EPIDEMIOLOGY AND PREVENTION

Undiagnosed HIV Infection Among New York City Jail
Entrants, 2006: Results of a Blinded Serosurvey
Elizabeth M. Begier, MD, MPH, Yussef Bennani, MPH, Lisa Forgione, MA, Amado Punsalang, PhD,
David B. Hanna, MS, Jeffrey Herrera, BA, Lucia Torian, PhD, Maria Gbur, MD,
Kent A. Sepkowitz, MD, and Farah Parvez, MD, MPH

Objective: Since 2004, when all New York City jail entrants began

Key Words: HIV infections/diagnosis, HIV infections/epidemiology,
prevalence, prisons, prisoners

being offered rapid testing at medical intake, HIV testing has increased
4-fold. To guide further service improvement, we determined HIV
prevalence among jail entrants, including proportion undiagnosed.

(J Acquir Immune Defic Syndr 2010;54:93–101)

Methods: Remnant serum from routine syphilis screening was
salvaged for blinded HIV testing in 2006. Using HIV surveillance
data and electronic clinical data, we ascertained previously diagnosed
HIV infections before permanently removing identifiers. We defined
‘‘undiagnosed’’ as HIV-infected entrants who were unreported to
surveillance and denied HIV infection.
Results: Among the 6411 jail entrants tested (68.9% of admissions),
HIV prevalence was 5.2% overall (males 4.7%; females: 9.8%).
Adjusting for those not in the serosurvey, estimated seroprevalence is
8.7% overall (6.5% males, 14% females). Overall, 28.1% of HIV
infections identified in the serosurvey were undiagnosed at jail entry;
only 11.5% of these were diagnosed during routine jail testing. Few
(11.1%) of the undiagnosed inmates reported injection drug use or
being men who have sex with men.
Conclusions: About 5%–9% of New York City jail entrants are HIV
infected. Of the infected, 28% are undiagnosed; most of whom denied
recognized HIV risk factors. To increase inmate’s acceptance of routine
testing, we are working to eliminate the required separate written
consent for HIV testing to allow implementation of the Centers for
Disease Control and Prevention–recommended opt out testing model.

Received for publication May 12, 2009; accepted October 21, 2009.
From the New York City Department of Health and Mental Hygiene, New
York, NY. Dr Farah Parvez was also associated with the National Center
for HIV/AIDS, Viral Hepatitis, sexually transmitted disease, and TB
Prevention, Centers for Disease Control and Prevention.
Supported in part by cooperative agreement 5U62PS001026-02 with the
Centers for Disease Control and Prevention.
Presented in part at the 15th Conference on Retroviruses and Opportunistic
Infections, February 2008, Boston, MA. Abstract #: V-203.
F.P., L.T., E.B., and K.A.S. contributed to conceptualization of the study. Y.B.,
L.F., D.H., and J.H. contributed to data management. Y.B., L.F., L.T., and
A.P. contributed to laboratory testing and oversight. Y.B., L.F., and D.B.H.
conducted the data analysis. E.M.B., Y.B., and M.G. contributed to
drafting the article. All authors contributed to the editing of the article.
Correspondence to: Elizabeth M. Begier, MD, MPH, Director for HIV
Epidemiology and Field Services, New York City Department of Health
and Mental Hygiene, 346 Broadway, Room 707, New York, NY 10013
(e-mail: ebegier@health.nyc.gov; bbegier@gmail.com).
Copyright Ó 2010 by Lippincott Williams & Wilkins

INTRODUCTION
Identification of undiagnosed persons with HIV infection
is a cornerstone of the Centers for Disease Control and
Prevention’s (CDC) approach to controlling the HIV epidemic in
the United States1–2 based on evidence that those who are aware
of their HIV status reduce transmission risk behaviors by about
half compared with the undiagnosed.3 Although CDC has only
recently advocated for routine testing in all medical settings,1
they have called for routine HIV testing in correctional settings
for over a decade because of the substantially higher HIV prevalence among inmates compared with the nonincarcerated.4–5
In New York City, elevated HIV seroprevalence in the
large city-run jail system has been well characterized through
blinded HIV serosurveys,6–7 that is, studies that involve HIV
testing of deidentified remnant serum drawn for routine clinical
purposes, such as syphilis testing. In 6 serosurveys during
1989–1998, seroprevalence in New York City jails decreased in
males from a high of 16.2% in 1989 to 7.6% in 1998 and from
25.1% in 1989 to 18.1% in 1998 among females.6–7 However,
the proportion of HIV-infected inmates who were unaware of
their serostatus has never been assessed. CDC estimates that
21% of persons with HIV are undiagnosed nationwide.8 In
New York City, population-based estimates of the proportion of
HIV-infected persons who are undiagnosed have ranged from
5%, with a wide 95% confidence interval (CI) of 0.7%–29.9%
from a 2006 blinded serosurvey of noninstitutionalized New
Yorkers,9 to 12%–29% based on calculated estimates of the
undiagnosed using census, research, and surveillance data.10
Since the last jail serosurvey in 1998, named HIV
reporting has been implemented and New York City’s jail HIV
testing program has been dramatically expanded. In March
2004, New York City’s Department of Health and Mental
Hygiene (DOHMH) began to routinely offer voluntary rapid
HIV testing to all city jail entrants during medical intake.
Because of the New York State requirements, a true opt out
testing model1,5 was not introduced because a separate written
consent was and is still necessary for each HIV test. Annual
jail testing volume increased from 6500 tests in 2003 to 25,000
tests in 2006. Rapid testing technology also allowed more
inmates to get their results before release.

