Financial Incentives for Linkage to Care and Viral Suppression Among HIV-Positive Patients

1Antiretroviral therapy resulting in viral suppression dramatically reduces human immunodeficiency virus (HIV)-related morbidity and risk of HIV transmission. To realize these benefits, gaps in the HIV care continuum must be minimized. In the United States, approximately 13% of HIV-positive individuals are unaware of their HIV infection and only 55% of patients diagnosed with HIV have achieved viral suppression.

2Financial incentives (FI) can assist in achieving desirable health behaviors. In a meta-analysis of clinical trials that used cash transfers for health behaviors, 10 of 11 studies demonstrated a positive effect, including 3 for medication adherence. Other studies have shown their effectiveness in decreasing risk for acquisition of HIV and other sexually transmitted infections.

3The potential for FI to close the gaps in the HIV care continuum motivated the HIV Prevention Trials Network (HPTN) study, HPTN 065, which evaluated the effectiveness of FI for improving linkage to care and viral suppression in the Bronx, New York, and Washington, DC, 2 communities severely affected by HIV.

4Bronx, New York, and Washington, DC, HIV test sites with the highest number of newly HIV diagnosed individuals and HIV care sites treating the largest number of HIV-positive patients, determined based on 2008 to 2009 HIV surveillance data, were approached for potential study participation. The trial protocol is available in Supplement 1. The CONSORT diagram in Figure 1 shows the cascade of site selection and randomization. HIV test and care sites were separately randomized to FI or standard of care (SOC). Test sites provided FI from April 2011 until December 2012, and care sites from February 2011 through January 2013. Data from US surveillance system for 2010 were considered as baseline data.

5Individuals who tested positive for HIV at an FI HIV test site received a coupon redeemable within 3 months for 2 cash-equivalent gift cards: $25 on getting blood drawn for HIV-related tests and $100 on meeting with a clinician and developing a care plan. Few HIV test sites offered coupons to known HIV-positive persons who had not been in care for at least 12 months.

6HIV-positive patients receiving antiretroviral therapy at FI care sites were eligible for incentives only if they were already engaged in care at that site (ie, had at least 1 prior viral load measurement there within the last 3 to 9 months). Such patients could qualify for a $70 gift card if their plasma viral load was suppressed (HIV RNA <400 copies/mL), for a maximum of once every 3 months for the duration of the FI component of the study (from February 2011 through January 2013). Clinicians received information on prevailing HIV treatment guidelines.

7The outcomes for linkage to care, viral suppression, and continuity in care were assessed using data routinely reported to each jurisdiction’s HIV surveillance system. All confirmed diagnoses of HIV from HIV test sites were reported by providers and specific HIV-related test results (positive findings on Western blot, HIV RNA viral load, and CD4+ cell count) were reported by laboratories to the local surveillance system. Site-aggregated data per calendar quarter from 2010 through 2013 were transmitted to the HPTN statistical and data management center for analysis.

8For linkage to care, the primary outcome was the proportion of individuals testing positive at each HIV test site who linked to care within 3 months, as indicated by those individuals’ record of CD4+ cell count or viral load test result in the surveillance system. HIV-positive patients were eligible if they were either newly diagnosed (based on no previous report in the surveillance system of a positive Western blot finding) or previously diagnosed but out of care (evidenced by no reported CD4+ cell count or viral load test result in the prior year).

9For viral suppression, the primary outcome was the proportion of established patients at each HIV care site with suppressed viral load, assessed at each calendar quarter. For each quarter, patients were considered established in care if HIV laboratory test results were reported from that site in 2 different calendar quarters over the prior 15 months. A patient was considered virally suppressed if a viral load had been measured within 6 months and the most recent viral load was less than 400 copies/mL. Patients with no viral load test results in the past 6 months were assumed not suppressed. Viral suppression was evaluated for each site based on all viral load test results, regardless of whether these were from patients on antiretroviral therapy (treatment status is not captured in the surveillance system). Consistent viral suppression at baseline was defined as having a viral load of less than 400 copies/mL in every quarter in 2010.

10For continuity in care (among established patients), the outcome was defined as the proportion of patients with CD4+ cell count or HIV viral load test results in the surveillance system during at least 4 of the prior 5 calendar quarters.

11Site randomization aimed to achieve balance between arms by site size and preintervention endpoints, as measured by prerandomization (2008-2009) surveillance data: for test sites, balance in both the number of HIV-positive individuals identified in the previous year and the proportion of those linked to care within 3 months of HIV diagnosis; for care sites, balance in both the number of HIV-positive patients in care in the previous year and the proportion with viral suppression.

12For the linkage to care intervention, a sample size of 40 HIV test sites (20 per arm) was estimated to provide 80% power to detect a 13% increase in linkage to care, assuming an average of 54 HIV diagnoses per site and an intraclass correlation of 0.18. For the viral suppression intervention, a sample size of 40 HIV care sites (20 per arm) was estimated to provide 90% power to detect a 6% increase in viral suppression, assuming an average of 220 patients per site and an intraclass correlation of 0.10.

13The linkage to care intervention effect was estimated using logistic regression for a binary outcome of linkage to care adjusted for baseline site proportion linked to care and arm, using a generalized estimating equation (GEE) with exchangeable correlation structure.

14The intervention effect for viral suppression was estimated using linear regression for the site-aggregated proportion of virally suppressed patients in each of the 5 quarters with arm as the primary covariate, adjusting for sites’ baseline proportion virally suppressed, weighted by the average number of patients at each site during the assessment period, and using a GEE approach with exchangeable correlation structure to account for 5 quarterly measures for each site. The same modeling approach was used for continuity in care, but adjusted for site’s baseline proportion of continuity in care. Thus, the model estimates an intervention effect corresponding to the increase in probability of viral suppression for a patient in care at a site implementing FI compared with SOC. Analysis of the peak intervention quarter (fourth quarter of 2012) for viral suppression, a prespecified sensitivity analysis, used simple weighted linear regression.

