symptoms were reported as a single category, but the researchers used a
low standard in coding psychotic symptoms. If there was any mention of hallucinations or delusionary thoughts,
even if not observed by the clinician or if denied by the offender, the crisis event was coded as having a psychotic
symptom. For example, one participant had threatened self‐injurious behavior but also reported the presence of
visual hallucinations in the past; this was also coded as a psychotic symptom even though the clinician stated that
there was no evidence for psychosis. It should be noted that some events involved both symptoms of psychosis
and self‐harming ideation/behaviors; therefore case numbers represent the same person on both graphs.
There are several limitations of these data. These include that self‐harming ideation/behavior or psychotic symp‐
toms could have occurred without staff’s knowledge, offenders may have discussed or exhibited thoughts or be‐
haviors on these dimensions during regularly scheduled mental health appointments, and offenders’ self‐harming
histories prior to study entry were unknown. For example, it was clear from the crisis notes that an individual
with numerous crisis events had a long history of self‐harming behavior and SCCF placements prior to his enroll‐
ment in the study; this is not reflected in the data. Furthermore, the reason for the self‐harming idea‐
tion/behavior is not captured in the graphed data, but the reasons vary widely. As an example, one offender
threatened suicide because he did not want to be removed from CSP to be placed in a new program for offenders
with mental illness located at a lower security facility. In another example, one person reported self‐harming be‐
havior due to a recent automobile accident where several family members died. Therefore, without more infor‐
mation, it is not possible to attribute the reasons for their mental health crisis to their confinement setting.
We were interested in including the crisis data as an outcome measure in the change over time analyses in order
to determine if the occurrence of crises was impacted by confinement conditions, mental status, and time. Be‐
cause the number of participants who experienced a crisis event was so small, it was not possible to include this
variable as an outcome measure in the change over time analyses. These data raise more questions than they
provide answers; it was determined that further case study of participants’ mental health histories was outside
the scope of the current research.
42
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 29. Crisis Events by Individual Participants who had SelfHarming Ideation or Behavior
•
36
Suicidal/self harm ideation
Self harming behavior
• Suicide attempt
•
35
34
•
33
32
31
30
29
28
27
26
17'
) 25
c 24
> 23
LLI
U) 22
20
SCCF
•
GP MI
12
CSP NMI
11
0
10
9
0
8
• •
CSP MI
0
•
•
0
• •
•
0
7
•
6
QS
5
CD
•
•
0
4
0
3
0
2
0
I
I
I
I
I
I
I
I
I
I
2 3 4 5 6 7 8 9 10 11 12 13
Months in Study
RESULTS
43
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 30. Crisis Events by Individual Participants who had a Psychotic Symptom
363534-
•
3332313029282726^
•
Tij 25.5 24-
•
CD
•
uj
> 23-
••
(/)
_ 220
.r: 21U
.0 20'- 19 -
•
•
• •
(I) 18^
4...
C
0 17a.
z 16E
03 15a' 14>,
13-
SCCF
•
Cl) 12-
•
GP NMI
•
GP MI
1110•
9-
CSP MI
8^
76-
•
•
••
•
5-
• •
4•
3-
•
2-
•
1I
0
I
1
1
2
1
3
1
4
1
5
1
6
1
7
i
8
1
9
1
10
1
11
1
12
1
13
Months in study
44
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
GROUP COMPARISONS
Offender Characteristics
Study groups differed from each other at entry into the study in a number of statistically significant ways
(see Table 9). Some differences were consistent with their AS placement or the mental health needs of the
groups. The two CSP groups were more likely to have a prior AS placement and to have higher anger needs.
The three mentally ill groups (CSP MI, GP MI, SCCF) had higher needs for medical and intellectual disability
services. The CSP MI and SCCF groups had higher self destruction needs and were less likely to be gang
members. The two GP groups had the lowest rates of sex offender treatment needs. Finally, ethnic/racial
composition was different for each of the groups, with more whites in the GP MI and SCCF groups and more
Hispanics in the CSP NMI group.
Table 9. Study Group Comparisons on Offender Characteristics
CSP MI
CSP NMI
GP MI
Demographics
Mean age (SD)
31.2 (9.7)
30.0 (9.9)
30.2 (7.8)
Ethnicity/Race
White
41%
19%
52%
Hispanic
33%
54%
39%
African American
19%
22%
9%
Other
8%
5%
0%
High school achievement
HS diploma
12%
10%
7%
HS equivalency
51%
54%
58%
Neither
37%
36%
36%
Test of Adult Basic Education
Mean reading score (SD)
8.7 (3.6)
7.8 (3.3)
8.7 (3.0)
Mean math score (SD)
6.7 (2.5)
6.7 (2.5)
6.9 (2.6)
Mean language score (SD)
7.7 (4.0)
7.2 (3.8)
7.9 (3.5)
Mean total score (SD)
7.7 (3.5)
7.4 (3.4)
7.8 (3.0)
Sentence and Criminal History
Mean prior incarcerations (SD) 0.5 (0.9)
0.4 (0.8)
0.4 (0.6)
Mean felony class 1 – 6 (SD)
3.4 (1.1)
3.1 (1.1)
3.6 (1.1)
Mean LSI‐R (SD)
35.3 (7.4)
33.1 (5.8)
35.4 (8.6)
% Sentenced for violent crime
67%
70%
61%
Institutional Behavior
# Disciplinary violations
22.0 (27.5)
13.2 (10.8)
17.2 (15.3)
% Prior AS placement
38%
32%
27%
% Gang member
30%
43%
21%
Need Levels (% scored 3‐5)
% Academic
42%
41%
39%
% Vocational
83%
87%
88%
% Medical
23%
10%
18%
% Substance abuse
83%
71%
91%
% Sex offender
44%
30%
24%
% Intellectual disability
11%
3%
9%
% Anger
69%
70%
56%
% Self‐destruction
34%
10%
16%
GP NMI
33.5 (7.5)
40%
33%
26%
2%
18%
63%
20%
10.2 (3.0)
7.1 (3.3)
7.8 (4.1)
8.3 (3.8)
0.5 (0.7)
3.5 (1.0)
31.8 (7.0)
54%
16.0 (15.5)
19%
33%
26%
77%
7%
78%
14%
0%
51%
10%
SCCF
33.9 (8.7)
55%
24%
15%
6%
p
n.s.
.01
n.s.
19%
50%
31%
7.7 (3.7)
6.5 (3.4)
7.0 (4.2)
7.1 (3.7)
0.7 (0.9)
3.6 (1.0)
33.7 (8.5)
54%
14.2 (16.6)
0%
10%
40%
85%
27%
71%
37%
20%
52%
39%
.01
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
n.s.
<.001
<.001
n.s.
n.s.
.02
n.s.
.02
.01
n.s.
<.001
RESULTS
45
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
NORMATIVE COMPARISONS
Because we used standardized assessments, normative data were available to compare to the study groups.
Normative means were based on non‐clinical male samples when available; a general adult sample was used
if a male sample was not available. If only clinical sample normative data were available then those were
used. Figures 3 to 10 provide the means over time for the measures with total scores for each study group
along with highlighted cutoff score ranges. Each graph shows the possible range of scores on the y axis. Fig‐
ures 3 to 10 are presented for visual reference only; analyses are conducted in later sections. Normative
comparisons for subscales used in this study are available in Appendix B.
Figure 3. Mean Scores over Time for the BHS Total Score by Group
20
15
10
CD
—
0 ................ .
................................. .0.
,..,8
v-.
o_
n
o
E,
•
................ A
•
a
.-.
A. - - ..
._,_,..::;;Lluo.-Pri
.........
c
ra
aJ
2
0
Test 1
Test 2
Test 3
Test 4
Test 5
Minimal
0
0
0
0
0
Mild
4
4
4
4
4
Moderate
9
9
9
9
9
Severe
46
.
15
15
15
15
15
—11—CSP MI
7.16
7.59
7.75
8.12
6.32
—M— CSP NMI
5.14
4.66
4.79
3.87
4.71
•• •k•• GP MI
7.09
5.56
5.38
5.62
4.22
••.A. • • GP NMI
2.26
2.35
2.56
2.60
1.58
• —0 • • SCCF
9.84
9.20
8.57
9.20
8.93
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 4. Mean Scores over Time for the BSI Global Symptom Index by Group
4
3.5
3
2.5
Mean Group Score
2
1.5
1
0.5
0
I
I
•
..............
...............k ..............
I
• .............. 4
A
V... -- ..............yi(
t
..............
■
■
K
Test 1
Test 2
Test 3
Test 4
Test 5
Low
0.00
0.00
0.00
0.00
0.00
Normal
0.06
0.06
0.06
0.06
0.06
Moderate
0.46
0.46
0.46
0.46
0.46
Severe
0.87
0.87
0.87
0.87
0.87
CSP MI
1.4
1.23
1.18
1.05
1.08
4.1
CSP NMI
0.82
0.58
0.62
0.64
0.61
•• .A• • GP MI
1.22
0.96
1.11
0.93
0.91
• • •)1(• •
GP NMI
0.50
0.37
0.36
0.31
0.27
SCCF
1.61
1.41
1.42
1.47
1.35
Figure 5. Mean Scores over Time for the PAS Total Score by Group
60
50
Mean Group Score
40
..............
.....
30
..............
20
10
0
Test 1
Test 2
Test 3
Test 4
Test 5
Low/Normal
0
0
0
0
0
Mild/Moderate
16
16
16
16
16
Marked
24
24
24
24
24
Extreme
45
45
45
45
45
CSP MI
31.9
32.4
31.7
31.2
30.1
..,CSP NMI
27.4
26.0
25.9
26.4
25.9
•• .A• • GP MI
32.5
30.5
31.0
28.5
29.0
•• ex • • GP NMI
23.8
23.2
23.0
22.2
22.5
••
35.8
34.2
34.1
34.4
33.6
• • SCCF
RESULTS
47
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 6. Mean Scores over Time for the POMS Total Score by Group
168
Mean Group Score
118
68
18
‐32
•••
:
■ee •••••••••••
•••
•
•••*
Test 1
Test 2
Test 3
Test 4
Test 5
Low
‐32.00
‐32.00
‐32.00
‐32.00
‐32.00
Normal
‐16.00
‐16.00
‐16.00
‐16.00
‐16.00
Moderate
53.00
53.00
53.00
53.00
53.00
Severe
83.00
83.00
83.00
83.00
83.00
68.48
61.77
58.21
58.67
52.02
40.88
29.61
29.68
32.06
27.18
-.-,CSP MI
-.-CSP NMI
•••A••
GP MI
70.68
59.33
64.17
55.81
50.01
•••)K• •
GP NMI
23.99
12.73
13.78
10.44
8.92
77.31
71.07
67.40
71.07
68.43
• ••••• SCCF
Figure 7. Mean Scores over Time for the STAI State Anxiety Score by Group
80
70
60
Mean Group Score
50
•
...
*
.
•
•
000000000000000
40
000000000000000000000000000000
X
X
30
20
Test 1
Test 2
Test 3
Test 4
Test 5
Low
20.00
20.00
20.00
20.00
20.00
Normal
25.32
25.32
25.32
25.32
25.32
Moderate
46.12
46.12
46.12
46.12
46.12
Severe
56.52
56.52
56.52
56.52
56.52
CSP MI
46.90
45.83
45.45
44.01
42.59
CSP NMI
42.05
38.43
37.39
37.50
37.46
•••A••
GP MI
47.64
44.53
47.29
45.08
43.77
•••X••
GP NMI
39.39
36.68
34.65
33.80
33.81
50.15
48.41
48.49
49.41
48.29
••••• SCCF
48
00000000000000
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 8. Mean Scores over Time for the STAI Trait Anxiety Score by Group
80
70
60
O. 00000000000000
Mean Group Score
50
0
.
•
•
4. 0000
•
40
)K•••••
••
••••••••• •
X
X
W
30
20
Test 1
Test 2
Test 3
Test 4
Test 5
Low
20.00
20.00
20.00
20.00
20.00
Normal
25.70
25.70
25.70
25.70
25.70
Moderate
44.08
44.08
44.08
44.08
44.08
Severe
53.27
53.27
53.27
53.27
53.27
CSP MI
48.41
47.89
47.48
45.75
44.64
II
CSP NMI
42.78
38.77
38.40
39.26
38.70
GP MI
49.70
46.59
47.59
45.77
44.07
GP NMI
37.82
35.92
34.44
34.06
32.74
SCCF
54.45
52.64
51.50
52.76
52.27
Il
Figure 9. Mean Scores over Time for the SLUMS Total Score by Group
30
25
••
K. ............. 0(
i
Mean Group Score
20
15
10
5
0
Test 1
Test 2
Test 3
Test 4
Test 5
Dementia
0
0
0
0
0
Mild NC Disorder
20
20
20
20
20
Normal
25
25
25
25
25
CSP MI
20.80
21.16
22.26
22.92
23.59
21.73
22.64
24.02
24.38
24.25
iMoCSP NMI
••.A••
GP MI
21.52
23.09
23.88
24.34
24.93
•••)1(••
GP NMI
23.38
24.12
24.49
24.79
24.82
SCCF
20.54
21.49
22.85
23.38
23.26
RESULTS
49
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 10. Mean Scores over Time for the Trails Task B Time (in Seconds) by Group
125
105
a.........
Mean Group Score
85
65
...s....
000000
_
25
5
1
2
3
4
5
Normal
1
1
1
1
1
Impaired
97
97
97
97
97
84.70
77.74
75.42
71.63
69.43
CSP NMI
83.46
71.09
66.33
62.78
62.27
GP MI
82.47
66.31
66.11
58.29
54.98
GP NMI
68.94
70.63
63.41
56.78
58.81
SCCF
96.29
81.57
78.26
74.69
72.71
—.— CSP MI
50
•••
)K
45
‐15
000000000 0000000000000000000
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
In this section, comparisons were made between each study groups’ mean and the normative mean using a
one sample t test. One sample t tests indicated that, in general, scores were elevated above the normative
data when entering the study and tended to stay that way for all groups except the GP NMI group. Table 10
provides a visual representation of the significant differences by group at each time period on each meas‐
ure.
Table 10. Significant Differences of Study Groups from Normative Means
Norm Norm
Measure
CSP MI
CSP NMI
GP MI
Mean Population
Time Interval: 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 1
BHS
2.32 College
Students
BHS
6.04 Psychiatric
Adult
BSI GSI
0.25 Adult
Males
PAS
16.66 Community
Sample
POMS
14.80 Adult
SLUMS
25.70
STAI‐S
35.72
STAI‐T
34.89
Trails
B/A
2.18
Males
Adult (<HS
education)
Working
Adults
Working
Adults
Adult (25 ‐
54 yrs old)
GP NMI
SCCF
2 3 4 5 1 2 3 4 5
Note. Red shading indicates that the group mean is significantly different from the normative mean in the direction of more psycho‐
logical or cognitive problems, whereas green shading indicates that the group mean is significantly better than the normative mean.
No shading indicates the groups were statistically similar to the normative data.
In addition to comparing group means to normative data, the percentage of participants within each group
who scored in the elevated range of a measure was computed, using cutoff scores from the manual for
moderately severe and above when available or using the percentage scoring beyond two standard devia‐
tions from the mean. Table 11 presents this data.
Table 11. Percentage of Participants Scoring above Cutoffs at Time 1
Measure “Abnormal” range CSP MI CSP NMI GP MI GP NMI
BHS
≥ 9
43%
17%
30%
9%
BSI GSI
≥ 0.46
28%
5%
19%
0%
PAS
≥ 16
92%
90%
90%
80%
POMS
≥ 53
58%
36%
66%
18%
STAI‐S
≥ 46.12
53%
44%
62%
18%
STAI‐T
≥ 44.08
68%
52%
66%
27%
SLUMS
< 20
36%
24%
28%
18%
Trails B/A
≥ 3
40%
50%
53%
33%
SCCF
57%
37%
98%
67%
78%
84%
40%
39%
RESULTS
51
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
CHANGE OVER TIME
To compare change over time on the psychological measures, three sets of analyses comparing mean
change over time were completed—comparisons between the two AS groups (CSP MI vs. CSP NMI) across
the six time periods, comparisons of the NMI groups (CSP NMI vs. GP NMI) across the five common time pe‐
riods, and comparisons of the MI groups (CSP MI vs. SCCF vs. GP MI) across the five common time periods.
Mixed design analysis of variance was used to analyze the data for all participants who had data on the
composite scores. Huyn‐Feldt correction factors were used to adjust the degrees of freedom due to lack of
sphericity for the within subject factors. Partial eta‐square, providing the percentage of variance explained,
was used as an effect size measure (represented by η2 in the tables). An effect was considered small if it ac‐
counted for 1% to 5% of the variance, medium if it accounted for 6% to 14% of the variance, and large if it
accounted for 15% or more of the variance. A significance level of .05 was used to determine a statistically
significant effect. In addition to mean comparisons over time, a slopes analysis was completed in which
slopes were computed for each individual to represent rate of change over time and then comparisons were
made between groups.
As a reminder, higher scores on self‐report composites, Trails derived scores, correctional staff ratings, and
clinician ratings indicate worse performance (e.g., more depression, more anxiety), whereas higher scores
indicate better cognitive performance on the SLUMS. Composites are standardized scores and indicate devi‐
ation from the first assessment period scores.
Comparisons between CSP Groups
A key purpose of the study was to compare segregated inmates with mental illness to those without mental
illness to determine if AS has a differential impact on participants with different mental health needs. Partic‐
ipants were compared across the six time periods. The first assessment was completed while participants
were awaiting a hearing for potential placement in AS. The second assessment occurred within 2 weeks of
being placed in CSP. The third through sixth assessments were completed approximately every three months
following placement in CSP, with the sixth assessment at one year post‐placement in CSP.
Comparisons on Self‐Report Measures. Comparisons between the two CSP groups were made on the seven
composite scores and the two cognitive measures. The summary statistics (mean and standard deviation)
for each group at each time period on the composites and cognitive measures are given in Table 12 and the
inferential statistics (F values and partial eta‐squared) are given in Table 13. Across all seven mental health
composites, the MI group scored statistically higher than the NMI group indicating that there was more psy‐
chological distress for the MI groups. The effect sizes for the differences between groups vary across com‐
posites with large effects for anxiety, depression‐hopelessness, and somatization composites. The NMI
group had significantly higher average scores on the SLUMS measure, although this was a small effect. There
was no significant difference between the groups on the Trails derived score.
52
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 12. Summary Statistics on SelfReport Measures across 6 Time Periods for the Two CSP Groups
Variable
M1
Anxiety
CSP MI (n = 48)
.30
CSP NMI (n = 53)
‐.43
Both Groups
‐.06
Depression‐Hopelessness
CSP MI (n = 48)
.20
CSP NMI (n = 53)
‐.37
Both Groups
‐.09
Hostility‐Anger Control
CSP MI (n = 48)
.20
CSP NMI (n = 53)
‐.15
Both Groups
.03
Hypersensitivity
CSP MI (n = 48)
.24
CSP NMI (n = 53)
‐.21
Both Groups
.01
Psychosis
CSP MI (n = 48)
.18
CSP NMI (n = 53)
‐.26
Both Groups
‐.04
Somatization
CSP MI (n = 48)
.23
CSP NMI (n = 53)
‐.45
Both Groups
‐.11
Withdrawal‐Alienation
CSP MI (n = 48)
.12
CSP NMI (n = 53)
‐.31
Both Groups
‐.10
SLUMS
CSP MI (n = 48)
20.75
CSP NMI (n = 53)
21.85
Both Groups
21.30
Trails B/A
CSP MI (n = 48)
2.98
CSP NMI (n = 53)
3.10
Both Groups
3.04
PBRS Anti‐authority
CSP MI (n = 43)
7.12
CSP NMI (n = 49)
8.06
Both Groups
7.59
PBRS Anxious‐Depressed
CSP MI (n = 41)
6.03
CSP NMI (n = 49)
3.34
Both Groups
4.68
PBRS Dull‐Confused
CSP MI (n = 42)
4.45
CSP NMI (n = 47)
1.91
Both Groups
3.18
PBRS Total
CSP MI (n = 41)
17.73
CSP NMI (n = 49)
13.33
Both Groups
15.53
SD1
.84
.53
.78
.76
.53
.71
.64
.60
.64
.84
.62
.76
.78
.71
.78
.80
.62
.78
.85
.59
.76
5.59
3.49
4.62
.95
1.54
1.30
7.22
7.44
7.32
6.21
4.36
5.42
4.46
2.42
3.74
16.22
12.42
14.36
M2
SD2
M3
SD3
M4
.19
‐.58
‐.19
.08
‐.46
‐.19
‐.01
‐.35
‐.18
.15
‐.53
‐.19
.06
‐.50
‐.22
.12
‐.54
‐.21
.37
‐.15
.11
20.88
22.55
21.70
2.61
2.78
2.70
5.68
5.56
5.62
3.05
2.00
2.52
2.14
1.32
1.73
11.32
9.04
10.18
.83
.56
.80
.72
.58
.70
.63
.52
.60
.88
.67
.84
.78
.61
.75
.79
.64
.78
.89
.84
.90
4.91
3.64
4.35
1.07
.79
.92
5.46
6.41
5.96
3.95
3.06
3.51
2.93
2.16
2.57
10.35
10.22
10.29
.20
‐.55
‐.18
.09
‐.46
‐.18
‐.01
‐.25
‐.13
.23
‐.42
‐.10
.17
‐.46
‐.15
.22
‐.48
‐.13
.29
‐.17
.06
22.35
24.04
23.20
2.61
2.94
2.77
6.67
4.00
5.34
3.44
1.49
2.46
2.71
.87
1.79
13.21
6.47
9.84
.88
.61
.84
.82
.59
.76
.68
.59
.64
.94
.74
.90
.96
.64
.89
.76
.59
.76
.81
.81
.84
4.72
3.28
4.10
.88
1.02
.97
7.78
4.95
6.53
4.53
2.51
3.68
3.15
1.21
2.49
13.72
7.63
11.29
.03
‐.54
‐.25
.07
‐.48
‐.21
‐.07
‐.24
‐.15
.02
‐.30
‐.14
‐.10
‐.40
‐.25
.05
‐.48
‐.21
.33
‐.02
.16
22.75
24.40
23.57
2.44
2.44
2.64
3.40
4.41
3.91
2.66
1.61
2.14
2.82
1.02
1.92
9.42
7.13
8.28
SD4
.89
.60
.80
.87
.56
.77
.74
.59
.69
.89
.75
.83
.88
.70
.80
.77
.55
.71
.81
.82
.83
4.50
2.94
3.83
.98
.98
1.05
5.04
4.87
4.95
3.37
3.45
3.44
3.70
1.50
2.89
9.94
7.63
8.78
M5
SD5
M6
SD6
‐.05
‐.58
‐.32
‐.09
‐.50
‐.30
‐.07
‐.25
‐.16
.02
‐.37
‐.18
.06
‐.42
‐.18
‐.03
‐.53
‐.28
.14
‐.08
.03
23.60
24.34
23.97
2.58
2.58
2.59
4.68
3.20
3.94
2.67
1.77
2.22
2.24
1.05
1.64
9.60
6.32
7.96
.88
.58
.78
.82
.57
.73
.73
.62
.67
.84
.72
.80
.97
.75
.89
.83
.61
.76
.79
.85
.82
4.07
3.23
3.65
.60
.60
.74
6.68
4.45
5.62
3.83
3.31
3.56
3.03
1.59
2.44
11.34
7.41
9.49
‐.17
‐.60
‐.38
‐.17
‐.56
‐.37
‐.19
‐.23
‐.21
‐.07
‐.40
‐.23
‐.09
‐.46
‐.27
‐.16
‐.55
‐.36
.18
‐.04
.07
23.85
25.26
24.56
2.34
2.70
2.52
3.70
2.52
3.11
2.90
1.51
2.21
2.42
1.38
1.90
8.57
5.42
6.99
.78
.59
.72
.73
.55
.67
.74
.68
.70
.80
.65
.74
.91
.64
.80
.69
.60
.67
.85
.78
.82
4.58
2.90
3.84
.74
.88
.84
5.02
4.46
4.74
3.75
2.96
3.93
3.03
2.68
2.88
9.80
8.31
9.10
RESULTS
53
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 13. F Statistics and Partial η2 Comparing AS Groups across 6 Time Periods
Variable
Self‐Report
Anxiety
Depression‐
Hopelessness
Hostility‐Anger
Control
Hypersensitivity
Psychosis
Somatization
Withdrawal‐
Alienation
SLUMS
Trails B/A
Staff Report
PBRS Anti‐
Authority
PBRS Anxious‐
Depressed
PBRS Dull‐
Confused
PBRS Total
BPRS Activity
BPRS Anxious‐
Depressed
BPRS Hostility‐
Suspiciousness
BPRS Thought
Disorders
BPRS Withdrawal
BPRS Total
Group Main Effect
Time Main Effect
2
F(1, 99) = 25.85, p < .001, η = .21
F(3.98, 393.87) = 8.13, p < .001, η = .08
2
F(1, 99) = 18.86, p < .001, η = .16
2
F(1, 99) = 4.08, p = .05, η = .04
2
F(1, 99) = 14.03, p < .001, η = .12
2
F(1, 99) = 13.51, p < .001, η = .12
2
F(1, 99) = 23.63, p < .001, η = .19
2
F(1, 99) = 7.10, p = .01, η = .07
2
F(1, 99) = 3.99, p = .05, η = .04
2
F(1, 91) = 2.74, p = .10, η = .03
2
F(1, 90) = .62, p = .43, η = .01
2
F(1, 88) = 9.46, p = .003, η = .10
2
F(1, 87) = 27.08, p < .001, η = .24
2
F(1, 88) = 7.28, p = .01, η = .08
2
F(1, 82) = 14.04, p < .001, η = .15
Interaction Effect
2
F(3.98, 393.87) = 2.97, p = .02, η = .03
2
F(4.10, 405.75) = 1.12, p = .35, η = .01
2
F(4.08, 403.72) = 2.37, p = .05, η = .02
F(4.10, 405.75) = 6.21, p < .001, η = .06
F(4.08, 403.72) = 4.58, p = .001, η = .04
2
F(4.81, 476.08) = 2.91, p = .02, η = .03
2
F(4.34, 430.18) = 2.79, p = .02, η = .03
2
F(4.34, 429.69) = 6.04, p < .001, η = .06
2
F(4.74, 469.56) = 3.62, p = .004, η = .04
2
F(4.56, 451.82) = 31.78, p < .001, η = .24
2
F(4.14, 376.44) = 4.91, p = .001, η = .05
2
F(4.74, 469.56) = 1.93, p = .09, η = .02
2
F(4.56, 451.82) = .71, p = .60, η = .01
2
F(4.14, 376.44) = .81, p = .52, η = .01
F(4.30, 378.62) = .96, p = .44, η = .01
2
F(3.90, 339.21) = 1.32, p = .26, η = .02
2
F(3.97, 349.18) = .75, p = .56, η = .01
2
F(1.55, 127.12) = .01, p = .99, η < .001
F(3.90, 339.21) = 4.28, p = .002, η = .05
F(3.97, 349.18) = 9.84, p < .001, η = .10
2
F(1.55, 127.12) = 2.46, p = .10, η = .03
F(1.56, 128.19) = 7.93, p = .002, η = .09
2
F(1.77, 145.16) = .33, p = .69, η = .004
2
F(1.99, 163.04) = .81, p = .45, η = .01
2
F(1.84, 151.31) = 2.82, p = .06, η = .03
F(1, 82) = 10.15, p = .002, η = .11
2
F(1, 82) = 36.90, p < .001, η = .31
2
F(4.81, 476.08) = 2.50, p = .03, η = .02
2
F(4.34, 430.18) = 1.49, p = .20, η = .02
2
F(4.34, 429.69) = 2.84, p = .02, η = .03
2
F(4.30, 378.62) = 7.63, p < .001, η = .08
2
F(1, 82) = 21.05, p < .001, η = .20
2
F(4.02, 361.54) = 1.80, p = .13, η = .02
F(2, 163.57) = .91, p = .40, η = .01
F(1, 82) = 18.12, p < .001, η = .18
2
2
F(4.02, 361.54) = 8.87, p = .001, η = .09
2
F(1, 82) = 19.34, p < .001, η = .19
2
2
2
2
2
2
2
2
F(2, 163.57) = 2.16, p = .12, η = .03
2
F(1.56, 128.19) = 1.71, p = .19, η = .02
2
F(1.77, 145.16) = .59, p = .54, η = .01
2
2
F(1.99, 163.04) = .17, p = .84, η = .002
2
F(1.84, 151.31) = 1.34, p = .26, η = .02
2
There were significant main effects of time on all variables; however, the results do not support the pre‐
dicted hypothesis of significant decline in psychological well‐being over time. Figure 11 provides the mean
change over time for each composite. The only variable showing decreased functioning over time was the
withdrawal‐alienation composite. Time 4 (6 months incarcerated in CSP) revealed the highest levels of with‐
drawal‐alienation, followed by a significant decline from time 4 to time 5. For the other psychological va‐
riables, there was improved functioning over time; however, when comparing sequential time periods, the
majority of the variables (i.e., anxiety, depression‐hopelessness, hostility‐anger control, hypersensitivity,
psychosis) only showed statistically significant improvement from the first to the second assessment period.
The exception to this basic pattern was for the somatization composite where statistically significant im‐
provement occurred between periods 5 and 6. For the cognitive variables there were also significant time
effects. The Trails derived score showed significant improvement from the first assessment to the second.
The SLUMS showed significant improvements in cognitive performance between times 2 and 3 and between
times 5 and 6.
54
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 11. Mean Scores over Time for the 7 Composites Summarized across the CSP Groups
1.00 - —0—Anxiety
0.80 0.60
Hostility-Anger Control
—0— Psychosis
—0—Withdrawal-Alienation
—M— Depression-Hopelessness
—0—Hypersensitivity
—0—Somatization
0.40 0.20 -
A 0.00 -0.20 rF
-0.40 -0.50 -0.s0 -1.00
1
2
3
4
5
Time
Although there were significant changes across time for all variables, we were particularly interested in the
group by time interaction to determine if there was differential change across times based on mental health
status. There were statistically significant interactions for the anxiety, hostility‐anger, hypersensitivity, and
somatization composites. Figures 12 to 15 provide graphical representations of these interactions. To fur‐
ther understand these interactions, simple main effects of time were examined for each group using Bonfer‐
roni pairwise comparisons of time periods.
RESULTS
55
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
For the anxiety composite, the CSP NMI group showed no significant change over time, but the CSP MI
group did. Mean scores at the fourth, fifth and sixth assessments were significantly lower than means at the
first three assessment periods, and the sixth assessment mean was significantly lower than the mean at the
fourth assessment.
Figure 12. Mean Scores over Time for the Anxiety Composite for each CSP Group
1.00
C SP MI (n= 48)
--M-- (SP NMI (1I = 53)
0.50 0.40 0.20 0.00 -0.20 -
T
-0.40 -0.60 -0.30 -1.00
1
2
3
4
5
6
Time
For the hostility‐anger control composite, the CSP NMI group showed no significant change over time. In
contrast, the CSP MI group showed significant improvement over time with mean hostility‐anger control
scores significantly elevated at the first assessment compared to all other time periods and the last assess‐
ment period significantly lower than the first three assessment periods.
Figure 13. Mean Scores over Time for the HostilityAnger Control Composite for each CSP Group
1.00 0.80 -
—•—05P MI (n = 48)
0.50 -
--.--05P NMI
(n = 53)
0.40 0.20 0.00 x
-0.20 -
-----411
-0.40 -0.50 -0.30 -1.00
1
2
3
4
5
Thee
56
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
For the hypersensitivity composite, both groups demonstrated significant change over time. The CSP NMI
group showed significant improvement from time 1 to time 2 but then scores worsened over time so that
scores at the fourth assessment were significantly worse than the scores at the second period. For the CSP
MI group, there was a significant decline in scores from the first to the second assessment periods, then an
increase in scores with an elevation occurring at time 3 (compared to time 2), and then a significant decline
in scores at the fourth assessment period.
Figure 14. Mean Scores over Time for the Hypersensitivity Composite for each CSP Group
1.00 .II. CSP MI (n= 48)
...•F CSP NMI (n = 53)
0.80 E 0.50 -
A
m
0.40 -
1
c2
i
0.20 0.00 -
-0.20 -0.40 -
iww wwww
-0.60 -0.80 -1.00
1
2
3
4
5
5
Time
For the somatization composite, there was significant change over time for the CSP MI group but not for the
CSP NMI group. Significant decreases in scores occurred from the third to the fourth assessment periods and
from the fifth to the sixth periods.
Figure 15. Mean Scores over Time for the Somatization Composite for each CSP Group
1.00
S CSP MI (n= 48)
...F..C.SF' NMI (11 =53)
0.80
0.50
2 0.40
c,§ 0.20
1 0.00
E -0.20
8
:-OAO
.....
i -0.50
-0.30
-1.00
1
3
4
Time
RESULTS
5
57
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Comparisons on Staff Report Measures. In addition to self‐report measures, data were collected from cor‐
rectional staff using the PBRS and from clinicians using the BPRS. The PBRS was given at each of the 6 as‐
sessment periods whereas the BPRS was given at 6 month intervals so that there were only three assess‐
ments. Table 12 provides the means and standard deviations for the PBRS scores, Table 14 provides the
summary statistics for the three BPRS assessments, and the inferential results for both variables are in Table
13.
Table 14. Summary Statistics on BPRS Scales across 3 Time Periods for All Study Groups
BPRS Subscale
M1
SD1
M3
SD3
M5
SD5
Activity
CSP MI (n = 49)
6.39 1.72
6.08 1.74
6.00 1.53
CSP NMI (n = 35)
5.60 1.14
5.26
.74
5.20
.53
GP MI (n = 25)
6.40 1.63
5.88 1.20
6.04 1.59
GP NMI (n = 25)
5.64 1.25
5.36
.86
5.28
.54
SCCF (n = 55)
6.85 2.67
6.45 1.48
6.24 1.98
Anxious‐Depressed
CSP MI (n = 49)
9.51 3.11
9.35 2.93
8.47 3.02
CSP NMI (n = 35)
7.37 2.07
6.74 2.24
7.42 3.14
GP MI (n = 25)
8.68 3.13
7.96 2.47
8.40 2.31
GP NMI (n = 25)
6.68 1.77
6.52 1.83
6.08 1.78
SCCF (n = 55)
10.54 3.28
8.87 2.65
8.85 2.98
Hostility‐Suspiciousness
CSP MI (n = 49)
5.51 2.42
5.35 2.80
4.41 1.94
CSP NMI (n = 35)
4.17 1.99
3.37
.69
3.31
.68
GP MI (n = 25)
4.84 1.84
4.52 2.29
4.36 1.93
GP NMI (n = 25)
3.96 1.97
3.60 1.53
3.72 1.67
SCCF (n = 55)
5.53 3.01
4.51 1.91
4.64 2.12
Thought Disorder
CSP MI (n = 49)
6.53 2.34
6.71 2.18
6.35 2.24
CSP NMI (n = 35)
5.43 1.04
5.14
.43
5.23
.55
GP MI (n = 25)
5.64
.99
5.40
.91
5.24
.91
GP NMI (n = 25)
5.20
.50
5.04
.20
5.44 1.44
SCCF (n = 55)
8.40 3.55
6.49 1.91
5.24
.83
Withdrawal
CSP MI (n = 49)
7.59 1.63
7.67 1.98
7.39 1.50
CSP NMI (n = 35)
6.68
.99
7.00 1.37
6.71 1.82
GP MI (n = 25)
7.00 1.55
6.80 1.32
7.16 1.34
GP NMI (n = 25)
6.44
.65
6.20
.50
6.32
.80
SCCF (n = 55)
8.56 2.48
7.69 1.75
7.53 1.49
Total
CSP MI (n = 49)
35.53 7.19
35.16 8.90
32.61 6.81
CSP NMI (n = 35)
29.26 4.85
27.51 3.71
27.89 4.92
GP MI (n = 25)
32.56 5.86
30.56 4.98
31.20 4.17
GP NMI (n = 25)
27.92 4.81
26.72 3.23
26.84 3.75
SCCF (n = 55)
39.89 9.97
34.02 5.55
33.84 6.80
For the correctional staff ratings, there were statistically significant group differences on the Anxious‐
Depressed, Dull‐Confused, and Total scales with the CSP MI group scoring significantly higher on each subs‐
cale compared to the NMI group. There were significant changes across time for both groups with the first
assessment showing higher ratings compared to the second assessment period ratings on all PBRS scales.
Additionally, there was a statistically significant drop in Anti‐Authority scores from the third rating to the
58
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
fourth rating. There were no group by time interactions, indicating that change over time was the same for
the two CSP groups.
As might be expected, the CSP MI group was elevated on each of the clinical rating scales of the BPRS com‐
pared to the CSP NMI group. There was only significant change across time on the Hostility‐Suspiciousness
subscale with scores at the last time period (M = 3.86, SE = .17) significantly lower than the first (M = 4.84,
SE = .25) and middle assessment (M = 4.36, SE = .24) period means. There were no statistically significant
group by time interaction effects.
Comparisons between NMI Groups
A significant advantage of this study is the use of comparison groups to determine if the CSP groups change
over time differentially compared to similar groups of inmates who are not placed in AS. In the following
analyses, participants without mental health issues are compared in order to determine if those in AS
change over time on the measures differentially compared to those in the general prison population (CSP
NMI vs. GP NMI). (A later section compares the participants who have been identified as mentally ill.) The
groups are compared on the five common time assessments. Mixed design analysis of variance was used to
compare change across time and between groups.
Comparisons on Self‐Report Measures. The summary statistics for the groups are provided in Table 15 and
the analysis of variance results and effect sizes are provided in Table 16. For anxiety, depression‐
hopelessness, hostility‐anger control, hypersensitivity, psychosis, and somatization composites, there were
statistically significant group differences between the groups with the CSP NMI scoring significantly higher
than the GP NMI group. For the withdrawal‐alienation composite, the SLUMS cognitive measure, and the
Trails derived score, there were no statistically significant differences between groups.