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

www.jaids.com |

93

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

Begier et al

To guide further service improvement, DOHMH conducted a blinded serosurvey of New York City jail entrants to
determine current HIV prevalence and to estimate the proportion of HIV-infected jail entrants who have not previously
been diagnosed.

METHODS
Study Population
New York City Department of Correction operates 11
correctional facilities/jails, including 9 Rikers Island facilities
and 2 borough houses of detention. During 2006, there were
93,327 male admissions among 64,383 unique individuals and
11,896 female admissions among 8073 unique individuals.
Average daily jail census was 13,000–14,000 inmates. Median
length of stay was 7 days; 25% were released within 3 days
(Dr. Farah Parvez, personal communication, June 2008).
DOHMH’s Bureau of Correctional Health Services (CHS)
coordinates all medical, mental health, and dental services for
New York City jail inmates, including routine, voluntary
health screening at intake. During the health screening referred
to as ‘‘medical intake’’, which a part of the processing before
housing inmates, nursing staff verbally offers HIV testing at
the beginning of intake process. They document in the
electronic medical record whether or not the patient consents
to testing and have the patient sign the required written consent
if they do consent. The test is conducted by nursing staff or
a patient care associate. Results are provided during medical
intake before the inmate moves on to the housing area.
Negative results are provided by nursing staff. Positive results
are provided by physicians. Syphilis testing is conducted on all
inmates by the jail-based health care personnel conducting the
medical intake process. This serosurvey was conducted using
remnant serum specimens drawn for this routine universal
syphilis testing.
All consecutive new admissions to New York City jails
beginning in May 1, 2006, were eligible for inclusion in the
sample. Sample targets were 4411 unique persons for men and
1791 for women (6202 overall). Sample size calculations were
done separately for men and women to ensure adequate power
to assess predictors within each sex, thus targets for women
represent a greater proportion of the study sample than of all
jail inmates. To reach target sample sizes, men’s study period
lasted through June 1, 2006 (31 days), and for women through
August 13, 2006 (104 days).

Definitions
‘‘Serosurvey testing’’ refers to blinded HIV testing
conducted for this study on all inmates using remnant serum
collected for routine syphilis testing. ‘‘Jail testing’’ refers to
routine, voluntary HIV testing as part of medical care that
inmates receive during their jail stay. ‘‘Diagnosed’’ HIV infection refers to persons with HIV-positive specimens during
serosurvey testing and previous evidence of HIV diagnosis
either from self-report at intake or from information in
New York City’s population-based HIV surveillance registry
[HIV/AIDS Reporting System (HARS)]. ‘‘Undiagnosed’’ HIV
infection refers to persons with HIV-positive specimens during
serosurvey testing but no previous evidence of HIV diagnosis.

94

| www.jaids.com

Sample Selection and Data Management
During the specimen collection period, New York City’s
Department of Correction provided DOHMH with a daily list
of newly admitted jail inmates, including name, date of birth,
sex, race, address, and admission facility. These records
provided the basis for identifying eligible persons for the
serosurvey and were linked to electronic data on remnant
serum specimens provided from the syphilis-testing laboratory. This process established which admissions had specimens drawn for syphilis testing and which specimens had been
exhausted during syphilis testing. Deduplication was conducted to ensure that all specimens were from unique
individuals. If repeat specimens were available, an individual’s
earliest specimen was used.
We obtained electronic medical records for intake exams
for all new jail admissions during the specimen collection
period. These records contained medical history data, including self-reported HIV infection status and other disease
history, medication, and treatment history (including history of
treatment for mental health disorders), and selected HIV risk
factors. This information was collected through a standardized
medical intake questionnaire in which jail-based health care
workers completed data elements based on their interview with
the patient. Regarding drug use, the inmates were asked: ‘‘Do
you use drugs?’’ and fields were provided to record which
drugs: barbiturates, marijuana, crystal meth, crack, methadone, heroin, cocaine, other; and how much of each drug used.
This intake also included a specific question regarding
a history of violence: ‘‘Have you ever been charged with
a violent act (rape, assault)?’’ Notably, this system documented
whether inmates consented to HIV testing at intake but not
whether they were actually tested. To determine which new
admissions had tested positive for HIV through jail HIV
testing, names of inmates testing positive during the specimen
collection period were obtained through routine HIV reporting
and linked with new admissions records.
Transmission risk categories were assigned based on the
following risk factors: a history of injection drug use (IDU),
men who have sex with men (MSM), and high-risk
heterosexual behavior. For this analysis, persons reporting
sex only with someone of the opposite sex who were also
noted in the medical record to have reported ‘‘multiple
partners’’, ‘‘unprotected sex’’, or a history of sexually
transmitted disease diagnosis were classified as ‘‘high-risk’’
heterosexual. Men with both IDU and MSM risks were
classified as injection drug users (IDUs). Otherwise, when
more than 1 risk factor was reported, persons were classified
using the CDC hierarchy of transmission categories.11 Persons
reporting none of the identified risks were reported classified
as ‘‘no identified risk.’’ Occupational exposure and perinatal
exposure were not assessed but would be expected to be very
rare in this population as they are in the general population.

Surveillance Registry Matching
Records for all new jail admissions were temporarily
linked with HARS. Linkages were performed using a 32match key hierarchy, consisting of components of first name,
last name, date of birth, and when available, Social Security
number. Exact matches of first name, last name, and date of
q 2010 Lippincott Williams & Wilkins

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

birth were automatically considered as true matches. All other
potential matches were manually reviewed by 2 independent
reviewers to determine if there was a true match. If 2 reviewers
disagreed, a third made the final match determination.
The 6% of jail admissions with a non–New York City zip
codes were included in the surveillance linkage process
because HIV surveillance unit in New York City investigates
and confirms all reports from New York City providers
regardless of the residence of the reported patient (more than
7% of confirmed cases have a non–New York City residence),
and most non–New York City jail admits (76%) were from
adjoining states.

