HIV Prevalence, Risk Behaviors, Health

Care Use, and Mental Health Status of

Transgender Persons: Implications for
Public Health Intervention

A B S T R A C T Kristen Clements-Nolle, MPH, Rani Marx, PhD, MPH, Robert Guzman, BA,
and Mitchell Katz, MD

Objectives. This study described Transgender is a term used to describe 392 male-to-female and 123 female-to-male
HIV prevalence, risk behaviors, health individuals who have a persistent and dis- transgender persons in San Francisco, Calif.
care use, and mental health status of tressing discomfort with their assigned gen- We describe HIV prevalence and risk be-
male-to-female and female-to-male trans- der.1 Such individuals were born anatomically haviors among male-to-female and female-to-
gender persons and determined factors as one biological sex but live their lives to male individuals and assess the independent
associated with HIV. varying degrees as the opposite sex. Qualita- predictors of HIV infection among male-to-
Methods. We recruited transgender tive research suggests that male-to-female and female transgender individuals. We also re-
persons through targeted sampling, female-to-male transgender individuals expe- port on health care use and mental health sta-
respondent-driven sampling, and agency rience severe employment, housing, and health tus of both populations.
referrals; 392 male-to-female and 123 care discrimination, and many engage in be-
female-to-male transgender persons were haviors that put them at risk for HIV.2–6
interviewed and tested for HIV. High HIV prevalence has been found in Methods
Results. HIV prevalence among small published studies of male-to-female
male-to-female transgender persons was sex workers recruited from street locations Subjects and Recruitment
35%. African American race (adjusted in Atlanta, Ga7 (68%), and Tel Aviv, Israel8
odds ratio [OR] = 5.81; 95% confidence (11%); a drug treatment center in Rome, Our targeted sampling plan20 was based
interval [CI] = 2.82, 11.96), a history of Italy9 (46%); and a clinic in Italy10 (57%). In on information gathered from focus groups
injection drug use (OR =2.69; 95% CI= addition, research comparing male-to-female with 100 transgender persons,6 30 key in-
1.56, 4.62), multiple sex partners (ad- sex workers with male and female sex work- formant interviews, and social mapping in
justed OR = 2.64; 95% CI = 1.50, 4.62), ers in the same neighborhoods has consis- neighborhoods thought to have high concen-
and low education (adjusted OR = 2.08; tently found higher HIV prevalence in trans- trations of transgender persons. We used par-
95% CI=1.17, 3.68) were independently gender samples.7,8,11–14 ticipatory research strategies whereby mem-
associated with HIV. Among female-to- Male-to-female transgender persons not bers of the target population were involved in
male transgender persons, HIV preva- currently involved in sex work were included development, implementation, and interpreta-
lence (2%) and risk behaviors were much in 4 published studies that sampled in clinics, tion of the study.21,22
lower. Most male-to-female (78%) and communities, and prison systems.15–19 HIV Three male-to-female transgender indi-
female-to-male (83%) transgender per- prevalence was lower than in studies of male- viduals (African American, Filipina, and La-
sons had seen a medical provider in the to-female sex workers, but sexual and injec- tina) and 3 female-to-male transgender per-
past 6 months. Sixty-two percent of the tion drug risk behaviors were common.15–17 sons (1 Vietnamese and 2 White) were hired
male-to-female and 55% of the female- Two of these studies compared male-to-female as interviewers. Staff received 35 training ses-
to-male transgender persons were de- transgender persons with males, females re- sions, including HIV counseling, interview-
pressed; 32% of each population had at- cruited in the same settings, or both, and found
tempted suicide. a higher prevalence of risk behaviors among The authors are with the San Francisco Department
Conclusions. High HIV prevalence male-to-female transgender persons.18,19 of Public Health, San Francisco, Calif. Mitchell
suggests an urgent need for risk reduction Studies of male-to-female transgender Katz is also with the University of California, San
Francisco.
interventions for male-to-female trans- individuals have relied on small convenience
Requests for reprints should be sent to Kris-
gender persons. Recent contact with samples and predominantly have reported ten Clements-Nolle, MPH, Epidemiology and Eval-
medical providers was observed, sug- data on individuals seeking HIV testing or uation Section, San Francisco Department of Pub-
gesting that medical providers could pro- self-reporting HIV seropositivity.9,10,15–19 lic Health, 25 Van Ness, #500, San Francisco, CA
vide an important link to needed pre- Three studies included street-based sampling 94102 (e-mail: [email protected]).
vention, health, and social services. (Am methods but lacked power to determine in- This article was accepted July 29, 2000.
J Public Health. 2001;91:915–921) dependent predictors of HIV infection.7,8,17 Note. All study protocols and materials received
approval from the Committee on Human Research
No studies have quantitatively assessed HIV at the University of California, San Francisco, and
risk among female-to-male transgender in- all participants provided informed consent. The views
dividuals. To address these limitations, we expressed are those of the authors and do not nec-
conducted interviews and HIV testing with essarily reflect those of the funding agencies.