RESULTS
59
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 15. Summary Statistics on Measures across 5 Time Periods for the NMI Groups
Variable
Anxiety
CSP NMI (n = 55)
GP NMI (n = 38)
Depression‐Hopelessness
CSP NMI (n = 55)
GP NMI (n = 38)
Hostility‐Anger Control
CSP NMI (n = 55)
GP NMI (n = 38)
Hypersensitivity
CSP NMI (n = 55)
GP NMI (n = 38)
Psychosis
CSP NMI (n = 55)
GP NMI (n = 38)
Somatization
CSP NMI (n = 55)
GP NMI (n = 38)
Withdrawal‐Alienation
CSP NMI (n = 55)
GP NMI (n = 38)
SLUMS
CSP NMI (n = 55)
GP NMI (n = 38)
Trails B/A
CSP NMI (n = 55)
GP NMI (n = 38)
PBRS Anti‐authority
CSP NMI (n = 51)
GP NMI (n = 22)
PBRS Anxious‐Depressed
CSP NMI (n = 51)
GP NMI (n = 20)
PBRS Dull‐Confused
CSP NMI (n = 49)
GP NMI (n = 20)
PBRS Total
CSP NMI (n = 51)
GP NMI (n = 20)
M1
SD1
‐.44
.53
‐.71
.47
‐.39
.52
‐.73
.35
‐.13
.61
‐.34
.49
‐.25
.65
‐.50
.68
‐.28
.71
‐.60
.72
‐.46
.62
‐.61
.51
‐.30
.60
‐.45
.78
21.74
3.46
23.16
3.97
3.11
1.52
2.82
.82
7.75
7.75
5.58
7.23
3.21
4.32
2.70
3.66
1.84
2.40
1.71
2.13
12.81 12.44
10.22 12.32
M2
SD2
M3
SD3
M4
‐.59
‐.82
.56
.40
‐.56
‐.85
.60
.40
‐.55
‐.86
‐.47
‐.82
.57
.34
‐.47
‐.84
.58
.39
‐.50
‐.81
‐.34
‐.45
.52
.47
‐.23
‐.48
.59
.54
‐.23
‐.54
‐.54
‐.66
.66
.62
‐.44
‐.73
.74
.68
‐.31
‐.64
‐.51
‐.81
.60
.53
‐.47
‐.81
.63
.64
‐.41
‐.77
‐.56
‐.77
.63
.41
‐.50
‐.77
.63
.41
‐.50
‐.72
‐.12
‐.32
.85
.68
‐.15
‐.42
.82
.70
.00
‐.32
22.53
23.92
3.66
3.26
23.98
24.47
3.30
3.55
24.34
24.71
2.84
3.07
.93
1.64
3.00
2.90
1.22
1.04
2.88
2.65
5.46
7.02
6.30
5.41
3.92
6.69
4.87
7.24
4.38
7.75
1.92
2.80
3.02
3.49
1.43
3.69
2.48
3.90
1.55
3.86
1.26
1.62
2.13
1.77
.86
.09
1.19
2.08
.98
2.10
8.80
11.39
10.09
7.23
6.31
12.82
7.52
11.48
6.99
13.52
SD4
.59
.35
M5
‐.58
‐.88
SD5
.55
.29
.64
.50
.75
.51
.69
.53
.56
.49
.82
.83
3.05
3.69
1.14
.76
4.79
5.91
4.40
5.28
1.48
2.65
7.51
10.98
.50
‐.85
‐.24
‐.50
‐.36
‐.73
‐.40
‐.86
‐.52
‐.76
‐.07
‐.32
24.18
24.82
2.71
2.80
3.40
7.93
1.84
3.10
1.12
2.07
6.64
13.55
.57
.30
.58
.40
.64
.51
.72
.56
.75
.64
.62
.49
.84
.80
3.33
3.24
1.02
.95
4.48
6.92
3.27
4.05
1.60
2.83
7.47
11.77
60
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 16. F Statistics and Partial η2 Comparing NMI Groups across 5 Time Periods
Variable
Self‐Report
Anxiety
Depression‐
Hopelessness
Hostility‐Anger
Control
Hypersensitivity
Psychosis
Somatization
Withdrawal‐
Alienation
SLUMS
Trails B/A
Staff Report
PBRS Anti‐
Authority
PBRS Anxious‐
Depressed
PBRS Dull‐
Confused
PBRS Total
BPRS Activity
BPRS Anxious‐
Depressed
BPRS Hostility‐
Suspiciousness
BPRS Thought
Disorder
BPRS Withdrawal
BPRS Total
Group Main Effect
Time Main Effect
2
F(1, 91) = 8.74, p = .004, η = .09
F(3.03, 275.73) = 4.52, p = .004, η = .05
F(3.03, 275.73) = .28, p = .84, η = .003
F(3.86, 351.74) = 4.74, p = .001, η2 = .05
F(3.86, 351.74) = 1.45, p = .22, η = .02
2
F(1, 91) = 15.24, p < .001, η = .14
Interaction Effect
2
2
F(3.50, 318.75) = 2.77, p = .03, η = .03
2
F(3.79, 345.07) = 4.70, p = .001, η = .05
2
F(3.09, 281.17) = 5.11, p = .001, η = .05
2
F(3.43, 312.15) = 2.42, p = .06, η = .03
2
F(3.88, 352.78) = 3.49, p = .01, η = .04
2
F(3.77, 343.26) = 15.33, p < .001, η = .14
2
F(3.53, 303.21) = 1.25, p = .29, η = .01
2
F(3.43, 243.59) = .60, p = .64, η = .01
2
F(3.62, 249.81) = .30, p = .86, η = .004
2
F(3.59, 240.56) = .23, p = .91, η = .003
2
F(3.37, 232.76) = .40, p = .77, η = .01
2
F(1.43, 83.11) = 4.60, p = .01, η = .07
2
F(1.76, 101.84) = .81, p = .43, η = .01
F(1, 91) = 4.69, p = .03, η = .05
F(1, 91) = 5.18, p = .02, η = .05
2
F(1, 91) = 8.67, p = .004, η = .09
2
F(1, 91) = 4.75, p = .03, η = .05
F(1, 91) = 2.79, p = .10, η = .03
F(1, 91) = 2.09, p = .15, η = .02
2
F(1, 86) = .13, p = .72, η = .001
F(1, 71) = 4.67, p = .03, η = .06
F(1, 69) = 4.63, p = .04, η = .06
F(1, 67) = 7.61, p = .01, η = .10
F(1, 69) = 6.54, p = .01, η = .09
2
F(1, 58) = .19, p = .67, η = .003
F(1, 58) = 2.66, p = .11, η = .04
2
2
2
2
2
F(3.88, 352.78) = .52, p = .72, η = .01
2
F(3.77, 343.26) = 1.17, p = .32, η = .01
2
F(3.53, 303.21) = 1.29, p = .28, η = .02
F(3.62, 249.81) = 1.92, p = .12, η = .03
2
F(3.59, 240.56) = 1.45, p = .22, η = .02
2
F(3.37, 232.76) = 3.22, p = .02, η = .04
2
F(1.43, 83.11) = .03, p = .93, η < .001
2
F(1.76, 101.84) = 1.59, p = .21, η = .03
2
F(1.47, 85.54) = 1.15, p = .31, η = .02
F(1.89, 109.66) = 2.54, p = .09, η = .04
2
2
F(3.79, 345.07) = 1.21, p = .31, η = .01
2
F(3.09, 281.17) = .65, p = .59, η = .01
2
F(3.43, 312.15) = .69, p = .58, η = .01
2
2
F(1, 58) = 5.95, p = .02, η = .09
2
F(1, 58) = 1.35, p = .25, η = .02
2
F(3.50, 318.75) = .13, p = .96, η = .001
F(3.43, 243.59) = 3.64, p = .01, η = .05
F(1.47, 85.54) = 4.83, p = .02, η = .08
F(1, 58) = .07, p = .78, η = .001
2
2
2
F(1, 58) = .22, p = .64, η = .004
2
2
2
F(1.73, 100.26) = .08, p = .90, η = .001
2
F(1.83, 106.28) = 3.69, p = .03, η = .06
2
2
2
2
2
2
2
F(1.89, 109.66) = 1.79, p = .17, η = .03
2
F(1.73, 100.26) = .97, p = .37, η = .02
2
F(1.83, 106.28) = .11, p = .88, η = .002
For all variables except the somatization composite and Trails derived score, there were statistically signifi‐
cant changes across time; however, there were not any significant group by time interactions, indicating
that the two groups changed similarly across time. Figure 16 gives the mean change over time on the com‐
posites (summarized across the two NMI groups). For all composite variables except withdrawal‐alienation,
the pattern of change was the same when examining differences between sequential time periods. There
were statistically significant improvements in reported psychological well‐being from the first to the second
assessment but no other significant differences between time periods. The withdrawal‐alienation composite
was the only variable that showed significantly higher scores over time, with statistically significant change
on average from the first to second assessments and from the third to the fourth assessments. The SLUMS
also showed change across time with significant improvement from the first (M = 22.45, SE = .39) to second
(M = 23.22, SE = .27) assessment and from the second to third (M = 24.23, SE = .36) assessment.
RESULTS
61
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 16. Mean Scores over Time for the 7 Composites Summarized across the NMI Groups
1.00
0.80
—.—Anxiety
Hostility-Anger Control
—•—Psychosis
—•—Withdrawal-Alienation
—M—Depresslon-Hopelessness
—Hypersensitivity
—0—Somatizaton
0.60
0.40
0.20
0.00
-0.20
-0.40
-0.60
-0.80
-1.00
1
3
Time
4
5
Comparisons on Staff Report Measures. The summary statistics for the correctional officer ratings on the
PBRS are given in Table 15, the summary statistics for the clinician ratings are given in Table 14, and the
analysis of variance results for the staff report comparisons are given in Table 16.
For the correctional officer ratings, there are significant group differences on all four PBRS scales. When av‐
eraged over time, the CSP NMI group scored significantly lower than the GP NMI group on each of the four
scales. There were no significant main effects of time on any of the scales and no significant interaction ef‐
fects for the Anxious‐Depressed or Dull‐Confused scales; however, there were significant interaction effects
for Anti‐Authority and Total scales. These interactions are displayed in Figures 17 and 18. For both scales,
the same basic pattern occurs with the CSP NMI scores tending to decrease across time with significant
drops from the first to the second assessment and with the GP NMI scores showing no significant change
across time (although scores tend to increase).
62
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 17. Mean Scores over Time for the PBRS AntiAuthority Subscale for each NMI Group
21 —4—GP NMI (n = 22)
—--05P NMI (n = 51)
18 -
1 15
1
1 12 -
3
t
•--------*
t
S........il. ...... ..5 ......se
i
1
2
3
4
5
Time
Figure 18. Mean Scores over Time for the PBRS Total Scale for each NMI Group
21
--•—.GP NMI (n = 20)
--M--05P NMI (n = 51)
.s...
ophouswIl<
IIAT41
.18....
1.1.1.1.1.1.1.1.°411
".. ....... .
uhuis qh
1
1
•
1
2
3
4
5
Time
For the clinician ratings, there was a significant group difference on the Withdrawal subscale of the BPRS
with the CSP NMI group (M = 6.80, SE = .13) rated significantly higher compared to the GP NMI group (M =
6.32, SE = .15). No other BPRS subscales had statistically significant group differences. There were significant
time effects on Activity, Hostility‐Suspiciousness, and Total scores but no significant interaction effects for
any of the BPRS subscales. For the Activity subscale, ratings at the first assessment (M = 5.62, SE = .16) were
significantly higher than ratings at the third assessment (M = 5.24, SE = .10) but not different from ratings at
RESULTS
63
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
the second assessment (M = 5.24, SE = .07). For the Hostility‐Suspiciousness subscale, the ratings at the first
assessment (M = 7.03, SE = .26) were significantly greater than ratings at the second (M = 6.63, SE = .27) and
the third (M = 6.75, SE = .35) assessment periods. The total score showed this same pattern with first as‐
sessment (M = 28.59, SE = .63) ratings significantly higher than second (M = 27.12, SE = .47) and third (M =
27.36, SE = .59) periods.
Comparisons between MI Groups
In the following analyses, the three groups with participants identified as mentally ill are compared. Like the
comparisons between the NMI groups, there is a CSP MI group and a GP MI group plus a third group of in‐
mates who have been placed in a psychiatric treatment facility (SCCF). Analyses were completed on the five
common time periods using mixed design analysis of variance techniques.
Comparisons on Self‐Report Measures. The summary statistics for the groups are provided in Table 17 and
the analysis of variance results and effect sizes are provided in Table 18. Significant group differences were
found on the anxiety, depression‐hopelessness, psychosis, somatization, and withdrawal‐alienation compo‐
sites. Using Bonferroni corrected pairwise comparisons, the SCCF group was always significantly higher than
the GP MI group on these composites. Additionally, the SCCF group was significantly higher than the CSP MI
group for the depression‐hopelessness, psychosis, and withdrawal‐alienation composites but not significant‐
ly different for anxiety and somatization composites. The GP MI and CSP MI groups did not show any statis‐
tically significant mean differences although the CSP MI group always had a higher mean. There were not
significant group differences on the hostility‐anger control and hypersensitivity composites or on the cogni‐
tive measures (SLUMS and Trails derived).
64
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 17. Summary Statistics on Measures across 5 Time Periods for the MI Groups
Variable
M1
Anxiety
CSP MI (n = 55)
.26
GP MI (n = 28)
.11
SCCF (n = 55)
.51
Depression‐Hopelessness
CSP MI (n = 55)
.19
GP MI (n = 28)
.00
SCCF (n = 55)
.64
Hostility‐Anger Control
CSP MI (n = 55)
.20
GP MI (n = 28)
.11
SCCF (n = 55)
.02
Hypersensitivity
CSP MI (n = 55)
.16
GP MI (n = 28)
.13
SCCF (n = 55)
.32
Psychosis
CSP MI (n = 55)
.16
GP MI (n = 28)
.02
SCCF (n = 55)
.46
Somatization
CSP MI (n = 55)
.23
GP MI (n = 28)
.07
SCCF (n = 55)
.46
Withdrawal‐Alienation
CSP MI (n = 55)
.15
GP MI (n = 28)
.08
SCCF (n = 55)
.43
SLUMS
CSP MI (n = 55)
20.80
GP MI (n = 28)
21.36
SCCF (n = 55)
20.96
PBRS Trails B/A
CSP MI (n = 55)
2.99
GP MI (n = 28)
3.23
SCCF (n = 55)
2.97
PBRS Anti‐Authority
CSP MI (n = 50)
7.04
GP MI (n = 16)
5.31
SCCF (n = 41)
2.85
PBRS Anxious‐Depressed
CSP MI (n = 49)
5.96
GP MI (n = 16)
2.31
SCCF (n = 41)
5.15
PBRS Dull‐Confused
CSP MI (n = 49)
3.94
GP MI (n = 16)
1.69
SCCF (n = 41)
3.50
PBRS Total
CSP MI (n = 49)
17.30
GP MI (n = 16)
9.38
Both Groups
11.70
RESULTS
SD1
.84
.77
.71
.78
.76
.89
.67
.68
.69
.85
.92
.78
.80
.80
.84
.82
.71
.67
.83
.88
.83
5.44
4.18
3.55
1.04
1.30
1.16
6.96
4.61
5.60
6.58
3.53
4.77
4.32
2.15
3.92
15.89
9.62
13.46
M2
SD2
.14
‐.09
.35
.07
‐.19
.47
.00
.01
‐.08
.08
‐.02
.17
.04
‐.29
.31
.14
‐.10
.40
.34
.03
.53
21.20
23.11
21.81
2.74
2.81
2.95
6.66
6.48
4.00
3.86
4.32
5.43
2.71
2.50
3.54
13.63
12.35
13.26
M3
.82
.65
.81
.74
.62
.93
.67
.65
.73
.87
.71
.78
.78
.73
.94
.83
.65
.68
.88
.78
.75
4.86
3.66
4.24
1.15
.68
1.13
6.34
5.05
4.73
5.23
6.30
5.05
3.57
3.14
2.85
13.63
12.68
10.59
.12
‐.06
.27
.04
‐.18
.35
‐.03
‐.04
‐.09
.11
.60
.14
.09
‐.31
.31
.21
‐.03
.29
.25
.12
.56
22.27
23.71
23.11
2.67
2.75
2.90
6.75
8.58
3.51
3.45
5.00
5.06
2.63
2.62
3.16
13.22
16.28
11.97
SD3
M4
.86
.59
.88
.80
.53
.98
.69
.58
.65
.94
.65
.91
.95
.65
.88
.79
.83
.81
.83
.68
.75
4.68
2.99
4.37
.89
.85
1.16
7.60
10.14
4.18
4.31
4.63
4.46
3.04
3.28
2.94
13.25
15.04
9.31
SD4
‐.01
‐.20
.32
.03
‐.22
.39
‐.09
‐.08
‐.05
‐.08
‐.14
.20
‐.14
‐.46
.33
.09
‐.24
.30
.30
‐.06
.50
22.84
24.32
23.52
2.60
2.76
2.80
3.56
7.00
5.67
2.94
2.89
6.84
2.82
1.81
4.20
9.20
11.72
16.99
M5
.86
.56
.86
.84
.59
.95
.74
.58
.79
.90
.66
.77
.87
.52
.95
.81
.53
.83
.81
.71
.75
4.38
3.73
4.07
1.05
.87
.98
5.40
6.75
6.64
3.44
3.83
6.35
3.64
2.95
3.99
9.89
11.77
15.90
SD5
‐.10
‐.22
.21
‐.11
‐.30
.33
‐.12
‐.08
‐.10
‐.05
‐.11
.20
‐.02
‐.39
.22
‐.01
‐.14
.31
.18
‐.02
.55
23.51
24.96
23.35
2.67
2.62
2.74
4.64
7.56
5.70
2.88
3.62
6.82
2.26
1.25
4.30
9.93
12.44
17.22
.86
.61
.76
.81
.54
.86
.72
.55
.71
.81
.69
.75
.94
.64
.86
.85
.59
.76
.80
.62
.73
4.03
3.29
4.04
.76
.92
1.29
6.35
9.77
6.94
3.90
5.20
6.01
3.12
1.95
4.24
10.98
15.05
15.58
65
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 18. F Statistics and Partial η2 Comparing MI Groups across 5 Time Periods
Variable
Self‐Report
Group Main Effect
Anxiety
Depression‐
Hopelessness
Hostility‐Anger
Control
Hypersensitivity
Psychosis
Somatization
Withdrawal‐
Alienation
SLUMS
Trails B/A
F(2, 135) = 3.97, p = .02, η = .06
F(3.67, 495.07) = 9.11, p < .001, η = .06
F(2, 135) = 8.44, p < .001, η2 = .11
F(3.52, 475.64) = 6.76, p < .001, η = .05
Staff Report
PBRS Anti‐
Authority
PBRS Anxious‐
Depression
PBRS Dull‐
Confused
PBRS Total
BPRS Activity
BPRS Anxious‐
Depressed
BPRS Hostility‐
Suspiciousness
BPRS Thought
Disorder
BPRS Withdrawal
BPRS Total
Time Main Effect
2
2
F(2, 135) = .13, p = .88, η = .002
Interaction Effect
2
F(7.34, 495.07) = .42, p = .90, η = .01
2
F(7.04, 475.64) = .24, p =.98, η = .004
2
F(6.90, 465.81) = .70, p = .67, η = .01
F(3.45, 465.81) = 5.49, p = .001, η = .04
2
F(3.92, 529.43) = 2.80, p = .02, η = .02
2
F(3.82, 148.17) = 6.55, p < .001, η = .05
2
F(3.87, 522.39) = 4.83, p = .001, η = .04
2
F(4.00, 539.26) = .79, p = .53, η = .01
F(2, 135) = 1.41, p = .25, η = .02
2
F(2, 135) = 7.28, p = .001, η = .10
2
F(2, 135) = 4.26, p = .02, η = .06
F(2, 135) = 5.51, p = .01, η = .08
2
F(2, 134) = 1.32 p = .27, η = .02
2
F(2, 128) = .47, p = .63, η = .01
2
F(2, 104) = 2.56, p = .08, η = .05
2
F(2, 103) = 5.92, p = .004, η = .10
2
F(2, 103) = 5.03, p = .01, η = .09
2
F(2, 103) = .40, p = .67, η = .01
2
F(2, 126) = 1.36, p = .26, η = .02
2
F(2, 126) = 2.32, p = .10, η = .04
2
F(2, 126) = .73, p = .48, η = .01
2
F(2, 126) = 9.91, p < .001, η = .14
2
F(2, 126) = 5.46, p = .005, η = .08
2
F(2, 126) = 7.10, p = .001, η = .10
2
2
2
F(3.69, 494.42) = 28.64, p < .001, η = .18
2
F(4, 512) = 4.50, p = .001, η = .03
2
2
2
2
F(7.84, 529.43) = .62, p = .76, η = .01
2
F(7.66, 148.17) = .97, p = .46, η = .01
2
F(7.74, 522.39) = .89, p = .52, η = .01
2
F(7.99, 539.26) = .77, p = .63, η = .01
2
F(7.38, 494.42) = .76, p = .63, η = .01
2
F(8, 512) = .62, p = .76, η = .01
2
F(7.17, 372.72) = 3.30, p = .002, η = .06
2
F(7.25, 373.16) = 2.97, p = .004, η = .05
2
F(6.91, 355.98) = 1.38, p = .21, η = .03
2
F(6.85, 352.95) = 3.46, p = .001, η = .06
2
F(3.63, 228.85) = .16, p = .95, η = .003
F(3.58, 372.72) = .63, p = .62, η = .01
F(3.62, 373.16) = .06, p = .99, η = .001
F(3.46, 355.98) = .23, p = .90, η = .002
F(3.43, 352.95) = .13, p = .96, η = .001
2
F(1.82, 228.85) = 2.78, p = .07, η = .02
2
F(2, 252) = 5.15, p = .01, η = .04
2
2
2
2
2
F(4, 252) = 1.66, p = .16, η = .03
2
F(3.84, 242.10) = 1.31, p = .27, η = .02
2
F(3.23, 203.27) = 4.54, p = .003, η = .07
F(1.92, 242.10) = 5.40, p = .01, η = .04
F(1.61, 203.27) = 5.50, p = .01, η = .04
2
F(1.99, 250.81) = 1.76, p = .17, η = .01
2
F(1.85, 232.85) = 8.94, p < .001, η = .07
2
2
2
F(3.98, 250.81) = 2.02, p = .09, η = .03
2
F(3.70, 232.85) = 2.95, p = .02, η = .04
There were significant changes across times for all composites except the withdrawal‐alienation composite.
The hostility‐anger control composite also showed a significant interaction indicating differential change
across time between groups. Figure 19 provides the mean plot demonstrating change across time for the
five composites that had a significant time effect but no interaction effect. For the anxiety and depression‐
hopelessness composites, there were significant decreases in mean scores from the first to second assess‐
ment periods. For the psychosis and somatization composites, there were significant decreases in mean rat‐
ings from the first to second and from the third to fourth assessment periods. The hypersensitivity compo‐
site had a significant time effect, but the comparison of sequential time periods showed no significant
change (pairwise comparisons indicated that the first assessment mean was significantly higher than mean
scores at the fourth and fifth periods).
66
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 19. Mean Scores for the 7 Composites over Time Summarized across MI Groups
Anxiety
Depression‐Hopelessness
.1 , Hypersensitivity
Psychosis
Hostility‐Anger
.111 Somatization
Mean Composite Score
Withdrawal‐Alienation
,.
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
‐0.05
‐0.1
‐0.15
-
....
.....,...,..
.:
-
1
2
3
4
5
Assessment Period
For the SLUMS cognitive assessment, there were significant increases in performance from the first (M =
21.04, SE = .40) to second (M = 22.04, SE = .39) and from the second to third (M = 23.03, SE = .38) assess‐
ment periods. For the Trails derived score there were significant improvements in performance from the
first (M = 3.08, SE = .10) to second (M = 2.83, SE = .10) assessment periods (indicated by a decrease in mean
scores).
The significant interaction for the hostility‐anger control composite is graphed in Figure 20. There was signif‐
icant change over time for the CSP MI group with time 1 scores significantly higher than all other assessment
periods. There were not significant changes over time for the SCCF and GP MI groups.
RESULTS
67
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 20. Mean Change over Time on the HostilityAnger Control Composite for each MI Group
1.0
tCSP MI (n = 55)
—M—GP MI (n = 28)
—A-5CCF(n = 55)
0.3
E
A
A
0.6
0.4
1 0.2
a 0.0
i
li
-0.2
llimm.--.N•diaimi,_. —__7_1
-0.4
-0.6
-0.8
-1.0
1
2
3
4
5
Time
Comparisons on Staff Report Measures. The summary statistics for the correctional staff ratings on the PBRS
are given in Table 17, the summary statistics for the clinician ratings are given in Table 14, and the analysis
of variance results for the staff report comparisons are given in Table 18.
For the correctional staff ratings, there were significant group differences for PBRS Anxious‐Depressed and
Dull‐Confused subscale scores. The SCCF group (Mad = 5.86, SE = .48; Mdc = 3.74, SE = .31) scored significantly
higher than the both the CSP MI (Mad = 3.82, SE = .44; Mdc = 2.87, SE = .28) and the GP MI (Mad = 3.63, SE =
.76; Mdc = 1.98, SE = .50) groups on both subscales. The Anti‐Authority rating scale did not show statistically
significant group differences (p = .08) but there was a small to moderate effect size (η2 = .05); the only signif‐
icant difference was between the GP MI and the SCCF groups (p = .04). There were no statistically significant
time effects for any of the subscales of the correctional officer ratings; however, there were significant
group by time interactions for the Anti‐Authority, Anxious‐Depressed, and Total scores. Figures 21 to 23
demonstrate the interaction for these three variables.
Further analyses of the interaction effects showed the same pattern of significance – there were significant
changes over time for the CSP MI group but not for the other two groups. Specifically, for the Anti‐Authority
subscale, the fourth assessment had lower scores than the first three assessment periods; for the Anxious‐
Depressed subscale, the first assessment was higher than all other assessment periods; and for the Total
PBRS scale, the first assessment was significantly higher than the fourth and fifth assessments. There were
not significant changes over time for the other two groups; however it is noteworthy that these scores
tended to increase over time.
68
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 21. Mean Scores over Time for the PBRS AntiAuthority Subscale for each MI Group
—•—051) Mr (n = 50)
—M—GP MI {n = 16)
SCCF (n 41)
Figure 22. Mean Scores over Time for the PBRS AnxiousDepressed Subscale for each MI Group
—•—05P MI (n = 49)
—M—GP MI (n = 16)
—A-5CCF (n = 41)
RESULTS
69
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 23. Mean Scores over Time for the PBRS Total Scale for each MI Group
tCSP MI (n = 49)
--GP MI (n = 16)
—'—SCCF(n = 41)
For clinician ratings on the BPRS, there were significant mean group differences for the Total scale, the
Thought Disorder subscale, and the Withdrawal subscale. The GP MI group (M = 31.44, SE = .99) had signifi‐
cantly lower means on the total scale compared to both the CSP MI (M = 34.44, SE = .70) and the SCCF (M =
35.92, SE = .66) groups, but there was not a significant difference between the CSP MI and SCCF groups. All
three groups were significantly different from each other on the Thought Disorder subscale with the GP MI
(M = 5.43, SE = .32) having the lowest scores followed by CSP MI group (M = 6.53, SE = .23) and then the
SCCF group (M = 7.16, SE = .22). For the Withdrawal subscale, the SCCF group had significantly higher means
(M = 7.93, SE = .16) compared to both the CSP MI (M = 7.55, SE = .17) and the GP MI (M = 6.99, SE = .24)
groups, but there was not a significant difference between the CSP MI and GP MI groups.
Time effects were statistically significant for all BPRS scales except Activity and Withdrawal subscales; how‐
ever, there were also significant interactions for Thought Disorder and Total scales. Figure 24 provides the
means for change over time for the Activity, Anxious‐Depressed, Hostility‐Suspiciousness, Though Disorder,
and Withdrawal subscales of the BPRS. For the Anxious‐Depressed and Hostility‐Suspiciousness scales, mean
ratings at the first assessment were greater than mean ratings at the third assessment for all three subs‐
cales. Additionally, for the Anxious‐Depressed subscale, the mean rating at the first assessment was also
significantly greater than the mean at the second assessment.
70
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 24. Mean Scores for the BPRS Subscales over Time Summarized across MI Groups
Activity
Anxious‐Depressed
Thought Disorder
Withdrawal2
Hostility‐Suspiciousness
Mean BPRS Subscale Score
12
10
8
6
4
2
0
1
2
3
Assessment Period
Figures 25 and 26 are mean plots to demonstrate the interaction effects for BPRS Thought Disorder subscale
and BPRS Total scores. Simple main effects for each group on the Thought Disorder subscale indicate that
there are significant changes over time for the SCCF group but not for the other two groups. The SCCF group
had significantly higher scores at the first assessment compared to the other two assessment periods.
Figure 25. Mean Scores for the BPRS Thought Disorder Subscale over Time for each MI Group
4
.,
7-
A
5
-
—•—05P MI (n = 49)
—M—GP MI (n = 25)
—A—SCCF(n = 55)
0
1
2
3
Time
For BPRS Total scores, the GP group does not change significantly over time but both of the other groups
have significant time effects. In particular, the last assessment scores for the CSP MI group were significantly
lower than the first two assessment periods, and the first assessment scores for the SCCF group were signifi‐
cantly higher than each of the other assessment periods.
RESULTS
71
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 26. Mean Scores for the BPRS Total Scale over Time for each MI Group
45
40
35
30
25
d 20
a 15
—0—05P MI (n = 49)
—M—GP MI (n = 25)
—A—SCCF(n = 55)
10
5
0
1
Z
3
Time
Slopes Analysis
To compare change over time in another way, slopes analyses were conducted in addition to the means ana‐
lyses. For these analyses, a slope and intercept were computed for each person on each composite using the
available time periods for anyone who completed two or more assessments. These slopes and intercepts
were then compared across groups. If AS was impacting change across time, we would expect slopes to be
different across study groups. We also computed an intercept value for each person on each self‐report va‐
riable; these intercepts were computed so that they represented an estimated value at initial assessment.
Thus differences in groups would indicate different starting points. As a reminder, for all dependent va‐
riables except the SLUMS, lower scores indicate better performance. Thus a positive slope would indicate a
worsening of psychological well‐being over time and a negative slope would indicate an improvement over
time. Similarly, a positive or larger intercept value indicates higher psychological distress (or lower cognitive
functioning) at the outset compared to lower (or negative) values for all measures except the SLUMS.
Table 19 gives the means and standard deviations for the slopes and intercepts for each group on each
composite and Table 20 provides the statistical results from a one‐way analysis of variance on each variable
comparing if there are group differences in mean slopes and intercepts.
72
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 19. Summary Statistics on Slopes and Intercepts for each SelfReport Variable for the Study Groups
CSP MI
M
SD
Anxiety
Intercept
.27
.82
Slope
‐.11
.17
Depression‐Hopelessness
Intercept
.17
.70
Slope
‐.07
.13
Hostility‐Anger Control
Intercept
.16
.64
Slope
‐.07
.14
Hypersensitivity
Intercept
.20
.78
Slope
‐.09
.18
Psychosis
Intercept
.16
.72
Slope
‐.08
.19
Somatization
Intercept
.25
.78
Slope
‐.08
.17
Withdrawal‐Alienation
Intercept
.26
.76
Slope
‐.02
.18
SLUMS
Intercept
20.74 4.71
Slope
.65
.97
Trails B/A
Intercept
2.94
.94
Slope
‐.10
.22
Variable
CSP NMI
M
SD
‐.48
.56
‐.05
.20
‐.38
.55
‐.04
.15
‐.17
.57
‐.02
.14
‐.33
.67
‐.05
.26
‐.34
.65
‐.05
.23
‐.44
.68
‐.03
.17
‐.19
.71
.05
.18
22.01 3.29
.78
.87
3.11 1.15
‐.06
.30
GP MI
M
SD
.10
.73
‐.06
.20
.003
.65
‐.05
.22
.06
.64
‐.04
.15
.16
.81
‐.04
.21
‐.03
.79
‐.08
.29
.08
.68
‐.05
.16
.15
.78
‐.01
.23
21.93 3.62
.81
.82
3.14
.87
‐.15
.36
GP NMI
M
SD
‐.68
.48
.03
.10
‐.71
.40
‐.01
.10
‐.37
.50
‐.03
.09
‐.48
.72
‐.05
.15
‐.63
.63
‐.04
.10
‐.59
.52
‐.04
.10
‐.37
.73
.02
.17
23.53 3.16
.38
.70
3.08 1.09
‐.07
.28
SCCF
SD
.44
.69
‐.07
.19
.54
.84
‐.21
.21
‐.01
.65
‐.04
.18
.18
.66
.03
.26
.38
.80
‐.06
.28
.46
.66
‐.06
.27
.37
.81
.05
.34
20.72 3.83
.83 1.42
3.05
.99
‐.98
.40
All
M
M
‐.02
‐.06
‐.02
‐.05
‐.05
‐.04
‐.04
‐.04
‐.05
‐.06
‐.01
‐.05
.07
.02
21.61
.70
3.05
‐.09
SD
.80
.18
.80
.17
.63
.14
.77
.23
.80
.23
.79
.19
.81
.24
3.93
1.04
1.02
.32
Table 20. F Statistics and Partial η2 Comparing Study Groups on Slopes and Intercepts
Self‐Report Measures
Intercept Comparisons
Slope Comparisons
Anxiety
F(4, 257) = 27.14, p < .001, η2 = .30
F(4, 257) = 1.39, p = .24, η2 = .02
2
Depression‐Hopelessness
F(4, 257) = 28.62, p < .001, η = .31
F(4, 257) = 1.30, p = .27, η2 = .02
2
Hostility‐Anger Control
F(4, 257) = 5.86, p < .001, η = .08
F(4, 257) = .88, p = .48, η2 = .01
2
Hypersensitivity
F(4, 257) = 10.23, p < .001, η = .14
F(4, 257) = 2.84, p = .02, η2 = .04
Psychosis
F(4, 257) = 16.34, p < .001, η2 = .20
F(4, 257) = .24, p = .92, η2 = .004
2
Somatization
F(4, 257) = 24.10, p < .001, η = .27
F(4, 257) = .58, p = .68, η2 = .01
2
Withdrawal‐Alienation
F(4, 257) = 9.00, p < .001, η = .12
F(4, 257) = 1.18, p = .32, η2 = .02
2
SLUMS
F(4, 257) = 4.50, p = .002, η = .07
F(4, 257) = 1.46, p = .21, η2 = .02
2
Trails B/A
F(4, 257) = .30, p = .88, η = .005
F(4, 257) = .54, p = .71, η2 = .008
As might be expected, there were significant group differences on intercepts for each self‐report variable
except for the Trails B/A derived task. Figure 27 provides the mean intercept values for each self‐report
composite for each group. In general, the MI groups had worse performance on these psychological va‐
riables compared to the NMI groups. The exceptions to this general finding is that the CSP NMI was not sig‐
nificantly different from SCCF and GP MI on hostility composite and was not different from GP MI on psy‐
chosis and withdrawal‐alienation composites. Similarly, the GP NMI group was not significantly different
from the GP MI on SLUMS. Another general finding for intercept differences is that groups with the same
mental health status (MI, NMI) tended to be similar to one another. The exceptions to this general finding
were that for the MI groups, the SCCF group was significantly higher than the GP MI group on depression‐
hopelessness, psychosis, and withdrawal‐alienation composites.
RESULTS
73
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 27. Mean Intercept Values for each Composite by Study Group
0.80
0.60
0.40
■ Anxiety
0.20
Depression‐Hopelessness
Hostility‐Anger Control
0.00
Hypersensitivity
Psychosis
‐0.20
Somatization
Withdrawal‐Alienation
‐0.40
‐0.60
‐0.80
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
In contrast to intercept analyses that showed many group differences, only hypersensitivity had significant
differences in slopes between groups (see Table 20). The only significant differences were between the CSP
MI group and the SCCF group. The CSP MI group had a negative slope indicating improvement over time and
the SCCF group had a positive slope indicating a worsening trend over time.
To better understand how change is occurring across groups, we identified participants as having positive,
negative, or no change over time. Participants were classified as positive changers if they had strong positive
change on at least one variable (i.e., slope was more than 2 standard deviations from mean and in the direc‐
tion of positive change) or had smaller positive change on three or more variables (i.e., slopes on at least
three variables were more than 1 standard deviation from mean and in the direction of positive change).
Likewise, participants were classified as negative changers if slopes were in negative direction. The remain‐
ing participants who did not meet the rules for either positive or negative change were classified as not
changing. Figure 28 provides the percentage of change types for each study group.
74
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Figure 28. Percentage of Change Types by Study Group
90%
Negative Change (%)
Positive Change (%)
No Change (%)
80%
70%
60%
50%
40%
30%
20%
10%
0%
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Groups were significantly different in the percentage of change types, χ2(8, N = 270) = 16.26, p = .04. Using
standardized residuals, the following conditions were found to be different from expectations: for the CSP
MI group, there was a lower than expected percentage of persons changing negatively (5% vs. 8%); for the
CSP NMI group, there was a lower percentage of people changing in positive direction than expected (14%
vs. 24%); for the GP MI, there was a higher percentage of people changing in the positive direction (39% vs.
24%) and fewer than expected stable patterns (52% vs. 68%); for the GP NMI group, there was a lower per‐
centage of people changing positively (14% vs. 24%) and more stable patterns than expected (81% vs. 68%);
and for the SCCF group, there was a higher percentage of persons changing negatively (13% vs. 8%).
PREDICTOR ANALYSES
The purpose of these analyses was to explore if there were predictors of the rate of change across time on
each composite. Using regression analyses to predict individual slopes as the dependent variable, we ex‐
amined if the variables listed in Table 23 could explain rate of change. These variables were identified by the
literature or the study advisory board as potential predictors. All study participants are used in these analys‐
es, and MI status and AS status are used as two of the predictors.
RESULTS
75
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table 23. Variables Used to Predict Change over Time
Predictor
Notes
MI status (study group membership)
0: NMI group; 1: MI group
AS status (study group membership)
0: Not AS; 1: AS
Demographics
Age (at start of study)
DCIS variable
Education
DCIS; 0: Less than HS; 1: HS diploma or GED
Minority status
DCIS; 0: Not a minority; 1: Minority
Criminal History
Offense degree
DCIS
Previous AS confinement
DCIS; 0: No; 1: Yes
Prior incarcerations
DCIS
Gang membership
DCIS; 0: No; 1: Known gang member
Psychological History
Anger needs
DCIS
Anti‐social personality disorder
CCI
Anxiety (Axis I)
CCI
Avoidant personality disorder
CCI
Borderline personality disorder
CCI
Dependent personality disorder
CCI
Depression (Axis I)
CCI
Depressive personality disorder
CCI
History of deliberate self harm
DSHI (life time incidence)
Histrionic personality disorder
CCI
Impulsivity
CCI
Narcissistic personality disorder
CCI
Obsessive‐compulsive personality disorder
CCI
Paranoid personality disorder
CCI
Passive‐aggressive personality disorder
CCI
Psychotic thinking (Axis I)
CCI
Post traumatic stress disorder (Axis I)
CCI
PSI Attitudes towards AS
PSI Time 1
PSI Fear Level
PSI Time 1
PSI Safety
PSI Time 1
Sadistic personality disorder
CCI
Schizoid personality disorder
CCI
Schizophrenia (Axis I)
CCI
Schizotypal personality disorder
CCI
Self‐defeating personality disorder
CCI
Self‐destruction needs
DCIS
Sex offender needs
DCIS
Social Phobia (Axis I)
CCI
Trauma symptoms
TSI
Withdrawal (Axis I)
CCI
To determine which variables are potential predictors of self‐report outcomes, a forward statistical regres‐
sion was used. The information in Table 24 presents the regression analysis results providing the adjusted R2
(proportion of variance explained in the slope variable by the predictors) and lists which variables were
found to be significant predictors, along with standardized regression coefficients. The sign of the regression
coefficient provides information about the direction of the relationship between the dependent variable and
the predictor. Recall that the dependent variable is rate of change (slope) with positive scores indicating
worsening of performance over time for all variables except on the SLUMS and negative scores indicating
improving performance over time for all variables except on the SLUMS. Thus, a negative relationship of a
76
RESULTS
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
predictor with a slope implies that high scores on a variable covary with lower slope scores and thus more
improvement for all variables (except SLUMS where a negative relationship implies higher scores on predic‐
tor goes with more decline in SLUMS performance over time).
Table 24. Regression Results: Significant Predictors of Rate of Change over Time in Composite Variables
Construct
Adj. R2
Significant Predictors
Anxiety
.11***
Schizophrenia (Axis 1; β = ‐0.32)
PSI Safety (β = ‐0.20)
TSI Total (β = 0.21)
Depression‐
.17***
Paranoid PD (β = ‐0.23)
Hopelessness
Sadistic PD (β = 0.19)
TSI Total (β = 0.30)
Schizophrenia (Axis I; β = ‐0.38)
Hostility‐
.12***
Passive‐Aggressive PD (β = ‐0.29)
Anger Control
TSI Total (β = 0.16)
Withdrawal (Axis I; β = ‐0.21)
PSI Safety (β = 0.19)
Anger Needs (β = 0.13)
Hypersensitivity .04***
DSHI (β = ‐0.22)
Psychosis
.03**
Narcissistic PD (β = ‐0.20)
SLUMS
.08***
Obsessive‐Compulsive PD (β = 0.30)
PSI Fear Level (β = ‐0.24)
Somatization
.09***
PSI Safety (β = ‐0.26)
Narcissistic PD (β = ‐0.14)
Withdrawal‐
.03**
DSHI (β = ‐0 .18)
Alienation
Trails B/A
.08***
PSI AS Attitude (β = ‐0.25)
Antisocial PD (β = 0.19)
Narcissistic PD (β = ‐0.22)
Depressive PD (β = 0.20)
*p < .05; **p < .01; ***p < .001
The study variables—MI status and AS confinement—were never significant predictors of outcomes. There
were 15 different significant predictors on at least one outcome.For a predictor to have practical meaning in
an applied setting, it would be important for predictors to be related to multiple outcome variables. There
were no predictors that were significantly related to a majority of outcomes; however there were predictors
that were significantly related to change over time on two or three multiple constructs. These were trauma
history (positive relationship with change), PSI Safety (both positive and negative relationships with change),
narcissistic personality disorder (negative relationship with change), schizophrenia scores ( negative rela‐
tionship with change), and history of self harm ( negative relationship with change). To provide an interpre‐
tive example, the positive relationships between trauma history and slopes for the anxiety composite indi‐
cate that higher scores on trauma co‐vary with higher anxiety slopes. This implies that more trauma leads to
worsening over time on the anxiety composite. Thus, generalizing to significant predictors, higher scores on
trauma and lower scores on narcissistic personality disorder, schizophrenia (axis I), and self harm, lead to
more negative change over time. The PSI Safety subscale had both positive and negative relationships with
outcome variables. Higher scores on the PSI Safety subscale (i.e., feeling safer in AS) was related to im‐
provements in anxiety and somatization but also to more hostility over time.