Deidentification of Data
After all data sources were obtained and combined, all
new admissions in the sample were assigned a random
identification number (ID). All other identifying information
was deleted. Specimens were stripped of their previous
laboratory accession number and relabeled with the corresponding study ID. All electronic and hardcopy documents
linking the study ID to any patient identifiers were destroyed,
leaving the study ID as a unique study identifier now unlinked
to any personal identifying information. After this point, there
was no way that investigators could relink inmates’ identifying
information with their subsequent serosurvey HIV test results.
Blinded serosurveys are specifically allowed by law in New
York State, and this study’s protocol was approved by the
DOHMH IRB (# 05-082) as exempt research.

Laboratory Testing
Once data and specimens were delinked from identifying data, specimens were HIV tested using Bio-Rad HIV-1/
HIV-2 EIA plus ‘‘O’’ (Bio-Rad Laboratories, Hercules, CA),
with reactive tests confirmed using Bio-Rad HIV-1 Western
blot. HIV results were added to each corresponding record
using the study ID to link records to HIV test results.

Data Analysis
We compared the distribution of demographic and HIV
status variables between inmates who did or did not have
specimens available for serosurvey testing using x2 analyses.
HIV prevalence and 95% CIs were calculated both overall and
for all demographic and risk strata after weighting by sex to
account for female inmate oversampling. Weights were
generated by dividing the sex ratio of all unique individuals
in the city jail system in 2006 by the sex ratio of those who
were included in this analysis. Prevalence calculation
denominators included any inmate with positive or negative
HIV serosurvey test results and excluded persons with
indeterminate results. Unadjusted odds ratios with 95% CIs
were calculated for all strata, based on logistic regression
models weighted by sex. All available descriptive variables
described above were included in univariate analyses. Factors
significant at the P , 0.20 level in univariate testing were
considered for inclusion in a multivariate logistic regression
model using backward elimination to determine predictors of
being HIV infected.
Using the same weighted modeling approach described
above, we examined predictors of being HIV infected, but
q 2010 Lippincott Williams & Wilkins

New York City Jail Entrant HIV Serosurvey

undiagnosed, among all inmates testing positive through
serosurvey testing. In addition, to evaluate if a targeted
approach to reoffering HIV testing after intake could be
utilized to identify a substantial portion of the undiagnosed, we
examined predictors of being HIV infected among all inmates
who presented without evidence of HIV diagnosis (ie,
‘‘undiagnosed’’ HIV-infected inmates versus uninfected based
on HIV serosurvey testing). To adjust for eligible inmates
exclusion because of lack of remnant serum, true seroprevalence estimates for the entire jail entrant population were
calculated by multiplying measured prevalence by the likelihood of being in HARS for all inmates divided by that in the
serosurvey sample [ie, (proportion matching to HARS among
all inmates/proportion matching to HARS among inmates with
specimen for serosurvey testing) 3 (measured serosurvey
seroprevalence)].
All analyses were initially conducted stratified by
gender; but because stratified results did not identify any
additional risk factors compared with combined results
controlling for gender, we present combined overall results
only. All analyses were performed using SAS 9.1 (SAS
Institute, Cary, NC).

RESULTS
Overall Serosurvey Results
Of 10,297 inmates admitted during specimen collection,
9405 (91.1%) had record of a medical intake exam, and of
these, 6411 (68.9%) had adequate remnant serum volume
from routine syphilis testing for HIV serosurvey testing
(Fig. 1). The most common reasons specimens were unavailable were (1) specimen drawn but exhausted during syphilis
testing (51%), (2) inmate had medical intake screening but no
sample drawn (24%), and (3) inmate admitted to jail system
but either discharged or not available for medical intake (eg,
left for court hearing) (23%).
Of 6411 specimens that were tested in the serosurvey,
389 (5.2%) tested HIV positive, 5977 (94.1%) tested HIV
negative, and 45 had indeterminate results (0.7%). Only 7 of
5977 inmates who tested negative (0.11%) self-reported being
HIV infected at intake. Among the 389 HIV-infected inmates,
232 (59.5%) were in HARS at jail admission. Of these, 169
(72.4%) self-reported as HIV infected and 63 (27.6%) did not.
Among the 157 (40.5%) HIV-infected inmates not in HARS,
53 (32.4%) self-reported as HIV positive, and likely largely
represent persons who were diagnosed but unreported to
HARS. The remaining 104 of those not in HARS (28.1% of all
HIV-positive inmates) did not self-report being HIV infected;
we considered this group to have undiagnosed HIV infection
at intake.

Comparison of Tested and Nontested
New Admissions
Admissions with a specimen available (n = 6411)
differed significantly with regard to race, age, and HIV risk
factors (all P , 0.001) compared with new admissions without
a specimen (n = 2994) (Table 1). Compared with those in the
serosurvey sample, new admissions without a specimen were
www.jaids.com |

95

Begier et al

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

FIGURE 1. 2006 New York City jail serosurvey specimen selection and results*. *Percentages weighted by sex based on study
design. †46 of these 892 had specimens available for testing but no medical intake data, making a total of 6,457 with remnant
serum available of the 10,297.

less likely to be black (55.3% vs. 58.2%, P , 0.0001); tended
to be older (median age range: 35–39 years vs. 30–34 years,
P , 0.0001); and were more likely to be IDUs (11.9% vs.
6.1%, P , 0.0001). Importantly, new admissions without
a specimen were more likely to self-report being HIV infected
(10.6% vs. 3.6%, P , 0.001) and be in HARS (9.3%, vs.
3.9%, P , 0.001). Only 1.0% of males reported a history of
sex with another man, but this proportion was similar for those
with or without a specimen (0.7% vs. 1.0%).