June 2001, Vol. 91, No. 6 American Journal of Public Health 915
ing, confidentiality, harm reduction, street out- defined so as to exclude the use of dildos and and methamphetamine use were excluded from
reach, suicide prevention, and referrals. From “sex toys.” the multivariate model because of collinearity
July through December 1997, staff recruited We measured the prevalence of lifetime with one another and with injection drug use
individuals from street settings, bars, and social and recent (past 6 months) use of marijuana, (Pearson r≥0.40).
gatherings catering to transgender persons, and cocaine, crack cocaine, amyl nitrite, heroin, Review of the correlation matrix of the
agencies referred clients to the study. We also hallucinogens, methamphetamine, and non- multivariate parameter estimates identified no
initiated respondent-driven sampling23; each hormonal injection drugs. Among nonhor- serious multicollinearity. Deviance and Pearson
study participant received an incentive for el- monal injection drug users, we determined the residuals from the logistic model were calcu-
igible persons recruited. prevalence of sharing syringes, using cookers, lated and plotted; more than 99% of the de-
Recruited individuals called a toll-free and backloading (using 1 syringe to mix drugs viance and more than 95% of the Pearson resid-
number and were screened for eligibility. In- and load another syringe). uals were between –2 and 2, indicating good
dividuals were eligible if they (1) were 18 Measures of physical and mental health model fit.29,30 Two-way interactions among
years or older; (2) lived, worked, or social- included emergency department and outpatient modeled variables were assessed, but none were
ized in San Francisco; (3) spoke English, medical visits in the past 6 months (di- significant (P≤.20).
Spanish, Vietnamese, or Tagalog; and (4) chotomized as none vs ≥1 visit), recent hor- Factors associated with HIV were not as-
stated that their primary gender was trans- mone use and injection, hormone syringe shar- sessed for female-to-male transgender persons
gender, transsexual, bigender, transvestite, ing, and source of hormones and syringes. We because few were HIV positive. All analyses
cross-dresser, intersexed, or the opposite sex also measured history of mental health hospi- were conducted with SAS, Version 6.0 (SAS
of that at birth. talization, suicide attempts, HIV testing, and Institute, Inc, Cary, NC); S-Plus, Version 4.5
Eligible subjects were scheduled for an use of HIV-related health care. The 20-item Professional (Insightful Corp, Seattle, Wash),
interview at 1 of 8 community-based or- Center for Epidemiologic Studies Depression was used to graph the relationship between age
ganizations and could request the gender or Scale was used to screen for depression.24 This and lifetime number of partners with HIV. All
race/ethnicity of their interviewer. Inter- scale has high sensitivity and specificity for P values were 2-sided.
viewers identified and prevented 3 partici- major depression among primary care pa-
pants from reenrolling. A composite vari- tients25; we used the standard cutoff score (≥16)
able of unique personal identifiers also was to classify depression.26
created; no duplicates were identified with HIV testing was performed with the Ora-
Results
this variable. Sure HIV-1 Oral Specimen Collection Device
Participation
Written informed consent was obtained (Epitope, Inc, Beaverton, Ore). Specimens were
before interviewing and HIV testing. Two screened with the Vironostika HIV-1 Microelisa
We screened 645 individuals, of whom
weeks later, participants received HIV test System Kit (Organon Teknika Corp, Durham,
586 (91%) were eligible for inclusion in the
results and further counseling and referrals NC). Specimens that repeatedly tested reac-
study. Of the eligible participants, 523 (89%)
from the same interviewer. Missed follow-up tive to enzyme-linked immunosorbent assay
completed the interview and HIV test. No de-
appointments could be rescheduled, but staff (ELISA) were confirmed with OraSure HIV-
mographic differences were found between el-
did not try to locate participants who missed 1 Western Blot Kits (Epitope, Inc, Beaverton,
igible subjects who did and did not complete
appointments. Participants received $40 for Ore). With confirmation, the OraSure test has
the interview. Eight intersexed individuals
the interview, $10 for the follow-up appoint- high sensitivity and specificity.27
(born with ambiguous or both male and female
ment, and $5 for up to 5 eligible persons they
genitalia) were excluded from this analysis,
referred. Statistical Methods
because they could not be classified as male-
to-female or female-to-male persons.The final
Instrument and Measures The χ2 or the Fisher exact test was used to
sample comprised 392 male-to-female and 123
determine differences in proportions; median
female-to-male transgender individuals.
Sociodemographic measures included tests were used to assess differences in contin-
The most common (non–mutually ex-
race/ethnicity, age, sexual orientation, educa- uous variables. Unadjusted smoothing tech-
clusive) ways that participants were recruited
tion, incarceration (ever and past 6 months), niques (spline and loess) were used to deter-
for the study were respondent-driven sam-
and current housing status (stable in a house mine whether age and lifetime number of
pling (39%), outreach by study staff (38%),
or apartment vs unstable in a single-room- partners had a linear relation to HIV preva-
flyer recruitment (14%), and referrals from
occupancy hotel, on the streets, in parks, in lence among male-to-female transgender per-
agencies (10%). Male-to-female transgender
parked cars, in shelters, or temporarily staying sons. An inverse U-shaped curve between age
persons were more likely to report respondent-
with others). and HIV existed; cutpoints for age categories
driven sampling recruitment (42% vs 32%,
We ascertained the number of lifetime sex (18–29, 30–39, ≥40) were chosen to preserve
P=.04), and female-to-male transgender per-
partners and the prevalence of unprotected (no this relationship.28 We created a dichotomous
sons were more likely to be recruited by in-
condom used) anal and vaginal sex (insertive variable for number of partners based on
terviewers (58% vs 32%, P<.001). There were
and receptive) with male, female, and trans- smoothing techniques that indicated a nonlin-
no other differences in demographics or HIV
gender sex partners. The number, gender, and ear relation with HIV and a clear cutpoint
prevalence by recruitment type. Seventy per-
type of sex partners (main, casual, and ex- (>200 partners).