RESULTS
77
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
DISCUSSION
The results of this study were largely inconsistent with our hypotheses and the bulk of literature that indi‐
cates AS is extremely detrimental to inmates with and without mental illness. We hypothesized that inmates
in segregation would experience greater psychological deterioration over time than comparison inmates,
who were comprised of similar offenders confined in non‐segregation prisons. Similar to other research, our
study found that segregated offenders were elevated on multiple psychological and cognitive measures
when compared to normative adult samples (Andersen et al, 2000; Haney, 2003; Suedfeld et al., 1982; Zin‐
ger et al., 2001). However, there were elevations among the comparison groups too, suggesting that high
degrees of psychological disturbances are not unique to the AS environment. The GP NMI group was the
only one that was similar to the normative group on a number of scales.
In examining change over time patterns, there was initial improvement in psychological well‐being across all
study groups, with the bulk of the improvements occurring between the first and second testing periods,
followed by relative stability for the remainder of the study. On only one measure – withdrawal – did of‐
fenders worsen over time, but this finding was only true for the two NMI groups, so it is not attributable to
AS. Even given the improvements that occurred within the study timeframe, the elevations in psychological
and cognitive functioning that were evident at the start of the study remained present at the end of the
study.
Another hypothesis was that offenders with mental illness would deteriorate over time in AS at a rate more
rapid and more extreme than for those without mental illness. Patterns indicated that the MI groups (CSP
MI, GP MI, SCCF) tended to look similar to one another but were significantly elevated compared to the NMI
groups (CSP NMI, GP NMI), regardless of their setting. For the AS offenders, the MI group scored worse than
the NMI group on all self‐report measures except the Trails test and all staff measures except the PBRS Anti‐
Authority scale. In addition to the changes over time described above, PBRS scores decreased significantly
for segregated inmates regardless of their mental health status, which would be an indicator that staff may
be perceiving improvements, but the significant differences were from the first to the second assessment
periods when the majority of participants changed facilities, which suggests this is perhaps a measurement
error rather than a true improvement. As hypothesized there was a differential time effect for the MI and
NMI groups on several composite measures (i.e., anxiety, hostility‐anger control, hypersensitivity, somatiza‐
tion), but the interactions were in the opposite direction of our hypothesis; on average, the CSP NMI group
did not change while the CSP MI group improved.
We stated that offenders in segregation would develop an array of psychological symptoms consistent with
the SHU syndrome. As already discussed, all of the study groups, with the exception of the GP NMI group,
showed symptoms that were associated with the SHU syndrome. These elevations were present from the
start and were more serious for the mentally ill than non‐mentally ill. In classifying people as improving, de‐
clining, or staying the same over time, the majority remained the same. There was a small percentage (7%)
who worsened and a larger proportion (20%) who improved. Therefore, this study cannot attribute the
presence of SHU symptoms to confinement in AS. The features of the SHU syndrome appear to describe the
most disturbed offenders in prison, regardless of where they are housed. In fact, the group of offenders who
were placed in a psychiatric care facility (SCCF) had the greatest degree of psychological disturbances and
the greatest amount of negative change.
78
DISCUSSION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Finally, in this study, we conducted some exploratory predictive analyses to determine if there were individ‐
ual characteristics that could identify who may be at greater risk of psychological harm from segregation.
There were no individual predictors that showed strong effects for predicting change. This could indicate
that we did not have the correct predictors or that patterns of decompensation are individualized (i.e., not
predictable), but it is more likely that the relative stability over time makes it difficult to predict change.
A review of the findings warrants a discussion of plausible alternative explanations that might account for
our results. The use of a repeated measures design enabled us to determine that change was occurring and
in which direction. Even given the debate about whether or not harmful effects resulted from AS, it was
never suggested that inmates might improve as this study found. The presence of comparison groups avoids
an attribution error; the changes, improvements in this case (i.e., 20%), are not due to segregation. These
conclusions replicate those drawn by Zinger and colleagues (2001) where there was a similar lack of evi‐
dence of harm. These studies suffered criticism for high refusal rates, high attrition rates, small sample sizes,
and short durations – limitations that were corrected in the present study (note, however, that no generali‐
zations should be made beyond the 1 year follow‐up period in this study). Furthermore, the use of reliable
and valid standardized measures enabled the present research study to assess psychological functioning in
an objective manner. Although the majority of these tests were not normed for prisoner populations, the
current reliability and validity findings increased our confidence in these measures.
The most difficult finding to interpret is the improvement that occurred between the first and second testing
sessions, which was significant for all groups except the GP NMI group. This effect may be due to reactivi‐
ty—the participants know they are in a study and respond in a particular way. Perhaps they have a need to
respond in a way that puts them in the most favorable light (e.g., ability to handle demands of prison con‐
finement). (Sometimes improvement in performance due to being observed is called the Hawthorne effect;
however this effect seems to be misunderstood and it was not merely the fact of being studied that led to
those original finding of improvement [Gottfredson, 1996]). It is also possible that there are demand charac‐
teristics introduced by the field researcher that cues the participants on how to respond; this seems unlikely
as the participants would be expected to respond in the hypothesized direction. Although a testing or prac‐
tice effect might explain the improvements on cognitive measures, we were unable to find support in the
literature or from the study advisory board that psychological measures should be influenced by testing ef‐
fects. Because the changes occurred in the AS and comparison groups, it is not possible to attribute the im‐
provements to the confinement conditions; however it may be that participating in the study produces
some unknown expectation. Although study demands may lead to positive ratings, it seems unlikely that
these response biases would overshadow the negative impacts of AS if they really existed. However, there is
not enough information in the data collected to understand the reason for the positive change. The most
likely explanation is that study participants were included in our study when they were in the midst of a cri‐
sis and, with time, the crisis dissipated.
LIMITATIONS
This study was able to incorporate several design features that improved upon the capability of previous
research to draw conclusions about the effects of AS. On the other hand, this study has several limitations
that affect its generalizability to other settings. First, this study included literate adult male offenders and
should therefore not be generalized to female offenders, illiterate offenders, or juveniles. Second, this study
can only be generalized to other prison systems to the extent that their conditions of AS confinement are
DISCUSSION
79
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
similar to Colorado’s. The same findings might not be found in other AS units that have different offender
populations or criteria for placement, more restrictive confinement, or fewer mental health services.
The duration of this study was limited to one year. We believed this time period to be adequate to detect
harmful effects because it was postulated in earlier research that the effects of segregation would be quickly
evident (Grassian & Friedman, 1986; Haney, 2003; Kupers, 2008; Lovell et al., 2007; Rhodes, 2004; Toch,
1992). Kupers (2008, p. 1006) stated “that for just about all prisoners, being held in isolated confinement for
longer than 3 months causes lasting emotional damage if not full‐blown psychosis and functional disability.”
Therefore, we expected that deleterious effects would become evident within a year, but it is possible they
do not appear until after longer periods of segregation.
This study used a moderate sample size because we anticipated moderate to large effects based on the lite‐
rature (e.g., Grassian, 1983; Haney, 2003; Pizarro & Stenius, 2004). It is possible that the true magnitude of
the negative effect is small and, therefore, larger sample sizes would be required to detect negative changes
and predict the types of offenders who might be harmed by segregation. In support of this postulate, the
present study found small to moderate effect sizes for change over time, however they were in the opposite
(positive) direction.
This study examined group averages. It was not designed to identify if certain individuals might be worsened
by the conditions of AS; rather the purpose was to examine whether offenders on the whole, both mentally
ill and non‐mentally ill, are harmed by long‐term segregation. Also, in the design of this study, we assumed a
general linear trend in the data and were not able to capture nonlinear changes over time that might have
occurred. It is possible that a person in segregation could have had one or more brief episodes, possibly
even severe episodes, of psychopathology that were not reflected in our data because testing occurred at
three month intervals and that would not have been reflected in trend analyses of their psychological func‐
tioning. This study was not designed to assess brief changes in psychological functioning, however serious.
This study attempted to triangulate data between inmate self‐report, staff observations, and official records.
In the research study, we had the largest degree of success in gathering self‐report data. Some may question
whether inmates’ self‐report is reliable because they may have reason to exaggerate their symptoms, but
our testing of the measures’ psychometric properties indicated that the participants responded in remarka‐
bly reliable and valid ways. The official record data, which was intended to help us understand the varying
degree of social isolation to which study participants were exposed, was inconsistent and incomplete. Be‐
cause our findings did not show negative change over time, the official record data would not have been as
useful as originally intended; however, it would have still been beneficial in describing the conditions of con‐
finement. Additionally, the data from the clinical staff suggested that there were issues with the BPRS data,
where clinicians were able to rank order groups, but they did not estimate elevations to similar heights as
those reported by inmates. These data also raise the possibility that clinical staff are aware of less distress
than inmates validly report.
FUTURE RESEARCH
The definition of AS varies greatly from state to state, so much so that it is difficult to define or count the
number of inmates held nationally in AS; therefore, replication is needed in other prison systems to deter‐
mine whether these findings still hold true when the conditions of confinement are varied. The present re‐
search was unable to determine which elements of CSP were essential to prevent harm from occurring, and
80
DISCUSSION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
it did not assess the interventions used at CSP to monitor and treat inmates. Ongoing research is needed to
better understand how the different components of segregation may impact offenders differentially. For
example, the type and intensity of psychiatric services provided to AS inmates may have a particularly strong
effect on whether they decompensate during extended periods of segregation. Research that incorporates
qualitative data, such as mental health records and historical patient records, may also help to understand
how individuals are impacted by their confinement conditions. It is important to study other high‐security
settings that permit more out of cell time or increased interpersonal contact (i.e., group treatment).
Similar research is needed with female offenders in AS. Although they represent a small percentage of the
AS population, there is a stunning lack of information about the pathways that lead women to segregation
or how they adapt to this environment. Women offenders have high rates of mental illness (James & Glaze,
2006; O’Keefe & Schnell, 2008). Trauma appears to be a major determinant of mental illness in female of‐
fender populations (Green, Miranda, Daroowalla, & Siddique, 2005; Zlotnik & Pearlstein, 1997); incarcerated
women report much higher rates of abuse than incarcerated men (McClellan, Farabee, & Crouch, 1997). In
examining coping mechanisms, the most unique gender differences exhibited by women are their strong
need for social interaction and their propensity to cope with the prison environment predominantly through
relationship formation (Severance, 2005). Given their higher rates of mental illness, trauma history, and
needs for social interaction, women may be particularly vulnerable to potentially harmful effects of segrega‐
tion.
One untapped topic in the area of segregation research is the role of staff, both how they affect the setting
and the effects of the setting on them (Haney, 2008; Mears, 2008; Pizarro & Narag, 2008). Recognizing that
correctional officers have the greatest amount of contact with offenders, Dvoskin and Spiers (2004) have
suggested they need a larger role in the treatment and psychiatric care of offenders, including, but not li‐
mited to, psychotherapeutic techniques to diffuse crisis situations, consultation with mental health profes‐
sionals, and monitoring inmates’ compliance and adjustment to psychotropic medications. Evaluation re‐
search would benefit the field in understanding the effectiveness of staff training and intervention programs
implemented in AS. Finally, assessing inmate perceptions of staff may have value in understanding the im‐
pact of long‐term segregation on inmates because how they are treated may have a significant impact on
their adjustment to AS.
There were some findings in this study that were difficult to interpret or that did not fit into the same gen‐
eral patterns described above. The Trails test did not differentiate between groups as did the other meas‐
ures. In contrast, the BPRS tended to differentiate between offenders with and without mental illness even
better than did the self‐report measures. The hypersensitivity construct showed more variability and more
differential changes over time than the other constructs; for offenders in AS, those with mental illness im‐
proved in hypersensitivity between times 1 and 2 but then showed a worsening trend while those without
mental illness significantly improved between times 1 and 2, but then worsened between times 2 and 3, and
then improved at the fourth interval. Further research may be needed to explore the reason for these dif‐
ferent patterns or to determine if these were spurious findings.
POLICY IMPLICATIONS
Does this study legitimize the use of segregation with offenders, including those with serious and persistent
mental illness? Because this study may not generalize to other prison systems, especially those that have
conditions of confinement dissimilar to CSP, it is not possible to conclude that AS is not detrimental for all
DISCUSSION
81
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
offenders. Systems that are more restrictive and have fewer treatment and programming resources should
not generalize these findings to their prisons. Replication is needed to understand how increased services,
privileges, and out of cell time ameliorate the unintended consequences of AS, and research needs to inform
prison systems about the standards and practices necessary to protect inmates in segregation from harmful
psychological effects.
It is also important to note that there may be other negative consequences of AS that we did not study. For
example, Lovell et al. (2007) found that inmates released directly from segregation to the streets had dra‐
matically higher rates and severity of detected recidivism than AS inmates who first released to GP (but see
Mears & Bales, 2009). We also did not study the degree to which AS met its purported goal of changing in‐
mate behavior for the better over time. The only questions addressed by this study were related to psycho‐
logical changes over time in segregation. Thus, we make no empirical or value judgments about whether and
to what degree the use of AS balances the benefits (e.g., a safer prison system) with costs (e.g., significant
reductions in freedom).
It is impossible to ignore the extremely disproportionate rate at which inmates with serious mental illness
are assigned to AS (Lovell, 2008; Metzner & Fellner, 2010; O’Keefe, 2008a), which has to some degree
“shocked the conscience” of the courts (see Jones ‘El v. Berge, 2001; Madrid v. Gomez, 1995; Ruiz v. John‐
son, 1999). In an era when prisons are expected to implement evidence‐based practices and to rehabilitate
offenders who will be releasing back to the community, is it enough to avoid harm? Must we ask ourselves
another question: what are the conditions required to improve inmates’ mental well‐being while in segrega‐
tion? Prison systems are held to a standard of treatment that is at least equivalent to community standards.
It is likely that this most difficult segment of society has failed at all levels of community treatment and ear‐
lier criminal justice interventions, but the quest to treat and improve services for the most needy is an im‐
portant reality facing corrections agencies.
Regarding their psychological functioning and levels of distress, these data suggest, although the differences
were small, that inmates with serious mental illness are less likely to improve in segregation and are less
likely to get worse compared to mentally ill inmates in GP. We do not assume that the reasons for these ap‐
parently contradictory findings are the same. For example, it is possible that fewer inmates with mental ill‐
ness get worse because segregation is a safer and more structured environment. On the other hand, hypo‐
theses regarding their unlikeliness to improve include the significant limitations that segregation places on
various types of therapeutic activities and services such as group therapy. Further, the data do not tell us
which aspects of AS prevent psychological improvement and deterioration, respectively, among inmates
with mental illness. However, since prisons have a constitutional duty to respond to serious medical (includ‐
ing psychiatric) needs, the possibility that segregation may prevent improvement is cause for concern and
further study.
There remain significant implications for mental health staff who work in prison systems that permit the
placement of mentally ill offenders in long‐term segregation. It is critical for mental health staff to screen
and assess offenders prior to AS placement to determine their vulnerability to harm that might occur as a
result of their segregation. While in segregation, it is important that the mental status of all offenders be
assessed on a frequent, regular basis through rounds and individual sessions. Prison systems need to have a
range of confinement options, such that offenders who are at risk of or are showing signs of decompensa‐
82
DISCUSSION
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
tion can be removed from segregation and placed in an alternative high security environment that permits
greater out of cell time and interaction with others.
Other systems have rejected confinement models that isolated offenders and held them in extremely re‐
strictive spaces. Even if the segregation models of the early 1900’s and the state psychiatric hospitals of the
mid‐19th century are viewed as “primitive” compared to modern‐day AS facilities, it is important to examine
and understand why these models failed and were ultimately dismantled. Although there are a number of
researchers who predict that there is no end in sight to the supermax model (King, 1999; Mears, 2008; Pizar‐
ro & Narag, 2008; Pizarro & Stenius, 2004), they have also raised empirical questions regarding their effica‐
cy. Questions about the efficacy of AS will be asked until more is known about whether the use of AS in pris‐
on systems improves conditions for the rest of the system, whether and how they improve inmate behavior
within and beyond the prison walls, whether they are cost‐effective, whether they increase risks to public
safety, and whether there are settings or individuals that are prone to psychological deterioration.
DISCUSSION
83
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
REFERENCES
American Correctional Association (2009). 2009 directory: Adult correctional departments, institutions,
agencies, and probation and parole authorities. Alexandria, VA: Author.
American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed).
Washington, DC: Author.
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text
revision). Washington, DC: Author.
American Psychological Association. (2009). Publication Manual of the American Psychological Association
(6th ed.). Washington, DC: Author.
Andersen, H. S., Sestoft, D., Lillebaek, T., Gabrielsen, G., & Hemmingsen, R. (2003). A longitudinal study of
prisoners on remand: Repeated measures of psychopathology in the initial phase of solitary versus
nonsolitary confinement. International Journal of Law and Psychiatry, 26, 165‐177. Retrieved from
http://sciencedirect.com
Andersen, H. S., Sestoft, D., Lillebaek, T., Gabrielsen, G., Hemmingsen, R., & Kramp, P. (2000). A longitudinal
study of prisoners on remand: Psychiatric prevalence, incidence, and psychopathology in solitary vs.
non‐solitary confinement. Acta Psychiatrica Scandinavica, 102, 19‐25.
Andrews, D. A., & Bonta, J. L. (1995). Level of Service Inventory—Revised. Toronto, ON: Multi‐Health Sys‐
tems.
Andrews, D. A., & Bonta, J. L. (2003). Level of Service Inventory‐Revised (LSI‐R): U.S. norms manual supple‐
ment. North Tonawanda, NY: Multi‐Health Systems, Inc.
Arrigo, B. A., & Bullock, J. L. (2008). The psychological effects of solitary confinement on prisoners in super‐
max units: Reviewing what we know and recommending what should change. International Journal
of Offender Therapy and Comparative Criminology, 52, 622‐641. doi: 10.1177/0306624X07309720
Atherton, E. (2001). Incapacitation with a purpose. Corrections Today, 63, 100‐103. Retrieved from
http://vnwebnhwwilsonweb.com
Austin, J., Alexander, J., Anuskiewicz, S., & Chin, L. (1995). Evaluation of the Colorado Objective Prison Classi‐
fication System. San Francisco, CA: National Council on Crime and Delinquency.
Beck, A. T., & Steer, R. A. (1993). Beck Hopelessness Scale: Manual. San Antonio, TX: The Psychological Cor‐
poration—Harcourt Brace & Company.
Benjamin, T. B., & Lux, K. (1975). Constitutional and psychological implications of the use of solitary con‐
finement: Experience at the Maine state prison. Clearinghouse Review, 9, 83‐84.
Briere, J. (1995). Trauma Symptom Inventory: Professional manual. Lutz, FL: Psychological Assessment Re‐
sources, Inc.
84
REFERENCES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Briggs, C. S., Sundt, J. L., & Castellano, T.C. (2003). The effects of supermaximum security prisons on aggre‐
gate levels of institutional violence. Criminology, 41, 1341‐1376. Retrieved from
http://www3.interscience.wiley.com/cgi‐bin/fulltext/118866214/PDFSTART
Brodsky, S., & Scogin, F. (1988). Inmates in protective custody: First data on emotional effects. Forensic Re‐
ports, 1, 267‐280.
Burger, K. G., Calsyn, R. J., Morse, G. A., Klinkenberg, W. D., & Trusty, M. L. (1997). Factor structure of the
Expanded Brief Psychiatric Rating Scale. Journal of Clinical Psychology, 53, 451‐454. doi:
10.1002/(SICI)1097‐4679(199708)53:5<451::AID‐JCLP5>3.0.CO;2‐Q
Cohen, F. (2008). Penal Isolation: Beyond the Seriously Mentally Ill. Criminal Justice and Behavior, 35, 1017‐
1047.
Collins, W. C. (2004). Correctional law for the correctional officer (4th ed.). Lanham, MA: American Correc‐
tional Association.
Cooke, D. J. (1998). The development of the Prison Behavior Rating Scale. Criminal Justice and Behavior, 25,
482‐506. doi: 10.1177/0093854898025004005
Coolidge, F. L. (2004). An introduction to the Coolidge Correctional Inventory (CCI). Retrieved from
http://www.uccs.edu/~faculty/fcoolidg//
CTB/McGraw‐Hill. (1994). Tests of Adult Basic Education (TABE): Version 7‐8. Monterey, CA: CTB/McGraw‐
Hill.
CTB/McGraw‐Hill. (2004). Tests of Adult Basic Education (TABE): TABE Form 7 to the 2002 GED Linking Re‐
port. Retrieved from
http://www.ctb.com/media/articles/pdfs/AdultEducation/TABE_Form7_2002_GED_Linking_Report.
pdf?FOLDER<>folder_id=59997&bmUID=1089131925485
Derogatis, L. (1993). Brief Symptom Inventory: Administration, scoring, and procedures manual. Minneapolis,
MN: NCS Pearson, Inc.
Ditton, P. M. (1999, July). Mental health and treatment of inmates and probationers. Washington, DC: U.S.
Department of Justice, Bureau of Justice Statistics.
Dvoskin, J. A., & Spiers, E. M. (2004). On the role of correctional officers in prison mental health. The Psy‐
chiatric Quarterly, 75, 41‐59.
Ecclestone, C. E. J., Gendreau, P., & Knox, C. (1974). Solitary confinement of prisoners: An assessment of its
effects on prisoners’ personal constructs and andrenocortical activity. Canadian Journal of Beha‐
vioral Science, 6, 178 – 191.
Gawande, A. (2009, March 30). Hellhole: The United States holds tens of thousands of inmates in long‐term
solitary confinement. Is this torture? The New Yorker, 36‐45. Retrieved from
http://www.newyorker.com/reporting/2009/03/30/090330fa_fact_gawande
REFERENCES
85
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Gendreau, P., & Bonta, J. (1984). Solitary confinement is not cruel and unusual punishment: People some‐
times are! Canadian Journal of Criminology, 26, 467‐478.
Gendreau, P. E., Freedman, N., Wilde, G. J. S., & Scott, G. D. (1968). Stimulation seeking after seven days of
perceptual deprivation. Perceptual and Motor Skills, 26, 547‐550.
Gendreau, P. E., Freedman, N., Wilde, G. J. S., & Scott, G. D. (1972). Changes in EEG alpha frequency and
evoked response latency during solitary confinement. Abnormal Psychology, 79, 54‐59.
Gendreau, P., McLean, R., Parsons, T., Drake, R., and Ecclestone, J. (1970). Effect of two days’ monotonous
confinement on conditioned eyelid frequency and topography. Perceptual and Motor Skills, 31, 291‐
293.
Gottfredson, G. D. (1996). The Hawthorne misunderstanding (and how to get the Hawthorne effect in action
research). Journal of Research in Crime and Delinquency, 33, 28‐48. doi:
10.1177/0022427896033001003
Grassian, S. (1983). Psychopathological effects of solitary confinement. American Journal of Psychiatry, 140,
1450‐1454.
Grassian, S., & Friedman, N. (1986). Effects of sensory deprivation in psychiatric seclusion and solitary con‐
finement. International Journal of Law and Psychiatry, 8, 49‐65.
Gratz, K. L. (2001). Measurement of deliberate self‐harm: Preliminary data on the deliberate self‐harm in‐
ventory. Journal of Psychopathology and Behavioral Assessment, 23, 253‐263.
Green, B. L., Miranda, J., Daroowalla, A., & Siddique, J. (2005). Trauma exposure, mental health functioning
and program needs of women in jail. Crime & Delinquency, 51, 133‐151. doi:
10.1177/0011128704267477
Haney, C. (1993). Infamous punishment: The psychological effects of isolation. National Prison Project Jour‐
nal, 8, 3‐21.
Haney, C. (2003). Mental health issues in long‐term solitary and "supermax" confinement. Crime and Delin‐
quency, 49, 124‐156.
Haney, C. (2008). A culture of harm: Taming the dynamics of cruelty in supermax prisons. Criminal Justice
and Behavior, 35, 956‐984.
Hedlund, J. L., & Vieweg, B. W. (1980). The Brief Psychiatric Rating Scale (BPRS): A comprehensive review.
Journal of Operational Psychiatry, 11, 48‐64.
Hewitt v. Helms, 459 U.S. 460 (1983).
Hodgins, S., & Côté, G. (1991). The mental health of penitentiary inmates in isolation. Canadian Journal of
Criminology, 33, 175‐182.
Human Rights Watch. (1997). Cold storage: Super‐maximum security confinement in Indiana. New York, NY:
Human Rights Watch.
86
REFERENCES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Human Rights Watch. (1999). Red Onion State Prison: Super‐maximum security confinement in Virginia.
Human Rights Watch, 11, 1‐24.
Human Rights Watch. (2000). Out of sight: Super‐maximum security confinement in the United States. Hu‐
man Rights Watch, 12, 1‐9.
James, D., & Glaze, L. (2006). Mental health problems of prisons and jail inmates. Washington, DC: Bureau of
Justice Statistics.
Jones v. Baker, 115 F.3d. 810 (6th Cir. 1998).
Jones ‘El v. Berge, 164 F. Supp. 2d 1096 (W.D. Wis. 2001).
King, R. D. (1999). The rise and rise of supermax: An American solution in search of a problem? Punishment
and Society, 1, 163‐186. doi: 10.1177/14624749922227766
Kupers, T. (2008). What to do with the survivors? Coping with the long‐term effects of isolated confinement.
Criminal Justice and Behavior, 35, 1005‐1016.
Kurki, L., & Morris, N. (2001). The purposes, practices, and problems of supermax prisons. Crime and Justice,
28, 1‐21.
Lovell, D. (2008). Patterns of disturbed behavior in a supermax population. Criminal Justice and Behavior, 35,
985‐1004.
Lovell, D., Johnson, L. C., & Cain, K. C. (2007). Recidivism of supermax prisoners in Washington state. Crime
and Delinquency, 53, 633‐656. doi: 10.1177/0011128706296466
Lukoff, D., Nuechterlein, K., & Ventura, A. (1986). Manual for the expanded brief psychiatric rating scale.
Schizophrenia Bulletin, 13, 261‐276.
Madrid v. Gomez, 899 F. Supp. 1146 (N.D. Cal. 1995).
McClary v. Kelly, 4 F. Supp. 2d 195 (W.D.N.Y. 1998).
McClellan, D. S., Farabee, D., & Crouch, B. M. (1997). Early victimization, drug use, and criminality: A com‐
parison of male and female prisoners. Criminal Justice and Behavior, 24, 455‐476.
McNair, D. M., Lorr, M., & Droppleman, L. F. (1992). Manual: Profile of Mood States, Revised. San Diego, CA:
Educational and Industrial Testing Service.
Mears, D. P. (2008). An assessment of supermax prisons using an evaluation research framework. The Prison
Journal, 88, 43‐68.
Mears, D. P., & Bales, W. D. (2009). Supermax incarceration and recidivism. Criminology, 47, 1131‐1166. doi:
10.1111/j.1745‐9125.2009.00171.x
Mears, D. P., & Watson, J. (2006). Towards a fair and balanced assessment of supermax prisons. Justice
Quarterly, 23, 232‐270. doi: 10.1080/07418820600688867
REFERENCES
87
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Metzner, J. L., & Dvoskin, J. A. (2006). An overview of correctional psychiatry. Psychiatric Clinics of North
America , 29, 761‐772. Retreived from
http://psych.theclinics.com/issues/contents?issue_key=S0193‐953X(06)X0019‐9
Metzner, J. L., & Fellner, J. (2010). Solitary confinement and mental illness in U.S. prisons: A challenge for
medical ethics. Journal of the American Academy of Psychiatry and the Law, 38, 104‐108. Retrieved
from http://www.jaapl.org/cgi/reprint/38/1/104.pdf
Morey, L. C. (1997). Personality Assessment Screener: Professional manual. Lutz, FL: Psychological Assess‐
ment Resources, Inc.
Motiuk, L. L., & Blanchette, K. (2001). Characteristics of administratively segregated offenders in federal cor‐
rections. Canadian Journal of Criminology, 43, 131‐144.
Motiuk, M. S., Motiuk, L. L., & Bonta, J. (1992). A comparison between self‐report and interview‐based in‐
ventories in offender classification. Criminal Justice and Behavior, 19, 143‐159. doi:
10.1177/0093854892019002003
Naday, A., Freilich, J. D., & Mellow, J. (2008). The elusive data on supermax confinement. The Prison Journal,
88, 69‐93.
National Institute of Corrections. (1997). Supermax housing: A survey of current practice, special issues in
corrections. Longmont, CO: U.S. Department of Justice, National Institute of Corrections.
National Institute of Corrections. (1999). Supermax prisons: Overview and general considerations. Longmont,
CO: US Department of Justice, National Institute of Corrections.
National Institute of Mental Health. (2010). The Numbers Count: Mental Disorders in America. Retrieved
from http://www.nimh.nih.gov/health/publications/the‐numbers‐count‐mental‐disorders‐in‐
america/index.shtml
O’Keefe, M. L. (2005). Analysis of Colorado’s administrative segregation. [Technical Report]. Colorado
Springs, CO: Department of Corrections.
O’Keefe, M. L. (2008a). Administrative segregation for mentally ill inmates. Journal of Offender Rehabilita‐
tion, 45, 149‐165.
O’Keefe, M. L. (2008b). Administrative segregation from within: A corrections perspective. The Prison Jour‐
nal, 88, 123‐143.
O’Keefe, M. L., Klebe, K., & Hromas, S. (1998). Validation of the Level of Supervision Inventory (LSI) for com‐
munity based offenders in Colorado: Phase II. [Technical Report]. Colorado Springs, CO: Colorado
Department of Corrections.
O’Keefe, M. L., & Schnell, M. (2008). Offenders with Mental Illness in the Correctional System. Journal of Of‐
fender Rehabilitation, 45, 81‐104.
Overall, J. E. & Gorham, D. R. (1962). The brief psychiatric rating scale. Psychological Reports, 10, 799‐812.
88
REFERENCES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Pizarro, J. M., & Narag, R., E. (2008). Supermax prisons: What we know, what we do not know, and where
we are going. The Prison Journal, 88, 23‐42.
Pizzaro, J. & Stenius, V. M. K. (2004). Supermax prisons: Their rise, current practices and effect on inmates.
The Prison Journal, 84, 248‐264.
Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic brain damage. Perceptual
and Motor Skills, 8, 271‐276.
Rhodes, L. (2004). Total confinement: Madness and reason in maximum security. Berkeley, CA: University of
California Press.
Ruiz v. Johnson, 37 F. Supp. 855 (S.D. Texas 1999).
Sandin v. Conner, 515 U.S. 473 (1995).
Severance, T. A. (2005). You know who you can go to: Cooperation and exchange between incarcerated
women. The Prison Journal, 85, 343‐367.
Smith, P. S. (2008). “Degenerate criminals:” Mental health and psychiatric studies of Danish prisoners in soli‐
tary confinement, 1870‐1920. Criminal Justice and Behavior, 35, 1048‐1064. doi:
10.1177/0093854808318782
Spielberger, C., Gorsuch, R., & Lushene, R. (1970). The State‐Trait Anxiety Inventory: Test manual for Form X.
Palo Alto, CA: Consulting Psychologists Press.
Suedfeld, P., Ramirez, C., Deaton, J., & Baker‐Brown, G. (1982). Reactions and attributes of prisoners in soli‐
tary confinement. Criminal Justice and Behavior, 9, 303‐340. doi: 10.1177/0093854882009003004
Sundt, J. L., Castellano, T. D., & Briggs, C. S. (2008). The sociopolitical context of prison violence and its con‐
trol: A case study of supermax and its effect in Illinois. The Prison Journal, 88, 94‐122.
Tariq, S. H., Tumosa, N., Chibnall, J. T., Perry, M. H., & Morley, J. E. (2006). Comparison of the Saint Louis
University Mental Status Examination and the Mini‐Mental State Examination for detecting demen‐
tia and mild neurocognitive disorder—a pilot study. American Journal of Geriatric Psychiatry, 14,
900‐910.
Toch, H. (1992). Living in prison: The ecology of survival. Washington, DC: American Psychological Associa‐
tion.
Toch, H. (2001). The future of supermax confinement. The Prison Journal, 81, 376‐388.
Turner v. Safley, 482 U.S. 78 (1987).
Ventura, J., Lukoff, D., Nuechterlein, K. H., Liberman, R. P., Green, M. F., & Shaner, A. (1993). Brief Psychia‐
tric Rating Scale (BPRS) expanded version (4.0): Scales, anchor points, and administration manual.
International Journal of Methods in Psychiatric Research, 3, 227‐244.
Ward, D. A., & Werlich, T. G. (2003). Alcatraz and Marion: Evaluating super‐maximum custody. Punishment
and Society, 5, 53‐75. doi: 10.1177/1462474503005001295
REFERENCES
89
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Widows, M. R., & Smith, G. P. (2005). Structured Inventory of Malingered Symptomatology: Professional ma‐
nual. Lutz, FL: Psychological Assessment Resources, Inc.
Wilkinson v. Austin, 545 U.S. 209 (2005).
Zinger, I., Wichman, C., & Andrews, D. A. (2001). The psychological effects of 60 days in administrative se‐
gregation. Canadian Journal of Criminology, 43, 47‐88.
Zlotnik, C., & Pearlstein, T. (1997). Posttraumatic stress disorder (PTSD) comorbidity, and childhood abuse
among incarcerated women. Journal of Nervous and Mental Disease, 185, 761‐763.
90
REFERENCES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
APPENDIX A
PSI
Please rate how often the following items have applied to you in the past week.
Never Rarely Sometimes Often
True
True
True
True
1. I find that even quiet noises are
0
1
2
3
loud and disturbing.
2. My heart races faster than normal
0
1
2
3
at times.
3. I am afraid for no reason.
0
1
2
3
4. I might go a day without brushing
0
1
2
3
my teeth.
5. I do cardiovascular activity (jogging,
0
1
2
3
running, speed walking, etc.).
6. I have difficulty catching my breath
0
1
2
3
even when I am not exercising.
7. There are smells here that make
0
1
2
3
me queasy.
8. I have pounding headaches that
0
1
2
3
make it hard to concentrate.
9. I comb or brush my hair daily.
0
1
2
3
10. I struggle to get air.
0
1
2
3
11. I have a lot of energy.
0
1
2
3
12. My fear prevents me from doing
0
1
2
3
things that I’d like to do.
13. I feel dizzy at times.
0
1
2
3
14. I look forward to getting back to
0
1
2
3
the general population.
15. I do strength training (weight lift‐
0
1
2
3
ing, pull‐ups, push‐ups, etc.).
16. I feel as though I am choking.
0
1
2
3
17. I feel lightheaded or like I am going
0
1
2
3
to faint.
18. I avoid shaving or grooming my fa‐
0
1
2
3
cial hair.
19. I sleep most of the day.
0
1
2
3
20. I find my whole body trembling for
0
1
2
3
no apparent reason.
APPENDICES
Usually
True
Always
True
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
4
5
N/A
4
5
4
5
4
5
4
5
4
5
4
5
91
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
21. I have episodes where I am certain I
will die soon.
22. I do not speak to anyone even
when they talk to me.
23. I shower every day that I am al‐
lowed.
24. I am troubled by physical pain or
aches.
25. I cannot stop myself from shaking.
26. I am bored to death.
27. Exercise is not important to me.
28. I sleep soundly at night.
29. It is important to me to keep good
hygiene.
30. I break out in a sweat when I am
not doing anything.
31. I find the quiet to be peaceful.
32. I start conversations with other
people.
33. It is unsafe for me in the general
population.
34. My cell temperature is comforta‐
ble.
35. I feel calm and relaxed.
36. I need a single cell for my own pro‐
tection.
37. This place makes me feel misera‐
ble.
38. I am not bothered by thoughts of
dying.
39. I prefer administrative segregation
to the general population.
Never
True
Rarely
True
Sometimes
True
Often
True
Usually
True
Always
True
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
N/A
92
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
PRISON BEHAVIOR RATING SCALE
Instructions: Based on your interactions and records such as Chronological Notes, disciplinary infractions or
other incidents, please rate the inmate’s behavior by circling your answer.
1. Tried but failed to follow instructions
2. Appeared tense and unable to relax
3. Appeared close to tears
4. Caused trouble during his free time
5. Cursed and swore (in an abusive manner)
6. Appeared easily upset
7. Appeared sluggish and drowsy
8. Been held out of normal circulation
(e.g. dry cell, mental health watch, special controls,
RFP, punitive segregation etc.)
9. Had trouble sleeping at night
10. Appeared lacking in energy
11. Sought reassurance
12. Appeared to be brooding on something
13. Victimized weaker inmates
14. Appeared dull and unintelligent
15. Fidgeted and been unable to sit still
16. Tried to con staff
17. Appeared frightened of other inmates
18. Complained about staff
19. Not been aware of what is going on around him
20. Been aggressive towards staff
APPENDICES
Never/
Rarely
Sometimes
Often
Most of
the Time
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
0
0
1
1
1
2
3
2
3
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
93
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Never/
Rarely
Sometimes
Often
Most of
the Time
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
21. Had a quick temper
22. Been on report (e.g., got a negative CHRON,
written up for COPD violation)
23. Appeared preoccupied/dreamy
24. Tried to play staff against each other
25. Openly defied rules
26. Appeared sad and depressed
27. Stirred up trouble among other inmates
28. Aided or abetted others to break the rules
29. Been out of touch with what is happening
around him
30. Been victimized by other inmates
31. Not understood orders
32. Appeared to be scared
33. Has few if any friends
34. Avoided other inmates
35. Given the impression of ignorance/inability
36. Appeared depressed, gloomy, or sulky
37. Had poor hygiene
Completed by _________________________________
Please Print Name
94
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
APPENDIX B
SUMMARY OF STUDY MEASURES
Most of the measures included in this study were self‐report pencil‐and‐paper tests; however, we also col‐
lected data from clinicians who completed the Brief Psychiatric Rating Scale (BPRS) and correctional staff
who completed the Prison Behavior Rating Scale (PBRS). Two additional measures assessing cognitive func‐
tioning (i.e., St. Louis University Memory Scale, Trail Making Test) were administered by a researcher. In‐
struments used in this study were selected to assess a broad range of symptoms believed to be associated
with long‐term segregation. We assessed eight constructs by means of 10 different measures (and/or ap‐
propriate subscales). The constructs of interest in this study were anxiety, cognitive impairment, depression‐
hopelessness, hostility‐anger control, hypersensitivity, psychosis, somatization, and withdrawal‐alienation.
In addition to these key variables, we measured other variables that may be predictors of outcomes, includ‐
ing trauma, personality disorders, malingering, and history of self‐harm.
Measures were selected for ease of administration and strength of psychometric properties. In this appen‐
dix, we describe the measures used in this study, provide results concerning the psychometric properties of
the measures, and describe the composites used for analyses in the report. Data are reported for the entire
sample at each time period. Table B1 provides a quick reference guide to the tests used in this study as well
as the constructs assessed by each of them.