(95% CI: 12.7% to 15.3%). Multivariate logistic regression
found that female and black inmates, and those with medical
insurance, were significantly more likely (P , 0.05) to be HIV
infected than others in the serosurvey. Those reporting history
of IDU, MSM sexual activity, syphilis, previous incarceration,
a prior HIV test, and hepatitis C were also more likely (P ,
0.0001) to be HIV infected. Odds of being HIV infected
increased with age (P , 0.0001).

‘‘Undiagnosed’’ HIV Infection
HIV Prevalence Based on Serosurvey Testing
HIV prevalence for all inmates in the serosurvey was
5.2%: 4.7% in men and 9.8% in women (Table 2). Adjusting
our serosurvey prevalence to reflect all new admissions (those
with and without remnant serum) based on proportion
matching to HARS, the estimated true New York City inmate
prevalence would be 8.7% overall (95% CI: 8.1% to 9.2%):
6.5% in males (95% CI: 6.0% to 7.1%) and 14% in females

96

| www.jaids.com

Of the 389 persons who were identified as HIV infected
by serosurvey testing, 104 (28.1%) appeared to have been not
diagnosed at the time of their admission. Undiagnosed inmates
were significantly less likely than other HIV-infected jail
entrants to self-report previous HIV test [adjusted odds ratio
(AOR) = 0.19; 95% CI: 0.08 to 0.45], history of mental health
treatment (AOR = 0.25; 95% CI: 0.10 to 0.64), hepatitis C
infection (AOR = 0.06; 95% CI: 0.01 to 0.47), and MSM
q 2010 Lippincott Williams & Wilkins

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

TABLE 1. Characteristics of Eligible New York City Jail
Entrants in 2006 Serosurvey, by Availability of Specimens for
Serosurvey HIV Testing (n = 9405)
Admissions
With
Remnant
Specimen
Available
Total
Sex
Male
Female
Race/ethnicity
Black
Hispanic
White
Other/unknown
Age group in yrs
16–19
20–29
30–39
40–49
50–59
60+
Transmission risk factor
MSM
Injection drug use history
High-risk heterosexual
No identified risk
Self-reported HIV status
Positive
Negative or unknown
Matched to HIV registry
Yes
No

Admissions
Without
Remnant
Specimen
Available
P†

n

%*

n

%*

6411

100.0

2994

100.0%

4669
1742

72.8
27.2

2066
928

69.0
31.0

—
—

3730
1997
501
183

58.2
31.1
7.8
2.9

1656
917
311
110

55.3
30.6
10.4
3.7

,0.0001

733
2020
1632
1615
340
71

11.4
31.5
25.5
25.2
5.3
1.1

184
826
801
883
266
34

6.1
27.6
26.8
29.5
8.9
1.1

,0.0001

47
390
2924
3050

0.7
6.1
45.6
47.6

31
356
1159
1448

1.0
11.9
38.7
48.4

0.13
,0.0001
,0.0001
0.48

229
6182

3.6
96.4

317
2677

10.6
89.4

,0.0001

253
6158

3.9
96.1

277
2717

9.3
90.7

,0.0001

0.0001

(95% CI: 552 to 934) would potentially remain undiagnosed if
testing conditions remain similar to the time of the serosurvey.

Should We Target Testing to Specific Entrants?
Just 13 (11.1%) of the 104 undiagnosed reported the
well-established HIV risk factors of MSM or IDU activity and
only an additional 41 (39.2%) reported a sexually transmitted
disease history, unprotected sex, and/or multiple sex partners.
We examined predictors of HIV infection among jail inmates
who had no evidence of a previous HIV diagnosis at admission
(n = 6126; ie, all HIV-negative entrants plus 104 undiagnosed
HIV-positive entrants) (Table 4). Women (AOR = 1.7; 95% CI:
1.0 to 3.0) and self-reported MSM (AOR = 5.2; 95% CI: 1.7 to
15.9) were more likely to be HIV infected (Table 4). Younger
persons were less likely to be HIV infected (16–29 compared
with 40–49 years: AOR = 0.55; 95% CI: 0.32 to 0.92).
However, only 31.2% of undiagnosed inmates had any of these
3 characteristics (ie, female sex, MSM, or age 40–49 years).
Notably, those who were HIV infected were equally likely to
consent to testing as those who were not HIV infected (30.2%
vs. 31.8%, P = 0.75).

DISCUSSION

*Weighted by sex based on study design.
†The P values for all comparisons based on x2 analysis.

sexual activity (AOR = 0.18; 95% CI: 0.04 to 0.82) (Table 3).
Undiagnosed inmates were also younger than the diagnosed
(AOR = 11.1 for 16–29 vs. 40–49 age group; 95% CI: 4.5
to 27.2).