cent of the participants returned for their HIV
change) in the past 6 months and the preva- To identify factors independently as-
test results.
lence of unprotected anal and vaginal sex by sociated with HIV prevalence among male-
partner type also were documented. An ex- to-female transgender persons, we entered life-
change partner was defined as “a person you time demographic and risk factors significantly Sociodemographics
had sex with in exchange for things you needed associated with HIV in the bivariate analyses
or they needed such as money, drugs, shelter, (P < .05) into a multiple logistic regression Most male-to-female and female-to-male
or food.” All reported sexual behaviors were model as simultaneous entries. Crack, cocaine, participants self-identified as transgender, the

916 American Journal of Public Health June 2001, Vol. 91, No. 6
 (63% of the African American male-to-female
TABLE 1—Demographics of Male-to-Female and Female-to-Male Transgender participants were HIV positive). Other factors
Persons: San Francisco, Calif, 1997
independently associated with HIV prevalence
Male-to-Female Female-to-Male included having less than a high school degree,
(n = 392), No. (%) (n = 123), No. (%) Pa having had more than 200 lifetime sexual part-
ners, and using nonhormonal injection drugs.
Gender identity
Transgender 145 (37) 46 (37) .09
We compared recent risk behaviors of the
Female 135 (34) ... male-to-female participants who were HIV in-
Male ... 41 (33) fected and uninfected (Table 3). About one
Transsexual 100 (26) 26 (21) fifth of the male-to-female participants reported
Otherb 12 (3) 10 (8) nonhormonal injection drug use in the past 6
Race/ethnicityc
African American 104 (27) 12 (10) < .001 months; this behavior was more prevalent
Latina/Latino 106 (27) 14 (11) among individuals who were HIV positive. Al-
White 106 (27) 82 (67) most half of the injectors shared syringes and
Asian and Pacific Islander 49 (13) 9 (7) backloaded, and 29% shared cookers. Syringes
Native American 24 (6) 5 (4)
Median age, y (range) 34 (18–67) 36 (19–61) .25
were typically obtained from needle exchange
Sexual orientationc sites (54%; n=39), the streets (22%; n=16),
Heterosexual 271 (69) 43 (35) < .001 pharmacies or clinics (15%; n=11), and friends
Bisexual 82 (21) 40 (33) (8%; n=6).
Lesbian/gay 21 (5) 15 (12) Eighty percent of the male-to-female par-
Homosexual/gay 15 (4) 22 (18)
Education ticipants had had anal, vaginal, or oral sex in the
<High school diploma 113 (29) 5 (4) < .001 past 6 months, and 37% had had more than 10
≥High school diploma 278 (71) 118 (96) sex partners (Table 3).Three fourths of the male-
Median monthly income, $ (range)c 744 (55–7346) 1100 (100–6000) <.001 to-female participants had had sex with males,
Current housing statusd
Stable 207 (53) 96 (78) < .001
6% with females, and 8% with other transgen-
Unstable 185 (47) 26 (21) der persons. Participants were more likely to en-
Incarceration, ever 255 (65) 36 (29) < .001 gage in receptive anal sex (with male or trans-
Incarceration, past 6 mo 119 (30) 6 (5) < .001 gender persons) than insertive anal sex (with
HIV positive (OraSure test result) 137 (35) 2 (2) < .001 male, female, or transgender persons) (63% vs
a
Associations between characteristics and gender (male-to-female vs female-to-male) 30%, P<.001).As shown inTable 3, unprotected
assessed with χ2 test, Fisher exact test (for HIV prevalence), and the median test (for receptive anal sex was more common with main
income and age). partners, followed by casual and exchange part-
b
Includes transvestite, cross-dresser, bigender. ners. Participants who were HIV positive were
c
Missing data: race/ethnicity (4), sexual orientation (6), income (16).
d
Stable housing = owns or rents home or apartment. Unstable housing = single-room-
more likely than uninfected participants to report
occupancy hotel; living on the streets, in parks, in parked cars, or in shelters; or receptive anal sex with main partners; no other
temporarily staying with others. differences were seen between the 2 groups in
sexual behaviors or condom use. Only 7% of
the male-to-female transgender persons had had
vaginal construction surgery, so unprotected re-
ceptive vaginal sex with male or transgender
opposite gender of that at birth, or transsex- HIV Risk Among Male-to-Female persons was rarely reported (2%).Two male-to-
ual (Table 1). The median age of male-to- Participants female transgender persons reported unprotected
female and female-to-male participants was insertive vaginal sex with a female.
similar, but the 2 populations differed signif- Thirty-five percent of the male-to-female
icantly on all other sociodemographic char- participants (n=137) had positive HIV test re- HIV Risk Among Female-to-Male
acteristics. Seventy-three percent of the male- sults, of whom 65% (n=89) knew they were in- Participants
to-female persons identified as non-White, fected, 20% (n=27) learned their status through
whereas two thirds of the female-to-male per- study participation, and 15% (n=21) did not Two female-to-male participants (2%) had
sons identified as White. Male-to-female trans- know their status and failed to return for test re- positive HIV test results; both knew their sta-
gender individuals were more likely than sults.Fifty-twopercent(n=11)oftheparticipants tus and were receiving HIV-related health care.