Table B1. Study Measures
Measure
Outcome Variables
Beck Hopelessness Scale (BHS)
Brief Psychiatric Rating Scale (BPRS)
Activity
Anxious‐Depressed
Hostility‐Suspiciousness
Thought Disorder
Withdrawal
Brief Symptom Inventory (BSI)
Anxiety
Depression
Hostility
Interpersonal Sensitivity
Obsessive‐Compulsive
Paranoid Ideation
Phobic Anxiety
Psychoticism
Somatization
Personality Assessment Screener (PAS)
Acting Out
Alienation
Anger Control
Health Problems
Hostile Control
Negative Affect
Psychotic Features
Social Withdrawal
Suicidal Thinking
APPENDICES
Construct
Administration
Depression‐Hopelessness
Self‐Report
Clinicians
Anxiety, Depression‐Hopelessness
Hostility‐Anger Control
Psychosis
Withdrawal‐Alienation
Self‐Report
Anxiety
Depression‐Hopelessness
Hostility‐Anger Control
Hypersensitivity
Anxiety
Psychosis
Anxiety
Psychosis
Somatization
Self‐Report
Hostility‐Anger Control
Withdrawal‐Alienation
Hostility‐Anger Control
Somatization
Hostility‐Anger Control
Anxiety, Depression‐Hopelessness
Psychosis
Withdrawal‐Alienation
Depression‐Hopelessness
Times Assessed
B, every 3 months
B, every 6 months
B, every 3 months
B, every 3 months
95
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Measure
Prison Behavior Rating Scale (PBRS)
Anti‐Authority
Anxious‐Depressed
Dull‐Confused
Prison Symptom Inventory (PSI)
Panic Disorder
Hypersensitivity‐External Stimuli
Physical Symptoms
Profile of Mood States (POMS)
Anger‐Hostility
Depression‐Dejection
Fatigue‐Inertia
Tension‐Anxiety
Saint Louis University Memory Scale
(SLUMS)
State‐Trait Anxiety Inventory (STAI)
State Anxiety
Trait Anxiety
Trail Making Test (TMT)
Time to Complete A Task
Time to Complete B Task
B – A Time
B/A Time
Predictor Variables
Coolidge Correctional Inventory (CCI)
Deliberate Self‐harm Inventory DSHI)
Structured Inventory of Malingered
Symptoms (SIMS)
Trauma Symptom Inventory (TSI)
Construct
Administration
Officers
Times Assessed
B, every 3 months
Self‐Report
B, every 3 months
Self‐Report
B, every 3 months
Researcher
B, every 3 months
Self‐Report
Researcher
B, every 3 months
B, every 3 months
B, every 3 months
B, every 3 months
Yes
Yes
Yes
B
B
B, every 3 months
Hostility‐Anger Control
Anxiety, Depression‐Hopelessness
Cognitive Impairment
Anxiety
Hypersensitivity
Somatization
Hostility‐Anger Control
Depression‐Hopelessness
Somatization
Anxiety
Cognitive Impairment
Anxiety
Anxiety
Cognitive Impairment
Cognitive Impairment
Cognitive Impairment
Cognitive Impairment
Personality Disorders
Self‐Harm
Malingering
Trauma
2nd
Note. Times assessed include the first time the test was administered as well as the interval at which it was given (unless it was only
conducted at specific testing periods). B stands for baseline test.
DESCRIPTION OF INDIVIDUAL MEASURES
In this section, descriptions of the measures and summary statistics about reliability and validity are pro‐
vided. Summary statistics include those published in the literature as well as those conducted with our study
population. Cronbach’s alpha is used to estimate internal consistency reliability at each time period for the
entire sample. Correlations between consecutive time periods are used to estimate test‐retest reliability.
Convergent validity is estimated by correlations of each measure with other measures of the same construct
for the entire sample at each time period. Tables for descriptive statistics on the measures are given with
the description of the measure; reliability and validity statistics are presented in the description of the
measure and/or with the description of the composites.
Beck Hopelessness Scale (BHS)
Designed to measure an individual’s degree of despair/depression, the BHS (Beck & Steer, 1993; Beck,
Weissman, Lester, & Trexler, 1974) is a 20‐item self‐report measure on which scores can range from 0 to 20,
with higher scores indicating a greater degree of despair about the future (Clum & Yang, 1995). Since this
measure does not have any subscales, one total score is derived. Respondents answer true or false to
96
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
statements about their attitudes over the past week (Beck & Steer, 1993). It takes approximately 5 to 10
minutes to complete this measure (Beck & Steer, 1993).
The psychometric properties of the BHS are solid, as it has demonstrated internal consistency estimates of
.65 to .89 for nonclinical samples of college students (e.g., Beck et al., 1974; Durham, 1982; Steed, 2001), .86
to .93 for clinical in‐ or out‐patient samples (e.g., Beck et al., 1974; Durham, 1982; Dyce, 1996), and .83 in a
forensic sample (Durham, 1982). Three‐week test‐retest reliability in a university sample was found to be
acceptable (r = .85) for the entire sample and was slightly higher for males (r = .94; Holden & Fekken, 1988).
In clinical samples, test‐retest correlations ranged from .66 (six‐week test‐retest correlation) to .69 (one‐
week test‐retest correlation; Beck & Steer, 1993).
BHS self‐report ratings have also been correlated to clinician ratings of hopelessness, with correlations rang‐
ing from .78 to .98 (Beck et al., 1974), which suggests that this measure possesses acceptable convergent
validity. Additionally, the BHS is considered to be a predictor of suicide risk in clinical populations (e.g., Beck,
1986; Beck, Brown, Berchick, Stewart, & Steer, 1990; Beck et al., 1974; Brown, Beck, Steer, & Grisham,
2000), with scores of 9 and above being predictive of suicidal ideation (Beck, Steer, Kovacs, & Garrison,
1985). Additionally, the correlation between the BHS and the Modified Scale for Suicide Ideation was found
to be moderate at .46 (Clum & Yang, 1995) in a nonclinical sample of college students. Further evidence for
the BHS’s convergent validity comes from a study on a clinical inpatient sample; the BHS was found to be
significantly correlated with the Beck Depression Inventory (BDI; r = .68) and the Current Suicidal Intent (r =
.68; Kovacs, Beck, & Weissman, 1975).
Summary statistics are available for several different groups, including criminal psychiatric inpatients (Dur‐
ham, 1982). Durham (1982) assessed college students, general psychiatric patients, and forensic psychiatric
patients on the BHS. He found that the mean for the nonclinical, college student sample was 2.32 (SD = 2.25,
n = 197), 6.04 (SD = 4.67, n = 118) for the clinical, general psychiatric sample, and 6.62 (SD = 4.88, n = 99) for
the clinical, forensic sample. In another study, including 2,067 psychiatric outpatients, the mean total BHS
score was 9.06 (SD = 5.61; Bieling, Beck, & Brown, 2000). Palmer and Connelly (2005) assessed BHS scores
for prisoners with (n = 24) and without (n = 24) a history of self‐harming behavior. They found that the mean
BHS score for prisoners with a history of self‐harming behavior was 10.13 (SD = 4.81); for offenders without
a history of self‐harming behavior the BHS mean score was found to be significantly lower at 6.29 (SD =
4.49).
Summary statistics on the BHS for the current study are given in Table B2 for each group. Internal consisten‐
cy estimates indicated excellent consistency across items at each time (mean Cronbach’s alpha = .93; range
= .92 to .94). Test‐retest correlation coefficients ranged between .66 and .79 (M = .71). Examination of the
validity coefficients, given in the composite section of this appendix, indicated that the BHS is correlated
with other self‐report measures of depression (mean r = .58, range = .43 to .77); however, the correlations
with relevant staff reports (BPRS Anxious‐Depressed and PBRS Anxious‐Depressed) were lower (mean r =
.18, range = ‐.02 to .33).
APPENDICES
97
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B2. Summary Statistics (M, SD, n) on BHS by Group and Time
Time
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
7.16 (5.63) 5.14 (4.52) 7.09 (5.38) 2.26 (3.16) 9.84 (6.28)
1
n = 63
n = 63
n = 33
n = 43
n = 67
7.59 (6.62) 4.66 (4.96) 5.56 (4.68) 2.35 (3.06) 9.20 (6.89)
2
n = 62
n = 59
n = 32
n = 41
n = 64
7.75 (6.37) 4.79 (4.88) 5.38 (4.58) 2.56 (3.92) 8.57 (6.85)
3
n = 60
n = 57
n = 32
n = 41
n = 61
8.12 (6.90) 3.87 (4.11) 5.62 (4.91) 2.60 (2.89) 9.20 (7.05)
4
n = 60
n = 56
n = 29
n = 39
n = 59
6.32 (6.31) 4.71 (5.03) 4.07 (4.22) 1.58 (2.06) 8.93 (6.88)
5
n = 56
n = 56
n = 29
n = 38
n = 57
5.86 (5.57) 3.51 (4.10)
6
NA
NA
NA
n = 51
n = 54
All
6.56 (5.75)
n = 269
6.24 (6.11)
n = 258
6.12 (5.99)
n = 251
6.22 (6.16)
n = 243
5.55 (5.92)
n = 236
4.61 (4.98)
n = 106
Brief Psychiatric Rating Scale (BPRS)
The expanded Brief Psychiatric Rating Scale (BPRS‐E; Ventura et al., 1993) is a 24‐item measure administered
by clinicians to assess patients with psychiatric disorders. It is designed to allow for the rapid review of psy‐
chological symptoms over time (e.g., Ventura et al., 1993). Ratings are made after a semi‐structured clinical
interview with a client. Clinicians rate the different items on the BPRS‐E by means of a 7‐point severity scale
(1‐ not present to 7‐ extremely severe); higher scores on this measure generally indicate greater severity of
psychopathology (Segal & Silverman, 2002; Thomas, Donnell, & Young, 2004). However, since clinicians also
have an option of using a not assessed (N/A) rating on any given item, scores may not accurately reflect the
degree of psychopathology (Ventura et al., 1993). The clinical interview takes approximately 10 to 40 mi‐
nutes, depending on familiarity with the client as well as presenting symptoms at the time of the assessment
(Thomas et al., 2004). Research has indicated that there are five factors to which the individual items of the
BPRS‐E are associated: thought disorder (directly reflecting psychosis), withdrawal, anxious‐depressed, hos‐
tility‐suspiciousness, and activity (Burger, Calsyn, Morse, Klinkenberg, & Trusty, 1997).
Internal consistency reliability for the total BPRS‐E was found to be between .74 and .79 for clinical popula‐
tions (Perlick, Rosenheck, Clarkin, Sirey, & Raue, 1999; Segal & Silverman, 2002; Thomas et al., 2004). Fur‐
thermore, when considering the internal consistency reliability for the 5‐factor structure, the coefficients for
four out of the five scales ranged from .73 (i.e., anxiety‐depression) to .81 (i.e., activity); the Cronbach’s al‐
pha for the hostility‐suspiciousness factor was found to be lower at .49 (Burger et al., 1997).
Mean total scores for clinical populations were found to be between 37.9 (SD = 11.1) and 61.6 (SD = 12.9;
Biancosino et al., 2004; Brown, Chhina, & Dye, 2008; Segal & Silverman, 2002), while the mean total for the
BPRS‐E among inmates with psychiatric problems in the prison population was found to be 49.29 (SD =
14.78; Gray, Bressington, Lathlean, & Mills, 2008). When individuals were tested over time in a clinical set‐
ting, mean scores significantly decreased at each testing interval (Biancosino et al., 2004; Brown et al.,
2008), which indicates that the test may be sensitive to change over time following an intervention.
The BPRS‐E has been shown to be correlated with the Brief Symptom Inventory (BSI), a self‐report measure
of psychological symptoms (Morlan & Tan, 1998), indicating convergent validity. Furthermore, a study on
female inmates showed that mental health referrals are more often done for those inmates who have re‐
ceived higher BPRS‐E scores, suggesting that this assessment tool is useful in detecting psychopathology in
inmates (Nicholls, Lee, Corrado, & Ogloff, 2004).
98
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B3 provides the summary statistics for the study groups on the BPRS scales at each time period. The
BPRS scores tend to be lower than normative data found with other clinical populations, indicating a poten‐
tial floor effect and potential rater bias.
Table B3. Summary Statistics (M, SD, n) on BPRS Scales by Group at each 6 month Time Period
Assessment CSP MI
CSP NMI
GP MI
GP NMI
SCCF
All
Activity
6.30 (1.74)
5.56 (1.05)
6.21 (1.54)
5.82 (1.34)
6.82 (2.58)
6.18 (1.84)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
5.92 (1.49)
5.26 (.67)
5.90 (1.27)
5.42 (1.18)
6.25 (1.44)
5.78 (1.31)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
6.17 (2.36)
5.18 (.49)
5.96 (1.56)
5.28 (.53)
6.27 (2.05)
5.82 (1.73)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Anxious‐Depressed
9.51 (3.03)
7.37 (2.22)
9.03 (2.90)
6.67 (1.81)
10.40 (3.34)
8.74 (3.08)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
9.07 (2.95)
6.62 (2.15)
8.20 (2.44)
6.50 (1.84)
8.91 (2.64)
7.97 (2.71)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
8.44 (2.85)
7.04 (2.95)
8.22 (2.31)
5.97 (1.68)
9.06 (3.13)
7.90 (2.92)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Hostility‐Suspiciousness
5.52 (2.46)
4.04 (1.76)
4.70 (1.72)
3.90 (1.83)
5.47 (3.01)
4.81 (2.39)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
5.17 (2.45)
3.34 (.64)
4.43 (2.16)
3.42 (1.26)
4.38 (1.54)
4.20 (1.86)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
4.54 (2.16)
3.36 (.71)
4.37 (1.86)
3.76 (1.62)
4.72 (2.17)
4.19 (1.88)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Thought Disorder
6.59 (2.35)
5.32 (.87)
5.64 (.99)
5.18 (.51)
8.29 (3.33)
6.38 (2.41)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
6.50 (1.87)
5.17 (.48)
5.33 (.84)
5.10 (.39)
6.59 (1.94)
5.85 (1.54)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
6.35 (2.45)
5.18 (.49)
5.44 (1.22)
5.38 (1.35)
6.61 (2.11)
5.89 (1.84)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Withdrawal
7.73 (1.76)
6.79 (1.18)
7.00 (1.41)
6.38 (.63)
8.61 (2.60)
7.43 (1.92)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
7.83 (1.96)
7.06 (1.40)
6.83 (1.26)
6.21 (.53)
7.59 (1.56)
7.20 (1.58)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
7.50 (1.62)
6.71 (1.74)
7.22 (1.42)
6.31 (.76)
7.59 (1.55)
7.14 (1.57)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Total
35.66 (7.60) 29.07 (4.71)
32.58 (5.38)
27.95 (4.88)
39.60 (9.69) 33.52 (8.28)
Time 1
n = 59
n = 57
n = 33
n = 39
n = 62
n = 250
34.50 (7.64) 27.45 (3.51)
30.70 (4.76)
26.66 (3.77)
33.71 (4.67) 31.00 (6.13)
Time 3
n = 54
n = 47
n = 30
n = 38
n = 56
n = 225
33.00 (8.56) 27.47 (4.59)
31.22 (4.20)
26.69 (3.57)
34.25 (7.12) 30.92 (6.93)
Time 5
n = 48
n = 45
n = 27
n = 29
n = 51
n = 200
Table B4 provides the Cronbach’s alpha estimates for the subscales at each assessment period. The internal
consistency estimates for the BPRS subscales (M = .55, range = .40 to .66) were lower than those found in
normative samples but similar across time periods; however, the internal consistency estimates for the total
score is similar to that found in normative samples. Correlations between sequential time periods (6 months
APPENDICES
99
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
apart) are provided in Table B5 and show low stability across time. It is possible that participants changed
facilities from one testing session to the next, causing a switch in their assigned clinicians. This change in cli‐
nicians could lower correlations between BPRS scores across time and present a picture of inmates’ psycho‐
pathological instability when, in fact, inter‐rater disparity might be causing the change in BPRS scores over
time. Correlations between the BPRS scales and relevant self‐report scales of the same construct ranged
between .15 and .35 (M = .28). Correlations between the BPRS scales and the relevant correctional officer
ratings (PBRS scales) ranged between .08 and .29 (M = .19). These convergent validity estimates are lower
than expected and are likely impacted by restriction of range (i.e., scores on the BPRS are averaging at the
low end of the possible scores and standard deviations are small).
Table B4. Internal Consistency Estimates (Cronbach’s
alpha) for BPRS Scales at each Time Period
BPRS Scale
Time 1 Time 3 Time 5
Activity
.58
.53
.64
Anxious‐Depressed
.55
.60
.66
Hostility‐Suspiciousness
.57
.61
.51
Thought Disorder
.64
.52
.57
Withdrawal
.47
.49
.40
Total Scale
.81
.80
.79
Table B5. Correlations between Con
secutive Time Periods for BPRS Scales
BPRS Scale
T1‐T3 T3‐T5
Activity
.36
.40
Anxious‐Depressed
.45
.43
Hostility‐Suspiciousness
.36
.48
Thought Disorder
.33
.58
Withdrawal
.30
.23
Total Scale
.41
.51
77
TT
Brief Symptom Inventory (BSI)
The BSI (Derogatis, 1993) is a 53‐item self‐report measure that is widely employed to assess a broad range
of psychological symptoms. It measures clinical symptoms across nine subscales (i.e., Somatization, Obses‐
sive‐Compulsive, Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation,
and Psychoticism) and three global scales (i.e., General Severity Index [GSI]; Positive Symptom Total; Positive
Symptom Distress Index; Boulet & Boss, 1991). Respondents are asked to rate the degree of distress expe‐
rienced over the last week, using a 5‐point rating scale (0 – not at all to 4 – extremely). Higher scores on the
BSI indicate a greater degree of psychopathology. Despite having different subscales, the BSI seems to be
better at providing information on the general degree of psychopathology instead of the nature of it (Boulet
& Boss, 1991). A minimum of 6th grade reading ability is required to complete this measure, and it generally
takes 10 minutes to complete (Boulet & Boss, 1991).
Internal consistency reliabilities across subscales are acceptable for clinical populations (range = .57 to .89;
Boulet & Boss, 1991; Broday & Mason, 1991; Hayes, 1997; Kellett, Beail, Newman, & Frankish, 2003). Addi‐
tionally, internal consistency reliabilities for nonclinical, community samples for the different subscales
ranged from .60 to .81 (Kellett et al., 2003), whereas they ranged from .52 to .86 for forensic populations
(Kellett et al., 2003; Zinger, Wichmann, & Andrews, 2001). Item‐total correlations for the scales ranged from
100
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
.57 to .79, with a median correlation of .69, for clinical populations (Hayes, 1997) and ranged from .73 to .91
for forensic populations (Boulet & Boss, 1991). Two‐week test‐retest reliability is acceptable for the subs‐
cales (range = .68 [Somatization] to .91 [i.e., Phobic Anxiety]) and the GSI (r = .90; Cundick, 1975; Derogatis,
1993; Kellett et al., 2003; Piersma, Reaume, Boes, 1994) across nonclinical, clinical, and forensic samples.
The BSI has been shown to be valid for studying change over time (Long, Harring, Brekke, Test, & Greenberg,
2007).
Normative data are widely available for psychiatric in‐ and out‐patients and the general population but not
for a prison population (Derogatis, 1993). Normative means for the different subscales ranged from .67 (SD
= .71) to 1.65 (SD = 1.11) in psychiatric outpatients, from .71 (SD = .97) to 1.26 (SD = 1.15) in psychiatric in‐
patients, and from .11 (SD = .25) to .37 (SD = .41) in nonclinical populations (Derogatis, 1993). Cochran and
Hale (1985) conducted a normative study on male and female college students at a 4‐year college. They
found that mean scores ranged from .29 (SD = .27) to 1.17 (SD = .77) for males (n = 143) and from .32 (SD =
.45) to 1.12 (SD = .66) for females (n = 204). Furthermore, normative data are available on the global scales
of the BSI; the normative mean for the BSI GSI was 1.20 (SD = .70) for psychiatric outpatients, .25 (SD = .24)
for nonclinical populations, and .97 (SD = .78) for psychiatric inpatients (Derogatis, 1993).
Convergent validity has been assessed by means of comparing dimensions of the BSI to clusters on the Min‐
nesota Multiphasic Personality Inventory (MMPI; Boulet & Boss, 1991; Cundick, 1975). Correlations of these
comparisons were between .30 and .72 in Cundick’s (1975) study. In Boulet and Boss’s (1991) study, correla‐
tions between the most relevant MMPI and BSI subscales were found to be moderate, ranging from .50
(MMPI Depression and BSI Depression) to .53 (MMPI Hypochondriasis and BSI Somatization). In a clinical
sample, some of the BSI subscales were significantly correlated with the associated subscales on the BPRS:
the depression scale on the BSI was significantly correlated to the depressive mood scale on the BPRS (r =
.69), the anxiety scales on the BSI and BPRS correlated as well (r = .49), and the two hostility scales of both
measures were also significantly correlated with one another (r = .49; Morlan & Tan, 1998). Overall, mod‐
erate to high correlations with other measures seem to indicate that the BSI does, indeed, have adequate
convergent validity.
Table B6 provides the summary statistics for the present study on the BSI scales at each time period. Inter‐
nal consistency estimates for the BSI subscales were strong with Cronbach’s alphas ranging between .71 and
.91 (M = .85). Test‐retest reliability estimates ranged between .53 and .79 (M = .72) indicating good stability
within three month testing intervals. The BSI subscales showed reasonable convergent validity as correla‐
tions with other self‐report measures of the same constructs ranged between .15 and .89 (M = .56) but
there were lower validity estimates with staff reports with correlations ranging between ‐.01 and .43 (M = .23).
Table B6. Summary Statistics (M, SD, n) on BSI Scales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Anxiety
7.87 (6.52) 3.43 (4.00) 7.30 (5.56) 2.30 (3.15) 9.69 (5.75)
1
n = 64
n = 63
n = 33
n = 43
n = 67
7.21 (6.01) 2.24 (3.71) 5.56 (5.58) 1.46 (2.65) 7.91 (6.29)
2
n = 62
n = 58
n = 32
n = 41
n = 64
6.69 (6.50) 2.68 (3.65) 6.25 (5.75) 1.85 (4.19) 7.92 (6.11)
3
n = 59
n = 57
n = 32
n = 41
n = 61
5.92 (6.39) 2.71 (3.36) 5.17 (5.52) 1.49 (2.21) 8.11 (6.27)
4
n = 60
n = 56
n = 29
n = 39
n = 59
APPENDICES
All
6.33 (5.92)
n = 270
5.14 (5.77)
n = 257
5.23 (5.88)
n = 250
4.91 (5.66)
n = 243
101
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
CSP MI
5.61 (6.25)
5
n = 56
4.84 (5.88)
6
n = 51
Depression
9.32 (6.43)
1
n = 63
7.92 (5.92)
2
n = 62
8.07 (6.94)
3
n = 59
7.18 (6.48)
4
n = 60
7.30 (6.41)
5
n = 56
6.29 (5.98)
6
n = 51
Hostility
6.74 (5.20)
1
n = 64
5.66 (4.56)
2
n = 62
5.78 (5.23)
3
n = 59
5.30 (5.50)
4
n = 60
5.61 (5.15)
5
n = 56
4.32 (4.77)
6
n = 51
Interpersonal Sensitivity
5.39 (4.67)
1
n = 63
5.14 (4.02)
2
n = 62
4.62 (4.44)
3
n = 60
4.00 (4.07)
4
n = 60
4.42 (4.21)
5
n = 56
4.13 (4.43)
6
n = 51
Obsessive‐Compulsive
9.76 (6.52)
1
n = 64
8.92 (6.48)
2
n = 62
8.60 (7.16)
3
n = 59
7.87 (6.54)
4
n = 60
8.07 (6.51)
5
n = 56
102
CSP NMI
2.43 (3.20)
n = 56
2.78 (3.56)
n = 54
5.94 (5.24)
n = 63
4.72 (5.05)
n = 58
4.64 (5.10)
n = 57
4.27 (4.63)
n = 56
4.05 (4.76)
n = 56
3.96 (4.70)
n = 54
4.02 (3.95)
n = 63
2.53 (3.72)
n = 58
3.10 (4.04)
n = 57
3.66 (4.54)
n = 56
3.45 (4.36)
n = 56
3.56 (4.64)
n = 54
3.43 (3.22)
n = 63
2.36 (2.81)
n = 58
2.44 (3.21)
n = 57
2.64 (3.11)
n = 56
2.52 (2.79)
n = 56
2.24 (2.72)
n = 54
5.41 (4.19)
n = 63
4.28 (4.09)
n = 58
4.46 (4.14)
n = 57
4.75 (96)
n = 56
4.27 (4.33)
n = 56
GP MI
4.79 (5.12)
n = 29
GP NMI
1.08 (1.85)
n = 38
SCCF
7.47 (5.77)
n = 57
NA
NA
NA
7.97 (6.34)
n = 33
6.44 (5.86)
n = 32
7.38 (6.01)
n = 32
6.62 (5.25)
n = 29
5.76 (4.98)
n = 29
3.24 (3.65)
n = 43
2.22 (3.49)
n = 41
2.49 (4.18)
n = 41
1.95 (2.62)
n = 39
1.68 (2.58)
n = 38
11.91 (6.22)
n = 67
10.46 (7.02)
n = 64
9.87 (7.24)
n = 61
10.20 (7.28)
n = 59
10.07 (7.16)
n = 57
NA
NA
NA
5.58 (5.01)
n = 33
4.34 (3.92)
n = 32
5.12 (4.16)
n = 32
5.31 (5.25)
n = 29
5.17 (4.23)
n = 29
1.93 (2.77)
n = 43
1.71 (1.98)
n = 41
1.71 (2.78)
n = 41
1.91 (2.64)
n = 39
1.21 (1.92)
n = 38
6.50 (5.47)
n = 67
5.76 (5.22)
n = 64
6.30 (5.12)
n = 61
6.64 (6.23)
n = 59
5.79 (4.96)
n = 57
NA
NA
NA
5.15 (3.99)
n = 33
3.72 (3.60)
n = 32
4.44 (3.99)
n = 32
3.90 (4.04)
n = 29
3.41 (3.63)
n = 29
2.40 (3.43)
n = 43
1.68 (2.36)
n = 41
1.27 (2.42)
n = 41
1.23 (2.04)
n = 39
0.89 (1.72)
n = 38
6.91 (3.96)
n = 67
5.98 (4.04)
n = 64
6.28 (4.34)
n = 61
6.39 (4.57)
n = 59
6.56 (4.33)
n = 57
NA
NA
NA
10.02 (6.54)
n = 33
8.31 (5.99)
n = 32
9.03 (5.86)
n = 32
8.14 (5.60)
n = 29
7.59 (5.21)
n = 29
3.02 (3.29)
n = 43
2.78 (3.35)
n = 41
2.66 (3.76)
n = 41
2.26 (3.19)
n = 39
2.23 (3.19)
n = 38
11.03 (5.36)
n = 67
10.22 (6.44)
n = 64
9.95 (6.66)
n = 61
10.54 (6.62)
n = 59
9.18 (6.36)
n = 57
All
4.47 (5.34)
n = 236
3.74 (4.91)
n = 106
8.03 (6.42)
n = 269
6.74 (6.34)
n = 257
6.73 (6.62)
n = 250
6.34 (6.32)
n = 243
6.10 (6.28)
n = 236
5.05 (5.46)
n = 106
5.14 (4.95)
n = 270
4.19 (4.47)
n = 257
4.54 (4.76)
n = 250
4.70 (5.31)
n = 243
4.38 (4.69)
n = 236
3.89 (4.69)
n = 106
4.80 (4.20)
n = 269
4.00 (3.85)
n = 257
3.96 (4.19)
n = 251
3.81 (4.10)
n = 243
3.80 (4.02)
n = 236
3.13 (3.75)
n = 106
8.02 (6.07)
n = 270
7.14 (6.18)
n = 257
7.06 (6.39)
n = 250
6.93 (6.32)
n = 243
6.44 (5.95)
n = 236
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
6
Paranoid Ideation
1
2
3
4
5
6
Phobic Anxiety
1
2
3
4
5
6
Psychoticism
1
2
3
4
5
6
Somatization
1
2
3
4
5
6
APPENDICES
CSP MI
6.96 (6.09)
n = 51
8.66 (4.78)
n = 64
7.71 (4.41)
n = 62
7.51 (5.35)
n = 59
6.43 (5.06)
n = 60
7.10 (5.23)
n = 56
6.78 (5.30)
n = 51
506 (5.18)
n = 64
4.60 (5.09)
n = 62
4.56 (5.31)
n = 59
3.98 (5.50)
n = 59
3.84 (4.53)
n = 56
3.20 (4.74)
n = 51
7.30 (5.16)
n = 63
6.84 (4.72)
n = 62
6.20 (5.16)
n = 59
5.06 (4.81)
n = 60
5.48 (5.04)
n = 56
5.27 (4.64)
n = 51
7.64 (7.35)
n = 64
5.76 (5.63)
n = 62
6.12 (6.59)
n = 59
5.55 (6.65)
n = 60
5.09 (5.18)
n = 56
3.74 (5.50)
n = 51
CSP NMI
4.23 (4.62)
n = 54
GP MI
GP NMI
SCCF
NA
NA
NA
6.68 (4.52)
n = 62
4.72 (3.95)
n = 58
4.95 (3.98)
n = 57
5.24 (4.45)
n = 56
5.18 (4.73)
n = 56
5.07 (4.48)
n = 54
7.94 (5.12)
n = 33
5.88 (4.93)
n = 32
6.75 (4.98)
n = 32
5.31 (3.87)
n = 29
5.45 (4.73)
n = 29
4.49 (4.71)
n = 43
3.56 (3.79)
n = 41
2.70 (3.63)
n = 41
3.02 (3.22)
n = 39
2.47 (3.83)
n = 38
9.29 (4.90)
n = 67
8.78 (5.32)
n = 64
9.26 (4.84)
n = 61
9.78 (5.39)
n = 59
8.56 (4.71)
n = 57
NA
NA
NA
1.89 (3.28)
n = 63
1.43 (2.70)
n = 58
1.22 (2.48)
n = 57
1.05 (2.57)
n = 56
1.27 (2.34)
n = 56
1.13 (2.84)
n = 54
3.82 (4.81)
n = 33
2.72 (3.87)
n = 32
3.59 (4.65)
n = 32
2.41 (3.64)
n = 29
2.3 (3.67)
n = 29
1.28 (2.37)
n = 43
0.83 (1.73)
n = 41
0.86 (1.67)
n = 41
0.69 (1.30)
n = 39
0.50 (1.11)
n = 38
6.15 (5.15)
n = 67
4.86 (5.21)
n = 64
5.41 (5.22)
n = 61
5.47 (4.84)
n = 59
4.74 (4.71)
n = 57
NA
NA
NA
4.83 (4.26)
n = 63
3.28 (3.31)
n = 58
3.32 (3.41)
n = 57
3.52 (3.34)
n = 56
3.54 (3.77)
n = 56
3.14 (3.72)
n = 54
6.27 (5.20)
n = 33
5.28 (4.62)
n = 32
5.44 (4.58)
n = 32
4.66 (3.70)
n = 29
4.76 (3.38)
n = 29
3.23 (3.51)
n = 43
2.07 (3.09)
n = 41
2.07 (3.75)
n = 41
1.90 (2.73)
n = 39
1.50 (3.16)
n = 38
9.05 (4.99)
n = 67
7.94 (4.93)
n = 64
7.90 (5.18)
n = 61
8.04 (5.19)
n = 59
7.60 (4.81)
n = 57
NA
NA
NA
3.89 (5.22)
n = 63
2.59 (4.27)
n = 58
3.37 (3.82)
n = 57
2.93 (4.38)
n = 56
3.05 (4.35)
n = 56
2.71 (3.82)
n = 54
5.21 (5.21)
n = 33
4.34 (4.45)
n = 32
5.59 (6.23)
n = 32
3.58 (3.87)
n = 29
4.52 (5.12)
n = 29
2.46 (3.31)
n = 43
1.61 (2.99)
n = 41
1.66 (2.97)
n = 41
1.20 (2.23)
n = 39
1.47 (3.34)
n = 38
8.09 (5.76)
n = 67
7.07 (5.62)
n = 64
6.67 (6.52)
n = 61
7.02 (6.50)
n = 59
6.30 (6.33)
n = 57
NA
NA
NA
All
5.50 (5.53)
n = 106
7.60 (5.03)
n = 269
6.41 (4.90)
n = 257
6.47 (5.11)
n = 250
6.29 (5.11)
n = 243
6.05 (5.10)
n = 236
5.85 (4.96)
n = 106
3.84 (4.75)
n = 270
3.11 (4.42)
n = 257
3.28 (4.59)
n = 250
2.95 (4.44)
n = 242
2.80 (3.97)
n = 236
2.11 (3.99)
n = 106
6.38 (5.08)
n = 269
5.36 (4.76)
n = 257
5.18 (4.96)
n = 250
4.87 (4.66)
n = 243
4.80 (4.67)
n = 236
4.14 (4.31)
n = 106
5.76 (6.07)
n = 270
4.53 (5.23)
n = 257
4.83 (5.77)
n = 250
4.37 (5.63)
n = 243
4.24 (5.53)
n = 236
3.18 (4.70)
n = 106
103
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Coolidge Correctional Inventory (CCI)
The Coolidge Correctional Inventory (CCI; Coolidge, 2004) is an adaptation of the Coolidge Axis II Inventory
(CATI; Coolidge, n.d. a; Coolidge, Segal, Klebe, Cahill, & Whitcomb, 2009) designed for use by CDOC with new
prison admissions to identify personality disorders and neuropsychological problems among inmates. The
assessment follows a self‐report format with a 4‐point scale (1—strongly false to 4—strongly true) across
250 items. Scores are obtained for a total of 33 different scales (Coolidge et al., 2009) based on the Ameri‐
can Psychiatric Association’s (2000) diagnostic criteria (DSM‐IV‐TR). The CCI can be used to assess 14 perso‐
nality disorders – 10 from the DSM‐IV‐TR Axis II, 2 from the DSM‐IV‐TR (American Psychiatric Association,
2000) appendix, and 2 from the DSM‐III Axis II (American Psychiatric Association, 1980; Coolidge et al.,
2009). The personality disorders assessed by the CCI are as follows: Antisocial, Avoidant, Borderline, Depen‐
dent, Depressive, Histrionic, Narcissistic, Obsessive‐Compulsive, Paranoid, Passive‐Aggressive, Sadistic, Schi‐
zoid, Schizotypal, and Self‐Defeating. Furthermore, the CCI is used to assess other psychological and neurop‐
sychological problems and syndromes (i.e., Introversion‐Extroversion, Maladjustment, Executive Functions,
Decision Difficulty, Planning Problems, Neuropsychological Dysfunction, Language, Memory, Neurosomatic
Issues, Hostility‐Anger, Hostility‐Danger, Hostility‐Impulsivity, Hypersensitivity, Drug and Alcohol Problems)
as well as five selected Axis I scales and associated subscales (i.e., ADHD, Post‐Traumatic Stress Disorder,
Psychotic Thinking, Schizophrenia, Social Phobia, Withdrawal, Anxiety, and Depression; Coolidge et al.,
2009). The CCI also has response validity scales available. For this study, the CCI personality disorders and
Axis I scales were used as potential predictors of outcomes.
The Cronbach’s alphas for the CCI’s subscales were found to be acceptable within prison populations, with a
median Cronbach’s alpha of .78 for the personality subscales (range = .65 ‐ .86) in a sample of 3,962 inmates
(Coolidge et al., 2009) and a median Cronbach’s alpha of .75 (range = .47 ‐ .84) in a sample of 3,090 inmates
(Whitcomb, 2006). Mean scores on the personality disorders subscales of the CCI ranged from 42.76 (SD =
8.59) to 54.27 (SD = 10.88), with a median of 47.67 (Coolidge et al., 2009). In another study, Whitcomb
(2006) found mean scores ranging from 41.25 (SD = 9.47) to 58.80 (SD = 9.97) for violent offenders and from
41.38 (SD = 9.50) to 58.71 (SD = 9.29) for nonviolent offenders.
Because the CCI is an adaptation of the CATI and little research has been done on the CCI, test‐retest reliabil‐
ity as well as convergent validity can, at the very least, be evaluated for the CATI as it measures many of the
same components as the CCI but was not designed for correctional populations (Coolidge, n.d. a, Coolidge,
n.d. b). One‐week test‐retest reliabilities were found to be strong, with an average correlation of .90 for the
personality disorders (Coolidge, n.d. b). Scores on personality disorder scales of the CATI were correlated
with scores on the respective Brief Millon Clinical Multiaxial Inventory II (MCMI‐II) scales (range = .10 to .87;
Mdn = .58; Coolidge, n.d. b).
Cronbach’s alphas for the single assessment period of the CCI were varied with values ranging between .46
and .88 (M = .74). The majority of the internal consistency estimates were greater than .70 with lower esti‐
mates for Histrionic (.66), Self‐defeating (.64), Schizoid (.55), and Impulsivity (.46) scales.
Deliberate Self‐Harm Inventory (DSHI)
In order to assess self‐harming behavior in inmates who participated in this study, we used the Deliberate
Self‐Harm Inventory (DSHI; Gratz, 2001; Gratz & Chapman, 2007). The DSHI is a 17‐item measure that ques‐
tions respondents about various self‐harming behaviors. Engagement in as well as frequency of engagement
104
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
in different self‐harming behaviors are assessed. At the first testing session in this study, participants were
asked about their lifetime history of deliberate self‐harm. Specifically, they were asked to indicate whether
they have ever engaged in the various self‐harming behaviors and, if so, how old they were when they first
engaged in the activity, how many times they engaged in the activity, when they most recently engaged in
the behavior, how many years they engaged in the behavior, and whether engaging in the activity ever led
to required medical treatment and/or hospitalization. Completion of this assessment takes most people less
than 5 minutes (Fliege et al., 2006). For this study, self‐harming behavior was coded as a dichotomous varia‐
ble; a self‐harm total score was computed by summing the yes/no responses across the 17 self‐harming be‐
haviors. Additionally, the DSHI was administered at the last testing period for some of the participants;
however, the number of participants was small and those assessments were not used in this report.
Internal consistency reliability for the DSHI in clinical populations was found to be .81, with a split‐half corre‐
lation of .78 (Fliege et al., 2006). Item‐total correlations in Fliege et al.’s (2006) study were between .23 and
.55. Item‐total correlations in a nonclinical sample ranged from .00 to .65, with a median item‐total correla‐
tion of .45 (Gratz, 2001). Two‐ to four‐week test‐retest reliability was acceptable at .91 in a clinical popula‐
tion (Fliege et al., 2006) and at .68 in a nonclinical population (Gratz, 2001). Among a nonclinical population,
the internal consistency coefficient was .82 for the DSHI (Gratz, 2001). Gratz also assessed the convergent
validity of the DSHI, finding significant moderate correlations with other self‐harm measures (range = .35 to
.49), such as the mental health history self‐harm item, Diagnostic Interview for Borderlines‐Revised self‐
harm item, and Suicide Behaviors Questionnaire self‐harm item.
For the current sample, the internal consistency estimate for the DSHI with dichotomous responses on the
17 items was acceptable at .84. Table B7 provides the proportion of people who responded yes to each item
and summary statistics for the total score.
Table B7. Summary Statistics for the DSHI Items and Total Score
Item
CSP MI CSP NMI
Cutting
47%
17%
Burn with cigarette
44%
19%
Burn with match/lighter
33%
14%
Carved words into skin
16%
14%
Carved pictures into skin
14%
14%
Purposefully scratched
14%
5%
Bitten self (broke skin)
9%
2%
Rubbed sandpaper on body
3%
3%
Dripped acid on skin
3%
0%
Used bleach, comet, oven cleaner to scrub skin
6%
2%
Stuck sharp objects into skin
30%
11%
Rubbed glass into skin
5%
0%
Broken own bones
5%
2%
Banged head
20%
5%
Punched self
11%
3%
Prevented wounds from healing
8%
0%
Any other self‐harm
34%
8%
3.05
1.19
Total M (SD)
(3.43)
(1.82)
GP MI
33%
24%
24%
21%
15%
6%
3%
0%
0%
0%
9%
0%
0%
18%
9%
3%
18%
1.85
(2.28)
GP NMI
5%
12%
9%
5%
2%
0%
0%
0%
2%
0%
2%
0%
0%
2%
2%
2%
0%
0.46
(1.44)
SCCF
61%
25%
22%
25%
18%
24%
18%
9%
3%
4%
28%
6%
6%
27%
13%
16%
42%
3.50
(3.48)
All
35%
26%
21%
17%
13%
11%
7%
4%
2%
3%
18%
3%
3%
15%
8%
7%
23%
2.17
(2.96)
Note. Item statistics are the percentage of persons who indicated a history of the self‐harm behavior over their lifetime prior to the
study.