Detection of Undiagnosed HIV Through
Jail Testing
Of the 104 persons who appeared not to have been
previously diagnosed, 32 (30.2%) consented to HIV testing at
intake based on the electronic intake record. Of these, 12 were
diagnosed by the serosurvey’s end. Thus, of 104 previously
undiagnosed, 12 (11.5%) were newly diagnosed by routine jail
testing by the end of the serosurvey, independent of the
serosurvey. Assuming the sample period is representative of
the entire calendar year, we estimate that 820 persons (95% CI:
619 to 1021) enter the New York City jail system each year
with a previously undiagnosed HIV infection. Of these, 743
q 2010 Lippincott Williams & Wilkins

New York City Jail Entrant HIV Serosurvey

HIV prevalence among New York City jail entrants
remains markedly elevated above the general population
(;2.5–3.5 times as high for men and ;14–20 times as high for
women) but appears to have decreased nearly by half since the
last serosurvey in 1998. Importantly, .25% of HIV-infected
jail entrants appear undiagnosed at admission. To our
knowledge, this is the first jail serosurvey to estimate the
proportion of HIV infections that are undiagnosed by linking
to serosurvey results with HIV surveillance and electronic
medical record data. Despite a 4-fold increase in jail testing,
most undiagnosed infections still are not identified through
routine, voluntary jail testing, largely because of the low
acceptance of HIV testing (;30% consenting). Despite the
fact that IDU and MSM continue to be at higher HIV risk than
others, most undiagnosed (;90%) reported neither of these
recognized HIV risk factor. This underscores the importance
of increasing the proportion of inmates tested through the New
York City jails’ routine testing program rather than relying on
a targeted testing approach based on self-reported risk factors.
Implementation of a true opt out model without a separate
written consent for each HIV test1,5 when this becomes legally
possible in New York State would likely result in a larger
proportion of inmates learning their HIV status.
Compared with prior New York City jail serosurveys,
HIV seroprevalence decreased in males from 7.6% in 1998 to
4.7% in 2006 and from 18.1% in 1998 to 9.8% among
females. This likely, in part, mirrors the decrease in new
diagnoses seen citywide during this period, particularly among
intravenous drug users12 and among correctional populations
nationwide.13 Despite decreases, jail prevalence remains
markedly elevated above estimates for New York City’s
general population (1.9% for men and 0.7% for women14),
particularly for women. This higher prevalence, which has
been seen in other correctional serosurveys in the United Sates
and elsewhere,15 is likely attributable to a higher prevalence of
www.jaids.com |

97

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

Begier et al

TABLE 2. Predictors of Testing Positive for HIV Infection Among All New York City Jail Entrants in 2006 Serosurvey (n = 6411)
Univariate Logistic Regression
Total
HIV-Infected % HIV
Admissions (N) Admissions (n) Positive* Odds Ratio
95% CI
P
Total
Sex
Male
Female
Race/ethnicity‡
Black
Hispanic
White
Other/unknown
Age group in yrs§
16–19
20–29
30–39
40–49
50–59
60+
Transmission risk factor
MSM
Injection drug use history
High-risk heterosexual
No identified risk
History of previous HIV test
Yes
No
History of syphilis
Yes
No
History of incarceration
Yes
No
Current medical insurance
Yes
No
History of marijuana use
Yes
No
History of hepatitis C
Yes
No

Multivariate Logistic Regression
95% CI

P

6411

389

5.2

—

—

—

—

—

—

4669
1742

218
171

4.7
9.8

1.00
2.22

Reference
(1.69 to 2.93)

—
,.0001

1.00
1.98

Reference
(1.47 to 2.68)

—
,0.0001

3730
1997
501
183

249
111
26
3

5.7
4.9
4.4
1.6

1.63
1.38
1.00
—

(1.05 to 2.52)
(0.87 to 2.18)
Reference
—

0.03
0.17
—
—

1.84
1.38
1.00
—

(1.13 to 3.00)
(0.83 to 2.29)
Reference
—

0.01
0.21
—
—

733
2020
1632
1615
340
71

6
44
127
164
41
7

0.4
1.9
6.5
9.1
12.4
8.7

—
0.16
0.70
1.00
1.33
—

—
(0.11 to 0.22)
(0.54 to 0.92)
Reference
(0.94 to 1.88)
—

—
,0.0001
0.01
—
0.10
—

—
0.24
0.76
1.00
1.25
—

—
(0.17 to 0.35)
(0.57 to 1.00)
Reference
(0.86 to 1.81)
—

—
,0.0001
0.05
—
0.24
—

47
390
2924
3050

16
73
158
142

34.0
16.9
4.6
4.0

11.97
4.70
1.16
1.00

(7.05 to 20.33) ,0.0001
(3.34 to 6.63) ,0.0001
(0.91 to 1.49)
0.24
Reference
—

9.28
1.97
1.03
1.00

(5.10 to 16.89) ,0.0001
(1.30 to 2.98)
0.00
(0.78 to 1.35)
0.86
Reference
—

4626
1785

351
38

6.6
1.8

3.77
1.00

(2.61 to 5.43)
Reference

,0.0001
—

2.70
1.00

(1.85 to 3.94)
Reference

,0.0001
—

340
6071

66
323

18.3
4.7

4.59
1.00

(3.30 to 6.38)
Reference

,0.0001
—

1.87
1.00

(1.27 to 2.75)
Reference

0.00
—

4772
1639

349
40

6.2
2.2

2.98
1.00

(2.08 to 4.27)
Reference

,0.0001
—

1.63
1.00

(1.11 to 2.39)
Reference

0.01
—

2564
3847

210
179

7.3
4.0

1.90
1.00

(1.53 to 2.37)
Reference

,0.0001
—

1.44
1.00

(1.14 to 1.83)
Reference

0.00
—

803
5608

24
365

2.4
5.7

0.41
1.00

(0.26 to 0.64)
Reference

,0.0001
—

0.60
1.00

(0.37 to 0.96)
Reference

0.03
—

257
6154

66
323

22.7
4.6

6.12
1.00

(4.40 to 8.49)
Reference

,0.0001
—

2.32
1.00

(1.55 to 3.47)
Reference

,0.0001
—

Odds Ratio

*Weighted by sex based on study design.
†Also adjusting for borough of residence.
‡Other/unknown race grouped with white in regression analyses.
§For age, 16–19 and 20–29 grouped as 16–29, and 50–59, and 60+ grouped as 50+, in regression analyses.