female-to-male transgender individuals to who were HIV positive but did not obtain test re- A history of nonhormonal injection drug use
identify as heterosexual, to report prior incar- sults thought that it was “unlikely” or “there was was reported by 18% of the female-to-male
ceration, and to have unstable housing, low no chance” that they were HIV infected. Only participants; 54% had injected hormones.
education, and low monthly income. The most 50% of all the male-to-female individuals who Female-to-male participants reported a median
common ways (non–mutually exclusive) that were HIV infected were receiving HIV-related of 27 lifetime sex partners. Twenty-seven per-
male-to-female transgender individuals ob- medicalcare.Ofthepersonswhoknewtheywere cent reported unprotected receptive anal sex
tained money in the past 6 months included infected,78%(n=69)werereceivingHIV-related with a male, whereas 6% had engaged in un-
part- or full-time employment (40%), sex work medical care, and 58% (n=52) were receiving protected receptive anal sex with a transgender
(32%), Supplemental Security Income and So- HIV antiretroviral or prophylactic therapy. person. Sixty-three percent reported unpro-
cial Security Disability Insurance (29%), and Lifetime risk factors associated with HIV tected receptive vaginal sex with a male; 25%
General Assistance (23%), whereas 81% of were common (Table 2). After control for other had engaged in this behavior with a transgen-
the female-to-male were employed (not shown covariates, African American race was found to der person. Almost one third (31%) of the fe-
in table). be the strongest risk factor for HIV infection male-to-male participants had a history of sex

June 2001, Vol. 91, No. 6 American Journal of Public Health 917
 work or survival sex, and 59% reported forced
TABLE 2—Factors Associated With HIV Prevalence Among Male-to-Female sex or rape.
Transgender Persons: San Francisco, Calif, 1997
Recent risk behaviors were uncommon
All among female-to-male participants overall.
Male-to-Female HIV+ Adjusted OR Only 5 female-to-male participants (4%) re-
(n = 392), (n = 137), (95% CI)a ported nonhormonal injection drug use in the
No. (%) No. (%) (n = 381) P
past 6 months, although 4 of these 5 individu-
Race/ethnicity als shared syringes and cookers and back-
White 106 (27) 23 (17) Reference loaded. All female-to-male injectors obtained
African American 104 (27) 65 (47) 5.81 (2.82, 11.96) <.001 syringes off the streets or from friends. Twenty
Latina/Latino 106 (27) 31 (23) 0.80 (0.37, 1.72) .56 percent of the female-to-male participants had
Otherb 73 (19) 18 (13) 1.10 (0.49, 2.48) .81
Age, y not had anal, vaginal, or oral sex in the past 6
18–29 133 (34) 35 (26) Reference months; 46% had 1 partner, 32% had 2 to 10
30–39 149 (38) 65 (47) 1.71 (0.95, 3.07) .07 partners, and 2% had more than 10 partners.
≥40 110 (28) 37 (27) 0.93 (0.47, 1.84) .83 Fifty-eight percent of the female-to-male par-
Education
≥High school diploma 278 (71) 86 (63) Reference
ticipants had sex with females, 18% with
<High school diploma 113 (29) 51 (37) 2.08 (1.17, 3.68) .01 males, and 15% with transgender individuals.
History of incarceration 255 (65) 108 (79) 1.12 (0.62, 2.03) .71 Fifty-nine percent had sex with main partners,
Lifetime risk behaviors 41% with casual partners, and 4% with ex-
No. of sex partnersc change partners. Ten percent of the female-to-
≤200 212 (54) 51 (37) Reference
>200 180 (46) 86 (63) 2.64 (1.50, 4.62) <.001 male participants had receptive vaginal sex
Nonhormonal injection drug use 134 (34) 69 (50) 2.69 (1.56, 4.62) <.001 with a male or transgender person, of whom
Hormonal injection drug use 256 (65) 108 (79) 1.67 (0.94, 2.97) .08 67% (n=8) did not always use condoms. Only
URAI with male 330 (84) 127 (93) 2.34 (0.95, 5.78) .06 7% had receptive anal sex with a male or trans-
UIAI with male 213 (54) 94 (69) 1.38 (0.82, 2.35) .23
Sex work or survival sex 312 (80) 124 (91) 0.82 (0.37, 1.82) .62
gender person, but 56% (n=5) of this group
Forced sex or rape 231 (59) 92 (67) 1.02 (0.59, 1.75) .95 used condoms inconsistently. Only 2% (n=2)
of the female-to-male participants had had
Note. OR = odds ratio; CI = confidence interval; URAI = unprotected receptive anal penis construction surgery, so insertive anal
intercourse; UIAI = unprotected insertive anal intercourse. and vaginal sex was rare.
a
Factors associated with HIV (P < .05) in bivariate analyses were entered in multivariate
model. Eleven subjects were deleted because of missing data.
b
Includes Asian and Pacific Islanders (67%) and Native Americans (33%). Health Care Use and Mental Health
c
Includes anal, vaginal, and oral intercourse. Status Among Male-to-Female and
Female-to-Male Transgender Individuals

A large proportion of male-to-female and

TABLE 3—Recent Risk Behaviors of Male-to-Female Transgender Persons Who female-to-male participants lacked health in-
Were HIV Positive and HIV Negative: San Francisco, Calif, 1997 surance (Table 4). Among those insured, male-
to-female participants relied on public insur-
All Male-to-Female HIV+ HIV− ance, whereas female-to-male participants
(n = 392), (n = 137), (n = 255),
No. (%) No. (%) No. (%) Pa were more likely to have private insurance.