APPENDICES
105
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Personality Assessment Screener (PAS)
The PAS is a quick and effective screening tool that gauges the social functioning of an individual globally
and across 10 subscales (viz., Negative Affect, Acting Out, Health Problems, Psychotic Features, Social With‐
drawal, Hostile Control, Suicidal Thinking, Alienation, Alcohol Problem, Anger Control; Harrison & Rogers,
2007). The alcohol problem subscale was not included in this study because it was not a construct of inter‐
est. The PAS is a 22‐item screening measure that was originally derived from the larger Personality Assess‐
ment Inventory (Morey, 1991). Respondents rate each statement on a 4‐point scale (F—false, ST—slightly
true, MT—mostly true, VT—very true); higher scores on this measure indicate greater severity of clinical
problems (Morey, 1997) or problems with impression management, as Holden, Book, Edwards, Wasylkiw,
and Starzyk (2003) termed it. In order to complete this test, a 4th grade reading level is required; it should
take no longer than 5 minutes to complete this assessment (Morey, 1997). For this study, raw scores rather
than P‐scores were evaluated.
Levels of internal consistency reliability are acceptable for the PAS total score as well as for subscale scores. In a
sample of county jail inmates (N = 100), the Cronbach’s alpha was .74 (Harrison & Rogers, 2007) for total scores
but lower for subscales, which is likely due to the small number of items (2) on each subscale. Despite the low
number of items on each subscale, 6 of the 10 subscales exhibited alpha coefficients of .60 or greater (i.e., Nega‐
tive Affect, Health Problems, Psychotic Features, Social Withdrawal, Suicidal Thinking, and Alienation; Harrison &
Rogers, 2007); alpha coefficients on the remaining subscales were not provided in Harrison and Rogers’ (2007)
study. Alpha coefficients for the PAS total score have also been assessed in clinical and nonclinical samples. In a
nonclinical sample, internal consistency was found to be .75 for the total score and ranged between .34 (i.e., Al‐
cohol Problem) and .68 (i.e., Suicidal Thinking) for subscales (Morey, 1997). In a clinical sample, internal consis‐
tency was found to be .79 for the total score and ranged between .48 and .84 for subscales (Morey, 1997).
Both the total score and subscale scores were assessed to have good test‐retest reliability. For a nonclinical
sample, 1‐month test‐retest reliability was .89 for the total PAS score and ranged between .66 and .92 for
the subscales (Morey, 1997), with a median test‐retest reliability of .77 across subscales. For a clinical sam‐
ple, 1‐month test‐retest reliability was .85 for the total PAS score and ranged between .47 and .81 for the
subscales (Morey, 1997), with a median test‐retest reliability of .66 across subscales.
Normative data on the PAS are available for both clinical and nonclinical populations. The mean raw PAS
total score for a nonclinical, community sample was found to be 16.66 (SD = 7.40; Morey, 1997). Mean raw
scores on subscales ranged from .37 (SD = .94) to 4.05 (SD = .54; Holden et al., 2003; Morey, 1997). The
mean raw PAS total score for a clinical sample was found to be 25.83 (SD = 9.99; Morey, 1997). Mean raw
scores on subscales ranged from 1.19 (SD = 1.53) to 4.99 (SD = 2.48; Morey, 1997). Additionally, the PAS has
shown good convergent validity (Morey, 1997). The total score on the PAS has been positively correlated
with scores on the PAI and MMPI (Gondolf, 2008; Morey, 1997). Furthermore, adequate convergent validity
has also been shown for the different subscales of the PAS (Gondolf, 2008; Morey, 1997).
Table B8 provides the summary statistics for the study groups on the PAS scales at each time period. Cron‐
bach’s alphas were computed to assess internal consistency reliability with coefficients ranging between .27
and .95 (M = .64). These estimates were somewhat lower than those found in the literature. The lowest re‐
liability estimates were for the Acting Out and Hostile Control subscales. Test‐retest correlation coefficients
ranged between .54 and .84 (M = .69). Correlations of the PAS subscales with other self‐report measures of
106
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
the same construct ranged between .34 and .67 (M = .50) and with correctional officer and clinician ratings
the correlations ranged between .08 and .34 (M = .21).
Table B8. Summary Statistics (M, SD, n) on PAS Scales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Acting Out
6.38 (1.94) 6.40 (2.06) 7.24 (1.85) 6.40 (1.90) 6.79 (1.74)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
6.21 (2.07) 6.01 (1.99) 6.66 (1.94) 6.29 (1.93) 6.26 (1.88)
Time 2
n = 62
n = 59
n = 32
n = 41
n = 64
5.98 (2.04) 5.79 (2.15) 6.81 (1.49) 5.95 (2.12) 6.03 (1.97)
Time 3
n = 60
n = 57
n = 32
n = 41
n = 61
5.80 (2.07) 5.84 (2.21) 6.48 (1.57) 5.51 (2.39) 6.23 (2.02)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
5.52 (2.25) 6.07 (2.21) 6.66 (1.26) 5.78 (1.89) 6.32 (2.15)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
5.81 (2.42) 6.15 (206)
Time 6
NA
NA
NA
n = 51
n = 54
Alienation
3.44 (1.96) 3.08 (1.71) 3.88 (1.71) 2.63 (1.83) 4.06 (1.94)
Time 1
n = 63
n = 63
n = 33
n = 43
n = 67
4.12 (1.73) 3.41 (1.92) 3.84 (1.48) 2.93 (1.65) 4.42 (1.73)
Time 2
n = 60
n = 59
n = 32
n = 41
n = 64
3.92 (1.82) 3.35 (1.88) 4.06 (1.41) 2.83 (1.67) 4.44 (1.68)
Time 3
n = 60
n = 57
n = 32
n = 41
n = 61
3.86 (1.81) 3.55 (1.90) 3.32 (1.47) 2.85 (1.71) 4.58 (1.45)
Time 4
n = 59
n = 56
n = 28
n = 39
n = 59
3.86 (1.66) 3.45 (1.84) 3.48 (1.30) 2.84 (1.79) 4.46 (1.42)
Time 5
n = 56
n = 56
n = 29
n = 37
n = 57
3.78 (1.79) 3.52 (1.79)
Time 6
NA
NA
NA
n = 51
n = 54
Anger Control
3.52 (1.82) 2.79 (1.63) 3.06 (1.69) 2.19 (1.30) 2.92 (1.78)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
3.39 (1.83) 2.47 (1.56) 2.91 (1.78) 2.10 (1.18) 3.03 (1.94)
Time 2
n = 62
n = 58
n = 32
n = 41
n = 64
3.38 (1.76) 2.70 (1.65) 2.81 (1.47) 2.12 (1.42) 3.02 (1.78)
Time 3
n = 60
n = 56
n = 32
n = 41
n = 61
3.24 (1.72) 2.61 (1.67) 2.66 (1.54) 2.08 (1.44) 2.90 (1.87)
Time 4
n = 59
n = 56
n = 29
n = 39
n = 59
3.09 (1.69) 2.55 (1.49) 2.96 (1.73) 2.18 (1.61) 2.89 (1.75)
Time 5
n = 56
n = 56
n = 28
n = 38
n = 57
2.98 (1.61) 2.52 (1.66)
Time 6
NA
NA
NA
n = 51
n = 54
Health Problems
2.00 (1.62) 1.24 (1.64) 2.30 (1.78) 1.28 (1.79) 2.94 (1.70)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
2.05 (2.02) 1.12 (1.66) 1.84 (1.72) 1.17 (1.32) 2.84 (1.64)
Time 2
n = 61
n = 58
n = 32
n = 41
n = 64
2.43 (1.97) 1.18 (1.69) 1.91 (1.75) 1.05 (1.24) 2.54 (1.53)
Time 3
n = 60
n = 56
n = 32
n = 41
n = 61
1.95 (1.66) 1.07 (1.26) 1.62 (1.29) 0.95 (1.10) 2.60 (1.67)
Time 4
n = 59
n = 56
n = 29
n = 39
n = 58
APPENDICES
All
6.59 (1.91)
n = 270
6.24 (1.96)
n = 258
6.05 (2.01)
n = 251
5.95 (2.10)
n = 243
6.02 (2.08)
n = 236
5.96 (2.25)
n = 106
3.43 (1.90)
n = 269
3.80 (1.80)
n = 256
3.76 (1.80)
n = 251
3.74 (1.78)
n = 241
3.70 (1.70)
n = 235
3.64 (1.78)
n = 106
2.93 (1.72)
n = 270
2.82 (1.75)
n = 257
2.86 (1.69)
n = 250
2.75 (1.72)
n = 242
2.75 (1.67)
n = 235
2.74 (1.63)
n = 106
1.98 (1.81)
n = 270
1.87 (1.83)
n = 256
1.88 (1.77)
n = 250
1.70 (1.58)
n = 241
107
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
CSP MI
2.02 (1.93)
Time 5
n = 54
1.96 (1.65)
Time 6
n = 50
Hostile Control
3.35 (1.76)
Time 1
n = 62
3.00 (1.67)
Time 2
n = 61
2.78 (1.61)
Time 3
n = 60
2.85 (1.63)
Time 4
n = 59
2.87 (1.60)
Time 5
n = 55
2.90 (1.78)
Time 6
n = 50
Negative Affect
4.89 (2.48)
Time 1
n = 64
4.79 (2.06)
Time 2
n = 62
4.50 (2.20)
Time 3
n = 60
4.68 (2.35)
Time 4
n = 60
4.09 (2.32)
Time 5
n = 56
3.94 (2.14)
Time 6
n = 51
Psychotic Features
1.95 (1.54)
Time 1
n = 64
1.90 (1.68)
Time 2
n = 61
2.05 (1.82)
Time 3
n = 60
1.80 (1.64)
Time 4
n = 60
2.04 (1.70)
Time 5
n = 56
1.74 (1.82)
Time 6
n = 51
Social Withdrawal
3.14 (1.74)
Time 1
n = 64
3.53 (1.82)
Time 2
n = 62
3.30 (1.82)
Time 3
n = 60
3.52 (1.69)
Time 4
n = 60
3.13 (1.77)
Time 5
n = 56
108
CSP NMI
0.94 (1.13)
n = 55
1.17 (1.38)
n = 54
3.44 (1.62)
n = 63
3.26 (1.50)
n = 58
3.32 (1.66)
n = 57
3.29 (1.50)
n = 55
3.18 (1.54)
n = 55
3.17 (1.60)
n = 53
3.40 (1.56)
n = 63
3.06 (1.88)
n = 59
3.17 (1.74)
n = 57
3.21 (1.99)
n = 56
3.18 (1.88)
n = 56
3.06 (1.73)
n = 54
1.46 (1.51)
n = 63
1.24 (1.28)
n = 59
1.23 (1.27)
n = 57
1.38 (1.45)
n = 56
1.45 (1.55)
n = 55
1.33 (1.24)
n = 54
2.28 (1.42)
n = 63
2.51 (1.68)
n = 59
2.43 (1.69)
n = 56
2.68 (1.65)
n = 56
2.55 (1.62)
n = 56
GP MI
2.00 (1.58)
n = 29
GP NMI
0.97 (1.28)
n = 38
SCCF
2.53 (1.59)
n = 57
NA
NA
NA
2.76 (1.58)
n = 33
2.91 (1.55)
n = 32
2.81 (1.45)
n = 32
3.17 (1.07)
n = 29
2.76 (1.33)
n = 29
3.28 (1.45)
n = 42
3.07 (1.37)
n = 41
3.17 (1.46)
n = 41
3.13 (1.32)
n = 39
3.16 (1.52)
n = 38
2.61 (1.75)
n = 67
2.11 (1.44)
n = 64
2.48 (1.51)
n = 61
2.51 (1.60)
n = 59
2.16 (1.36)
n = 55
NA
NA
NA
4.73 (2.30)
n = 33
4.56 (2.50)
n = 32
4.59 (2.11)
n = 32
4.07 (1.93)
n = 29
4.17 (1.98)
n = 29
2.64 (1.64)
n = 43
2.68 (1.88)
n = 41
2.34 (1.77)
n = 41
2.38 (1.60)
n = 39
2.21 (1.51)
n = 38
5.60 (2.32)
n = 67
5.08 (2.16)
n = 64
4.95 (2.33)
n = 61
5.36 (2.31)
n = 59
5.12 (2.25)
n = 57
NA
NA
NA
1.91 (1.55)
n = 33
1.62 (1.74)
n = 32
1.59 (1.62)
n = 32
1.00 (.92)
n = 29
1.28 (1.13)
n = 29
0.95 (1.25)
n = 43
0.85 (1.11)
n = 41
0.93 (1.33)
n = 41
1.03 (1.20)
n = 39
0.79 (1.23)
n = 38
2.31 (1.88)
n = 65
2.17 (2.05)
n = 64
2.38 (1.72)
n = 60
2.30 (1.87)
n = 59
2.14 (1.92)
n = 57
NA
NA
NA
2.91 (1.88)
n = 33
2.75 (1.87)
n = 32
2.91 (1.61)
n = 32
2.72 (1.41)
n = 29
2.66 (1.37)
n = 29
2.05 (1.68)
n = 43
2.20 (1.42)
n = 41
2.05 (1.38)
n = 40
2.31 (1.73)
n = 39
2.26 (1.62)
n = 38
3.46 (1.88)
n = 67
3.61 (1.81)
n = 64
3.73 (1.66)
n = 60
3.48 (1.66)
n = 58
3.82 (1.75)
n = 56
All
1.72 (1.66)
n = 233
1.53 (1.56)
n = 105
3.10 (1.68)
n = 267
2.84 (1.56)
n = 256
2.90 (1.57)
n = 251
2.95 (1.50)
n = 241
2.81 (1.52)
n = 232
3.04 (1.68)
n = 104
4.34 (2.35)
n = 270
4.10 (2.28)
n = 258
3.97 (2.26)
n = 251
4.06 (2.34)
n = 243
3.83 (2.26)
n = 236
3.50 (1.98)
n = 106
1.76 (1.64)
n = 268
1.61 (1.69)
n = 257
1.70 (1.66)
n = 250
1.60 (1.59)
n = 243
1.63 (1.66)
n = 235
1.52 (1.56)
n = 106
2.82 (1.78)
n = 270
3.01 (1.81)
n = 258
2.96 (1.76)
n = 248
3.02 (1.71)
n = 242
2.96 (1.74)
n = 2356
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
CSP MI
CSP NMI
3.16 (1.76) 2.57 (1.60)
Time 6
n = 50
n = 53
Suicidal Thinking
0.95 (1.64) 0.35 (1.12)
Time 1
n = 62
n = 62
0.93 (1.63) 0.29 (.90)
Time 2
n = 61
n = 58
0.83 (1.59) 0.21 (.86)
Time 3
n = 58
n = 57
1.03 (1.83) 0.21 (.97)
Time 4
n = 60
n = 56
0.87 (1.62) 0.20 (.80)
Time 5
n = 55
n = 56
0.86 (1.71) 0.26 (1.08)
Time 6
n = 51
n = 53
GP MI
GP NMI
SCCF
NA
NA
NA
0.53 (1.08)
n = 32
0.53 (1.22)
n = 32
0.62 (1.36)
n = 32
0.52 (1.30)
n = 29
0.52 (1.30)
n = 29
0.05 (.21)
n = 43
0.05 (.31)
n = 41
0.10 (.49)
n = 41
0.02 (.16)
n = 39
0.10 (.51)
n = 38
2.64 (2.27)
n = 67
2.41 (2.36)
n = 64
2.31 (2.40)
n = 61
2.39 (2.24)
n = 59
2.02 (2.00)
n = 57
NA
NA
NA
All
2.88 (1.70)
n = 104
1.04 (1.81)
n = 266
0.96 (1.78)
n = 256
0.90 (1.77)
n = 249
0.95 (1.80)
n = 243
0.82 (1.58)
n = 235
0.55 (1.44)
n = 105
Prison Behavior Rating Scale (PBRS)
The PBRS was developed by Cooke (1998) for correctional staff to rate inmates’ behaviors in prison. While
the use of the PBRS in U.S. prisons has been limited, we were unable to find another rating scale that could
be easily used by correctional staff to record direct observations of inmates’ behaviors. The PBRS is a 36‐
item measure comprising three subscales: Anti‐Authority, Anxious‐Depressed, and Dull‐Confused. Higher
scores on the PBRS indicate worse behavior assessments of inmates by officers. Correctional staff use a 4‐
point rating scale (0—never/rarely, 1—sometimes, 2—often, 3—most of the time) to rate the inmates’ beha‐
viors within the 4 weeks preceding the assessment. The PBRS was modified to use language that was more
relevant for a United States sample; the questionnaire is given in Appendix A.
The PBRS demonstrated adequate internal consistency reliability across the three subscales in a sample of
467 male prisoners: .91 for Anti‐Authority, .84 for Anxious‐Depressed, and .72 for Dull‐Confused (Cooke,
1998). Cooke also demonstrated good test‐retest reliability over 2 to 3 weeks, with .76 for Anti‐Authority,
.86 for Anxious‐Depressed, and .82 for Dull‐Confused.
Table B9 provides the summary statistics for the study groups on the PBRS scales at each time period. Inter‐
nal consistency estimates for the PBRS scales at each time period are provided in Table B10. These alphas
indicate strong internal consistency with a mean alpha of .90. Correlations between sequential time periods
are provided in Table B11. Correlations between the first and second testing intervals tended to have the
weakest correlation coefficients (M = .16); this period is when many of the participants switched facilities so
there was a change in raters who may lack familiarity with the participants. Correlations between PBRS
scales and relevant clinician ratings were low (range = .08 to .24, M = .19) as they were with self‐report as‐
sessments (range = ‐.07 to .27, M = .10). Although the PBRS shows strong internal consistency estimates and
some evidence for test‐retest reliability, it does not relate well with other measures of similar constructs.
APPENDICES
109
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B9. Summary Statistics (M, SD, n) on PBRS Scales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
Anti‐Authority
7.62 (7.37)
7.68 (7.33)
6.05 (5.67)
6.28 (7.39)
Time 1
n = 63
n = 63
n = 26
n = 36
6.21 (7.40)
5.48 (6.67)
7.06 (5.07)
8.02 (6.96)
Time 2
n = 61
n = 59
n = 28
n = 35
6.17 (7.40)
3.73 (4.75)
9.01 (9.05)
6.92 (6.29)
Time 3
n = 57
n = 56
n = 28
n = 34
4.02 (5.84)
4.56 (4.88)
6.85 (7.96)
8.66 (6.32)
Time 4
n = 59
n = 55
n = 26
n = 31
4.88 (6.42)
3.32 (4.38)
7.30 (9.92)
7.63 (7.37)
Time 5
n = 56
n = 54
n = 27
n = 36
3.90 (5.11)
2.48 (4.27)
Time 6
NA
NA
n = 49
n = 54
Anxious‐Depressed
6.87 (7.26)
3.56 (4.61)
4.08 (5.73)
2.89 (3.59)
Time 1
n = 64
n = 63
n = 26
n = 36
4.39 (5.81)
2.00 (3.42)
5.24 (6.58)
3.59 (4.08)
Time 2
n = 61
n = 59
n = 28
n = 34
3.14 (4.10)
1.51 (2.72)
5.81 (3.91)
3.97 (3.64)
Time 3
n = 57
n = 56
n = 28
n = 33
3.41 (4.49)
1.51 (3.29)
3.59 (4.37)
8.66 (5.06)
Time 4
n = 59
n = 55
n = 26
n = 30
3.45 (4.57)
1.74 (3.21)
3.70 (4.90)
3.51 (5.39)
Time 5
n = 55
n = 54
n = 27
n = 35
4.00 (5.24)
1.63 (3.06)
Time 6
NA
NA
n = 48
n = 54
Dull‐Confused
3.80 (4.04)
2.14 (2.90)
6.05 (5.67)
6.28 (7.39)
Time 1
n = 64
n = 63
n = 26
n = 36
3.08 (4.01)
1.38 (2.44)
7.06 (5.07)
8.02 (6.96)
Time 2
n = 61
n = 58
n = 28
n = 34
2.35 (2.93)
0.93 (1.23)
9.01 (9.05)
6.92 (6.29)
Time 3
n = 57
n = 56
n = 28
n = 33
2.78 (3.55)
0.96 (1.41)
6.85 (7.96)
8.66 (6.32)
Time 4
n = 59
n = 55
n = 26
n = 31
2.64 (3.43)
1.05 (1.56)
7.30 (9.92)
7.63 (7.37)
Time 5
n = 55
n = 53
n = 27
n = 35
3.12 (3.64)
1.37 (2.64)
Time 6
NA
NA
n = 49
n = 54
Total Score
18.53 (15.94) 13.53 (12.84) 13.17 (13.80) 11.06 (11.93)
Time 1
n = 64
n = 63
n = 26
n = 36
13.98 (14.41) 9.00 (11.15) 15.12 (12.58) 13.64 (10.66)
Time 2
n = 61
n = 59
n = 28
n = 34
11.90 (13.52) 6.23 (7.60)
17.77 (12.88) 13.27(9.58)
Time 3
n = 57
n = 56
n = 28
n = 33
10.55 (11.75) 7.10 (7.44)
12.63 (12.74) 14.45 (11.97)
Time 4
n = 59
n = 55
n = 26
n = 30
11.10 (12.04) 6.38 (7.35)
12.26 (14.76) 13.54 (14.87)
Time 5
n = 55
n = 54
n = 27
n = 35
10.94 (12.11) 6.38 (7.35)
Time 6
NA
NA
n = 48
n = 54
110
SCCF
2.90 (5.51)
n = 67
4.59 (5.34)
n = 64
4.75 (5.62)
n = 51
5.34 (6.24)
n = 55
6.00 (7.14)
n = 54
NA
5.25 (5.06)
n = 67
5.25 (5.06)
n = 256
6.31 (5.46)
n = 246
5.73 (4.73)
n = 224
6.36 (6.20)
n = 224
NA
3.51 (3.94)
n = 67
4.18 (4.05)
n = 64
3.84 (3.63)
n = 51
4.01 (3.96)
n = 54
4.69 (4.82)
n = 54
NA
11.90 (13.52)
n = 67
15.43 (12.16)
n = 64
14.65 (11.77)
n = 51
15.96 (15.24)
n = 54
18.39 (16.48)
n = 54
NA
All
6.05 (6.99)
n = 255
5.97 (6.19)
n = 247
5.71 (6.68)
n = 226
5.43 (6.21)
n = 225
5.50 (6.98)
n = 227
3.15 (4.72)
n = 103
4.79 (5.65)
n = 256
4.30 (5.32)
n = 246
3.76 (4.18)
n = 224
3.73 (5.04)
n = 224
3.99 (5.40)
n = 225
2.74 (4.37)
n = 102
2.94 (3.50)
n = 256
2.78 (3.50)
n = 245
2.36 (2.85)
n = 225
2.45 (3.29)
n = 225
2.54 (3.65)
n = 224
2.20 (3.26)
n = 103
13.97 (14.01)
n = 256
13.25 (12.54)
n = 246
12.05 (11.58)
n = 225
11.77 (12.34)
n = 224
12.24 (13.74)
n = 225
8.06 (10.53)
n = 102
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B10. Internal Consistency Estimates (Cronbach’s alpha) for PBRS
Scales at each Time Period
PBRS Scale
Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Anti‐Authority
.94
.93
.94
.94
.95
.90
Anxious‐Depressed
.90
.90
.94
.90
.90
.90
Dull‐Confused
.84
.84
.78
.85
.87
.82
Total
.95
.94
.94
.95
.95
.94
Table B11. Testretest Correlations between Consecutive
Time Periods for each PBRS Subscale
Measure
T1–T2 T2–T3 T3–T4 T4–T5 T5–T6
Anti‐Authority
.24
.36
.33
.55
.59
Anxious‐Depressed
.14
.33
.46
.58
.48
Dull‐Confused
.08
.38
.39
.55
.31
Total
.16
.38
.42
.66
.51
Prison Symptom Inventory (PSI)
The PSI was created by the research staff for this study to measure variables that were not assessed by oth‐
er existing psychological measures but were thought to be important in association with long‐term segrega‐
tion. Using the literature concerning the impact of AS on psychological functioning (e.g., Grassian, 1983; Ha‐
ney, 2003), questions were written to assess symptoms associated with this form of confinement, including
nervousness, headaches, lethargy, chronic tiredness, trouble sleeping, a sense of impending breakdown,
perspiring hands, heart palpitations, dizziness, nightmares, trembling hands, and fainting. Furthermore,
questions about exercise, grooming, and safety issues within administration segregation were included in
the PSI. The scale has 39 items, rated on a 6‐point scale (0‐ never true to 5‐ always true). Questions were
grouped into the following nine areas: fear level, safety, panic disorder, sensitivity to external stimuli, physi‐
cal hygiene, physical well‐being and exercise, mental well‐being, mutism, and attitudes about administrative
segregation. The questionnaire is given in Appendix A.
Three subscales were used as part of the major constructs of interest: panic disorder as a measure of anxie‐
ty, sensitivity to external stimuli as a measure of hypersensitivity, and physical well‐being and exercise as a
measure of somatization. Fear level, safety, and attitudes about segregation subscales were used as predic‐
tors of how people changed over time, rather than as outcome variables. Analyses comparing groups on all
of the PSI scales are included in Appendix C. Table B12 provides the summary statistics for the study groups
on PSI scales at each time period.
Table B12. Summary Statistics (M, SD, n) on PSI Scales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
Attitudes about Segregation
2.95 (3.19)
1.68 (2.57)
2.46 (3.00)
1.00 (2.10)
Time 1
n = 57
n = 56
n = 28
n = 30
2.97 (3.54)
1.68 (2.63)
2.24 (2.76)
1.54 (2.42)
Time 2
n = 61
n = 56
n = 25
n = 26
3.02 (3.36)
1.04 (2.39)
2.04 (2.30)
1.22 (1.60)
Time 3
n = 55
n = 56
n = 25
n = 27
2.60 (3.21)
1.29 (2.43)
1.91 (2.45)
1.83 (2.58)
Time 4
n = 55
n = 55
n = 22
n = 24
3.12 (3.51)
1.45 (2.32)
2.54 (2.67)
1.30 (1.98)
Time 5
n = 52
n = 55
n = 22
n = 20
APPENDICES
SCCF
4.58 (3.47)
n = 67
5.55 (3.30)
n = 60
4.62 (3.45)
n = 55
4.96 (3.42)
n = 46
5.24 (3.57)
n = 45
All
2.81 (3.24)
n = 238
3.09 (3.42)
n = 228
2.58 (3.18)
n = 218
2.61 (3.20)
n = 202
2.89 (3.32)
n = 194
111
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
CSP MI
2.57 (3.45)
Time 6
n = 49
Fear Level
6.25 (4.62)
Time 1
n = 64
5.50 (4.18)
Time 2
n = 62
5.71 (4.10)
Time 3
n = 58
5.22 (3.83)
Time 4
n = 59
5.50 (3.50)
Time 5
n = 58
5.43 (3.43)
Time 6
n = 51
Hypersensitivity to External Stimuli
10.54 (4.02)
Time 1
n = 64
10.10 (4.52)
Time 2
n = 62
10.71 (4.65)
Time 3
n = 58
9.99 (4.64)
Time 4
n = 58
9.54 (4.09)
Time 5
n = 56
9.37 ( 4.33)
Time 6
n = 51
Mental Well‐Being
4.95 (2.48)
Time 1
n = 64
4.88 (2.24)
Time 2
n = 61
4.33 (2.42)
Time 3
n = 58
4.71 (2.43)
Time 4
n = 58
4.13 (2.24)
Time 5
n = 55
4.02 (2.01)
Time 6
n = 51
Mutism
3.67 (2.19)
Time 1
n = 64
4.44 (2.21)
Time 2
n = 62
4.14 (2.29)
Time 3
n = 57
4.41 (2.44)
Time 4
n = 58
4.09 (2.08)
Time 5
n = 55
4.00 (2.19)
Time 6
n = 50
112
CSP NMI
1.45 (2.32)
n = 52
GP MI
GP NMI
SCCF
NA
NA
NA
4.17 (3.46)
n =63
3.51 (2.54)
n = 59
3.91 (2.73)
n = 57
3.91 (2.96)
n = 56
3.87 (2.75)
n = 56
4.28 (2.72)
n = 54
4.94 (3.78)
n = 33
4.53 (3.37)
n = 32
4.88 (3.40)
n = 32
4.91 (3.08)
n = 29
4.79 (3.21)
n = 29
3.51 (3.03)
n = 43
3.46 (2.60)
n = 41
3.53 (2.07)
n = 41
3.26 (2.53)
n = 39
3.39 (2.49)
n = 38
7.63 (4.15)
n = 67
7.14 (4.96)
n = 64
7.66 (4.00)
n = 61
6.76 (4.09)
n = 59
6.81 (4.30)
n = 57
NA
NA
NA
9.62 (3.92)
n =63
7.81 (3.86)
n = 59
8.33 (4.11)
n = 57
9.22 (4.49)
n = 56
9.03 (4.09)
n = 56
9.02 (3.59)
n = 54
11.00 (5.38)
n = 33
11.06 (3.83)
n = 32
11.34 (3.95)
n = 32
10.15 (3.78)
n = 29
10.65 (4.43)
n = 29
8.44 (3.70)
n = 43
8.20 (4.09)
n = 41
7.76 (4.13)
n = 41
8.00 (3.20)
n = 39
7.60 (3.62)
n = 38
9.61 (3.94)
n = 67
10.11 (3.96)
n = 64
9.72 (4.63)
n = 61
9.22 (4.49)
n = 59
9.03 (4.09)
n = 57
NA
NA
NA
4.48 (2.48)
n =63
3.69 (2.55)
n = 59
3.98 (2.41)
n = 57
3.96 (2.26)
n = 56
3.59 (2.25)
n = 56
3.68 (2.52)
n = 54
5.39 (2.54)
n = 33
5.09 (2.61)
n = 32
4.97 (2.47)
n = 32
4.25 (2.78)
n = 28
4.66 (2.54)
n = 29
4.00 (2.43)
n = 43
3.24 (2.34)
n = 41
3.22 (2.31)
n = 41
2.77 (1.56)
n = 39
2.66 (2.29)
n = 38
5.19 (2.39)
n = 67
5.33 (2.53)
n = 63
5.44 (2.61)
n = 61
5.58 (2.44)
n = 59
5.21 (2.24)
n = 56
NA
NA
NA
2.65 (1.70)
n =63
2.98 (1.97)
n = 59
2.98 (1.81)
n = 57
2.96 (1.74)
n = 56
3.00 (1.80)
n = 55
2.75 (1.69)
n = 52
3.61 (2.07)
n = 33
3.16 (1.87)
n = 32
3.19 (1.89)
n = 32
3.28 (1.74)
n = 28
3.21 (1.76)
n = 29
2.40 (1.50)
n = 43
2.20 (1.50)
n = 41
2.32 (1.56)
n = 41
2.59 (1.44)
n = 39
2.26 (1.60)
n = 38
3.81 (1.96)
n = 67
4.11 (1.72)
n = 64
4.44 (2.28)
n = 61
3.95 (2.05)
n = 57
3.60 (1.94)
n = 57
NA
NA
NA
All
1.92 (2.98)
n = 102
5.51 (4.18)
n = 270
5.01 (4.00)
n = 258
5.31 (3.74)
n = 250
4.94 (3.63)
n = 241
5.00 (3.58)
n = 236
4.84 (3.11)
n = 106
9.82 (4.16)
n = 270
9.40 (4.22)
n = 258
9.52 (4.50)
n = 249
9.55 (4.10)
n = 241
9.32 (4.10)
n = 236
9.20 (3.93)
n = 106
4.80 (2.48)
n = 270
4.48 (2.56)
n = 256
4.42 (2.55)
n = 249
4.38 (2.48)
n = 240
4.08 (2.43)
n = 234
3.86 (2.27)
n = 106
3.26 (1.98)
n = 270
3.51 (2.04)
n = 258
3.52 (2.16)
n = 248
3.53 (2.06)
n = 238
3.31 (1.95)
n = 234
3.38 (2.04)
n = 103
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Panic Disorder
CSP MI
9.09 (10.18)
Time 1
n = 64
7.87 (8.62)
Time 2
n = 62
8.76 (9.17)
Time 3
n = 60
7.32 (7.72)
Time 4
n = 60
6.50 (8.80)
Time 5
n = 56
5.34 (7.25)
Time 6
n = 51
Physical Hygiene
5.39 (5.01)
Time 1
n = 64
5.80 (4.75)
Time 2
n = 61
5.67 (4.80)
Time 3
n = 58
5.48 (5.43)
Time 4
n = 58
5.73 (5.01)
Time 5
n = 56
4.90 (4.75)
Time 6
n = 51
Physical Well‐being and Exercise
15.89 (7.76)
Time 1
n = 64
17.10 (7.70)
Time 2
n = 62
17.14 (7.05)
Time 3
n = 58
16.13 (7.47)
Time 4
n = 58
15.39 (7.49)
Time 5
n = 56
13.95 (7.09)
Time 6
n = 51
CSP NMI
GP MI
GP NMI
3.89 (5.49)
n =63
3.71 (7.72)
n = 59
3.46 (4.20)
n = 57
3.79 (5.44)
n = 56
3.45 (4.34)
n = 56
3.60 (5.78)
n = 54
5.03 (4.88)
n = 33
5.43 (5.12)
n = 32
6.11 (5.02)
n = 32
4.99 (5.06)
n = 29
4.08 (5.13)
n = 29
2.77 (3.77)
n = 43
2.37 (3.45)
n = 41
3.00 (4.68)
n = 41
2.22 (3.28)
n = 39
1.82 (3.49)
n = 38
SCCF
10.98 (7.72)
n = 67
9.61 (8.70)
n = 64
9.55 (8.68)
n = 61
9.94 (8.68)
n = 59
8.07 (7.45)
n = 57
NA
NA
NA
4.00 (4.03)
n =63
4.06 (3.87)
n = 59
3.44 (3.59)
n = 57
3.25 (3.92)
n = 56
3.12 (3.57)
n = 56
3.18 (3.68)
n = 54
4.64 (2.69)
n = 33
3.66 (3.26)
n = 32
4.19 (4.10)
n = 32
3.21 (3.92)
n = 29
3.62 (3.70)
n = 29
2.07 (2.54)
n = 43
1.74 (2.69)
n = 41
2.17 (3.38)
n = 41
1.92 (2.67)
n = 39
1.08 (1.99)
n = 38
8.26 (4.64)
n = 67
7.59 (5.60)
n = 64
6.19 (4.55)
n = 61
5.61 (4.99)
n = 59
5.22 (5.07)
n = 57
NA
NA
NA
10.43 (5.55)
n =63
9.85 (6.23)
n = 59
10.30 (6.18)
n = 57
10.28 (6.14)
n = 56
9.54 (6.25)
n = 56
9.26 (6.72)
n = 54
15.79 (6.28)
n = 33
13.68 (7.37)
n = 32
13.84 (6.47)
n = 32
13.07 (5.92)
n = 29
13.58 (6.78)
n = 29
9.21 (5.61)
n = 43
7.26 (5.53)
n = 41
8.12 (5.02)
n = 41
7.44 (4.36)
n = 39
7.08 (4.19)
n = 38
18.93 (6.44)
n = 67
18.90 (6.01)
n = 64
18.98 (6.78)
n = 61
18.46 (6.67)
n = 59
17.60 (7.01)
n = 57
NA
NA
NA
All
6.84 (7.84)
n = 270
6.18 (7.62)
n = 258
6.47 (7.65)
n = 251
6.04 (7.19)
n = 243
5.10 (6.78)
n = 236
4.41 (6.54)
n = 106
5.16 (4.67)
n = 270
4.93 (4.77)
n = 257
4.52 (4.40)
n = 249
4.14 (4.64)
n = 241
3.98 (4.46)
n = 236
4.12 (4.41)
n = 106
14.29 (7.40)
n = 270
13.90 (7.91)
n = 258
14.12 (7.60)
n = 249
13.56 (7.48)
n = 241
12.97 (7.57)
n = 236
12.97 (7.57)
n = 106
Internal consistency estimates are provided in Table B13 for all the PSI subscales. Cronbach’s alphas ranged
between ‐.02 and .90 for the three scales related to the study constructs. Panic Disorder had strong reliabili‐
ty estimates ranging between .88 and .90. Hypersensitivity to External Stimuli demonstrated poor internal
consistency with alpha estimates ranging between .27 and .41 (M = .34). The internal consistency was ade‐
quate for the Physical Well‐being and Exercise subscale with values ranging from .72 to .76 (M = .74). Test‐
retest correlation coefficients are provided in Table B14 and indicate stability between time periods across
all subscales with correlations ranging between .45 and .83 (M = .67).
APPENDICES
113
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B13. Internal Consistency Estimates at each Time Period for each PSI Subscale
Measure
# Items Items*
T1
T2
T3
Attitudes about Segregation
2
r14, 39
.78 .74 .76
Fear Level
4
3, 12, 21, r38
.60 .58 .48
Hypersensitivity to External Stimuli
5
1, 7, r31, r34, 37
.41 .43 .49
Mental Well‐being
2
26, r35
.31 .38 .36
Mutism
2
22, r32
.47 .37 .50
Panic Disorder
9
2, 6, 10, 13, 16, 17, 20, 25, 30 .89 .89 .89
Physical Hygiene
5
4, r9, 18, r23, r29
.62 .65 .61
Physical Well‐being and Exercise
8
r5, 8, r11, r15, 19, 24, 27, r28 .72 .76 .73
Safety
2
33, 36
.82 .84 .76
Total
39
all
.90 .90 .90
T4
.75
.50
.35
.33
.52
.88
.68
.75
.81
.90
T5
.71
.46
.35
.30
.37
.90
.66
.76
.84
.89
T6
.67
.37
.27
‐.02
.31
.90
.60
.74
.90
.88
*r before a number indicates that the item is reversed scored.
Table B14. Testretest correlations between consecutive time periods for
each PSI Subscale
Measure
T1–T2 T2–T3 T3–T4 T4–T5 T5–T6
Attitudes about Segregation
.66
.78
.78
.80
.79
Fear Level
.46
.64
.64
.57
.60
Hypersensitivity to External Stimuli
.45
.62
.58
.58
.53
Mental Well‐being
.57
.62
.62
.57
.51
Mutism
.50
.59
.65
.64
.67
Panic Disorder
.68
.71
.78
.75
.74
Physical Hygiene
.67
.68
.73
.65
.64
Physical Well‐being and Exercise
.73
.78
.83
.77
.77
Safety
.72
.71
.74
.82
.69
For the three subscales that related to study constructs, correlations with other measures were calculated
as assessments of convergent validity. The PSI subscales demonstrated adequate validity estimates with
other self‐report measures of the constructs (i.e., anxiety, hypersensitivity, and somatization) with correla‐
tions ranging between .41 and .61 (M = .50) and had lower correlations with staff reports (range = .18 to .39,
M = .29).
Profile of Mood States (POMS)
Developed by McNair, Lorr, and Droppleman (1971, 1992), the POMS is intended to assess respondents
across six mood factors: Tension‐Anxiety (heightened musculoskeletal tension), Anger‐Hostility (anger and
antipathy towards others), Fatigue‐Inertia (weariness, inertia, low energy), Depression‐Dejection (depres‐
sion and sense of inadequacy), Vigor‐Activity (vigorousness, ebullience, high energy), and Confusion‐
Bewilderment (bewilderment and muddle‐headedness). The POMS is a 65‐item self‐report measure; higher
scores on the POMS indicate more negative feelings held over the past week (McNair & Heuchert, 2006).
Respondents rate each item on how well it describes them in the past week, using a 5‐point rating scale (0 –
not at all to 4 – extremely). Completion of the POMS takes approximately 15 to 20 minutes and requires an
8th grade reading level (Lorr, McNair, Heuchert, & Droppleman, 2003; McNair et al., 1992).