risk factors among inmates, including intravenous drug use
and sexually transmitted infections.15
As has been seen in other correctional serosurveys,16
women were twice as likely to be HIV infected, but here we are
able to document that, if infected, they were slightly less likely
to be undiagnosed (23% undiagnosed vs. 29% for men), likely
because of widespread prenatal testing. Women’s disproportionately higher prevalence in our population is likely because
a higher proportion of female inmates than men are jailed on
drug-related crimes (DOHMH corrections report, Farah
Parvez, MD, MPH; NYC Department of Health and Mental

98

| www.jaids.com

Hygiene, unpublished data, 2010), placing them at higher HIV
risk from either IDU or exchange sex (ie, trading sex for
drugs/money) performed to support their drug use. Along
these lines, a history of drug use was the only significant
predictor of HIV infection among female jail entrants without
an HIV diagnosis at entry. Also, sexual networks and
partnership dynamics have been shown to affect HIV
risk.16–17 These factors may be playing a role here, but we
are not able to explore their influence because this information
was not among that available to us from the electronic medical
record or other data sources.
q 2010 Lippincott Williams & Wilkins

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

New York City Jail Entrant HIV Serosurvey

TABLE 3. Predictors of Being Undiagnosed* Among HIV-Infected New York City Jail Entrants in 2006 Serosurvey (n = 389)
% of HIV-Infected
Admissions that
Univariate Logistic
Total
Undiagnosed
Regression
HIV-infected HIV-infected were Undiagnosed†
Admissions
Admissions
Odds
(N)
(n)
Column % Row % Ratio
95% CI
P
Total
Sex
Male
Female
Age group in yrs‡
16–19
20–29
30–39
40–49
50–59
60+
Transmission risk factor
MSM
Injection drug use history
High-risk heterosexual
No identified risk
History of previous HIV test
Yes
No
History of mental health treatment
Yes
No
History of hepatitis C
Yes
No

Multivariate
Logistic
Regression
Odds
Ratio

95% CI

P

—

—

—

389

104

100.0

28.1

—

—

—

218
171

64
40

82.7
17.3

29.4
23.4

1.00
0.74

Reference
(0.40 to 1.36)

—
0.32

6
44
127
164
41
7

2
29
34
33
5
1

1.7
29.7
31.9
30.6
4.7
1.3

49.9
69.0
29.0
20.9
10.5
17.6

—
7.90
1.55
1.00
0.49
—

—
—
—
—
—
(3.72 to 16.76) ,0.0001 11.06 (4.51 to 27.17) ,0.0001
(0.86 to 2.79)
0.15
1.48 (0.76 to 2.86)
0.25
Reference
—
1.00
Reference
—
(0.19 to 1.29)
0.15
0.56 (0.19 to 1.60)
0.28
—
—
—
—
—

16
73
158
142

3
10
41
50

3.9
6.9
39.2
50.0

18.8
11.4
27.7
37.7

0.31
0.23
0.64
1.00

(0.10 to 1.00)
(0.09 to 0.56)
(0.38 to 1.07)
Reference

0.05
0.001
0.09
—

0.18 (0.04 to 0.82)
0.75 (0.24 to 2.34)
0.70 (0.38 to 1.28)
1.00
Reference

0.03
0.62
0.24
—

351
38

77
27

76.3
23.7

23.8
68.8

0.14
1.00

(0.06 to 0.31)
Reference

,0.0001
—

0.19 (0.08 to 0.45)
1.00
Reference

0.0002
—

101
288

11
93

9.1
90.9

12.0
32.5

0.28
1.00

(0.13 to 0.61)
Reference

0.001
—

0.25 (0.10 to 0.64)
1.00
Reference

0.004
—

66
323

3
101

1.3
98.7

2.3
32.9

0.05
1.00

(0.01 to 0.29)
Reference

0.001
—

0.06 (0.01 to 0.47)
1.00
Reference

0.01
—

1.00
Reference
0.83 (0.39 to 1.77)

—
0.63

*Tested positive in serosurvey but did not self-report being HIV infected and were not in HIV surveillance registry.
†Weighted by sex based on study design.
‡For age, 16–19 and 20–29 grouped as 16–29, and 50–59 and 60+ grouped as 50+, in regression analyses.

More than a quarter of HIV-infected jail entrants appear
to be undiagnosed at admission. This is substantially higher
than the 5% estimated from a 2004 New York City household
serosurvey9 and more aligned with the 12%–29% estimate
calculated for New York City overall10 and CDC’s nationwide
21% estimate.8 The vast majority (;90%) of undiagnosed
inmates did not self-report recognized HIV risk factors (ie,
MSM or IDU activity). MacGowan et al18 reported similar
results: only 15% of new diagnoses in Florida, Louisiana,
upstate New York, and Wisconsin jails reported IDU or MSM
activity. Likewise, among Los Angeles jail entrants, Harawa
et al19 found that 68% of men and 55% of woman reported no
HIV risk factors.
Our analysis assessing which inmates were most likely
to have undiagnosed HIV infection among those without an
established HIV diagnosis at intake identified women, MSM,
and persons 40–49 years as at increased risk. However, these
groups account for only a third of the undiagnosed. Taken
together, this information confirms that increasing acceptance
of routine HIV testing (ie, offering all inmates testing
regardless of risk) is likely the best approach to diagnosing
more of the jail entrants with undiagnosed infection.
q 2010 Lippincott Williams & Wilkins