Most male-to-female and female-to-male par-
Injection drug use behaviors, past 6 mo ticipants received care at a clinic or doctor’s
Injected street drugs 72 (18) 40 (29) 32 (13) <.001
Shared syringesb 34 (47) 17 (43) 17 (53) .37 office, and about one fifth had visited an emer-
Backloadedb 35 (49) 18 (45) 17 (53) .58 gency department in the past 6 months. Male-
Shared cookersb 21 (29) 11 (28) 10 (31) .95 to-female participants had a higher prevalence
Sexual behaviors, past 6 mo of recent hormone use and were more likely to
No. of sex partnersc obtain hormones from a nonmedical source
0 80 (20) 27 (20) 53 (21) .10 but were less likely to inject their hormones
1 69 (18) 24 (18) 45 (18)
2–10 99 (25) 41 (30) 58 (23) than were female-to-male participants. Most
>10 144 (37) 45 (33) 99 (39) female-to-male (92%) and male-to-female
RAI with main partnerd 146 (37) 66 (48) 80 (31) <.001 (82%) hormone injectors obtained their sy-
URAIe 90 (62) 36 (55) 54 (68) .11 ringes from a medical source; only 3 individ-
RAI with casual partnerd 144 (37) 55 (40) 89 (35) .30
URAIe 63 (44) 25 (45) 38 (43) .75
uals reported sharing hormone syringes in the
RAI with exchange partnerd 128 (33) 46 (34) 82 (32) .78 past 6 months.
URAIe 36 (28) 11 (24) 25 (30) .43 Some type of sexual reassignment surgery
(typically breast augmentation for male-to-
a
Associations between characteristics and HIV prevalence assessed with χ2 test. female and reduction for female-to-male trans-
b
Among those who injected drugs in the past 6 months.
c
Includes anal, vaginal, and oral sex. gender individuals) was reported by 22% of the
d
RAI = receptive anal intercourse with male or transgender main, casual, or exchange male-to-female participants and one third of the
partners, respectively. female-to-male participants. Most male-to-
e
URAI = unprotected receptive anal intercourse among those who had RAI with male or female (72%) and female-to-male (85%) par-
transgender main, casual, or exchange partners, respectively.
ticipants planned to have some type of surgery
in the future. Slightly fewer than one third of both

918 American Journal of Public Health June 2001, Vol. 91, No. 6
 cult to maintain an erection,16 and playing the re-
TABLE 4—Health Care Use and Mental Health Status of Male-to-Female and ceptive role is viewed as more feminine.3
Female-to-Male Transgender Persons: San Francisco, Calif, 1997
Inconsistent condom use during receptive
Male-to-Female Female-to-Male anal sex was commonly reported, particularly
(n = 392), No. (%) (n = 123), No. (%) Pa with main partners. This finding is consistent
with past research on male-to-female trans-
Health care use gender persons7,38 and calls for interventions
Health insurance, current that decrease barriers to condom use with main
None 202 (52) 51 (41) <.001
Public 132 (34) 14 (11) partners. Risk reduction interventions also are
Private 56 (14) 58 (47) needed for female-to-male transgender per-
Clinic or doctor visit, past 6 mo 306 (78) 102 (83) .25 sons who have sex with men, because condom
Emergency department visit, past 6 mo 98 (25) 22 (18) .10 use was infrequent.
Hormone use, past 6 mo 288 (73) 65 (53) <.001
Injected hormones, past 6 mo b
121 (42) 63 (97) <.001
According to qualitative research, many
Source of hormones, past 6 mob male-to-female transgender persons turn to sex
Medical provider 204 (71) 63 (97) <.001 work because they face severe employment
Streets, black market, friends 84 (29) 2 (3) discrimination2–4,6; this may account for the
Mental health status high number of sexual partners and prevalence
Mental health hospitalization, ever 87 (22) 24 (20) .50 of sex work among male-to-female transgen-
Suicide attempt, ever 127 (32) 39 (32) .89
Depression: CES-D score ≥16c 242 (62) 68 (55) .17
der persons in our study. Street outreach and
more intensive HIV prevention interventions,
Note. CES-D = Center for Epidemiologic Studies Depression Scale.24 such as prevention case management, are ur-
a
Associations between characteristics and gender (male-to-female vs female-to-male) gently needed for sex workers and should in-
assessed with χ2 test. clude education, job training, and job place-
b
Of those who used hormones in the past 6 months. Fisher exact test used.
c
Seven missing values. ment. Given the high rate of incarceration
among our participants, jails and prisons may
be good settings for HIV prevention interven-
tions for sex workers.