Acceptable levels of internal consistency (as measured by Kuder‐Richardson Formula 20) for the subscales
were found in a sample of 350 male psychiatric outpatients, ranging from .86 to .95 (McNair & Heuchert,
2006; Norcross, Guadagnoli, & Prochaska, 1984). Test‐retest reliability was assessed in psychiatric outpa‐
tients over the course of 3 to 110 days, with a median number of 20 days between tests. Stability coeffi‐
cients were found to range between .65 and .74, with a median of .69 (McNair & Heuchert, 2006).
114
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Normative data are available for a variety of populations but not for a prison population. Means for outpa‐
tients ranged from 10.0 (SD = 6.5) to 26.0 (SD = 15.8; McNair & Heuchert, 2006; Norcross et al., 1984). In a
nonclinical, community sample comprised of males only, mean scores across subscales ranged from 5.6 (SD
= 4.1) to 19.8 (SD = 6.8; McNair & Heuchert, 2006; Nyenhuis, Yamamoto, Luchetta, Terrien, & Parmentier,
1999). In a nonclinical, male college student sample, mean scores across subscales ranged from 8.6 (SD =
4.6) to 15.6 (SD = 6.0; McNair & Heuchert, 2006; Nyenhuis et al., 1999). Convergent validity has also been
assessed for the POMS and found to be acceptable for the total and subscale scores (Nyenhuis et al., 1999).
Table B15 provides the summary statistics for the study groups on the POMS subscales at each time period.
Estimates of internal consistency reliability were strong with Cronbach’s alphas ranging between .89 and .96
(M = .93). Correlations between sequential time periods indicated stability over time with coefficients rang‐
ing between .54 and .80 (M = .68). Convergent validity estimates with other self‐report measures of the
same construct ranged between .35 and .81 (M = .57), and with staff reports the coefficients ranged be‐
tween .14 and .38 (M = .25).
Table B15. Summary Statistics (M, SD, n) on POMS Subscales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Anger‐Hostility
20.92 (11.86) 14.78 (9.21) 19.28 (11.10) 10.37 (8.80) 18.03 (12.51)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 66
17.46 (11.29) 11.07 (9.36) 17.74 (10.34)
7.54 (6.76)
17.80 (12.07)
Time 2
n = 61
n = 58
n = 32
n = 41
n = 64
16.88 (12.16) 12.23 (10.88) 18.61 (10.48)
7.67 (8.05)
17.74 (10.69)
Time 3
n = 59
n = 57
n = 31
n = 41
n = 61
17.22 (13.07) 12.18 (10.75) 16.87 (11.56)
7.54 (7.46)
18.64 (12.34)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
16.02 (12.01) 12.02 (10.64) 15.17 (9.80)
6.71 (7.61)
17.63 (11.84)
Time 5
n = 56
n = 55
n = 29
n = 38
n = 57
14.03 (11.11) 11.65 (10.62)
Time 6
NA
NA
NA
n = 51
n = 54
Depression‐Dejection
25.13 (15.64) 17.17 (13.77) 25.34 (16.57) 12.27 (11.39) 28.92 (14.33)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
22.25 (14.17) 13.55 (12.14) 21.56 (13.46)
8.28 (8.66)
26.33 (15.39)
Time 2
n = 61
n = 58
n = 32
n = 41
n = 64
20.69 (14.91) 13.27 (13.42) 23.18 (15.04) 8.56 (12.16) 25.17 (15.23)
Time 3
n = 59
n = 57
n = 31
n = 41
n = 61
21.18 (14.85) 13.90 (12.77) 20.38 (15.02)
7.36 (7.20)
8.56 (12.16)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
19.35 (14.48) 12.28 (11.24) 19.03 (13.43)
6.90 (7.73)
7.36 (7.20)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
18.49 (13.01) 11.86 (11.94)
Time 6
NA
NA
NA
n = 51
n = 54
Fatigue‐Inertia
10.10 (6.92)
5.99 (6.18)
10.17 (7.26)
4.56 (4.78)
11.26 (6.76)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 66
9.36 (6.65)
4.60 (4.36)
9.44 (5.52)
3.39 (4.15)
10.19 (6.28)
Time 2
n = 61
n = 58
n = 32
n = 41
n = 64
8.84 (7.23)
4.82 (5.32)
10.45 (7.28)
3.63 (4.86)
9.94 (7.02)
Time 3
n = 59
n = 57
n = 31
n = 41
n = 61
9.07 (6.99)
5.20 (5.06)
8.21 (6.98)
3.00 (3.49)
10.55 (7.26)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
APPENDICES
All
16.88 (11.41)
n = 269
14.54 (11.06)
n = 256
14.72 (11.30)
n = 249
14.81 (12.01)
n = 243
13.86 (11.32)
n = 235
12.80 (10.88)
n = 105
22.19 (15.51)
n = 270
18.98 (14.65)
n = 256
18.40 (15.52)
n = 249
18.19 (14.80)
n = 243
16.97 (14.21)
n = 236
15.08 (12.85)
n = 105
8.54 (6.92)
n = 269
7.54 (6.21)
n = 256
7.54 (6.92)
n = 249
7.46 (6.73)
n = 243
115
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Time 5
Time 6
Tension‐Anxiety
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
CSP MI
7.59 (6.84)
n = 56
6.08 (5.14)
n = 51
15.83 (8.31)
n = 64
14.78 (8.28)
n = 62
14.26 (8.84)
n = 60
13.04 (8.08)
n = 60
12.46 (8.12)
n = 56
11.47 (7.73)
n = 51
CSP NMI
5.18 (5.41)
n = 56
4.42 (4.85)
n = 54
GP MI
7.94 (6.87)
n = 29
GP NMI
3.05 (4.39)
n = 38
SCCF
11.06 (6.98)
n = 57
NA
NA
NA
10.36 (7.26)
n = 63
8.06 (6.01)
n = 59
8.40 (7.06)
n = 57
8.37 (7.13)
n = 56
7.73 (6.64)
n = 56
7.85 (6.34)
n = 54
17.09 (8.71)
n = 33
13.97 (7.16)
n = 32
14.48 (7.63)
n = 31
13.09 (9.11)
n = 29
12.45 (7.25)
n = 29
8.60 (6.64)
n = 43
6.48 (4.72)
n = 41
6.66 (6.02)
n = 41
6.51 (4.99)
n = 39
6.10 (5.10)
n = 38
17.21 (8.49)
n = 67
15.74 (8.46)
n = 64
15.47 (8.29)
n = 61
16.16 (8.82)
n = 59
6.51 (4.99)
n = 57
NA
NA
NA
All
7.17 (6.75)
n = 236
5.23 (5.04)
n = 105
13.90 (8.61)
n = 270
12.06 (8.12)
n = 258
12.00 (8.47)
n = 250
11.68 (8.49)
n = 243
10.91 (7.74)
n = 236
9.56 (7.24)
n = 106
Saint Louis University Mental Status (SLUMS) Examination
The SLUMS Examination (Tariq, Tumosa, Chibnall, Perry, & Morley, 2006) is an 11‐item screening tool de‐
signed to assess mild neurocognitive impairment and dementia. It assesses orientation, memory, attention,
and executive functions. Scores on the SLUMS Examination can range from 0 to 30, with higher scores indi‐
cating better cognitive functioning. Cut‐offs for mild neurocognitive impairment and dementia are provided
for persons with varying degrees of education (i.e., more than high school, less than high school; Tariq et al.,
2006). Administration takes approximately 7 minutes. While the SLUMS Examination is similar to the Mini
Mental Status Exam (i.e., both measures screen for cognitive impairment), the SLUMS Examination may be
better for assessing milder cognitive problems, because it is a more sensitive measure (Tariq et al., 2006).
Due to its more sensitive nature and its associated ability to detect very mild forms of neurocognitive prob‐
lems, the SLUMS Examination was selected for this study.
Summary statistics are available for several different populations, including nonclinical populations and old‐
er individuals. The mean for the total scale was found to be between 25.7 (SD = 2.8) and 26.9 (SD = 2.00;
Tariq et al., 2006) for a nonclinical adult population, while means ranged from 26.9 (SD = 2.5) to 28.1 (SD =
2.3; Heeter, Winn, Winn, & Bozoki, 2008) for older adults between 60 and 80 years of age.
Table B16 provides the summary statistics for the study groups on the SLUMS test at each time period. The
SLUMS’ internal consistency estimates for the present study were low (range = .48 to .60, M = .52) which
may be reasonable given that this is a screening measure and assesses several cognitive functions. Test‐
retest reliability estimates were stronger with correlations ranging between .63 and .78 (M = .71). Conver‐
gent validity was estimated by assessing the relationship of the SLUMS to the Trails B/A task and correction‐
al staff’s ratings on the PBRS Dull‐Confused subscale. Convergent validity coefficients were small with corre‐
lations to the Trails task ranging from .13 to .31 (M = .21) and to the PBRS Dull‐Confused subscale ranging
between .03 and .18 (M = .10).
116
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B16. Summary Statistics (M, SD, n) on SLUMS Score by Group and Time
Time
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
All
20.80 (5.43) 21.73 (3.34) 21.52 (4.04) 23.38 (3.73) 20.54 (3.73) 21.45 (4.24)
1
n = 64
n = 63
n = 33
n = 43
n = 67
n = 270
21.16 (4.77) 22.64 (3.63) 23.09 (3.68) 24.12 (3.23) 21.49 (4.33) 22.30 (4.16)
2
n = 62
n = 59
n = 32
n = 41
n = 63
n = 257
22.26 (4.59) 24.02 (3.25) 23.88 (2.88) 24.49 (3.49) 22.85 (4.37) 23.37 (3.95)
3
n = 60
n = 57
n = 32
n = 41
n = 61
n = 251
22.92 (4.31) 24.38 (3.03) 23.34 (3.67) 24.79 (3.68) 23.38 (3.91) 23.84 (3.80)
4
n = 60
n = 56
n = 29
n = 39
n = 59
n = 243
23.59 (4.04) 24.25 (3.34) 24.93 (3.24) 24.82 (3.24) 23.26 (4.05) 24.03 (3.69)
5
n = 56
n = 56
n = 29
n = 38
n = 57
n = 236
23.94 (4.57) 25.30 (2.88)
24.62 (3.82)
6
NA
NA
NA
n = 49
n = 54
n = 104
State‐Trait Anxiety Inventory (STAI)
The STAI (Spielberger, Gorsuch, & Lushene, 1970) partitions anxiety into that which is attributable to the
condition one is in (i.e., state) and into the inherent anxiety of an individual (i.e., trait). It is a 40‐item self‐
report inventory that includes two 20‐item subscales. The first subscale assesses state anxiety and is ans‐
wered on a 4‐point scale (1—not at all, 2—somewhat, 3—moderately so, 4—very much so); the second
subscale assesses trait anxiety and is also answered on a 4‐point scale (1—almost never, 2—sometimes, 3—
often, 4—almost always).
Internal consistency is acceptable for the STAI with coefficients between .81 and .92 and a median .84
(Metzger, 1976). In a variety of nonclinical samples (i.e., college students, high school students, military re‐
cruits, working adults), the median alpha coefficient was .60 (Novy, Nelson, Goodwin, & Rowzee, 1993).
Across males of three different ethnicities (i.e., White, Black, Latino), alpha coefficients were found to be
between .93 and .95 for state anxiety and between .92 and .95 for trait anxiety (Novy et al., 1993). Internal
consistency measures are high in prison populations (.83; Zinger et al., 2001). Overall, this inventory is valu‐
able in its ability to distinguish between types of anxiety and because normative data exist for a prisoner
population (Spielberger et al., 1970).
Test‐retest reliability has been variable for the two subscales on the STAI. In a replication study by Joisting
(1976), the STAI was administered both before and after a class examination. Correlations between the two
tests were .66 for trait anxiety and .60 for state anxiety (Joesting, 1976). For a 104‐day test‐retest assess‐
ment, test‐retest reliability ranged from .73 to .84 (Spielberger et al., 1970). Furthermore, test‐retest relia‐
bility in another nonclinical sample was found to be .16, .26, and .15 for state anxiety assessed for different
intervals (3 months, 8 months, 11 months; Nixon & Steffeck, 1977). Test‐retest reliability was also assessed
for trait anxiety in the same nonclinical sample and was found to be .48, .54, and .29 for trait anxiety as‐
sessed for the same three intervals (3 months, 8 months, 11 months; Nixon & Steffeck, 1977). In another
study with college students, test‐retest reliability was found to be .97 for trait anxiety and .45 for state an‐
xiety (Metzger, 1976).
Means for state anxiety in nonclinical populations seem to range from 32.90 (SD = 11.10) to 49.20 (SD =
11.89), while means for trait anxiety in nonclinical populations ranged from 35.60 (SD = 9.90) to 45.89 (SD =
12.96; Joesting, 1976; Nixon & Steffeck, 1977; Novy et al., 1993; Nyenhuis et al., 1999).
APPENDICES
117
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
The STAI has shown to be a valid measure, demonstrating convergent validity from .52 to .85 and good dis‐
criminant validity (Spielberger, 1983). Novy et al. (1993) found moderate to high correlations between the
BDI and the STAI State/Trait scale (range = .59 to .81 for BDI and STAI State; range = .44 to .71 for BDI and
STAI Trait) and between the BHS and the STAI State/Trait (range = .67 to .92 for BHS and STAI State; range =
.26 to .76 for BHS and STAI Trait).
Table B17 provides the summary statistics for the study groups on the STAI scales at each time period. Cron‐
bach’s alpha coefficients ranged between .93 and .95 (M = .94) for the two subscales, which indicates strong
internal consistency estimates. Correlations between sequential time periods (range = .65 to .82, M = .73)
suggest good stability over 3 month intervals with trait anxiety showing slightly stronger correlations (M =
.79) than state anxiety (M = .68). Convergent validity with other self‐report measures of anxiety indicated
good validity with coefficients ranging between .37 and .85 (M = .64); however, correlations with staff re‐
ports of anxiety were lower (ranging from .07 to .49 with a mean of .23).
Table B17. Summary Statistics (M, SD, n) on STAI Subscales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
State Anxiety
46.90 (12.16) 42.05 (11.42) 47.64 (13.52) 39.39 (12.04) 50.14 (12.72)
Time 1
n = 62
n = 63
n = 33
n = 43
n = 67
45.83 (12.43) 38.43 (10.39) 44.53 (13.14) 36.68 (10.98) 48.42 (13.39)
Time 2
n = 62
n = 59
n = 32
n = 41
n = 64
45.45 (12.40) 37.89 (12.05) 47.29 (11.68) 34.65 (7.35) 48.49 (12.78)
Time 3
n = 60
n = 57
n = 32
n = 41
n = 61
44.01 (13.38) 37.50 (10.89) 45.08 (11.90) 33.80 (8.69) 49.41 (12.48)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
42.60 (12.98) 37.46 (11.54) 43.76 (12.07) 33.81 (9.40) 48.29 (11.70)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
43.28 (12.23) 36.89 (9.98)
Time 6
NA
NA
NA
n = 51
n = 54
Trait Anxiety
48.41 (12.36) 42.78 (11.11) 49.70 (12.79) 37.82 (9.99) 54.45 (11.48)
Time 1
n = 62
n = 63
n = 33
n = 43
n = 67
47.89 (11.91) 38.77 (10.25) 46.59 (11.84) 35.93 (10.76) 52.64 (12.28)
Time 2
n = 62
n = 59
n = 32
n = 41
n = 64
47.49 (12.33) 38.40 (10.82) 47.59 (10.05) 34.44 (10.15) 51.50 (12.94)
Time 3
n = 60
n = 57
n = 32
n = 41
n = 61
45.75 (13.11) 39.27 (10.10) 45.78 (10.69) 34.06 (8.86) 52.77 (11.10)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
44.65 (12.85) 38.70 (11.02) 44.06 (10.01) 32.74 (8.98) 52.28 (11.65)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
43.95 (12.10) 37.54 (10.62)
Time 6
NA
NA
NA
n = 51
n = 53
All
45.46 (12.80)
n = 268
43.16 (12.86)
n = 258
42.83 (12.78)
n = 251
42.31 (12.93)
n = 243
41.48 (12.65)
n = 236
40.09 (11.52)
n = 106
47.06 (12.79)
n = 269
44.92 (12.97)
n = 258
44.27 (13.10)
n = 250
44.09 (12.70)
n = 243
43.09 (12.93)
n = 236
40.70 (11.71)
n = 105
Structured Inventory of Malingered Symptomatology (SIMS)
The SIMS (Widows & Smith, 2005) is a 75‐item screening measure intended to detect feigned symptoms of
psychopathology and cognitive functioning in clinical and forensic settings. A total score and scores on five
subscales—Psychosis (bizarre or unusual psychotic symptoms), Neurologic Impairment (illogical or highly
atypical neurological symptoms), Amnestic Disorders (symptoms of memory impairment), Low Intelligence
(general cognitive incapacity or intellectual deficit), and affective disorders (atypical symptoms of depression
and anxiety)—are obtained (Widows & Smith, 2005). The subscales are comprised of 15 items each; comple‐
118
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
tion of this measure takes approximately 10 to 15 minutes (Widows & Smith, 2005). Participants answer
whether statements are applicable to them or are generally considered true (T); if a statement does not de‐
scribe them or cannot be considered true, an F is circled as the answer choice. The SIMS assesses whether
respondents endorse atypical, improbable, inconsistent, or illogical symptoms. Scores above the cutoff mark
suggest probable malingering but may also suggest genuine psychopathology. For this study, we used eleva‐
tions above these cutoff scores as an indicator of possible malingering.
Internal consistency estimates have ranged from .24 to .86 for subscales and were found to be .72 to .88 for
total scores (Merckelbach & Smith, 2003; Smith, as cited in Widows & Smith, 2005). Three‐week test‐retest
reliability in honest responders was found to be .72 in a Dutch sample.
Also reported in the manual are validity studies that indicated the SIMS to be a valid screening device for
malingering. The SIMS total score correlated strongly with validity scales of the MMPI, including the F scale
(“faking bad”; r = .84) and F‐K index scores (“honesty”; r = .81; Widows & Smith, 2005). A moderate correla‐
tion (r = .45) was found between the SIMS total score and the 16 Personality Factor Questionnaire Faking
Bad scale (Widows & Smith, 2005). Furthermore, the SIMS total score was highly correlated with other
commonly used indexes of malingering, such as the MMPI‐2 validity scales (range = .44 to .51), the Struc‐
tured Interview of Report Symptoms (SIRS) scales (.43 < r < .80), and the M Test (.46 < r < .67; Heinze & Pu‐
risch, 2001). A study by Edens, Poythress, and Watkins‐Clay (2009) indicated that the SIMS correlated highly
with the SIRS (.81) and Personality Assessment Inventory NIM (.84) and that it correlated moderately with
the Personality Assessment Inventory MAL (.68) and RDF (.45) scales. Furthermore, the SIMS total scores
have been found to correlate significantly (p < .01) with the BDI (.64) as well as the STAI Trait (.55). Thus, the
SIMS scales seem to be related to both validity and psychopathology measures.
Mean scores on the SIMS were given in Lewis, Simcox, and Berry’s (2002) study. Mean scores across subs‐
cales for a forensic sample ranged from 1.2 (SD = 2.1) to 5.2 (SD = 2.6) and the mean total score for the sam‐
ple was found to be 14.5 (SD = 8.8; Lewis et al., 2002). Edens, Poythress, and Watkins‐Clay (2009) also found
that the SIMS nearly always correctly classified non‐malingering inmates but that there were more errors
with mentally ill, such that caution against classification of inmates with mental illness as malingerers is war‐
ranted. While it is suggested to administer follow‐up tests once an elevated score has been found on the
SIMS, this measure by itself is yet another way to gain a more comprehensive picture of the inmates in this
study, be it in regards to the degree of their malingering or their psychopathology.
Table B18 provides the summary statistics for the study groups on the SIMS scales at each time period. In‐
ternal consistency estimates ranged between .50 and .93 (M = .76) with the lowest alphas for the Affective
Disorder subscale (M = .55) and the Low Intelligence subscale (M = .59). Table B19 provides the internal con‐
sistency estimates for the SIMS scales at each time period. Test‐retest coefficients were quite variable rang‐
ing between .06 and .83 (M = .48) with total scores showing the least variability (range = .54 to .79, M = .68).
There were not correlations with other malingering variables to assess validity; however, correlations be‐
tween subscales were computed for each time period. Correlations between subscales ranged between .34
and .98 with variability in which measures demonstrated the weakest and strongest correlations at each
time period.
APPENDICES
119
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B18. Summary Statistics (M, SD, n) on POMS Subscales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Affective Disorders
5.97 (2.53)
4.53 (2.13)
6.12 (2.60)
3.52 (1.89)
6.88 (2.25)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
6.07 (2.31)
5.94 (12.44)
8.90 (16.38)
3.17 (1.85)
8.31 (11.74)
Time 2
n = 62
n = 60
n = 33
n = 41
n = 65
6.59 (2.52)
6.36 (12.59)
6.31 (2.26)
3.60 (2.07)
8.05 (12.09)
Time 3
n = 60
n = 58
n = 32
n = 41
n = 62
6.09 (2.57)
4.94 (2.42)
6.27 (2.78)
3.51 (1.83)
6.19 (2.60)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
5.62 (2.42)
4.87 (2.38)
5.95 (2.29)
3.34 (1.82)
6.33 (2.47)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
6.05 (2.31)
4.96 (2.57)
Time 6
NA
NA
NA
n = 51
n = 54
Amnestic Disorders
3.16 (3.78)
1.27 (1.53)
2.88 (2.94)
0.70 (1.12)
4.30 (2.60)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
3.51 (3.67)
2.97 (12.74)
5.42 (17.02)
1.01 (2.50)
5.98 (12.24)
Time 2
n = 62
n = 60
n = 33
n = 41
n = 65
3.32 (3.63)
2.76 (12.95)
2.60 (2.74)
0.49 (.90)
5.66 (12.59)
Time 3
n = 60
n = 58
n = 32
n = 41
n = 62
2.92 (3.52)
1.70 (2.54)
2.66 (3.38)
0.80 (2.31)
3.95 (3.73)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
2.63 (3.29)
1.27 (2.37)
2.62 (3.45)
0.66 (.85)
3.46 (3.53)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
2.48 (3.07)
1.37 (2.63)
Time 6
NA
NA
NA
n = 51
n = 54
Low Intelligence
4.04 (12.24)
2.33 (1.59)
2.28 (1.44)
1.77 (1.56)
2.97 (2.17)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
2.53 (2.19)
4.13 (12.56)
5.45 (16.90)
1.60 (1.94)
4.14 (12.10)
Time 2
n = 62
n = 60
n = 33
n = 41
n = 65
3.09 (2.53)
4.09 (12.80)
2.06 (1.72)
1.34 (1.51)
4.27 (12.43)
Time 3
n = 60
n = 58
n = 32
n = 41
n = 62
2.55 (2.33)
2.23 (2.03)
2.52 (2.01)
1.59 (1.98)
2.56 (1.95)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
2.53 (2.34)
2.58 (1.94)
2.18 (1.97)
1.50 (1.50)
2.61 (2.17)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
2.85 (2.44)
2.37 (2.10)
Time 6
NA
NA
NA
n = 51
n = 54
Neurological Impairment
4.85 (12.36)
2.24 (2.09)
2.88 (2.42)
1.44 (1.45)
4.24 (2.92)
Time 1
n = 64
n = 63
n = 33
n = 43
n = 67
3.04 (3.23)
3.70 (12.66)
5.58 (16.89)
1.54 (2.18)
5.74 (12.33)
Time 2
n = 62
n = 60
n = 33
n = 41
n = 65
3.27 (3.08)
3.80 (12.88)
3.13 (2.79)
1.30 (1.77)
5.37 (12.54)
Time 3
n = 60
n = 58
n = 32
n = 41
n = 62
2.97 (3.10)
2.54 (2.79)
2.59 (2.10)
1.31 (1.56)
3.90 (3.52)
Time 4
n = 60
n = 56
n = 29
n = 39
n = 59
2.68 (2.98)
1.93 (2.21)
3.14 (2.86)
1.11 (1.41)
3.83 (3.72)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
2.87 (2.83)
2.15 (2.82)
Time 6
NA
NA
NA
n = 51
n = 54
120
All
5.49 (2.56)
n = 270
6.50 (10.36)
n = 261
6.37 (8.72)
n = 253
5.46 (2.64)
n = 243
5.29 (2.51)
n = 236
5.49 (2.48)
n = 106
2.58 (2.93)
n = 270
3.85 (10.79)
n = 261
3.21 (9.12)
n = 253
2.51 (3.34)
n = 243
2.19 (3.07)
n = 236
1.89 (2.89)
n = 106
2.80 (6.17)
n = 270
3.52 (10.50)
n = 261
3.19 (8.82)
n = 253
2.32 (2.09)
n = 243
2.35 (2.06)
n = 236
2.62 (2.27)
n = 106
3.31 (6.44)
n = 270
3.95 (10.70)
n = 261
3.57 (9.00)
n = 253
2.78 (2.94)
n = 243
2.58 (2.95)
n = 236
2.49 (2.82)
n = 106
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Psychosis
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
Total
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
CSP MI
2.87 (3.52)
n = 64
2.67 (3.06)
n = 62
2.68 (3.06)
n = 60
2.55 (3.10)
n = 60
2.07 (2.84)
n = 56
2.26 (3.11)
n = 51
18.79 (15.41)
n = 64
17.82 (11.97)
n = 62
18.94 (12.00)
n = 60
17.07 (12.16)
n = 60
15.53 (10.87)
n = 56
16.53 (10.90)
n = 51
CSP NMI
GP MI
GP NMI
1.10 (1.66)
n = 63
2.42 (12.73)
n = 60
2.66 (12.94)
n = 58
1.18 (2.12)
n = 56
0.86 (1.54)
n = 56
1.02 (2.34)
n = 54
1.97 (2.47)
n = 33
4.48 (17.06)
n = 33
1.38 (1.62)
n = 32
1.79 (2.37)
n = 29
1.76 (3.11)
n = 29
0.42 (.66)
n = 43
0.80 (2.09)
n = 41
0.69 (.94)
n = 41
0.51 (1.02)
n = 39
0.50 (.86)
n = 38
SCCF
4.55 (3.59)
n = 67
5.43 (12.35)
n = 65
5.37 (12.67)
n = 62
3.73 (3.92)
n = 59
3.44 (3.76)
n = 57
NA
NA
NA
11.46 (5.85)
n = 63
12.55 (12.69)
n = 60
12.83 (12.82)
n = 58
12.59 (8.97)
n = 56
11.51 (7.43)
n = 56
11.88 (9.33)
n = 54
16.13 (8.36)
n = 33
17.83 (16.58)
n = 33
15.47 (7.88)
n = 32
15.81 (9.60)
n = 29
15.63 (10.48)
n = 29
7.85 (4.38)
n = 43
8.13 (9.43)
n = 41
7.41 (4.78)
n = 41
7.72 (5.17)
n = 39
7.11 (4.06)
n = 38
22.96 (9.56)
n = 67
23.49 (15.63)
n = 65
22.30 (16.32)
n = 62
20.33 (12.53)
n = 59
19.68 (13.31)
n = 57
NA
NA
NA
All
2.37 (3.12)
n = 270
3.24 (10.75)
n = 261
2.85 (9.06)
n = 253
2.11 (3.03)
n = 243
1.82 (2.87)
n = 236
1.63 (2.79)
n = 106
16.05 (11.25)
n = 270
16.50 (14.33)
n = 261
16.06 (13.15)
n = 253
15.18 (11.17)
n = 243
14.23 (10.78)
n = 236
14.13 (10.29)
n = 106
Table B19. Internal Consistency Estimates (Cronbach’s alpha) for SIMS Scales
at each Time Period
SIMS Scale
Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Affective disorder
.55
.60
.58
.56
.52
.50
Amnesia
.80
.86
.86
.88
.87
.87
Low Intelligence
.52
.52
.62
.63
.60
.65
Neurological Impairment
.75
.82
.79
.81
.83
.83
Psychosis
.85
.85
.84
.87
.86
.87
Total
.90
.92
.92
.93
.92
.91
Trail Making Test (TMT)
The TMT (Reitan, 1958) measures neurocognitive deficits related to attention, speed, and mental flexibility.
There are two tasks (A and B) and the length of time to complete each task was recorded as total score for
each task. Completion time on this measure varies widely but is generally around 5 to 10 minutes for both
tasks (Strauss, Sherman, & Spreen, n.d.). While individuals connect only numbers in ascending order on
Trails A, they have to connect numbers and letters alternately in ascending order for Trails B (Tombaugh,
2004). We computed two derived scores—ratio of times on the two tasks (B/A) and the difference between
times on the two tasks (B – A); these derived scores provide an indication of the time difference between
Trails A and Trails B (Tombaugh, 2004); however, for the analysis in the report the Trails ratio (B/A) was used
to assess change over time. The TMT has been shown to be sensitive to neurocognitive deficits (Sherrill‐
Pattison, Donders, & Thompson, 2000); it is important to consider age and education of participants when
interpreting scores though.
APPENDICES
121
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
The TMT has demonstrated adequate test‐retest reliability over a 14‐ to 24‐week period (Mdn = 20 weeks;
Trails A r = .46; Trails B r = .44; Matarazzo, Wiens, Matarazzo, & Goldstein, 1974) although stability may be
impacted by population groups and time intervals. Practice effects might be a problem (McCaffrey, Ortega,
& Haase, 1993) although research has shown that practice effects between administrations separated by at
least 3 months may be negligible (e.g., Basso, Bornstein, & Lang, 1999). Another study found that the TMT’s
3‐week test‐retest reliability was moderate to high, with Trails A having a correlation of .55 and Trails B hav‐
ing a correlation of .75 (Bornstein, Baker, & Douglass, 1987).
Normative data are available on nonclinical populations, separated by age group (Tombaugh, 2004). For
people aged 18 to 59, mean times on the Trails A ranged from 22.93 (SD = 6.87; 18‐24 years) to 35.10 (SD =
10.94; 55‐59 years); mean times on the Trails B ranged from 48.97 (SD = 12.69; 18‐24 years) to 78.84 (SD =
19.09; 55‐59 years; Tombaugh, 2004). Additionally, Matarazzo et al. (1974) found means for the Trails A and
B to be 21.76 (SD = 5.65) and 54.17 (SD = 12.54), respectively. Descriptive statistics were also provided on
the two scores that will be derived in the current study. While B‐A was found to have a mean of 39.7 (SD =
21.5), B/A was found to have a mean of 2.1 (SD = .6) for an older adult, community‐dwelling sample (Sán‐
chez‐Cubillo et al., 2009). Convergent validity is adequate for the total scores on each of the trials as well as
for the B‐A score (Sánchez‐Cubillo et al., 2009). The ratio score of B/A did not show significant correlations
with any of the other assessed cognitive measures in Sánchez‐Cubillo et al.’s (2009) study but Periáñez et al.
(2007) suggested that the B/A might be a purer measure of executive functioning.
After consultation with a neuropsychologist, it was decided to use a derived score by taking the ratio of time
to complete Task B with time to complete Task A; however, information about all Trails scores are provided
in this section. Table B20 provides the summary statistics for the study groups on the Trails tasks at each
time period. Correlations between sequential time periods for the entire sample are given in Table B21. Cor‐
relations between the Trails tasks at each time period ranged between .21 and .97 (absolute values of corre‐
lations are given; the Trails B/A was always negatively correlated with Trails A time). Table B21 also provides
the mean correlation of each task with the other tasks over time. The Trails tasks were correlated with per‐
formance on the SLUMS and the PBRS Dull‐Confused subscale. Correlation coefficients were small with the
SLUMS (range = .13 to .31, M = .21) and with the correctional officer ratings on the PBRS Dull‐Confused
(range = ‐.12 to .09, M = ‐.01), indicating that these measures are assessing distinct aspects of cognitive
functioning.
Table B20. Summary Statistics (M, SD, n) on TMT Scores by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Task A Time
29.36 (14.18) 27.20 (8.76) 25.65 (7.70)
24.28 (6.00) 32.36 (19.80)
Time 1
n = 62
n = 61
n = 33
n = 43
n = 67
29.70 (17.34) 24.46 (7.00) 24.09 (8.22)
22.25 (4.84) 29.46 (13.36)
Time 2
n = 61
n = 59
n = 32
n = 41
n = 64
29.45 (21.46) 23.06 (7.36) 24.52 (6.79)
21.70 (5.97) 28.13 (15.10)
Time 3
n = 60
n = 57
n = 32
n = 41
n = 61
29.78 (19.01) 22.63 (7.94) 21.99 (7.32)
21.04 (4.17) 26.64 (11.00)
Time 4
n = 59
n = 56
n = 29
n = 39
n = 59
27.54 (15.32) 23.48 (8.91) 21.34 (6.67)
20.97 (3.94) 27.57 (15.45)
Time 5
n = 56
n = 56
n = 29
n = 38
n = 57
27.27 (15.95) 20.91 (5.59)
Time 6
NA
NA
NA
n = 48
n = 54
122
All
28.34 (13.51)
n = 266
26.54 (12.10)
n = 256
25.78 (14.01)
n = 251
25.02 (12.27)
n = 242
24.78 (12.11)
n = 235
23.81 (12.02)
n = 103
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Task B Time
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
B – A Time
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
B/A Ratio
Time 1
Time 2
Time 3
Time 4
Time 5
Time 6
CSP MI
84.70 (55.39)
n = 61
77.74 (44.98)
n = 60
75.42 (45.66)
n = 60
71.63 (37.92)
n = 59
69.43 (32.73)
n = 56
62.75 (32.75)
n = 48
55.27 (47.23)
n = 61
48.82 (35.29)
n = 60
45.98 (31.68)
n = 60
41.84 (27.51)
n = 59
41.89 (23.20)
n = 56
34.48 (22.88)
n = 48
2.95 (1.11)
n = 61
2.84 (1.19)
n = 60
2.68 (.82)
n = 60
2.59 (1.00)
n = 59
2.64 (.74)
n = 56
2.40 (.74)
n = 48
CSP NMI
83.46 (42.82)
n = 62
71.09 (29.85)
n = 59
66.33 (26.20)
n = 57
62.78 (25.66)
n = 56
62.27 (30.01)
n = 56
55.84 (20.71)
n = 54
56.33 (40.60)
n = 60
46.63 (26.48)
n = 59
43.27 (24.97)
n = 57
40.15 (23.39)
n = 56
38.78 (25.33)
n = 56
34.93 (18.53)
n = 54
3.19 (1.54)
n = 60
2.94 (1.01)
n = 59
3.02 (1.24)
n = 57
2.92 (1.15)
n = 56
2.73 (1.00)
n = 56
2.73 (.86)
n = 54
GP MI
GP NMI
SCCF
82.46 (34.64)
n = 33
66.30 (27.77)
n = 31
66.11 (27.25)
n = 32
58.29 (22.17)
n = 29
54.98 (20.41)
n = 29
68.94 (23.83)
n = 43
70.63 (35.41)
n = 40
63.41 (24.79)
n = 41
56.78 (19.63)
n = 38
58.81 (19.96)
n = 38
96.29 (58.26)
n = 67
81.57 (34.51)
n = 63
78.26 (38.12)
n = 61
74.69 (37.81)
n = 59
72.71 (41.64)
n = 57
NA
NA
NA
56.81 (31.14)
n = 33
42.40 (21.97)
n = 31
41.59 (23.35)
n = 32
36.30 (19.15)
n = 29
33.64 (18.59)
n = 29
44.66 (21.75)
n = 43
48.41 (34.89)
n = 40
41.71 (22.39)
n = 41
35.81 (18.01)
n = 38
37.58 (18.81)
n = 37
63.93 (46.29)
n = 67
52.11 (27.20)
n = 63
50.13 (30.18)
n = 61
48.04 (30.67)
n = 59
45.14 (32.39)
n = 57
NA
NA
NA
3.30 (1.21)
n = 33
2.80 (.74)
n = 31
2.71 (.84)
n = 32
2.74 (.84)
n = 29
2.67 (.90)
n = 29
2.89 (.88)
n = 43
3.27 (1.84)
n = 40
2.97 (1.00)
n = 41
2.72 (.82)
n = 38
2.87 (1.05)
n = 37
3.04 (1.12)
n = 67
2.80 (.74)
n = 63
2.71 (.84)
n = 61
2.74 (.84)
n = 59
2.67 (.90)
n = 57
NA
NA
NA
All
84.50 (47.67)
n = 266
74.62 (35.86)
n = 253
70.90 (34.93)
n = 251
66.38 (31.78)
n = 241
65.04 (32.04)
n = 236
59.09 (27.13)
n = 102
56.17 (40.64)
n = 264
48.28 (29.82)
n = 253
45.12 (27.45)
n = 251
41.35 (25.45)
n = 241
40.24 (25.25)
n = 235
35.19 (20.59)
n = 102
3.06 (1.20)
n = 264
2.95 (1.20)
n = 253
2.87 (1.05)
n = 251
2.77 (.99)
n = 241
2.72 (.99)
n = 235
2.57 (.82)
n = 102
Table B21. TestRetest Correlation Coefficients for Trails Tasks at Consecutive Testing Intervals
Trails Task Time 1 to 2 Time 2 to 3 Time 3 to 4 Time 4 to 5 Time 5 to 6
r*
Trails A
.70
.65
.83
.81
.87
.41
Trails B
.66
.70
.73
.76
.75
.71
Trails B – A
.58
.63
.58
.63
.59
.69
Trails B/A
.36
.37
.44
.44
.39
.50
*means correlation of task with other Trails tasks over time.
Trauma Symptom Inventory (TSI)
The TSI (Briere, 1995) is a 100‐item self‐report assessment of posttraumatic stress and other psychological
consequences of traumatic events, including but not limited to rape, child abuse, spouse abuse, physical as‐
sault, combat, major accidents, and natural disasters. Respondents use a 4‐point rating scale (0—never to
APPENDICES
123
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
3—often) to report on the experience of 100 events that could have occurred within the last 6 months.
Scores are obtained on three validity scales (Atypical Response, Response Level, and Inconsistent Response)
and 10 clinical symptom domains (Anxious Arousal, Depression, Anger/Irritability, Intrusive Experiences, De‐
fensive Avoidance, Dissociation, Sexual Concerns, Dysfunctional Sexual Behavior, Impaired Self‐Reference,
and Tension Reduction Behavior). Greater scores indicate more symptoms associated with trauma (Fernan‐
dez, 1998). It has been found that a fifth‐ to seventh‐grade reading level is required to complete the TSI; it
takes approximately 20 minutes to complete this assessment (Fernandez, 1998).
The TSI’s clinical subscales have demonstrated internal reliability with different samples. Alpha coefficients
ranged between .74 and .91 for the standardization sample (nonclinical) with a median alpha coefficient of
.88; alphas ranged from .69 to .90 with a median alpha of .86 for the university sample; furthermore, inter‐
nal consistency reliability ranged from .74 to .90 with a median alpha coefficient of .89 for a clinical sample
(Briere, 1995). The three validity scales were also found to have internal consistency reliabilities between .51
and .80 across the standardization as well as a military sample (Briere, 1995). There are no known test‐retest
reliability estimates or studies completed with prison populations.
Normative data are available by gender and age groups for the general population as well as for a clinical
sample separated by gender (Briere, 1995). Means on the subscales ranged from 2.32 (SD = 4.20) to 7.69 (SD
= 6.03) for nonclinical, younger males (i.e., 18‐54; Briere, 1995). Clinical samples were separated by trauma
history; means for males in the group without trauma history ranged from 2.24 (SD = 3.17) to 9.45 (SD =
5.90), whereas means for males in the group with trauma history ranged from 6.30 (SD = 7.44) to 16.32 (SD
= 5.27; Briere, 1995).