At the time of this study, most undiagnosed infections
were not identified through jail testing. Since DOHMH’s
establishment of the jail rapid testing program in 2004, testing
has increased 4-fold to 25,000 tests in 2006. However, there
were .105,000 admissions to New York City jails in 2006,
representing .72,000 unique individuals. Two thirds of
inmates did not consent to HIV testing. Those with
undiagnosed HIV infections were just as likely to consent to
testing as HIV-negative inmates, suggesting that the undiagnosed are not deliberately avoiding testing because they
suspect their HIV-positive status. DOHMH CHS has been
conducting qualitative interviews with inmates to better
understand reasons for declining HIV testing at intake.
It is likely that more inmates would consent to testing
with a more streamlined opt out approach that does not include
a separate written consent for each HIV test conducted. A
separate written consent is still required by New York State
despite CDC recommendations for their elimination,1 including in correctional settings.5 Prison systems in Wisconsin
and Rhode Island have implemented opt out testing with great
success20–21 and have achieved testing rates as high as 86%.20
New York City DOHMH continues to advocate for reform in
www.jaids.com |

99

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

Begier et al

TABLE 4. Predictors of Having Undiagnosed HIV Infection Among All New York City Jail Entrants Without an HIV Diagnosis at
Admission in 2006 Serosurvey (n = 6126)
Total Inmates HIV-Infected
% of Total Inmates
Entering
Inmates Entering Entering Without
Univariate Logistic
Without an
Without an
an HIV Diagnosis
Regression
HIV Diagnosis HIV Diagnosis that were Found to be
(N)
(n)
HIV Infected†
Odds Ratio
95% CI
Total
Sex
Male
Female
Age group in yrs‡
16–19
20–29
30–39
40–49
50–59
60+
Transmission risk factor
MSM
Injection drug use history
High-risk heterosexual
No identified risk
History of violence
Yes
No

Multivariate Logistic Regression
P

Odds Ratio

95% CI

P

6126

104

1.5

4515
1611

64
40

1.4
2.5

1.00
1.77

Reference
—
(1.03 to 3.04) 0.04

1.00
1.74

Reference
—
(1.00 to 3.00) 0.05

729
2005
1539
1484
304
65

2
29
34
33
5
1

0.2
1.4
2.0
2.0
1.5
1.7

—
0.51
0.98
1.00
0.73
—

—
—
(0.30 to 0.85) 0.01
(0.58 to 1.64) 0.93
Reference
—
(0.30 to 1.81) 0.50
—
—

—
0.55
1.01
1.00
0.77
—

—
—
(0.32 to 0.92) 0.02
(0.60 to 1.70) 0.98
Reference
—
(0.31 to 1.92) 0.58
—
—

34
327
2807
2958

3
10
41
50

8.8
2.3
1.3
1.6

5.02
1.52
0.86
1.00

(1.67 to 15.11) 0.004
(0.66 to 3.47) 0.32
(0.56 to 1.32) 0.48
Reference
—

5.16
1.27
0.90
1.00

(1.70 to 15.68) 0.004
(0.55 to 2.94) 0.57
(0.58 to 1.39) 0.62
Reference
—

500
5626

2
102

0.3
1.6

0.19
1.00

(0.04 to 0.90) 0.04
Reference
—

0.20
1.00

(0.04 to 0.94) 0.04
Reference
—

*Tested positive in serosurvey but did not self-report being HIV infected and were not in HIV surveillance registry.
†Weighted by sex based on study design.
‡For age, 16–19 and 20–29 grouped as 16–29, and 50–59 and 60+ grouped as 50+, in regression analyses.
§History of violence is affirmative response to the following question during medical intake: ‘‘Have you ever been charged with a violent act (rape, assault)?’’

HIV consent laws to increase the number of inmates and other
New Yorkers tested through the elimination of a separate
written consent for HIV testing.
To attempt further improvement within current legal
constraints, DOHMH CHS has investigated the provider side
of HIV testing via a recent correctional clinical staff survey.
Many staff still incorrectly believe they are unqualified to offer
or test for HIV because they are not specialized HIV testing
counselors. CHS has renewed efforts to emphasize to clinical
staff that HIV testing is a routine part of medical care. Staff
have been retrained on a standardized, nonjudgmental approach
to offering HIV testing such as ‘‘I recommend HIV screening
for all my patients,’’ or ‘‘it is important for every person to
know his/her HIV status.’’ A system of at least 1 HIV testing
reoffer during postintake medical visits has also been introduced. Ongoing monitoring of these and other programmatic
interventions is in place by an established clinical quality
improvement team.
During our study, some inmates did consent to testing
but were not tested because of operational issues such as jail
release before testing or other competing medical issues
during medical intake. Since this serosurvey, substantial
strides have been made in ensuring that nearly all inmates who
consent to HIV testing are tested: in 2008, 98% of those
accepting testing had a rapid HIV test at intake.
Our study’s primary limitation is that new admissions
without a specimen showed evidence of higher HIV

100

| www.jaids.com

prevalence than those in the serosurvey based on our linkage
to HIV surveillance data. Many specimens were entirely
consumed during routine syphilis screening, whereas for
others, phlebotomy was not performed. This suggests that our
prevalence results are conservative estimates; prevalence may
be nearly twice as high as measured (;8.7% versus 5.2%
overall). Also, the low proportion reporting HIV risk factors,
including IDU and MSM activity, suggests that risk factors
may have been underreported. However, these results are what
actually available to clinicians caring for inmates. Finally,
HARS matching process is inexact complicated by multiple
pseudonyms sometimes used by incarcerated persons.
In conclusion, HIV prevalence appears to have substantially decreased among New York City jail entrants; however,
over one quarter of HIV-infected jails entrants are undiagnosed, representing more than 100 persons in our sample alone.
Despite a 4-fold increase in jail testing, most undiagnosed
infections are not identified during voluntary jail testing,
largely due to low testing acceptance rates. Most undiagnosed inmates did not report recognized HIV risk factors,
reinforcing the need to improve inmate acceptance of the
jails’ current routine testing program rather than focus on
increasing efforts among inmates reporting specific behaviors. To increase inmate’s acceptance of routine testing, we
are working to eliminate the required separate written consent
for HIV testing to allow implementation of the CDCrecommended opt out testing model.
q 2010 Lippincott Williams & Wilkins