We found very low levels of HIV risk as-
male-to-female and female-to-male participants derpersons who were HIV positive in our study sociated with hormone injection, probably be-
had attempted suicide, and about one fifth of were not receiving HIV-related health care, and cause several clinics in San Francisco offer hor-
each group had been hospitalized for a mental many individuals who were HIV positive and mone therapy. However, nonhormonal injection
health condition.Almost two thirds of the male- did not return for their test results thought it un- drug use was associated with HIV, and the prev-
to-female and 55% of the female-to-male par- likely that they were infected. These findings alence of sharing syringes among injectors was
ticipants were classified as depressed (Table 4). highlight the importance of counseling male-to- higher than that in studies with out-of-treatment
female transgender persons about HIV and the injection drug users in San Francisco.34,39,40 Pub-
benefits of early intervention. lic health providers need to ask transgender
Discussion There is a particular need to intervene clients about hormonal and nonhormonal in-
with African American male-to-female trans- jection drug practices and refer them to harm
To our knowledge, this is the largest quan- gender persons, two thirds of whom had pos- reduction and treatment interventions. In addi-
titative study to describe HIV risk, health care itive HIV test results in our study. Similar as- tion, needle exchange programs should reach
use, and mental health status of male-to-female sociations between African American ethnicity out to transgender populations and provide both
and female-to-male transgender individuals. and higher HIV prevalence have been found hormone needles (22 g, 11/2 in) and nonhor-
Our data show the diversity that exists within among transgender persons in Atlanta7 and mone needles (27 g, 5/8 in).
the transgender community. For example, gen- among gay men (older and younger), injection Our study confirms that many transgen-
der identification does not determine sexual drug users, and heterosexual men and women der persons enter the medical system in pursuit
orientation. Male-to-female study participants in San Francisco.31,32,35–37 As with previous of hormones.1,15,16 In addition to providing hor-
primarily had sex with the opposite gen- studies, socioeconomic and behavioral differ- mone therapy, health care providers should
der (males) and self-identified as heterosex- ences did not account for the disproportionate counsel and appropriately refer transgender
ual, whereas female-to-male participants were level of HIV infection among African Ameri- clients who are in need of HIV, substance abuse,
more likely to have sex with other males and cans in our study. and mental health services. There is a particu-
transgender individuals and self-identified as We found high levels of current risk be- lar need to assess the potential for depression
gay or bisexual. We also found major differ- haviors among both HIV-positive and HIV- and suicide and intervene appropriately.
ences between male-to-female and female-to- negative male-to-female transgender persons. The prevalence of suicide attempts among
male participants with respect to demographic We do not know, however, whether transgender male-to-female and female-to-male transgender
and HIV risk characteristics. In particular, populations serve as a “bridge group” in the persons in our study was much higher than that
male-to-female transgender persons were more HIV epidemic, because we did not determine found in US household probability samples
socioeconomically disadvantaged and had the sexual orientation and HIV status of partic- and a population-based sample of adult men
higher HIV prevalence. ipants’ sex partners. As with past research, we reporting same-sex partners.41,42 This finding
Our estimate of HIV prevalence among found that male-to-female transgender persons supports a previous study of male-to-female
male-to-female transgender persons is higher were more likely to engage in receptive rather transgender persons in the Netherlands, which
than estimates from studies with gay men and than insertive anal sex.3,11,16 Some researchers found that the number of suicides among male-
injection drug users of the same age in San Fran- have suggested that receptive anal sex is more to-female transgender persons was much higher
cisco.31–34 Half of the male-to-female transgen- common because hormone use makes it diffi- than the expected mortality for the Dutch male

June 2001, Vol. 91, No. 6 American Journal of Public Health 919
population.43 Although we did not assess fac- view and prepared the initial draft of the manuscript. foreign people involved in HIV-related risk ac-
tors associated with suicide attempts, re- All authors contributed to the final manuscript. tivities and attending an HIV reference centre
searchers have hypothesized that higher rates in Rome: the possible role of counseling in re-
ducing risk behavior. AIDS Care. 1998;10:
of attempted suicide among gay youths com-
473–480.
pared with heterosexual youths may partially Acknowledgments 10. Galli M, Esposito R, Antinori S, et al. HIV-1 in-
be the result of increased discrimination and This study was supported by Centers for Disease fection, tuberculosis, and syphilis in male trans-
victimization44–46—tragically, common expe- Control and Prevention Cooperative Agreement sexual prostitutes in Milan, Italy. J Acquir Im-
riences for transgender individuals.2–4,6 U62CCU902017-12 and California Department of mune Defic Syndr. 1991;4:1006–1007.
Health Services, Office of AIDS Grant 97-10787. 11. Tirelli U, Vaccher E, Bullian P, et al. HIV-1 se-
The primary limitation of our research was
We would like to thank Ari Bachrach, Nikki roprevalence in male prostitutes in northeast
the use of nonprobability sampling. Our find- Calma, Carla Clynes, Nashanta Stanley, Matt Rice, Italy. J Acquir Immune Defic Syndr. 1988;1:
ings may not generalize to other urban areas, and Doan Thai for conducting data collection, coun- 414–415.
and there may be threats to internal validity if seling, testing, and referrals; Vince Guilin and Scott 12. Tirelli U, Rezza G, Giuliani M, et al. HIV sero-
certain sampling methods were more likely to Ikeda for field coordination; the Transgender Com- prevalence among 304 female prostitutes from
recruit high-risk individuals. Most traditional munity Advisory Board for assistance with protocol, four Italian towns. AIDS. 1989;3:547–548.
survey, and materials development; the San Fran-
random sampling approaches would not pro- 13. Elifson KW, Boles J, Sweat M. Risk factors as-
cisco Department of Public Health Laboratory for sociated with HIV infection among male pros-
duce reliable samples, however, because the HIV test result processing; Art DeGuzman, Stella
transgender population has strong privacy con- titutes. Am J Public Health. 1993;83:79–83.