Reasonable convergent validity was found between the TSI and other measures, such as the BSI, Symptom
Checklist, Impact of Event Scale, and the Personality Assessment Inventory (Briere, 1995). More specifically,
three of the TSI’s clinical subscales that are most closely associated with subscales of the BSI were found to
have high correlations: Anxious Arousal (TSI) and Anxiety (BSI) had a high correlation of .75, Anger/Irritability
(TSI) and Hostility (BSI) correlated at .77, and Defensive Avoidance (TSI) and Depression (BSI) had a correla‐
tion of .82 (Briere, 1995). The correlations of the TSI subscales and the subscales from two posttraumatic
stress scales (i.e., Impact of Event Scale, Symptom Checklist) were found to range between .35 and .74, indi‐
cating convergent validity (Briere, 1995). The Sexual Concern subscale of the TSI was moderately correlated
(r = .53) with a measure tapping into sexual concerns (Briere, 1995). Furthermore, the Dysfunctional Sexual
Behavior subscale was moderately correlated (r = .32) with a question on another measure on how many
sexual partners the person had over the course of the past 12 months as well as with four questions assess‐
ing the likelihood of the participant engaging in sex with an attractive stranger (r = .32), sex with any stran‐
ger (r = .19), posing for pornography (r = .22), and sex for money (r = .17; Briere, 1995). The Borderline Per‐
sonality Disorder subscale of the PAI was found to correlate moderately with the Impaired Self‐Reference
scale (r = .65) as well as the Tension Reduction Behavior scale (r = .54; Briere, 1995).
Table B22 provides the summary statistics for the study groups on the TSI Total Score. The internal consis‐
tency estimates were similar across groups and high (M = .97).
124
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B22. Summary Statistics (M, SD, n) on TSI Total Score by Group and Time
Statistic
CSP MI CSP NMI
GP MI
GP NMI
SCCF
All
M
98.43
53.66
87.86
40.64
119.55
83.12
SD
58.85
42.04
50.06
30.34
58.40
58.17
n
62
60
33
41
60
262
Cronbach’s α
.98
.97
.96
.96
.97
.98
NORMATIVE COMPARISONS
Because we used standardized assessments in this study, it is possible to compare scores for the study sam‐
ple to normative data. In this section, comparisons were made between each study groups’ mean and the
normative mean using a one sample t test. Normative values were taken from the test manuals when avail‐
able or were gathered from the literature. Normative data from general adult populations were typically
used; if male norms were available they were used. If only normative data for clinical samples were available
then outpatient norms were used. One sample t tests indicated that in general, for all groups except the GP
NMI group, scores were elevated above the normative data when participants entered the study and tended
to stay that way. There were also fewer elevations on the SIMS malingering subscales with study groups fre‐
quently scoring similarly to “honest responders.” Table B23 provides a visual representation of the signifi‐
cant differences by group at each time period on each measure. Red shading indicates that the group mean
is significantly different from the normative mean in the direction of more psychological or cognitive prob‐
lems, whereas green shading indicates that the group mean is significantly better than the normative mean.
No shading indicates the groups were statistically similar to the normative data.
APPENDICES
125
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
•
•
•• • •
EMMEN
•••
•••
=MN=
MIN=
=NM=
MIN= •
'MN
MN MEN
.MEN
•
MEN
Ir•
MEN
M
MN
M
MN •
NE _EN
MEN
MEN
MEM
MN=
ME
MENNEMEMENEMEN MENEM...EMMEN =MENNEN=
•• •••
••
MENNEMEMENEMEN MENEM...MN=
=MENEM=
MENNEMEMENEMEN MENNEMENNEN • =MENEM=
MENNEMEMENEMENEMENEMENNEMEN • =MENEM=
•
MENI • MN ••1117
MN
11111111
MEN MN MN
MN
MEN • MN
MN
MEM
1.1
__AIP
MI1 l'i III
•1 I__
••w••••III1
•
• •••
1•
1- II•
••
•• ••
•
••
••
MMEMMEMMEMMEMMEMMEMEM
MN
• ••
MEN •
• ••
MN
N
••
MN
N
=MENEM=
MENNEMEMENEMEN =MN
MENNEMEMENEMEN MENEM
MENNEMENEMENEMENEMEN
•
••
••
•
Table B23. Significant Differences of Study Groups from Normative Means
Norm
Measure
CSP MI
CSP NMI
GP MI
GP NMI
SCCF
Mean
Time Interval*:
1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
BHS‐college
2.32
BHS‐clinical
6.04
BPRS Total
49.29
BSI Anxiety
1.56
BSI Depression
1.26
BSI Hostility
1.70
BSI Interpersonal Sensitivity
0.96
BSI Obsessive‐Compulsive
2.22
BSI Paranoid Ideation
1.65
BSI Phobic Anxiety
0.55
BSI Psychotic
0.75
BSI Somatization
1.61
BSI GSI
0.25
PAS Acting Out
2.45
PAS Alienation
1.70
PAS Anger Control
1.73
PAS Health Problems
1.13
PAS Hostile Control
2.52
PAS Negative Affect
2.84
PAS Psychotic Features
0.71
PAS Social Withdrawal
2.17
PAS Suicidal Thinking
0.37
PAS Total
16.66
POMS Anger‐Hostility
7.10
POMS Depression‐Dejection
7.50
POMS Fatigue‐Inertia
7.30
POMS Tension‐Anxiety
7.10
POMS Vigor
<19.80
POMS Total
14.80
SIMS Affective Disorders
5.2
SIMS Amnestic Disorders
2.5
SIMS Low Intelligence
3.2
SIMS Neurological Impairment 2.4
SIMS Psychosis
1.2
SIMS Total
14.5
SLUMS
<25.70
STAI‐State
35.72
STAI‐Trait
34.89
Trails A Time
22.93
Trails B Time
48.97
Trails B – A
29
Trails B/A
2.18
*There are six assessments for CSP groups and five for the other three groups.
COMPOSITE SCORES
A composite score was developed for seven of the eight primary constructs by standardizing scores from the
scales on the self‐report assessments. Clinician and correctional officer ratings are not included in compo‐
sites so that comparisons between self‐report and staff reports can be made. Self‐report scores were stan‐
126
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
dardized so that comparisons between different measures could be made more easily and to create a single
measure for constructs assessed by multiple self‐report assessments. Scores were standardized by centering
on the mean of the entire sample at the first assessment and dividing by the standard deviation. A compo‐
site score was computed by standardizing each assessment and averaging the standardized scores across
the individual assessments as the composite score. Internal consistency reliability estimates, test‐retest cor‐
relations, and validity coefficient estimates for these composites and associated subscales are presented
within the discussion of each construct. Composite means and standard deviations are reported in the main
body of the report.
Anxiety Construct
Anxiety was measured by eight self‐report variables assessed at each time period. The self‐report measures
used to create the anxiety composite score were the State and Trait subscales of the STAI; the Anxiety, Ob‐
sessive‐Compulsive, and Phobic Anxiety subscales of the BSI; the Negative Affect subscale of the PAS; the
Tension‐Anxiety subscale of the POMS; and the Panic Disorder subscale of the PSI. This construct was also
assessed with ratings by correctional staff (PBRS Anxious‐Depressed) and clinicians (BPRS Anxiety‐
Depression).
Internal consistency reliabilities were computed for each assessment period for the entire sample. The in‐
ternal consistency estimates are provided in Table B24. The reliability estimates were strong, indicating good
internal consistency at each time period for the composite. Reliabilities for individual scales were similar
across testing intervals, and they were similar to internal consistency estimates from normative samples;
only the BPRS showed low internal consistency estimates.
Table B24. Internal Consistency Estimates for Anxiety Construct
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Anxiety Composite
8
.89
.90
.91
.91
.90
.89
BSI Anxiety
6
.86
.90
.91
.91
.90
.90
BSI Obsessive‐Compulsive
6
.88
.89
.90
.91
.88
.88
BSI Phobic Anxiety
5
.83
.86
.86
.87
.82
.88
PAS Negative Affect
3
.68
.65
.65
.70
.65
.60
POMS Tension‐Anxiety
9
.91
.91
.92
.92
.91
.89
PSI Panic Disorder
9
.89
.89
.89
.89
.90
.90
STAI State
10
.93
.94
.94
.94
.94
.93
STAI Trait
10
.93
.94
.94
.94
.95
.93
PBRS Anxious‐Depressed
14
.90
.90
.94
.90
.90
.90
BPRS Anxious‐Depressed
5
.55
NA
.60
NA
.66
NA
Test‐retest correlations for the anxiety composite are provided in Table B25 and indicate stable constructs
for three month assessments.
Table B25. TestRetest Correlation Coefficients for Anxiety
Composite at Consecutive Time Periods for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2
.76
.57
.71
.80
.49
.56
Time 2‐3
.73
.86
.82
.84
.71
.82
Time 3‐4
.86
.86
.75
.70
.73
.84
Time 4‐5
.83
.76
.77
.76
.71
.83
Time 5‐6
.75
.83
NA
NA
NA
.80
APPENDICES
127
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
The validity coefficients between anxiety measures are provided in Table B26 and indicate good convergent
validity for all measures except for the PBRS Anxious‐Depressed subscale. The pattern of relationships is sim‐
ilar across time periods.
Table B26. Correlations between Anxiety Construct Measures at each Time Period
Time 1
BSI OC BSI PA PAS NA POMS TA PSI PD STAI‐S STAI‐T
BSI Anxiety
.79
.75
.65
.81
.71
.60
.68
BSI Obsessive‐Compulsive
.66
.69
.70
.64
.55
.66
BSI Phobic Anxiety
.50
.58
.56
.46
.50
PAS Negative Affect
.71
.46
.62
.76
POMS Tension‐Anxiety
.52
.70
.76
PSI Panic Disorder
.37
.41
STAI‐State
.78
STAI‐Trait
PBRS Anxious Depressed
BPRS Anxious‐Depressed
Time 2
BSI OC BSI PA PAS NA POMS TA PSI PD STAI‐S STAI‐T
BSI Anxiety
.78
.78
.62
.82
.73
.63
.67
BSI Obsessive‐Compulsive
.69
.60
.69
.68
.59
.68
BSI Phobic Anxiety
.55
.59
.65
.54
.56
PAS Negative Affect
.66
.39
.63
.79
POMS Tension‐Anxiety
.56
.72
.78
PSI Panic Disorder
.44
.46
STAI‐State
.83
STAI‐Trait
PBRS Anxious‐Depressed
Time 3
BSI OC BSI PA PAS NA POMS TA PSI PD STAI‐S STAI‐T
BSI Anxiety
.83
.83
.69
.83
.73
.67
.70
BSI Obsessive‐Compulsive
.75
.69
.73
.72
.66
.70
BSI Phobic Anxiety
.59
.64
.69
.55
.58
PAS Negative Affect
.73
.48
.68
.80
POMS Tension‐Anxiety
.58
.74
.79
PSI Physical Symptoms
.48
.48
STAI‐State
.85
STAI‐Trait
PBRS Anxious‐Depressed
BPRS Anxiety‐Depression
Time 4
BSI OC BSI PA PAS NA POMS TA PSI PD STAI‐S STAI‐T
BSI Anxiety
.81
.78
.63
.82
.69
.69
.72
BSI Obsessive‐Compulsive
.66
.64
.75
.69
.71
.72
BSI Phobic Anxiety
.52
.59
.58
.54
.63
PAS Negative Affect
.70
.48
.66
.77
POMS Tension‐Anxiety
.60
.77
.78
PSI Physical Symptoms
.51
.51
STAI‐State
.84
STAI‐Trait
PBRS Anxious‐Depressed
Time 5
BSI OC BSI PA PAS NA POMS TA PSI PD STAI‐S STAI‐T
BSI Anxiety
.80
.79
.66
.79
.72
.62
.65
BSI Obsessive‐Compulsive
.74
.69
.77
.67
.63
.70
BSI Phobic Anxiety
.62
.60
.58
.51
.58
PAS Negative Affect
.68
.47
.62
.76
POMS Tension‐Anxiety
.61
.78
.78
128
PBRS AD
.05
.12
.05
.12
.10
.06
.07
.11
PBRS AD
.10
.04
.09
.13
.11
.06
.18
.17
PBRS AD
.04
.04
.06
.14
.16
‐.02
.12
.14
PBRS AD
.20
.12
.15
.13
.17
.19
.15
.18
PBRS AD
.20
.22
.19
.21
.27
BPRS AD
.42
.35
.33
.35
.37
.35
.36
.37
.33
BPRS AD
BPRS AD
.41
.39
.36
.40
.40
.34
.40
.49
.12
BPRS AD
BPRS AD
.33
.33
.26
.38
.40
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
PSI Physical Symptoms
STAI‐State
STAI‐Trait
PBRS Anxious‐Depressed
BPRS Anxiety‐Depression
Time 6
BSI Anxiety
BSI Obsessive‐Compulsive
BSI Phobic Anxiety
PAS Negative Affect
POMS Tension‐Anxiety
PSI Physical Symptoms
STAI‐State
STAI‐Trait
PBRS Anxious‐Depressed
BSI OC BSI PA
.77
.74
.67
.44
PAS NA
.62
.55
.49
POMS TA
.81
.64
.52
.69
PSI PD
.74
.66
.62
.45
.63
STAI‐S
.57
.55
.37
.65
.76
.44
.47
.20
.83
.19
.21
STAI‐T PBRS AD
.64
.12
.61
.14
.49
.05
.77
.13
.77
.18
.49
‐.01
.85
.09
.08
.32
.36
.46
.26
BPRS AD
Note: Only Times 1, 3, and 5 had the BPRS administered. Time 6 includes only the CSP NMI and CSP MI groups.
Cognitive Impairment Construct
Cognitive impairment was assessed by two individually administered tests and ratings by the researcher. The
SLUMS was used to assess orientation, memory, attention, and executive function. The TMT was used to
assess attention. The time required to complete the Trails A (connect sequential numbers) and B (connect
alternating numbers and letters) tasks were collected, and the ratio of times (B/A) was used as the attention
measure. Because these measures were not correlated (see descriptions of measures above), we did not
combine these scores into a composite measure of cognitive impairment and each was used individually.
The correctional staff completed ratings on the PBRS Dull‐Confused scale.
Internal consistency estimates for the SLUMS were provided earlier in the description of the measure. Table
B27 provides the correlations between consecutive testing periods and Table B28 has the correlations be‐
tween the cognitive assessments, including the correctional officer ratings. The correlations between con‐
secutive time periods are moderate in size and indicate stability across 3 month assessment periods. There
are some variations in size of coefficients across groups. The validity coefficients are small and indicate that
these assessments are likely measuring unique aspects of cognitive function.
Table B27. TestRetest Correlation Coefficients for Cognitive Impair
ment Measures at Consecutive Time Intervals by Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
SLUMS Scores
Time 1‐2
0.72
0.55
0.64
0.38
0.59
0.63
Time 2‐3
0.63
0.74
0.65
0.59
0.78
0.70
Time 3‐4
0.75
0.77
0.67
0.70
0.76
0.75
Time 4‐5
0.75
0.42
0.67
0.64
0.75
0.67
Time 5‐6
0.84
0.67
NA
NA
NA
0.78
Trails B/A Scores
Time 1‐2
0.51
0.32
0.29
0.48
0.40
0.36
Time 2‐3
0.20
0.63
0.40
0.32
0.33
0.37
Time 3‐4
0.31
0.55
0.58
0.64
0.28
0.44
Time 4‐5
0.59
0.33
0.66
0.70
0.30
0.44
Time 5‐6
0.47
0.35
NA
NA
NA
0.39
APPENDICES
129
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B28. Correlations between Cognitive Impairment Measures at each Time Period
Time 1
TMT B/A PBRS DC Time 4
TMT B/A PBRS DC
SLUMS
‐.26
‐.12
SLUMS
‐.14
‐.10
TMT B/A
.09
TMT B/A
‐.01
PBRS Dull‐Confused
PBRS Dull‐Confused
Time 2
TMT B/A PBRS DC
Time 5
TMT B/A PBRS DC
SLUMS
‐.170
‐.081
SLUMS
‐.30
‐.18
TMT B/A
‐.019
TMT B/A
‐.05
PBRS Dull‐Confused
PBRS Dull‐Confused
Time 3
TMT B/A PBRS DC
Time 6
TMT B/A PBRS DC
SLUMS
‐.26
‐.11
SLUMS
‐.13
‐.03
TMT B/A
.04
TMT B/A
‐.12
PBRS Dull‐Confused
PBRS Dull‐Confused
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6 includes only the CSP NMI and CSP MI groups.
Depression/Hopelessness Construct
The depression construct was assessed using five self‐report measures. The subscales used to create this
composite were the BHS, the BSI Depression subscale, the PAS Negative Affect and Suicidal subscales, and
the POMS Depression‐Dejection subscale. Table B29 provides estimates of internal consistency reliability for
each subscale and the composite at each time period. There is evidence for adequate internal consistency
for the composite (M = .75; range = .71 to .77). Internal consistency estimates for the subscales were similar
to reliabilities found with normative samples with subscales with fewer items demonstrating lower alphas.
Table B29. Internal Consistency Estimates for Depression Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Depression Composite
5
.74
.76
.76
.77
.76
.71
BHS Total
20
.92
.94
.94
.94
.94
.92
BSI Depression
6
.87
.89
.90
.89
.90
.86
PAS Negative Affect
3
.68
.65
.65
.70
.65
.60
PAS Suicidal Thinking
2
.86
.91
.94
.95
.90
.94
POMS Depression‐Dejection
15
.95
.95
.96
.95
.95
.93
PBRS Anxious‐Depressed
14
.90
.90
.94
.90
.90
.90
BPRS Anxiety‐Depression
5
.55
NA
.60
NA
.66
NA
Table B30 provides estimates of test‐retest reliability. The correlations between consecutive assessments for
the depression composite were strong (M = .76, range = .57 to .90) indicating good stability over time. Al‐
though there was some variability in estimates across groups and times, there is reasonable stability esti‐
mates for each group.
Table B30. TestRetest Correlation Coefficients for Depression
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2
.77
.69
.57
.79
.65
.59
Time 2‐3
.68
.82
.74
.85
.73
.80
Time 3‐4
.88
.90
.73
.69
.72
.84
Time 4‐5
.77
.74
.82
.71
.77
.83
Time 5‐6
.70
.79
NA
NA
NA
.76
Table B31 provides the correlations between the measures of the depression construct. The validity coeffi‐
cients between assessments of the depression construct indicate good convergent validity with the excep‐
130
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
tion of the PBRS Anxious Depressed subscale. The magnitude and general pattern of relationships between
measures were similar across assessment periods.
Table B31. Correlations between DepressionHopelessness Measures at each Time Period
Time 1
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.71
.61
.47
.60
.12
.32
BSI Depression
.71
.59
.85
.10
.43
PAS Negative Affect
.46
.65
.12
.35
PAS Suicidal Thinking
.44
‐.01
.34
POMS Depression‐Dejection
.45
.13
PBRS Anxious Depressed
‐.03
BPRS Anxiety Depression
Time 2
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.70
.54
.45
.61
.10
BSI Depression
.63
.62
.88
.14
PAS Negative Affect
.46
.67
.13
PAS Suicidal Thinking
.54
.18
POMS Depression‐Dejection
.12
PBRS Anxious Depressed
Time 3
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.74
.61
.43
.65
‐.02
.33
BSI Depression
.69
.60
.88
.07
.37
PAS Negative Affect
.41
.71
.15
.40
PAS Suicidal Thinking
.53
.11
.30
POMS Depression‐Dejection
.10
.44
PBRS Anxious Depressed
.12
BPRS Anxiety Depression
Time 4
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.77
.61
.43
.69
.13
BSI Depression
.67
.58
.88
.18
PAS Negative Affect
.42
.66
.13
PAS Suicidal Thinking
.46
.23
POMS Depression‐Dejection
.18
PBRS Anxious Depressed
Time 5
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.75
.61
.39
.66
.19
.27
BSI Depression
.69
.51
.89
.25
.42
PAS Negative Affect
.39
.68
.21
.38
PAS Suicidal Thinking
.48
.29
.39
POMS Depression‐Dejection
.27
.42
PBRS Anxious Depressed
.26
BPRS Anxiety Depression
Time 6
BSI Dep PAS NA PAS ST POMS DD PBRS AD BPRS AD
BHS
.58
.46
.45
.45
.14
BSI Depression
.62
.51
.84
.20
PAS Negative Affect
.35
.62
.13
PAS Suicidal Thinking
.42
.41
POMS Depression‐Dejection
.12
PBRS Anxious Depressed
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6 includes only the CSP NMI and CSP MI groups.
APPENDICES
131
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Hostility/Anger Control Construct
The hostility/anger control construct was assessed using five self‐report measures: the BSI Hostility subscale;
the Anger Control, Hostile Control, and Acting Out subscales on the PAS; and the POMS Anger‐Hostility
subscale. Ratings by correctional staff (PBRS Anti‐authority) and clinicians (BPRS Hostility) also assess the
hostility construct. Table B32 provides the internal consistency reliabilities for the subscales and composite.
The composite internal consistency was low (M = .57) for the six time periods. Subscale reliabilities were
lower than expected for the PAS subscales and the BPRS, although the smaller internal consistency esti‐
mates were for scales with a small number of items and these reliability estimates are similar to other litera‐
ture.
Table B32. Internal Consistency Estimates for Hostility Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Hostility Composite
5
.54
.56
.57
.61
.55
.60
BSI Hostility
5
.85
.84
.87
.90
.87
.88
PAS Anger Control
2
.53
.60
.51
.56
.52
.33
PAS Hostile Control
2
.52
.47
.42
.37
.36
.61
PAS Acting Out
3
.27
.28
.30
.39
.39
.46
POMS Anger‐Hostility
12
.92
.93
.94
.94
.94
.94
PBRS Anti‐Authority
13
.94
.93
.94
.94
.95
.90
BPRS Hostility
3
.57
NA
.61
NA
.51
NA
The correlations between consecutive time periods are given in Table B33. These estimates of test‐retest
reliability indicate that the hostility composite is stable between 3 month assessment periods. Groups are
fairly similar in the magnitude of correlation coefficients between testing periods.
Table B33. TestRetest Correlation Coefficients for Hostility
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2
.73
.75
.83
.78
.69
.71
Time 2‐3
.67
.82
.76
.77
.80
.77
Time 3‐4
.80
.84
.76
.80
.77
.80
Time 4‐5
.76
.78
.56
.79
.77
.76
Time 5‐6
.66
.72
NA
NA
NA
.69
Table B34 provides the correlations between measures of the hostility‐anger control construct for each time
period. The validity coefficients were quite variable across all measures and time periods (ranging between
‐.11 and .84). Although scores on these measures tend to be stable, the different assessments may be tap‐
ping into quite different aspects of hostility given these variable and lower correlations. Examination of the
content of the PAS items suggested that these items tapped into useful domains for understanding hostile
and acting out behavior of the participants and thus these measures were kept, even though this leads to
lower internal consistency estimates. Removing these items from the composite did increase internal consis‐
tency estimates but did not substantially change the study results, thus all subscales were kept as part of the
composite (additional results are available from the authors upon request).
132
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B34. Correlations between HostilityAnger Control Measures at each Time Period
Time 1
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.20
.54
.17
.69
.06
.27
PAS Acting Out
.18
.14
.10
‐.11
.09
PAS Anger Control
.28
.44
.14
.10
PAS Hostile Control
.17
.12
.19
POMS Anger‐Hostility
.09
.19
PBRS Anti‐Authority
.15
BPRS Hostility
Time 2
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.15
.60
.19
.76
‐.01
PAS Acting Out
.15
.14
.12
.08
PAS Anger Control
.24
.56
.06
PAS Hostile Control
.14
.12
POMS Anger‐Hostility
.04
PBRS Anti‐Authority
Time 3
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.15
.45
.15
.50
.02
.36
PAS Acting Out
.15
.15
.16
.06
.05
PAS Anger Control
.24
.48
.11
.29
PAS Hostile Control
.11
.08
.13
POMS Anger‐Hostility
.05
.28
PBRS Anti‐Authority
.26
BPRS Hostility
Time 4
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.24
.55
.33
.84
.14
PAS Acting Out
.10
.13
.18
.03
PAS Anger Control
.33
.51
‐.02
PAS Hostile Control
.27
.09
POMS Anger‐Hostility
.12
PBRS Anti‐Authority
Time 5
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.19
.47
.22
.77
.18
.25
PAS Acting Out
.18
.14
.16
‐.02
.04
PAS Anger Control
.32
.46
.15
.20
PAS Hostile Control
.16
.003
.04
POMS Anger‐Hostility
.14
.27
PBRS Anti‐Authority
.27
BPRS Hostility
Time 6
PAS AO PAS AC PAS HC POMS AH PBRS AA BPRS H
BSI Hostility
.22
.46
.36
.81
.07
PAS Acting Out
.23
.34
.10
.03
PAS Anger Control
.39
.47
.08
PAS Hostile Control
.24
‐.06
POMS Anger‐Hostility
.13
PBRS Anti‐Authority
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6 includes only the CSP NMI and CSP MI groups.
Hypersensitivity Construct
The hypersensitivity construct was measured by two self‐report measures—the External Stimulus subscale
of the PSI and the Interpersonal Sensitivity subscale of the BSI. This construct is assessing two different as‐
pects of hypersensitivity—environmental and interpersonal. Internal consistency reliabilities for the subs‐
APPENDICES
133
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
cales computed for each assessment period for the entire sample indicate that there is substantial variability
in the internal consistency estimates (see Table B35); however, examination of each scale shows that the BSI
has strong internal consistency estimates whereas the PSI has low estimates. The PSI was created by the
researchers, and its purpose was to capture variables not measured by existing measures, thus it may not be
a unidimensional construct. The internal estimates of the composite are lower than might be hoped for and
are evidence for the lack of a homogeneous construct.
Although internal consistency estimates were low, the composite demonstrated modest estimates of test‐
retest reliability (see Table B36) and the correlations between these two subscales provided evidence of
convergent validity (see Table B37). Thus these scales were analyzed as a composite for the major analyses
completed in the report. Results for analyses done on each individual variable are available from the re‐
searchers upon request.
Table B35. Internal Consistency Estimates for Hypersensitivity Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Hypersensitivity Composite
2
.55
.58
.61
.51
.48
.47
BSI Interpersonal Hypersensitivity
4
.81
.71
.86
.84
.84
.83
PSI External Stimulus
5
.22
.32
.28
.39
.34
.27
Table B36. TestRetest Correlation Coefficients for Hypersensitivity
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2
.56
.59
.56
.67
.21
.46
Time 2‐3
.59
.74
.78
.80
.60
.71
Time 3‐4
.71
.71
.63
.65
.57
.70
Time 4‐5
.68
.64
.75
.72
.60
.71
Time 5‐6
.67
.65
NA
NA
NA
.68
Table B37. Correlations between Hypersensitivity
Measures at each Time Period
Time: 1
2
3
4
5
6
BSI IS with PSI ES .38 .41 .44 .34 .32 .31
Note: Time 6 includes only the CSP NMI and CSP MI groups.
Psychosis Construct
The psychosis construct was assessed by three self‐report measures—the Paranoid Ideation and Psychotic
subscales of the BSI and the Psychotic Features subscale of the PAS—and clinician ratings (BPRS Thought
Disorder). Table B38 provides the Cronbach’s alphas for each subscale and the psychosis composite. Internal
consistency estimates were good for this composite and its components. Internal consistency estimates for
the subscales were similar to those found with normative samples.
Table B38. Internal Consistency Estimates for Psychosis Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Psychosis Composite
3
.73
.78
.80
.79
.80
.76
BSI Paranoid Ideation
5
.78
.80
.82
.83
.82
.82
BSI Psychoticism
5
.77
.78
.80
.77
.79
.75
PAS Psychotic Features
2
.62
.72
.71
.71
.79
.73
BPRS Thought Disorder
5
.64
NA
.52
NA
.57
NA
134
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B39 provides correlations between sequential time periods. Examination of these test‐retest reliability
estimates indicates good stability between assessment periods. The study groups are similar in magnitude of
correlations.
Table B39. TestRetest Correlation Coefficients for Psychosis
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2 .59
.61
.67
.74
.55
.55
Time 2‐3 .52
.83
.75
.87
.73
.75
Time 3‐4 .76
.74
.70
.81
.77
.80
Time 4‐5 .71
.69
.78
.82
.68
.75
Time 5‐6 .64
.79
NA
NA
NA
.71
Table B40 provides estimates of validity coefficients between the measures of the psychosis construct. Cor‐
relations indicate reasonable convergent validity for this sample, although correlations are stronger be‐
tween self‐report assessments than between self‐ and staff‐report assessments.
Table B40. Correlations between Psychosis Measures at
each Time Period
Time Measure
BSI Psy PAS Psy BPRS TD
BSI Paranoid Ideation
.71
.57
.22
BSI Psychoticism
.35
.27
1
PAS Psychotic Features
.32
BPRS Thought Disorder
BSI Paranoid Ideation
.74
.66
BSI Psychoticism
.48
2
PAS Psychotic Features
BSI Paranoid Ideation
.77
.76
.31
BSI Psychoticism
.56
.30
3
PAS Psychotic Features
.25
BPRS Thought Disorder
BSI Paranoid Ideation
.78
.71
4
BSI Psychoticism
.47
PAS Psychotic Features
BSI Paranoid Ideation
.79
.72
.15
BSI Psychoticism
.53
.13
5
PAS Psychotic Features
.14
BPRS Thought Disorder
BSI Paranoid Ideation
.72
.64
6
BSI Psychoticism
.42
PAS Psychotic Features
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6 includes
only the CSP NMI and CSP MI groups.
Somatization Construct
The somatization construct was measured by four self‐report assessments, including the Somatization sub‐
scale of the BSI, the Health Problems subscale of the PAS, the POMS Fatigue‐Inertia subscale, and the Physi‐
cal Well‐Being subscale of the PSI. The mean Cronbach’s alpha across somatization measures and time pe‐
riods was .79 (see Table B41) and for the composite the mean alpha was .77, indicating adequate internal
consistency for this sample. Estimates were similar across time periods.
APPENDICES
135
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Table B41. Internal Consistency Estimates for Somatization Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Somatization Composite
3
.78
.79
.78
.78
.78
.73
BSI Somatization
7
.87
.85
.88
.89
.88
.88
PAS Health Problems
2
.56
.65
.65
.59
.65
.59
POMS Fatigue‐Inertia
7
.91
.90
.94
.92
.92
.90
PSI Physical Well‐Being
8
.72
.76
.73
.75
.76
.74
Table B42 provides test‐retest reliability estimates. Correlations between consecutive time periods indicate
strong stability across time. Reliability estimates are similar across study groups. Table B43 provides esti‐
mates of convergent validity. The correlations between the measures of somatization are reasonable for
both self‐report assessments and clinician ratings. Correlations show the same basic pattern and magnitude
at each time period.
Table B42. TestRetest Correlation Coefficients for Somatization
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI SCCF
All
Time 1‐2
.81
.69
.82
.79
.60
.62
Time 2‐3
.74
.80
.81
.86
.77
.83
Time 3‐4
.80
.74
.74
.76
.81
.84
Time 4‐5
.81
.71
.67
.82
.58
.77
Time 5‐6
.77
.67
NA
NA
NA
.85
Table B43. Correlations between Somatization Measures at each Time Period
Time Measure
PSI PE
PAS HP POMS F BPRS AD
BSI Somatization
.54
.54
.59
.34
PSI Physical Exercise
.55
.61
.43
1
PAS Health Problems
.42
.35
POMS Fatigue
.40
BPRS Anxiety Depression
BSI Somatization
.62
.55
.61
PSI Physical Exercise
.63
.64
2
PAS Health Problems
.49
POMS Fatigue
BSI Somatization
.56
.53
.65
.37
PSI Physical Exercise
.59
.56
.39
3
PAS Health Problems
.46
.44
POMS Fatigue
.40
BPRS Anxiety Depression
BSI Somatization
.58
.52
.67
PSI Physical Exercise
.62
.60
4
PAS Health Problems
.40
POMS Fatigue
BSI Somatization
.56
.54
.65
.27
PSI Physical Exercise
.57
.62
.44
5
PAS Health Problems
.47
.28
POMS Fatigue
.34
BPRS Anxiety Depression
BSI Somatization
.44
.40
.64
PSI Physical Exercise
.54
.54
6
PAS Health Problems
.38
POMS Fatigue
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6 includes only the CSP NMI and CSP MI groups.
136
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Withdrawal/Alienation Construct
The withdrawal/alienation construct was assessed using two PAS subscales—Alienation and Social With‐
drawal—and clinicians’ ratings on the BPRS Withdrawal subscale. Internal consistency reliabilities were
computed for each assessment period for the entire sample and are provided in Table B44. The Cronbach’s
alphas indicate adequate internal consistency estimates for the self‐report measures but are lower for the
clinicians’ ratings. Estimates are of similar magnitude across time periods.
Table B44. Internal Consistency Estimates for WithdrawalAlienation Construct by Time Interval
Measure
# of items Time 1 Time 2 Time 3 Time 4 Time 5 Time 6
Withdrawal Composite
2
.63
.71
.67
.70
.71
.62
PAS Alienation
2
.79
.74
.75
.72
.72
.72
PAS Social Withdrawal
2
.69
.74
.72
.78
.75
.83
BPRS Withdrawal
6
.47
NA
.49
NA
.40
NA
Table B45 provides the estimates of test‐retest reliability for the withdrawal composite. Correlations be‐
tween sequential time periods were strong indicating good stability. Reliabilities were similar across testing
intervals, although there was some variability.
Table B45. TestRetest Correlation Coefficients for Withdrawal
Composite at Consecutive Time Intervals for each Study Group
Interval
CSP MI CSP NMI GP MI GP NMI
SCCF
All
Time 1‐2
.65
.67
.55
.64
.87
.60
Time 2‐3
.72
.83
.75
.65
.50
.73
Time 3‐4
.76
.83
.61
.69
.52
.73
Time 4‐5
.63
.81
.60
.49
.69
.71
Time 5‐6
.67
.76
NA
NA
NA
.72
The convergent validity estimates are provided in Table B46. The correlations between the self‐report as‐
sessments (both PAS subscales) indicate strong correlation coefficients; however, the correlations of the
self‐report with the clinician reports are low. The same pattern is shown across all time periods.
Table B46. Correlations between Withdrawal
Alienation Measures at each Time Period
Time Measure
PAS SW BPRS W
PAS Alienation
.46
.17
1
PAS Social Withdrawal
.16
BPRS Withdrawal
PAS Alienation
.55
2
PAS Social Withdrawal
PAS Alienation
.50
.06
3
PAS Social Withdrawal
.14
BPRS Withdrawal
PAS Alienation
.53
4
PAS Social Withdrawal
PAS Alienation
.54
.18
5
PAS Social Withdrawal
.19
BPRS Withdrawal
PAS Alienation
.45
6
PAS Social Withdrawal
Note: Only times 1, 3, and 5 had the BPRS administered. Time 6
includes only the CSP NMI and CSP MI groups.
APPENDICES
137
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
SUMMARY
Using standardized measures allowed us to assess if the scores we obtained in this sample were reliable and
valid; we were also able to compare scores from our sample to known values. The self‐report assessment
variables in general tended to perform similarly across study groups, across time, and to normative samples
when examining internal consistency and test‐retest reliability. Self‐report measures of similar constructs
tended to demonstrate convergent validity. In order to combine measures, a composite was developed for
each construct of interest, except cognitive impairment, by standardizing scores from the first assessment
and computing the mean across measures of the same construct. These composites demonstrated adequate
research reliability (internal consistency and test‐retest).
This study included self‐report and staff report information to allow for convergence using different sources
of information. Clinician ratings were gathered using the well‐known BPRS. The scores from this measure
had a floor effect with scores much lower than normative data (used with clinical populations, including pa‐
tients who were not in crisis). These scores had lower than expected reliability and validity estimates, al‐
though some subscales correlated modestly with self‐report measures. There are several possible reasons
for this, none of which have been tested in this study—participants were not forthcoming with clinicians,
changing of locations may lead to unfamiliarity between participant and clinicians; lack of familiarity with
measures by clinicians; clinicians have been desensitized to extreme behaviors in a prison setting so partici‐
pants seem to be functioning well; clinicians are accurate but participants are exaggerating. Although the
floor effect is a concern, because we hypothesized the scores to increase over time on the BPRS, the meas‐
ure should be able to assess if the mean scores are getting worse over time.
We also used the only measure we could find that allowed for ratings by correctional staff. The only study
available on the PBRS as a reference described the development of this measure. The PBRS assessments
showed good internal consistency reliability; however, test‐retest reliabilities were low and there was little
evidence of convergent validity.
Mean scores for each sample were compared to means from normative samples or published research.
These comparisons to general adult populations (typically, but sometimes psychiatric populations were
used) tended to show that all groups except the general prison participants without mental illness (GP NMI)
had statistically significant elevations for the majority of measures across the study assessment periods. This
finding suggests that participants entering administrative segregation have significant mental health issues
to start, which underscores the importance of assessing individuals over time to explore changes that condi‐
tions of confinement might engender.
In summary, the demonstration of good psychometric properties of the data would suggest that responses
are not given randomly or haphazardly and that participants are responding in a consistent fashion.
138
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
APPENDIX B REFERENCES
American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed).
Washington, DC: Author.
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text
revision). Washington, DC: Author.
Basso, M. R., Bornstein, R. A., & Lang, J. M. (1999). Practice effects on commonly used measures of executive
function across twelve months. The Clinical Neuropsychologist, 13, 283‐292. doi:
10.1076/clin.13.3.283.1743
Beck, A. T. (1986). Hopelessness as a predictor of eventual suicide. Annals of the New York Academy of
Sciences, 487, 90‐96. doi: 10.1111/j.1749‐6632.1986.tb27888.x
Beck, A. T., Brown, G., Berchick, R. J., Stewart, B. L., & Steer, R. A. (1990). Relationship between hopelessness
and ultimate suicide: A replication with psychiatric outpatients. American Journal of Psychiatry, 147,
190‐195. Retrieved from http://ajp.psychiatryonline.org/
Beck, A. T., & Steer, R. A. (1993). Beck Hopelessness Scale: Manual. San Antonio, TX: The Psychological Cor‐
poration: Harcourt Brace & Company.
Beck, A. T., Steer, R. A., Kovacs, M., & Garrison, B. (1985). Hopelessness and eventual suicide: A 10 year
prospective study of patients hospitalized with suicidal ideation. American Journal of Psychiatry,
142, 559‐563. Retrieved from http://www.lexisnexis.com/
Beck, A. T., Weissman, A., Lester, D., & Trexler, L. (1974). The measurement of pessimism: The hopelessness
scale. Journal of Consulting and Clinical Psychology, 42, 861‐865. doi: 10.1037/h0037562
Biancosino, B., Barbui, C., Pera, V., Osti, M., Rocchi, D., Marmai, L., & Grassi, L. (2004). Patient opinions on
the benefits of treatment programs in residential psychiatric care. The Canadian Journal of Psychia‐
try, 49, 613‐619. Retrieved from http://server03.cpa‐
apc.org:8080/Publications/Archives/CJP/2004/september/biancosino.pdf
Bieling, P. J., Beck, A. T., & Brown, G. K. (2000). The Sociotropy‐Autonomy Scale: Structure and implications.
Cognitive Therapy and Research, 24, 763‐780. doi: 10.1023/A:1005599714224
Bornstein, R. A., Baker, G. B., & Douglass, A. B. (1987). Short‐term retest reliability of the Halstead‐Reitan
Battery in a normal sample. The Journal of Nervous and Mental Disease, 175, 229‐232. doi:
10.1097/00005053‐198704000‐00007
Boulet, J., & Boss, M. W. (1991). Reliability and validity of the Brief Symptom Inventory. Psychological As‐
sessment, 3, 433‐437. doi: 10.1037/1040‐3590.3.3.433
Briere, J. (1995). Trauma Symptom Inventory: Professional manual. Lutz, FL: Psychological Assessment Re‐
sources, Inc.