J Acquir Immune Defic Syndr  Volume 54, Number 1, May 1, 2010

ACKNOWLEDGMENTS
The authors would like to acknowledge the following
persons for their assistance: Scott Kellerman for organization
support for this project in its initial phase; Louise Cohen, and
Jason Hershberger for organization support for this project
and helpful comments regarding the presentation and
interpretation of this data; Eric Sorenson and Erik Berliner
for their assistance in providing admission data from the
Department of Correction; and Kevin Konty for assistance
with structuring the weighted procedures used for data
analysis. We also appreciate the assistance of the New York
City DOHMH Public Health Laboratory, including Scott Kent,
Erica DeBernardo, Mona El-Fishawy, William Oleszko, and
Eliza Wilson. We thank Isaac Weisfuse, James Hadler, Thomas
Frieden, and Timothy Dondero for helpful comments in an
earlier version of this article.
REFERENCES
1. Branson BM, Handsfield HH, Lampe MA, et al, and Centers for Disease
Control and Prevention (CDC). Revised recommendations for HIV testing
of adults, adolescents, and pregnant women in health-care settings.
MMWR Recomm Rep. 2006;55(RR-14):1–17.
2. Centers for Disease Control and Prevention (CDC). Advancing HIV
prevention: new strategies for a changing epidemic—United States, 2003.
MMWR Recomm Rep. 2003;52:329–332.
3. Marks G, Crepaz N, Senterfitt JW, et al. Meta-analysis of high-risk sexual
behavior in persons aware and unaware they are infected with HIV in the
United States: implications for HIV prevention programs. J Acquir
Immune Defic Syndr. 2005;39:446–453.
4. CDC. HIV/AIDS education and prevention programs for adults in prisons
and jails and juveniles in confinement facilities—United States, 1994.
MMWR Recomm Rep. 1996;45:268–271.
5. Centers for Disease Control and Prevention. HIV Testing Implementation
Guidance for Correctional Settings. 2008:1–38. Available at: http://
www.cdc.gov/hiv/topics/testing/resources/guidelines/correctional-settings/
authors.htm. Accessed December 22, 2009.
6. Weisfuse IB, Greenberg BL, Back SD, et al. HIV-1 infection among New
York City inmates. AIDS. 1991;5:1133–1138.

q 2010 Lippincott Williams & Wilkins

New York City Jail Entrant HIV Serosurvey

7. New York City Department of Health, AIDS Research/HIV Serosurvey
Program. HIV Seroprevalence Update, 1999. New York City, NY; 2001.
8. CDC. HIV prevalence estimates—United States, 2006. MMWR Morb
Mortal Wkly Rep. 2008;57:1073–1076.
9. Nguyen TQ, Gwynn RC, Kellerman SE, et al. Population prevalence of
reported and unreported HIV and related behaviors among the household
adult population in New York City, 2004. AIDS. 2008;22:281–287.
10. New York City Department of Health and Mental Hygiene. HIV
Prevalence in New York City, 2002: Estimating the Total Number of
People Living with HIV and AIDS in NYC. New York City, NY: New York
City Department of Health and Mental Hygiene; 2004.
11. Lee LM, McKenna MT, Janssen RS. Classification of transmission risk in
the national HIV/AIDS surveillance system. Public Health Rep. 2003;
118:400–407.
12. HIV Epidemiology and Field Services Program. Semiannual Report. Vol
2 No. 2. New York City, NY: New York City Department of Health and
Mental Hygiene; October 2007.
13. Maruschak LM, US Department of Justice, Bureau of Justice Statistics.
HIV in Prisons, 2006. NCJ 222179. Washington, DC: Bureau of Justice
Statistics Bulletin; 2008.
14. New York City Department of Health and Mental Hygiene. New York City
HIV/AIDS Annual Surveillance Statistics 2006. New York City, NY: New
York City Department of Health and Mental Hygiene; 2007.
15. Weinbaum CM, Sabin KM, Santibanez SS. Hepatitis B, hepatitis C, and
HIV in correctional populations: a review of epidemiology and
prevention. AIDS. 2005;19(Suppl 3):S41–S46.
16. Khan MR, Wohl DA, Weir SS, et al. Incarceration and risky sexual
partnerships in a southern US city. J Urban Health. 2008;85:
100–113.
17. Doherty IA, Padian NS, Marlow C, et al. Determinants and consequences
of sexual networks as they affect the spread of sexually transmitted
infections. J Infect Dis. 2005;191(Suppl 1):S42–S54.
18. MacGowan R, Eldridge G, Sosman J, et al. HIV counseling and testing of
young men in prison. J Correctional Health Care. 2006;12:203–213.
19. Harawa NT, Bingham TA, Butler QR, et al. Using arrest charge to screen
for undiagnosed HIV infection among new arrestees: a study in Los
Angeles County. J Correct Health Care. 2009;15:105–117.
20. Hoxie NJ, Chen MH, Prieve A, et al. HIV seroprevalence among male prison
inmates in the Wisconsin Correctional System. WMJ. 1998;97:28–31.
21. Desai AA, Latta ET, Spaulding A, et al. The importance of routine HIV
testing in the incarcerated population: the Rhode Island experience. AIDS
Educ Prev. 2002;14(5 Suppl B):45–52.

www.jaids.com |

101