Cheung, Irene Lee, and Belinda Van for data entry
14. Leonard T, Sacks J, Franks A, Sikes RK. The
cerns and has never been counted, and because and management; Aida Flandez for assistance with
prevalence of human immunodeficiency virus,
many transgenderpersons are marginally housed manuscript preparation; and Nancy Padian, PhD,
hepatitis B, and syphilis among female prosti-
or homeless. Mike Pendo, MPH, and Doug Sebesta, PhD, for man-
tutes in Atlanta. J Med Assoc Ga. 1988;77:
Our use of social mapping and targeted uscript review. We would also like to express our ap-
162–164.
preciation to the following agencies who assisted
sampling helped ensure that important sub- in recruitment and provided space to conduct inter-
15. Moriarty HJ, Thiagalingam A, Hill PD. Audit of
populations were sampled. Respondent-driven service to a minority client group: male to fe-
views and HIV testing: Asian & Pacific Islander
sampling capitalizes on the ability of members male transsexuals. Int J STD AIDS. 1998;9:
Wellness Center, Center for Special Problems, Fil-
238–240.
of a hidden population to recruit their peers ipino Task Force on AIDS, FTM International, Glide-
16. Valenta LJ, Elias AN, Domurat ES. Hormone
and reduces many biases associated with tra- Goodlett HIV/AIDS Program, Instituto Familiar de
pattern in pharmacologically feminized male
ditional snowball sampling.23 Despite the use la Raza, New Village, Proyecto ContraSIDA Por
transsexuals in the California State prison sys-
Vida, Southeast Asian Community Center, Tender-
of such methods, female-to-male participants tem. J Natl Med Assoc. 1992;84:241–250.
loin AIDS Resource Center, and Tom Waddell Clinic.
were difficult to recruit. Female-to-male trans- 17. Sykes DL. Transgendered people: an “invisible”
gender individuals were more likely to be re- population. Calif HIV/AIDS Update. 1999;12:
cruited through personal contacts with study 80–85.
18. Stephens T, Scott C, Braithwaite RL. Transsex-
staff than through respondent-driven sampling. References ual orientation in HIV risk behaviours in an adult
This may partially account for the demographic 1. White JC, Townsend MH. Transgender medi- male prison. Int J STD AIDS. 1999;10:28–31.
differences we observed between male-to- cine: issues and definitions. J Gay Lesbian Med 19. Nemoto T, Luke D, Mamo L, Ching A, Patria J.
female and female-to-male participants but is Assoc. 1998;2:1–3. HIV risk behaviors among male-to-female trans-
unlikely to account for the large difference in 2. Bockting WO, Robinson BE, Rosser BR. Trans- genders in comparison with homosexual or bi-
HIV prevalence. gender HIV prevention: a qualitative needs as- sexual males and heterosexual females. AIDS
sessment. AIDS Care. 1998;10:505–525. Care. 1999;11:297–312.
Despite these limitations, our data describe
3. Boles J, Elifson KW. The social organization of 20. Watters JK, Biernacki P. Targeted sampling: op-
an important population in San Francisco with tions for the study of hidden populations. Soc
transvestite prostitution and AIDS. Soc Sci Med.
many needs. Similar studies in other US cities 1994;39:85–93. Probl. 1989;36:416–430.
are needed to assess the role of transgender in- 4. Pang H, Pugh K, Catalan J. Gender identity dis- 21. Schwab M, Syme SL. On paradigms, commu-
dividuals in local HIV epidemics. Our ability order and HIV disease. Int J STD AIDS. 1994;5: nity participation, and the future of public health.
to recruit this hard-to-reach population was 130–132. Am J Public Health. 1997;87:2049–2051.
likely the result of engaging in participatory re- 5. Yep GA, Pietri M. In their own words: commu- 22. Cornwall A, Jewkes R. What is participatory re-
search and hiring a diverse group of male-to- nication and the politics of HIV education for search? Soc Sci Med. 1995;41:1667–1676.
transgenders and transsexuals in Los Angeles. 23. Heckathorn DD. Respondent-driven sampling: a
female and female-to-male transgender staff. new approach to the study of hidden popula-
In: Elwood WN, ed. Power in the Blood: A
Similar peer-based approaches could be used to Handbook on AIDS, Politics, and Communica- tions. Soc Probl. 1997;44:174–199.
provide culturally appropriate HIV, substance tion. Mahwah, NJ: Lawrence Erlbaum Associ- 24. Radloff LS. The CES-D scale: a self-report de-
use, mental health, education, employment, and ates Inc; 1998:199–214. pression scale for research in the general popu-
other social services for transgender persons. In 6. Clements-Nolle K, Wilkinson W, Kitano K, lation. Appl Psychol Meas. 1977;1:385–401.
addition, because transgender individuals ac- Marx R. HIV prevention and health service 25. Mulrow CD, Williams JW, Geretty MB, Ramirez
cess the health care system for hormone ther- needs of the transgender community in San G, Montiel OM, Kerber C. Case-finding in-
Francisco. In: Bockting W, Kirk S, eds. Trans- struments for depression in primary care set-
apy, medical providers can provide an impor- gender and HIV: Risks, Prevention, and Care. tings. Ann Intern Med. 1995;122:913–921.
