Community-based rapid HIV testing in homeless and marginally housed adults in San Francisco

Abstract

Background

Standard two-step HIV testing is limited by poor return-for-results rates and misses high-risk individuals who do not access conventional testing facilities.

Methods

We describe a community-based rapid HIV testing programme in which homeless and marginally housed adults recruited from shelters, free meal programmes and single room occupancy hotels in San Francisco received OraQuick Rapid HIV-1 Antibody testing (OraSure Technologies, Bethlehem, PA, USA).

Results

Over 8 months, 1614 adults were invited to participate and 1213 (75.2%) underwent testing. HIV seroprevalence was 15.4% (187 of 1213 individuals) overall and 3.5% (37 of 1063) amongst high-risk individuals reporting no previous testing, a prior negative test, or previous testing without result disclosure. All 1213 participants received their results. Of 30 newly diagnosed persons who received confirmatory results, 26 (86.7%) reported at least one contact with a primary healthcare provider in the 6 months following diagnosis.

Conclusions

We conclude that community-based rapid testing is feasible, acceptable and effective based on the numbers of high-risk persons tested over a short period, the participation rate, the prevalence of new infection, the rate of result disclosure, and the proportion linked to care.

Introduction

Counselling and testing for HIV is an integral component of prevention efforts and linkage to lifesaving medical care [1]. However, poor return-for-results rates decrease the effectiveness of standard two-step testing [2], and inadequate testing opportunities for high-risk individuals contribute to late diagnoses and increased morbidity [3].

We developed a community-based rapid HIV testing programme to address these issues in San Francisco's homeless and marginally housed persons, a population with an HIV prevalence of 10.5% [4] who may avoid conventional facilities and neglect to return for results [5,6]. Traditional two-step testing, with a lag of 4 days between testing and result disclosure, found that only 67% of patients agreed to testing and 91% received results despite use of modest cash incentives [4].

In the current programme, homeless and marginally housed adults recruited from homeless shelters, free meal programmes, and single room occupancy (SRO) hotels in San Francisco were provided with single-visit counselling, rapid HIV testing, result disclosure, and referral to medical and social services. The number of high-risk persons tested was used as an indicator of feasibility, while the participation rate determined acceptability. The effectiveness of rapid testing was evaluated using the number of new diagnoses, the rate of result disclosure, and the proportion of newly diagnosed individuals linked to medical care. Additionally, we calculated the false-positive rate of the rapid test to assess its accuracy when used in a field setting.

Methods

Over a period of 8 months (August 2003 to March 2004), individuals were recruited from all homeless shelters and free meal programmes in San Francisco and from a sampling of SRO hotels with preferential selection of hotels offering services or subsidies for homeless persons, women and HIV-infected persons. The shelter and free meal programme directors and hotel managers were approached by the study coordinator and all agreed to participate, many providing private testing space.

Teams of four to six staff members visited the venues at times determined by hours of operation, approached eligible individuals, explained HIV testing, and provided appointment cards for testing. People were excluded if they were less than 18 years of age, did not speak English, or were grossly intoxicated or floridly psychotic as determined by the supervisor.

HIV testing occurred at the shelter, meal line, or hotel if the facility had space for confidential testing and at a nearby community research site if an appropriate space was not available. Confidential, code-based testing was conducted between 8 a.m. and 3 p.m. on the Monday following recruitment, a 1–3 day time lag. The programme utilized seven or eight HIV counsellors, two intake/discharge staff members, two phlebotomists or other staff trained in administering the OraQuick Rapid HIV-1 Antibody test (OraSure Technologies, Bethlehem, PA, USA) and one co-ordinator. Informed consent was obtained and participants received HIV pre- and post-test counselling, OraQuick Rapid HIV-1 Antibody testing of a blood sample obtained by fingerstick, and structured interviews for demographic and HIV risk information. OraQuick results were interpreted on site by two staff members; positive results were confirmed with an enzyme immunoassay (HIV-1 Antibody EIA; Quest Diagnostics, San Jose, CA, USA) and western blot (HIV-1 Western Blot; Unilab, Tarzana, CA, USA). Negative results were not confirmed, in accordance with Centers for Disease Control and Prevention (CDC) recommendations for rapid testing [7]. Individuals with positive OraQuick results who did not have a primary healthcare provider were immediately referred to local primary care clinics and social services and asked to return in 1 week for confirmatory results. Participants received $15 for rapid testing and $15 upon return for confirmatory results. Confirmed HIV-infected individuals were invited to participate in a quarterly assessment of participant-reported healthcare utilization over the following 6 months. The protocol was approved by the University of California at the San Francisco Committee on Human Research. Descriptive statistics, 95% confidence intervals, and results of a χ² test of proportions were obtained using the SAS Statistical Analysis package (SAS Institute, Cary, NC, USA).