Broday, S. F., & Mason, J. L. (1991). Internal consistency of the Brief Symptom Inventory for counseling‐
center clients. Psychological Reports, 68, 94. doi: 10.2466/PR0.68.1.94‐94
APPENDICES
139
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Brown, G. K., Beck, A. T., Steer, R. A., & Grisham, J. R. (2000). Risk factors for suicide in psychiatric outpa‐
tients: A 20‐year prospective study. Journal of Consulting and Clinical Psychology, 68, 371‐377. doi:
10.1037/0022‐006X.68.3.371
Brown, S., Chhina, N., & Dye, S. (2008). The psychiatric intensive care unit: A prospective survey of patient
demographics and outcomes at seven English PICUs. Journal of Psychiatric Intensive Care, 4, 17‐27.
doi: 10.1017/S1742646408001295
Burger, K. G., Calsyn, R. J., Morse, G. A., Klinkenberg, W. D., & Trusty, M. L. (1997). Factor structure of the
Expanded Brief Psychiatric Rating Scale. Journal of Clinical Psychology, 53, 451‐454. doi:
10.1002/(SICI)1097‐4679(199708)53:5<451::AID‐JCLP5>3.0.CO;2‐Q
Clum, G. A., & Yang, B. (1995). Additional support for the reliability and validity of the Modified Scale for Sui‐
cide Ideation. Psychological Assessment, 7, 122‐125. doi: 10.1037/1040‐3590.7.1.122
Cochran, C. D., & Hale, W. D. (1985). College norms on the Brief Symptom Inventory. Journal of Clinical Psy‐
chology, 41, 777‐779. doi: 10.1002/1097‐4679(198511)41:6<777::AID‐JCLP2270410609>3.0.CO;2‐2
Cooke, D. J. (1998). The development of the Prison Behavior Rating Scale. Criminal Justice and Behavior, 25,
482‐506. doi: 10.1177/0093854898025004005
Coolidge, F. L. (n.d. a). An introduction to the Coolidge Correctional Inventory (CCI). Retrieved from
http://www.uccs.edu/~faculty/fcoolidg//
Coolidge, F. L. (n.d. b). Coolidge Axis Two Inventory‐Revised (CATI+): Manual. Retrieved from
http://www.uccs.edu/~faculty/fcoolidg//
Coolidge, F. L. (2004). Coolidge Correctional Inventory. Colorado Springs, CO: Author.
Coolidge, F. L., Segal, D. L., Klebe, K. J., Cahill, B. S., & Whitcomb, J. M. (2009). Psychometric properties of the
Coolidge Correctional Inventory in a sample of 3,962 prison inmates. Behavioral Sciences and the
Law, 27, 713‐726. doi: 10.1002/bsl.896
Cundick, B. P. (1975). Review of the Brief Symptom Inventory. In J. Close Conoley & J. J. Kramer (Eds.), The
tenth mental measurement yearbook (pp. 111‐112). Lincoln, NE: University of Nebraska Press.
Derogatis, L. (1993). Brief Symptom Inventory: Administration, scoring, and procedures manual. Minneapo‐
lis, MN: NCS Pearson, Inc.
Durham, T. W. (1982). Norms, reliability, and item analysis of the Hopelessness Scale in general psychiatric,
forensic psychiatric, and college populations. Journal of Clinical Psychology, 38, 597‐600. doi:
10.1002/1097‐4679(198207)38:3<597::AID‐JCLP2270380321>3.0.CO;2‐6
Dyce, J. A. (1996). Factor structure of the Beck Hopelessness Scale. Journal of Clinical Psychology, 52, 555‐
558. doi: 10.1002/(SICI)1097‐4679(199609)52:5<555::AID‐JCLP10>3.0.CO;2‐D
Edens, J. F., Poythress, N. G., & Watkins‐Clay, M. M. (2009). Detection of malingering in psychiatric unit and
general population prison inmates: A comparison of the PAI, SIMS, and SIRS. Journal of Personality
Assessment, 88, 33‐42. doi: 10.1207/s15327752jpa8801_05
140
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Fernandez, E. (1998). Review of the Beck Hopelessness Scale [revised]. In J. C. Impara & B. S. Plake (Eds.), The thir‐
teenth mental measurement yearbook (pp. 123‐125). Lincoln, NE: University of Nebraska Press.
Fliege, H., Kocalevent, R. D., Walter, O. B., Beck, S., Gratz, K. L., Guiterrez, P. M., & Klapp, B. F. (2006). Three
assessment tools for deliberate self‐harm and suicide behavior: Evaluation and psychopathological
correlates. Journal of Psychosomatic Research, 61, 113‐121. doi: 10.1016/j.jpsychores.2005.10.006
Gondolf, E. W. (2008). Supplemental mental health treatment for batterer program participants (Document
no. 223030). Retrieved from http://www.ncjrs.gov/pdffiles1/nij/grants/223030.pdf
Grassian, S. (1983). Psychopathological effects of solitary confinement. American Journal of Psychiatry, 140,
1450‐1454. Retrieved from http://ajp.psychiatryonline.org/index.dtl
Gratz, K. L. (2001). Measurement of deliberate self‐harm: Preliminary data on the deliberate self‐harm in‐
ventory. Journal of Psychopathology and Behavioral Assessment, 23, 253‐263. doi:
10.1023/A:1012779403943
Gratz, K. L., & Chapman, A. L. (2007). The role of emotional responding and childhood maltreatment in the
development and maintenance of deliberate self‐harm among male undergraduates. Psychology of
Men & Masculinity, 8, 1‐14. doi: 10.1037/1524‐9220.8.1.1
Gray, R., Bressington, D., Lathlean, J., & Mills, A. (2008). Relationship between adherence, symptoms, treat‐
ment attitudes, satisfaction, and side effects in prisoners taking antipsychotic medication. The Jour‐
nal of Forensic Psychiatry & Psychology, 19, 335‐351. doi: 10.1080/14789940802113493
Haney, C. (2003). Mental health issues in long‐term solitary and "supermax" confinement. Crime and Delin‐
quency, 49, 124‐156. doi: 10.1177/0011128702239239
Harrison, K. S., & Rogers, R. (2007). Axis I screens and suicide risk in jails: A comparative analysis. Assess‐
ment, 14, 171‐180. doi: 10.1177/1073191106296483
Hayes, J. A. (1997). What does the Brief Symptom Inventory measure in college and university counseling
center clients? Journal of Counseling Psychology, 44, 360‐367. doi: 10.1037/0022‐0167.44.4.360
Heeter, C., Winn, B., Winn, J., & Bozoki, A. (2008). The challenge of challenge: Avoiding and embracing diffi‐
culty in a memory game (Unpublished manuscript). Retrieved from
http://meaningfulplay.msu.edu/proceedings2008/mp2008_paper_31.pdf
Heinze, M. C., & Purisch, A. D. (2001). Beneath the mask: Use of psychological tests to detect and subtype
malingering in criminal defendants. Journal of Forensic Psychology Practice, 1, 23‐52. doi:
10.1300/J158v01n04_02
Holden, R. R., Book, A. S., Edwards, M. J., Wasylkiw, L. , & Starzyk, K. B. (2003). Experimental faking in self‐
reported psychopathology: Unidimensional or multidimensional? Personality and Individual Differ‐
ences, 35, 1107‐1117. doi: 10.1016/S0191‐8869(02)00321‐5
Holden, R. R., & Fekken, C. (1988). Test‐retest reliability of the Hopelessness Scale and its items in a universi‐
ty population. Journal of Clinical Psychology, 44, 40‐43. doi: 10.1002/1097‐
4679(198801)44:1<40::AID‐JCLP2270440108>3.0.CO;2‐P
APPENDICES
141
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Joesting, J. (1976). Test‐retest reliabilities of state‐trait anxiety inventory in an academic setting: replication.
Psychological Reports, 38, 318. Retrieved from http://ejournals.ammonsscientific.com/
Kellett, S., Beail, N., Newman, D. W., & Frankish, P. (2003). Utility of the Brief Symptom Inventory in the as‐
sessment of psychological distress. Journal of Applied Research in Intellectual Disabilities, 16, 127‐
134. doi: 10.1046/j.1468‐3148.2003.00152.x
Kovacs, M., Beck, A. T., & Weissman, A. (1975). Hopelessness: An indicator of suicidal risk. Suicide, 5, 98‐103.
Lewis, J. L., Simcox, A. M., & Berry, D. T. R. (2002). Screening for feigned psychiatric symptoms in a forensic
sample by using the MMPI‐2 and the Structured Inventory of Malingered Symptomatology. Psycho‐
logical Assessment, 14, 170‐176. doi: 10.1037/1040‐3590.14.2.170
Long, J. D., Harring, J. R., Brekke, J. S., Test, M.A., & Greenberg, J. (2007). Longitudinal construct validity of
Brief Symptom Inventory subscales in schizophrenia. Psychological Assessment, 19, 298‐308. doi:
10.1037/1040‐3590.19.3.298
Lorr, M., McNair, D. M., Heuchert, JW. P., & Droppleman, L. F. (2003). POMS: profile of mood states. Toron‐
to, Canada: Multi‐Health Systems.
Matarazzo, J., Wiens, A., Matarazzo, R., & Goldstein, S. (1974). Psychometric and clinical test‐retest reliabili‐
ty of the Halstead Impairment Index in a sample of healthy, young, normal men. Journal of Nervous
and Mental Disease, 158, 37‐49. doi: 10.1097/00005053‐197401000‐00006
McCaffrey, R. J., Ortega, A., & Haase, R. F. (1993). Effects of repeated neuropsychological assessments. Arc‐
hives of Clinical Neuropsychology, 8, 519‐524. doi: 10.1016/0887‐6177(93)90052‐3
McNair, D. M., & Heuchert, J. W. P. (2006). POMS: Profile of Mood States, technical update. North Tonawan‐
da, NY: Multi‐Health Systems Inc.
McNair, D. M., Lorr, M., & Droppleman, L. F. (1971). Manual: Profile of Mood States. San Diego, CA: Educa‐
tional and Industrial Testing Service.
McNair, D. M., Lorr, M., & Droppleman, L. F. (1992). Manual: Profile of Mood States, Revised 1992. San Di‐
ego, CA: Educational and Industrial Testing Service.
Merckelbach, H., & Smith, G. P. (2003). Diagnostic accuracy of the structured inventory of malingered symp‐
tomatology (SIMS) in detecting instructed malingering. Archives of Clinical Neuropsychology, 18,
145‐152. doi: 10.1093/arclin/18.2.145
Metzger, R. L. (1976). A reliability and validity study of the State‐Trait Anxiety Inventory. Journal of Clinical
Psychology, 32, 276‐278. doi: 10.1002/1097‐4679(197604)32:2<276::AID‐
JCLP2270320215>3.0.CO;2‐G
Morey, L. C. (1991). Personality Assessment Inventory: Professional manual. Odessa, FL: Psychological As‐
sessment Resources, Inc.
Morey, L. C. (1997). Personality Assessment Screener: Professional manual. Lutz, FL: Psychological Assess‐
ment Resources, Inc.
142
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Morlan, K. K., & Tan, S. (1998). Comparison of the Brief Psychiatric Rating Scale and the Brief Symptom In‐
ventory. Journal of Clinical Psychology, 54, 885‐894. doi: 10.1002/(SICI)1097‐
4679(199811)54:7<885::AID‐JCLP3>3.0.CO;2‐E
Nicholls, T. L., Lee, Z., Corrado, R. R., & Ogloff, J. R. P. (2004). Women inmates’ mental health needs: Evi‐
dence of the validity of the Jail Screening Assessment Tool (JSAT). International Journal of Forensic
Mental Health, 3, 167‐184. Retrieved from http://www.iafmhs.org/iafmhs.asp
Nixon, G. F., & Steffeck, J. C. (1977). Reliability of the State‐Trait Anxiety Inventory. Psychological Reports,
40, 357‐358. Retrieved from http://ejournals.ammonsscientific.com/
Norcross, J. C., Guadagnoli, E., & Prochaska, J. O. (1984). Factor structure of the Profile of Mood States
(POMS): Two partial replications. Journal of Clinical Psychology, 40, 1270‐ 1277. doi: 10.1002/1097‐
4679(198409)40:5<1270::AID‐JCLP2270400526>3.0.CO;2‐7
Novy, D. M., Nelson, D. V., Goodwin, J., & Rowzee, R. D. (1993). Psychometric comparability of the State‐
Trait Anxiety Inventory for different ethnic subpopulations. Psychological Assessment, 5, 343‐349.
doi: 10.1037/1040‐3590.5.3.343
Nyenhuis, D. L., Yamamoto, C., Luchetta, T., Terrien, A., & Parmentier, A. (1999). Adult and geriatric norma‐
tive data and validation of the Profile of Mood States. Journal of Clinical Psychology, 55, 79‐86. doi:
10.1002/(SICI)1097‐4679(199901)55:1<79::AID‐JCLP8>3.0.CO;2‐7
Palmer, E. J., & Connelly, R. (2005). Depression, hopelessness and suicide ideation among vulnerable prison‐
ers. Criminal Behaviour and Mental Health, 15, 164‐170. doi: 10.1002/cbm.4
Periáñez, J. A., Ríos‐Lago, M., Rodríguez‐Sánchez, J. M., Adrover‐Roig, D., Sánchez‐Cubillo, I., Crespo‐Facorro,
B., Quemade, J. I., … (2007). Trail Making Test in traumatic brain injury, schizophrenia, and normal
ageing: Sample comparisons and normative data. Archives of Clinical Neuropsychology, 22, 433‐447.
doi: 10.1016/j.acn.2007.01.022
Perlick, D. A., Rosenheck, R. A., Clarkin, J. F., Sirey, J., & Raue, P. (1999). Symptoms predicting inpatient ser‐
vice use among patients with bipolar affective disorder. Psychiatric Services, 50, 806‐812. Retrieved
from http://ps.psychiatryonline.org/index.dtl
Piersma, H. L., Reaume, W. M., & Boes, J. L. (1994). The Brief Symptom Inventory (BSI) as an outcome meas‐
ure for adult psychiatric inpatients. Journal of Clinical Psychology, 50, 555‐563. doi: 10.1002/1097‐
4679(199407)50:4<555::AID‐JCLP2270500410>3.0.CO;2‐G
Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic brain damage. Perceptual
and Motor Skills, 8, 271‐276. doi: 10.2466/PMS.8.7.271‐276
Sánchez‐Cubillo, I., Periáñez, J. A., Adrover‐Roig, D., Rodriguez‐Sánchez, J. M., Rios‐Lago, M., Tirapu, J., &
Barceló, F. (2009). Construct validity of the Trail Making Test: Role of task‐switching, working memo‐
ry, inhibition/inference control, and visuomotor abilities. Journal of International Neuropsychologi‐
cal Society, 15, 438‐450. doi: 10.1017/S1355617709090626
APPENDICES
143
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Segal, S. P., & Silverman, C. (2002). Determinants of client outcomes in self‐help agencies. Psychiatric Servic‐
es, 53, 304‐309. doi: 10.1176/appi.ps.53.3.304
Sherrill‐Pattison, S., Donders, J., & Thompson, E. (2000). Influence of demographic variables on neuropsy‐
chological test performance after traumatic brain injury. The Clinical Neuropsychologist, 14, 496‐
503. doi: 10.1076/clin.14.4.496.7196
Spielberger, C., Gorsuch, R., & Lushene, R. (1970). The State‐Trait Anxiety Inventory: Test manual for Form X.
Palo Alto, CA: Consulting Psychologists Press.
Spielberger, C., Gorsuch, R., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). The State‐Trait Anxiety Inventory
for adults: Sampler set. Palo Alto, CA: Consulting Psychologists Press.
Steed, L. (2001). Further validity and reliability evidence for Beck Hopelessness Scale scores in a nonclinical
sample. Educational and Psychological Measurement, 61, 303‐316. doi:
10.1177/00131640121971121
Strauss, E., Sherman, E. M. S., & Spreen, O. (Eds.). A compendium of neuropsychological tests: Administra‐
tion, norms, and commentary (3rd ed., pp. 3‐28). New York, NY: Oxford University Press.
Tariq, S. H., Tumosa, N., Chibnall, J. T., Perry, M. H., & Morley, J. E. (2006). Comparison of the Saint Louis
University Mental Status Examination and the Mini‐Mental State Examination for detecting demen‐
tia and mild neurocognitive disorder—a pilot study. American Journal of Geriatric Psychiatry, 14,
900‐910. doi: 10.1097/01.JGP.0000221510.33817.86
Thomas, A., Donnell, A. J., & Young, T. R. (2004). Factor structure and differential validity of the Expanded
Brief Psychiatric Rating Scale. Assessment, 11, 177‐187. doi: 10.1177/1073191103262893
Tombaugh, T. N. (2004). Trail Making Test A and B: Normative data stratified by age and education. Archives
of Clinical Neuropsychology, 19, 203‐214. doi: 10.1016/S0887‐6177(03)00039‐8
Ventura, J., Lukoff, D., Nuechterlein, K. H., Liberman, R. P., Green, M. F., & Shaner, A. (1993). Brief Psychia‐
tric Rating Scale (BPRS) expanded version (4.0): Scales, anchor points, and administration manual.
International Journal of Methods in Psychiatric Research, 3, 227‐244.
Whitcomb, J. M. (2006). Psychometric Properties of the Coolidge Correctional Inventory (CCI) in a sample of
4872 prison inmates (Unpublished master’s thesis). University of Colorado at Colorado Springs, CO.
Widows, M. R., & Smith, G. P. (2005). Structured Inventory of Malingered Symptomatology: Professional ma‐
nual. Lutz, FL: Psychological Assessment Resources, Inc.
Zinger, I., Wichmann, C., & Andrews, D. A. (2001). The psychological effects of 60 days in administrative se‐
gregation. Canadian Journal of Criminology, 43, 47‐83. Retrieved from http://www.ccja‐
acjp.ca/en/cjc.html
144
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
APPENDIX C
PRISON SYMPTOM INVENTORY ANALYSES
The Prison Symptom Inventory (PSI) was developed by the researchers to assess potential responses to AS
confinement that were not covered by other measures. These variables were identified through examina‐
tion of the professional literature and include nervousness, headaches, lethargy, chronic tiredness, trouble
sleeping, a sense of impending breakdown, perspiring hands, heart palpitations, dizziness, nightmares,
trembling hands, and fainting. Additionally, items about exercise, grooming, and safety issues within AS
were included. The 39‐item inventory is given in Appendix A and information about the psychometric prop‐
erties of the scale are provided in Appendix B.
PSI items were grouped by the researchers into subscales which were thought to measure specific con‐
structs. Three of these subscales (Hypersensitivity to External Stimuli, Panic Disorder, and Physical Well‐
Being and Exercise) related to constructs assessed by the composites and were included in the composite
analyses reported in the main body of the report. In this appendix we provide results from the analyses
comparing the study groups on all of the PSI subscales to address the major hypotheses. Table C1 gives a list
of the subscales along with items and possible range of scores. Higher scores on the PSI subscales indicate
more negative behaviors except on the Attitudes about Segregation subscale where higher scores indicate a
preference for AS. Table C2 (a replication of Table B12) provides the summary statistics on each subscale for
each study group.
Table C1. PSI Subscales and Range of Possible Scores
Subscale
# Items
Items
Attitudes about Segregation
2
r14, 39
Fear Level
4
3, 12, 21, r38
Hypersensitivity to External Stimuli
5
1, 7, r31, r34, 37
Mental Well‐Being
2
26, r35
Mutism
2
22, r32
Panic Disorder
9
2, 6, 10, 13, 16, 17, 20, 25, 30
Physical Hygiene
5
4, r9, 18, r23, r29
Physical Well‐Being and Exercise
8
r5, 8, r11, r15, 19, 24, 27, r28
Safety
2
33, 36
Range of Possible Scores
0 – 10
0 – 20
0 – 25
0 – 10
0 – 10
0 – 45
0 – 25
0 – 40
0 – 10
Note. Item numbers with an r indicate that the item is reversed coded.
Table C2. Summary Statistics (M, SD, n) on PSI Scales by Group and Time
Assessment
CSP MI
CSP NMI
GP MI
GP NMI
Attitudes about Segregation
2.95 (3.19)
1.68 (2.57)
2.46 (3.00)
1.00 (2.10)
1
n = 57
n = 56
n = 28
n = 30
2.97 (3.54)
1.68 (2.63)
2.24 (2.76)
1.54 (2.42)
2
n = 61
n = 56
n = 25
n = 26
3.02 (3.36)
1.04 (2.39)
2.04 (2.30)
1.22 (1.60)
3
n = 55
n = 56
n = 25
n = 27
2.60 (3.21)
1.29 (2.43)
1.91 (2.45)
1.83 (2.58)
4
n = 55
n = 55
n = 22
n = 24
3.12 (3.51)
1.45 (2.32)
2.54 (2.67)
1.30 (1.98)
5
n = 52
n = 55
n = 22
n = 20
2.57 (3.45)
1.45 (2.32)
6
NA
NA
n = 49
n = 52
APPENDICES
SCCF
4.58 (3.47)
n = 67
5.55 (3.30)
n = 60
4.62 (3.45)
n = 55
4.96 (3.42)
n = 46
5.24 (3.57)
n = 45
NA
All
2.81 (3.24)
n = 238
3.09 (3.42)
n = 228
2.58 (3.18)
n = 218
2.61 (3.20)
n = 202
2.89 (3.32)
n = 194
1.92 (2.98)
n = 102
145
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Fear Level
CSP MI
CSP NMI
6.25 (4.62)
4.17 (3.46)
1
n = 64
n =63
5.50 (4.18)
3.51 (2.54)
2
n = 62
n = 59
5.71 (4.10)
3.91 (2.73)
3
n = 58
n = 57
5.22 (3.83)
3.91 (2.96)
4
n = 59
n = 56
5.50 (3.50)
3.87 (2.75)
5
n = 58
n = 56
5.43 (3.43)
4.28 (2.72)
6
n = 51
n = 54
Hypersensitivity to External Stimuli
10.54 (4.02)
9.62 (3.92)
1
n = 64
n =63
10.10 (4.52)
7.81 (3.86)
2
n = 62
n = 59
10.71 (4.65)
8.33 (4.11)
3
n = 58
n = 57
9.99 (4.64)
9.22 (4.49)
4
n = 58
n = 56
9.54 (4.09)
9.03 (4.09)
5
n = 56
n = 56
9.37 ( 4.33)
9.02 (3.59)
6
n = 51
n = 54
Mental Well‐Being
4.95 (2.48)
4.48 (2.48)
1
n = 64
n =63
4.88 (2.24)
3.69 (2.55)
2
n = 61
n = 59
4.33 (2.42)
3.98 (2.41)
3
n = 58
n = 57
4.71 (2.43)
3.96 (2.26)
4
n = 58
n = 56
4.13 (2.24)
3.59 (2.25)
5
n = 55
n = 56
4.02 (2.01)
3.68 (2.52)
6
n = 51
n = 54
Mutism
3.67 (2.19)
2.65 (1.70)
1
n = 64
n =63
4.44 (2.21)
2.98 (1.97)
2
n = 62
n = 59
4.14 (2.29)
2.98 (1.81)
3
n = 57
n = 57
4.41 (2.44)
2.96 (1.74)
4
n = 58
n = 56
4.09 (2.08)
3.00 (1.80)
5
n = 55
n = 55
4.00 (2.19)
2.75 (1.69)
6
n = 50
n = 52
146
GP MI
GP NMI
4.94 (3.78)
n = 33
4.53 (3.37)
n = 32
4.88 (3.40)
n = 32
4.91 (3.08)
n = 29
4.79 (3.21)
n = 29
3.51 (3.03)
n = 43
3.46 (2.60)
n = 41
3.53 (2.07)
n = 41
3.26 (2.53)
n = 39
3.39 (2.49)
n = 38
SCCF
7.63 (4.15)
n = 67
7.14 (4.96)
n = 64
7.66 (4.00)
n = 61
6.76 (4.09)
n = 59
6.81 (4.30)
n = 57
NA
NA
NA
11.00 (5.38)
n = 33
11.06 (3.83)
n = 32
11.34 (3.95)
n = 32
10.15 (3.78)
n = 29
10.65 (4.43)
n = 29
8.44 (3.70)
n = 43
8.20 (4.09)
n = 41
7.76 (4.13)
n = 41
8.00 (3.20)
n = 39
7.60 (3.62)
n = 38
9.61 (3.94)
n = 67
10.11 (3.96)
n = 64
9.72 (4.63)
n = 61
9.22 (4.49)
n = 59
9.03 (4.09)
n = 57
NA
NA
NA
5.39 (2.54)
n = 33
5.09 (2.61)
n = 32
4.97 (2.47)
n = 32
4.25 (2.78)
n = 28
4.66 (2.54)
n = 29
4.00 (2.43)
n = 43
3.24 (2.34)
n = 41
3.22 (2.31)
n = 41
2.77 (1.56)
n = 39
2.66 (2.29)
n = 38
5.19 (2.39)
n = 67
5.33 (2.53)
n = 63
5.44 (2.61)
n = 61
5.58 (2.44)
n = 59
5.21 (2.24)
n = 56
NA
NA
NA
3.61 (2.07)
n = 33
3.16 (1.87)
n = 32
3.19 (1.89)
n = 32
3.28 (1.74)
n = 28
3.21 (1.76)
n = 29
2.40 (1.50)
n = 43
2.20 (1.50)
n = 41
2.32 (1.56)
n = 41
2.59 (1.44)
n = 39
2.26 (1.60)
n = 38
3.81 (1.96)
n = 67
4.11 (1.72)
n = 64
4.44 (2.28)
n = 61
3.95 (2.05)
n = 57
3.60 (1.94)
n = 57
NA
NA
NA
All
5.51 (4.18)
n = 270
5.01 (4.00)
n = 258
5.31 (3.74)
n = 250
4.94 (3.63)
n = 241
5.00 (3.58)
n = 236
4.84 (3.11)
n = 106
9.82 (4.16)
n = 270
9.40 (4.22)
n = 258
9.52 (4.50)
n = 249
9.55 (4.10)
n = 241
9.32 (4.10)
n = 236
9.20 (3.93)
n = 106
4.80 (2.48)
n = 270
4.48 (2.56)
n = 256
4.42 (2.55)
n = 249
4.38 (2.48)
n = 240
4.08 (2.43)
n = 234
3.86 (2.27)
n = 106
3.26 (1.98)
n = 270
3.51 (2.04)
n = 258
3.52 (2.16)
n = 248
3.53 (2.06)
n = 238
3.31 (1.95)
n = 234
3.38 (2.04)
n = 103
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Assessment
Panic Disorder
1
2
3
4
5
6
CSP MI
9.09 (10.18)
n = 64
7.87 (8.62)
n = 62
8.76 (9.17)
n = 60
7.32 (7.72)
n = 60
6.50 (8.80)
n = 56
5.34 (7.25)
n = 51
CSP NMI
GP MI
GP NMI
3.89 (5.49)
n =63
3.71 (7.72)
n = 59
3.46 (4.20)
n = 57
3.79 (5.44)
n = 56
3.45 (4.34)
n = 56
3.60 (5.78)
n = 54
5.03 (4.88)
n = 33
5.43 (5.12)
n = 32
6.11 (5.02)
n = 32
4.99 (5.06)
n = 29
4.08 (5.13)
n = 29
2.77 (3.77)
n = 43
2.37 (3.45)
n = 41
3.00 (4.68)
n = 41
2.22 (3.28)
n = 39
1.82 (3.49)
n = 38
SCCF
10.98 (7.72)
n = 67
9.61 (8.70)
n = 64
9.55 (8.68)
n = 61
9.94 (8.68)
n = 59
8.07 (7.45)
n = 57
NA
NA
NA
4.00 (4.03)
n =63
4.06 (3.87)
n = 59
3.44 (3.59)
n = 57
3.25 (3.92)
n = 56
3.12 (3.57)
n = 56
3.18 (3.68)
n = 54
4.64 (2.69)
n = 33
3.66 (3.26)
n = 32
4.19 (4.10)
n = 32
3.21 (3.92)
n = 29
3.62 (3.70)
n = 29
2.07 (2.54)
n = 43
1.74 (2.69)
n = 41
2.17 (3.38)
n = 41
1.92 (2.67)
n = 39
1.08 (1.99)
n = 38
8.26 (4.64)
n = 67
7.59 (5.60)
n = 64
6.19 (4.55)
n = 61
5.61 (4.99)
n = 59
5.22 (5.07)
n = 57
NA
NA
NA
10.43 (5.55)
n =63
9.85 (6.23)
n = 59
10.30 (6.18)
n = 57
10.28 (6.14)
n = 56
9.54 (6.25)
n = 56
9.26 (6.72)
n = 54
15.79 (6.28)
n = 33
13.68 (7.37)
n = 32
13.84 (6.47)
n = 32
13.07 (5.92)
n = 29
13.58 (6.78)
n = 29
9.21 (5.61)
n = 43
7.26 (5.53)
n = 41
8.12 (5.02)
n = 41
7.44 (4.36)
n = 39
7.08 (4.19)
n = 38
18.93 (6.44)
n = 67
18.90 (6.01)
n = 64
18.98 (6.78)
n = 61
18.46 (6.67)
n = 59
17.60 (7.01)
n = 57
NA
NA
NA
Physical Hygiene
5.39 (5.01)
1
n = 64
5.80 (4.75)
2
n = 61
5.67 (4.80)
3
n = 58
5.48 (5.43)
4
n = 58
5.73 (5.01)
5
n = 56
4.90 (4.75)
6
n = 51
Physical Well‐Being and Exercise
15.89 (7.76)
1
n = 64
17.10 (7.70)
2
n = 62
17.14 (7.05)
3
n = 58
16.13 (7.47)
4
n = 58
15.39 (7.49)
5
n = 56
13.95 (7.09)
6
n = 51
All
6.84 (7.84)
n = 270
6.18 (7.62)
n = 258
6.47 (7.65)
n = 251
6.04 (7.19)
n = 243
5.10 (6.78)
n = 236
4.41 (6.54)
n = 106
5.16 (4.67)
n = 270
4.93 (4.77)
n = 257
4.52 (4.40)
n = 249
4.14 (4.64)
n = 241
3.98 (4.46)
n = 236
4.12 (4.41)
n = 106
14.29 (7.40)
n = 270
13.90 (7.91)
n = 258
14.12 (7.60)
n = 249
13.56 (7.48)
n = 241
12.97 (7.57)
n = 236
12.97 (7.57)
n = 106
The analyses in this section follows the same structure as in the main body of the report: (1) comparisons
between the two CSP groups are made on the six time periods; (2) comparisons between the two NMI
groups are completed on the five common time periods; and (3) comparisons between the three MI groups
are completed on the five common time periods. Mixed design analysis of variance (ANOVA) techniques are
used to compare group differences, change over time, and the interaction between groups and time. We
were most interested in whether there are significant interactions which would imply that groups are chang‐
ing differentially over time. The analyses in this section help address goals 2 and 3 of the project—to assess
whether offenders with mental illness decompensate differentially in AS compared to offenders without
mental illness and to compare psychological functioning of participants AS to relevant comparisons groups.
APPENDICES
147
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Comparisons between CSP Groups
Table C3 gives the results from the ANOVA analyses comparing mean change over time and mean differenc‐
es between the two AS groups. For each PSI subscale, the MI group scored significantly higher than the NMI
group although the strength of the difference varied over the subscales (.05 ≤ η2 ≤ .23). For the Physical
Well‐Being and Exercise and Mental Well‐Being subscales there were significant time effects, but not signifi‐
cant interaction effects. For both subscales, there was a general decrease over time with scores at the last
time period showing significantly lowered mean scores compared to scores at earlier assessment periods.
For the Panic Disorder and Hypersensitivity to External Stimuli subscales, there were significant interaction
effects. For the Panic Disorder subscale, there was a significant decrease over time for the MI group; how‐
ever, the NMI group did not change significantly over time. Figure C1 provides a graphical display of this in‐
teraction.
Table C3. F Statistics and Partial η2 Comparing AS Groups across 6 Time Periods
Subscale
Group Main Effect
Attitudes about
Segregation
F(1, 77) = 3.35, p = .07, η = .04
Fear Level
F(1, 97) = 17.82, p < .001, η = .16
Hypersensitivity to
External Stimuli
F(1, 96) = 8.11, p = .005, η = .08
Mental Well‐Being
F(1, 94) = 5.10, p = .03, η = .05
2
2
2
2
Time Main Effect
Interaction Effect
2
F(3.82, 294.12) = 1.93, p = .11, η = .02
2
F(4.36, 422.41) = 1.54, p = .18, η = .02
F(3.82, 294.12) = 0.63, p = .48, η = .01
F(4.36, 422.41) = 0.73, p = .58, η = .01
2
F(5, 480) = 2.65, p = .02, η = .03
2
F(4.70, 441.37) = 3.25, p = .01, η = .03
F(5, 460) = 1.93, p = .09, η = .02
2
F(4.00, 396.25) = 2.75, p = .03, η = .03
F(1, 92) = 17.80, p < .001, η = .16
Panic Disorder
F(1, 99) = 12.60, p = .001, η = .11
Physical Hygiene
F(1, 95) = 8.76, p = .004, η = .08
Physical Well‐Being
and Exercise
F(1, 96) = 27.30, p < .001, η = .22
2
2
2
2
F(5, 480) = 1.55, p = .17, η = .02
2
Mutism
2
2
2
F(4.70, 441.37) = 2.44, p = .53, η = .01
2
F(5, 460) = 0.26, p = .94, η = .003
2
2
2
F(4.00, 396.25) = 3.10, p = .02, η = .03
2
F(5, 475) = 1.84, p = .10, η = .02
F(5, 475) = 1.46, p = .20, η = .02
2
F(4.60, 441.67) = 2.45, p = .03, η = .02
2
F(4.60, 441.67) = 1.00, p = .42, η = .01
Figure C1. Mean Scores over Time for AS Groups on the PSI Panic Disorder Scale
45
40
35
30
25
CSP MI
20
CSP NMI
15
10
5
0
1
148
2
3
4
5
6
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
For the Hypersensitivity to External Stimuli Scale, the interaction is displayed in Figure C2. There was a signif‐
icant change over time for the NMI group but the MI group did not change significantly over time. For the
NMI group, there was a significant decrease in mean scores from time 1 to time 2 but scores were signifi‐
cantly higher at times 4, 5, and 6 than time 2.
Figure C2. Mean Scores over Time for AS Groups on the PSI Hypersensitivity to External Stimuli Scale
24
21
18
15
CSP MI
12
CSP NMI
"4%.
9
6
3
0
1
2
3
4
5
6
Comparisons between NMI groups
Comparisons were made between the CSP NMI and GP NMI groups on the five common time periods. Table
C4 provides the results from the mixed design ANOVA analyses comparing mean change over time and
mean differences between groups.
Table C4. F Statistics and Partial η2 Comparing NMI Groups across 5 Time Periods
Subscale
Group Main Effect
Attitudes about
Segregation
F(1, 55) = 0.01, p = .93, η < .001
Fear Level
F(1, 91) = 1.88, p = .17, η = .02
Hypersensitivity to
External Stimuli
F(1, 91) = 1.74, p = .19, η = .02
Mental Well‐Being
F(1, 91) = 4.32, p = .04, η = .04
Mutism
F(1, 90) = 5.76, p = .02, η = .06
Panic Disorder
Time Main Effect
Interaction Effect
2
F(4, 220) = 1.66, p = .16, η = .03
2
F(4, 364) = 0.28, p = .89, η = .003
2
F(4, 364) = 2.59, p = .04, η = .03
2
F(3.64, 330.90) = 3.97, p = .005, η = .04
2
F(4, 360) = 0.70, p = .59, η = .01
2
2
F(1, 91) = 4.01, p = .05, η = .04
2
Physical Hygiene
F(1, 91) = 9.47, p = .003, η = .09
Physical Well‐Being
and Exercise
F(1, 91) = 6.36, p = .01, η = .06
2
2
F(4, 220) = 1.01, p = .40, η = .02
2
2
F(4, 364) = 0.32, p = .85, η = .003
2
F(4, 364) = 1.01, p = .40, η = .01
2
2
2
F(3.52, 319.86) = 0.28, p = .87, η = .003
2
F(4, 364) = 2.20, p = .07, η = .02
2
2
F(3.84, 349.56) = 1.73, p = .14, η = .02
2
F(3.64, 330.90) = 1.42, p = .23, η = .02
2
F(4, 360) = 1.24, p = .29, η = .01
2
F(3.52, 319.86) = 0.59, p = .65, η = .01
2
F(4, 364) = 1.56, p = .18, η = .02
2
F(3.84, 349.56) = 0.90, p = .46, η = .01
The CSP NMI group had significantly higher mean scores than the GP NMI group for all PSI subscales except
Fear Level, Hypersensitivity to External Stimuli, and Attitudes about Segregation. There were significant
main effects of time for Hypersensitivity to External Stimuli and for Mental Well‐Being. Follow‐up tests for
changes in sequential time periods indicated that the first assessment period scores were higher than the
APPENDICES
149
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
second assessment period scores for both scales with significant time effects. There were no statistically sig‐
nificant interactions between groups and time implying that scores over time were similar in the groups.
Comparisons between MI groups
Comparisons were made between the CSP MI, GP NMI, and SCCF groups on the five common time periods.
Table C5 provides the results from the mixed design ANOVA analyses comparing mean change over time and
mean differences between groups.
Table C5. F Statistics and Partial η2 Comparing MI Groups across 5 Time Periods
Subscale
Group Main Effect
Time Main Effect
Attitudes about
Segregation
F(2, 91) = 12.56, p < .001, η = .22
Fear Level
F(2, 132) = 6.86, p = .001, η = .09
Hypersensitivity to
External Stimuli
F(1, 131) = 0.65, p = .52, η = .01
Mental Well‐Being
F(2, 126) = 3.46, p = .03, η = .05
Mutism
F(2, 127) = 2.11, p = .12, η = .03
Panic Disorder
Interaction Effect
2
2
F(3.60, 327.62) = 1.34, p = .26, η = .02
F(7.20, 327.62) = 1.12, p = .35, η = .02
2
F(3.72, 491.72) = 1.27, p = .28, η = .01
2
F(7.45, 491.72) = 0.49, p = .86, η = .01
2
F(4, 524) = 0.77, p = .55, η = .01
2
F(4, 504) = 2.06, p = .08, η = .02
2
F(4, 508) = 0.62, p = .65, η = .005
2
F(2, 135) = 4.65, p = .01, η = .06
2
Physical Hygiene
F(2, 130) = 4.05, p = .02, η = .06
Physical Well‐
Being and Exercise
F(2, 131) = 5.73, p = .004, η = .08
2
2
2
2
F(4, 524) = 0.58, p = .80, η = .01
2
F(8, 504) = 1.03, p = .41, η = .02
2
2
2
2
F(8, 508) = 1.32, p = .23, η = .02
2
F(7.24, 489.00) = 0.69, p = .69, η = .01
2
F(7.82, 508.46) = 2.87, p = .004, η = .04
2
F(7.41, 485.27) = 1.02, p = .42, η = .02
F(3.62, 489.00) = 3.33, p = .01, η = .02
F(3.91, 508.46) = 3.13, p = .02, η = .02
F(3.70, 485.27) = 2.42, p = .05, η = .02
2
2
2
There were significant group differences on all of the subscales except Hypersensitivity to External Stimuli
and Mutism. For the Fear Level and Attitudes toward Segregation subscales, the SCCF group scored signifi‐
cantly higher than the other two groups. For the Panic Disorder, Physical Hygiene, Physical Well‐Being and
Exercise subscales, the GP MI group scored significantly lower than the other two groups. For the Mental
Well‐Being subscale, the SCCF group scored significantly higher than the CSP MI group but there were no
other significant differences.
The Panic Disorder and Physical Well‐Being and Exercise subscales showed statistically significant changes
over time; however for both variables the changes showed improvement over time. Just one of the subs‐
cales showed a statistically significant time effect and interaction effect—Physical Hygiene. For this subscale,
the SCCF group showed significant decreases across time (i.e., improved hygiene over time) whereas the CSP
MI and GP MI groups did not show significant change over time.
SUMMARY
Although there were statistically significant findings, the results did not support the hypotheses of the study.
We expected that there would be a worsening over time in reported behavior/sensations and that this
change would be worse for inmates with mental illness in AS. However, we found that when significant
changes over time occurred, they tended to be in the direction of improvement and this improvement
tended to occur more frequently for inmates with mental illness. When making comparisons of the AS
groups to the relevant comparison groups, there was no indication that the segregation groups behavior and
attitudes declined over time in comparison to the non‐segregated groups.
150
APPENDICES
This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.
Ps
APPENDICES
151