tant link to needed services. Binghamton, NY: The Haworth Press Inc; 2001. 26. Weissman MM, Sholomskas D, Pottenger M,
7. Elifson KW, Boles J, Posey E, Sweat M, Dar- Prusoff BA, Locke BZ. Assessing depressive
row W, Elsea W. Male transvestite prostitutes symptoms in five psychiatric populations: a val-
Contributors and HIV risk. Am J Public Health. 1993;83: idation study. Am J Epidemiol. 1977;106:
K. Clements-Nolle, R. Marx, and M. Katz conceived 260–262. 203–214.
and designed the study. K. Clements-Nolle and R. 8. Modan B, Goldschmidt R, Rubinstein E, et al. 27. Gallo D, George JR, Fitchen JH, Goldstein AS,
Guzman implemented the study protocols and su- Prevalence of HIV antibodies in transsexual and Hindahl MS. Evaluation of a system using oral
pervised data collection, coding, and entry. K. female prostitutes. Am J Public Health. 1992; mucosal transudate for HIV-1 antibody screen-
Clements-Nolle analyzed the study data. R. Marx and 82:590–592. ing and confirmatory testing. JAMA. 1997;277:
M. Katz assisted with data analysis and interpreta- 9. Spizzichino L, Casella P, Zaccarelli M, Rezza 254–258.
tion. K. Clements-Nolle conducted the literature re- G, Venezia S, Gattari P. HIV infection among 28. Mansergh G, Marks G. Age and risk of HIV in-

920 American Journal of Public Health June 2001, Vol. 91, No. 6
 fection in men who have sex with men. AIDS. Francisco/Berkeley Young Men’s Survey. JAMA. tempts in nine countries. Psychol Med. 1999;
1998;12:1119–1128. 1994;272:449–454. 29:9–17.
29. Hosmer DW, Lemeshow S. Applied Logistic Re- 36. Osmond DH, Page K, Wiley J, et al. HIV infec- 42. Cochran SD, Mays VM. Lifetime prevalence of
gression. New York, NY: John Wiley & Sons tion in homosexual and bisexual men 18 to 29 suicide symptoms and affective disorders among
Inc; 1989:135–175. years of age: the San Francisco Young Men’s men reporting same-sex sexual partners: results
30. Menard S. Applied Logistic Regression Analysis. Health Study. Am J Public Health. 1994;84: from NHANES III. Am J Public Health. 2000;
Thousand Oaks, Calif: Sage Publications Inc; 1933–1937. 90:573–578.
1995:71–72. 37. Wilson MJ, Marelich WD, Lemp GF, Ascher 43. van Kesteren PJ, Asscheman H, Megens JA,
31. McFarland W, Kellogg T, Nieri G, et al. San MS, Kerndt P, Kizer KW. HIV-1 seroprevalence Gooren LJ. Mortality and morbidity in trans-
Francisco HIV Epidemiology Report: Data among women attending sexually transmitted sexual subjects treated with cross-sex hormones.
Available to 1998. San Francisco, Calif: San disease clinics in California. West J Med. 1993; Clin Endocrinol. 1997;47:337–342.
Francisco Department of Public Health, HIV 158:40–43. 44. Garofalo R, Wolf RC, Wissow LS, Woods ER,
Seroepidemiology Unit; December 1998. Goodman E. Sexual orientation and risk of sui-
38. Tirelli U, Vaccher E, Covre P, Corsco C, Ser-
32. Moss AR, Vranizan K, Gorter R, Bacchetti P, cide attempts among a representative sample of
raino D, Rezza G. Condom use among trans-
Watters J, Osmond D. HIV seroconversion in youth. Arch Pediatr Adolesc Med. 1999;153:
vestites in Italy. J Acquir Immune Defic Syndr.
intravenous drug users in San Francisco, 1985– 487–493.
1991;4:302–303.
1990. AIDS. 1994;8:223–231. 45. Waldo CR, Hesson-McInnis MS, D’Augelli AR.
33. Watters JK, Bluthenthal RN, Kral AH. HIV se- 39. Watters JK, Estilo MJ, Clark GL, Lorvick J. Sy- Antecedents and consequences of victimization
roprevalence in injection drug users. JAMA. ringe and needle exchange as HIV/AIDS pre- of lesbian, gay, and bisexual young people: struc-
1995;273:1178. vention for injection drug users. JAMA. 1994; tural model comparing rural university and
34. Watters JK. Trends in risk behavior and HIV se- 271:115–120. urban samples. Am J Community Psychol. 1998;
roprevalence in heterosexual injection drug users 40. Kral AH, Lorvick J, Edlin BR. Sex- and drug- 26:307–334.
in San Francisco, 1986–1992. J Acquir Immune related risk among younger and older injection 46. D’Augelli AR, Hershberger SL, Pilkington NW.
Defic Syndr. 1994;7:1276–1281. drug users in San Francisco. J Acquir Immune Lesbian, gay, and bisexual youth and their fam-
35. Lemp GF, Hirozawa AM, Givertz D, et al. Se- Defic Syndr. 2000;24:162–167. ilies: disclosure of sexual orientation and its con-
roprevalence of HIV and risk behaviors among 41. Weissman MM, Bland RC, Canino GJ, et al. sequences. Am J Orthopsychiatry. 1998;68:
young homosexual and bisexual men: the San Prevalence of suicide ideation and suicide at- 361–371.

June 2001, Vol. 91, No. 6 American Journal of Public Health 921