Results

Patient characteristics

Of 1614 individuals approached, 1213 individuals (75.2%) were tested. Of those who participated, 532 (43.9%) were from SRO hotels, 333 (27.5%) from free meal programmes, and 348 (28.7%) from homeless shelters. Participants were predominately male and African American. Forty per cent [95% confidence interval (CI)=37.2–42.8%] of individuals reported sleeping on the street or in homeless shelters during the previous night. Forty-six per cent of participants reported injection drug use ever, 38.1% of men had a history of homosexual sex, and 33.0% had ever participated in sex trade for money or drugs (Table 1).

Table 1.   Patient characteristics

	Number	%	95% confidence interval
			Lower	Upper
Gender
Male	919	75.8	73.4	78.2
Female	259	21.3	19.0	23.6
Transgender male-to-female	26	2.1	1.3	2.9
Transgender female-to-male	9	0.7	0.2	1.2
Race/ethnicity
African American	547	45.4	42.6	48.2
Hispanic	91	7.5	6.0	9.0
Asian	25	2.1	1.3	2.9
Native American	37	3.1	2.1	4.1
Pacific Islander	19	1.6	0.9	2.3
Other/mix	98	8.1	6.6	9.6
White	389	32.3	29.7	34.9
Age (years)
18–29	85	7.0	5.6	8.4
30–39	215	17.7	15.6	19.9
40–49	483	39.8	37.1	42.6
≥50	430	35.5	32.8	38.1
Education
<12th grade	199	16.9	14.8	19.0
≥12th grade	978	83.1	81.0	85.2
Ever in prison
Yes	266	22	19.7	24.3
No	943	78	75.7	80.3
Currently homeless
Yes	478	40	37.2	42.8
No	716	60	57.2	62.8
Housing
Homeless ≥1 year	832	68.9	66.3	71.5
Homeless <1 year	374	31.1	28.5	33.7
Injection drug use ever			0.0	0.0
Yes	552	45.5	42.7	48.3
No	660	54.5	51.7	57.3
Men who have sex with men (ever)
Yes	360	38.1	35.0	41.3
No	585	61.9	58.7	65.0
Unprotected sex in past year (sexually active participants only)
Yes	340	43.7	40.2	47.2
No	439	56.4	52.8	59.9
Sex trade for money or drugs ever
Yes	396	33	30.3	35.7
No	804	67	64.3	69.7
Prior HIV testing (ever)
Yes	1039	85.7	83.6	88.6
No	174	14.3	12.4	16.4

Acceptability of rapid testing

Seventy-five per cent (95% CI=73.0–77.3%) of individuals agreed to rapid testing. Ninety-two per cent (95% CI=88.7–94.7%) of potential participants agreed to the test when testing occurred at the site of recruitment vs 70.9% (95% CI=68.3–73.4%) when participants were tested at a nearby community research site (P<0.0001). The acceptance rate also varied by the site of recruitment: 87.2 (shelters), 72.9 (meal lines) and 70.5% (SROs) (P<0.0001).

HIV prevalence

The HIV seroprevalence was 15.4% (95% CI=13.4–17.4%) for the entire sample and 9.4% (95% CI=7.2–11.5%) when individuals recruited from hotels with HIV services were excluded. The HIV seroprevalence at SROs, meal lines and shelters was 22.9, 13.2 and 6.0%, respectively, reflecting oversampling of SROs offering HIV services.

Of the 1213 individuals tested, 150 reported a previous positive test. Among the 1063 individuals reporting no prior testing, a negative test or prior testing without result disclosure, 37 (3.5%; 95% CI=2.54–4.76) tested positive. Nine of the newly diagnosed persons had never been tested, 19 reported a prior negative test, and nine had prior testing but had not received results. All 1213 individuals received same-day test results.

Of the 187 HIV-positive individuals, 85.5% were biologically male, 54.6% were 40–49 years of age, 17.4% were currently homeless, 65.2% reported injection drug use ever, and 80.4% of men reported a history of homosexual sex.

Successful referral to medical care

Of the 37 persons newly diagnosed with HIV infection, seven did not return for confirmatory results and had no further follow-up. Of the remaining 30 new positives, 26 (86.7%; 95% CI=69.3–96.2) reported at least one contact with a primary healthcare provider in the 6 months following diagnosis and had a median CD4 count of 385 cells/μL (range 83–979 cells/μL). Seven persons (22.6%; 95% CI=8.0–35.2%) had CD4 counts less than 200 cells/μL.

Performance of rapid testing

Of 1213 individuals screened, 189 (15.6%) tested OraQuick-positive (Table 2). Two individuals with positive rapid tests were confirmed HIV-negative by western blot, giving a false-positive rate of 1.1% (95% CI=0.13–3.8%) and a positive predictive value (PPV) of 98.9% (95% CI=96.3–99.9). Nine individuals declined confirmatory testing but were assumed to be positive because of the high PPV of the test [5].

Table 2.   Performance of rapid testing

	Number	Per cent of total (n=1213)	Per cent of OraQuick positives (n=189)
^* The nine unconfirmed positive OraQuick tests were classified as true positives because of the high positive predictive value of the test [5].
OraQuick negative	1024	84.4
OraQuick positive	189	15.6
Confirmed positive	178	14.7	94.2
Confirmed negative	2	0.16	1.1
Unconfirmed	9^*	0.74	4.8

Discussion

Community-based rapid HIV testing is a feasible, acceptable and effective means to provide HIV testing to a high-risk homeless and marginally housed population, as shown by the large numbers of high-risk persons tested over a short period, the 100% rate of result disclosure, and the high participation rate, prevalence of new infections and proportion linked to care.

An important component of feasibility, as determined by the ability to access and test large numbers of high-risk individuals over a relatively short period, was the establishment of working relationships with staff at the recruitment sites: SRO hotel managers, and shelter and meal line directors. We found that initial concerns about hidden agendas and maintaining confidentiality were allayed by face-to-face meetings with venue directors and addressing the common goal of providing medical and social services to transient populations.

Testing at the site of recruitment produced a higher acceptance rate (92.1%) than testing at a nearby community research site (70.9%), suggesting that the need to travel to an unfamiliar location was a barrier to HIV testing. The difference in acceptance rate by recruitment venue (87.2% in shelters, 72.9% in meal lines and 70.5% in SROs) may reflect the same phenomenon, because all testing of shelter participants was completed at the shelters while SRO residents were asked to test at the research site because of the lack of private space at SROs. When possible, it is preferable to test at the site of recruitment, although decisions on testing venues should factor in the need for private screening locations to ensure patient confidentially and controlled settings with stable temperatures and bright lighting to assure the accuracy of the OraQuick test.

The effectiveness of community-based rapid HIV testing is determined by the numbers of new diagnoses, the rate of result disclosure, and the proportion linked to care. We identified 37 undiagnosed HIV-infected individuals, representing potential increased survival, reduced healthcare costs, and decreased transmission as a result of early diagnosis. Nine of these persons reported previous testing without receipt of results, highlighting the importance of single-visit testing and result disclosure in transient populations. All 1213 persons tested in our programme received their OraQuick results and all persons with positive results were immediately referred to care, a key component of any community-based testing programme. Developing prompt and durable linkages to ongoing medical care is essential to improve clinical outcomes in marginalized populations.

We calculated a false-positive rate of 1.1% (95% CI=0.13–3.8%), while the manufacturer reported no false positives (95% CI=0–0.3%) in clinical trials [7,8]. The two false-positive tests occurred early in our programme and may have been a result of operator error as staff adjusted to testing in a potentially chaotic field setting. Regardless of cause, the nontrivial false-positive rate is generalizable to other community-based testing programmes. The false positives caused trepidation amongst the staff about providing ‘preliminary positive’ results. As a result, we revised our result disclosure protocol to emphasize the need to return for confirmatory results. While we included prior positives in this analysis to evaluate the performance characteristics of the OraQuick assay, future implementation of rapid testing should focus on high-risk HIV-negative and HIV-unknown individuals.

An important strength of this programme is its community-based recruitment of a wide cross-section of homeless and marginally housed individuals in an HIV endemic setting. Limitations include small numbers of young people and African American women (populations with large numbers of new infections in the USA [9]), limited generalizability to settings with lower HIV seroprevalence and programmes without cash incentives, and lack of data on individuals who refuse testing.

In summary, rapid HIV testing may be an effective tool in high-risk, marginalized populations for the diagnosis of HIV-positive individuals and subsequent linkage to medical care.

Acknowledgements

This work was supported by the NIH RO-1 MH54907, AA015287, and The Doris Duke Charitable Foundation.