15Four prespecified subgroup site analyses were planned: study community (Bronx vs Washington, DC), smaller vs larger sites (≤/> median number of patients), hospital vs community-based sites and lower vs higher percent with viral suppression at baseline (≤/> median percent). In addition, the effect of FI on patients not consistently virally suppressed prior to the intervention was assessed by excluding from the analysis all patients who achieved viral suppression in every quarter of 2010.

16The study involved minimal risk. Because outcomes were reported using only quarterly site-aggregated data already reported to surveillance, no additional individual data were collected from sites. Each site’s affiliated institutional review board approved the study with a waiver of patient informed consent granted under 45 CFR 46.116 (d). The protocol is available in Supplement 1.

17Study site selection is shown in Figure 1 and baseline site characteristics are shown in eTable 1 in Supplement 2. A total of 37 test sites (18 in the Bronx and 19 in Washington, DC) participated in the linkage to care component and 34 were included in the analysis (3 were excluded because no HIV-positive patients were identified). During the baseline year, the mean (SD) number of HIV-positive cases was 35 (45), and sites’ mean (SD) linkage to care rate was 74% (29%). Characteristics of HIV-positive patients at baseline by arm were not statistically different (eTable 2 in Supplement 2).

18A total of 39 HIV care sites participated in the viral suppression component (20 in the Bronx and 19 in Washington, DC), with 37 included in the analysis: 1 site was excluded owing to viral load test results not reported electronically to the surveillance system, and 2 sites within the same facility randomized to the same arm were combined owing to the inability to disaggregate their data in the surveillance system. At baseline, the mean (SD) number of patients in care was 374 (478), with a mean (SD) of 62% (16%) of patients virally suppressed. Characteristics of patients at HIV care sites by arm were not statistically different (eTable 3 in Supplement 2).

19A total of 1159 HIV-positive individuals were identified from test sites during the assessment; 389 in the Bronx and 770 in Washington, DC. The mean number of HIV-positive individuals diagnosed per site was 34. Among 914 patients identified in 2012 (303 in the Bronx and 611 in Washington, DC), 641 (70%) were men, 384 (42%) were men who have sex with men (99 [33%] in the Bronx and 285 [47%] in Washington, DC), 630 (69%) were black (151 [50%] in the Bronx and 479 [78%] in Washington, DC), 198 (22%) were Hispanic (140 [46%] in the Bronx and 58 [9%] in Washington, DC), and 150 (16%) were younger than 25 years (43 [14%] in the Bronx and 107 [17%] in Washington, DC). Characteristics were similar by study arm.

20A total of 1061 coupons were dispensed, 238 in the Bronx and 823 in Washington, DC, and 79% were redeemed for both $25 and $100 gift cards (Table 1).

21Financial incentives did not significantly increase linkage to care compared with SOC with adjusted odds ratio 1.10 (95% CI, 0.73-1.67; P =.65), with no effects noted in predefined subgroups (Table 2).

22At baseline, 16 208 patients were established in care at the sites (9703 in the Bronx and 6505 in Washington, DC). At baseline in 2010, most (10 201, 63%) were men (5485 [57%] in the Bronx and 4716 [72%] in Washington, DC), 4518 (28%) were men who have sex with men (1631 [17%] in the Bronx and 2887 [44%] in Washington, DC), 9355 (58%) were black (4434 [46%] in the Bronx and 4921 [76%] in Washington, DC), 5231 (32%) were Hispanic (4826 [50%] in the Bronx and 405 [6%] in Washington, DC), and 945 (6%) were younger than 25 years (575 in the Bronx and 370 in Washington, DC).

23A total of 9641 patients (5275 in the Bronx and 4366 in Washington, DC) were eligible for gift cards at FI care sites, with 41 530 visits that potentially qualified for a gift card and 39 359 (95%) gift cards dispensed.

24At baseline, mean (SD) overall viral suppression was 62% (16%) (eTable 1 and 4 in Supplement 2) and increased during the study at both FI and SOC sites (Table 3). The proportions of patients with viral suppression over time by site and study arm are shown in eFigure 1 in Supplement 2. Financial incentives had a statistically significant overall effect on viral suppression with a 3.8% (95% CI, 0.7%-6.8%; P =.01) higher proportion of virally suppressed patients at FI compared with SOC (Table 3) (Figure 2). The effect of FI was statistically significant at sites in Washington, DC (6.6% higher; 95% CI, 1.9%-11.3%; P =.006), at hospital-based sites (4.9% higher; 95% CI, 1.4%-8.5%; P =.007), as well as at sites with lower and higher baseline viral suppression (5.6% higher; 95% CI, 0.0%-11.3%; P =.05 and 3.6% higher; 95% CI, 0.3%-7.0%; P =.03, respectively), but not in the Bronx community-based sites or by size of the site. At the peak of the intervention, overall viral suppression was significantly higher by 4.6% (95% CI, 0.4%-8.8%; P =.031) at FI sites (eTable 5 in Supplement 2). Among the subpopulation of patients with viral load not consistently suppressed at baseline, viral suppression was significantly higher by 4.9% (95% CI, 1.4%-8.5%; P =.007) at FI sites (Table 3).

25The proportion of patients with continuity in care was higher by 8.7% (95% CI, 4.2%-13.2%; P <.001) at FI compared with SOC sites. This finding was consistent in both cities; at hospital and community-based sites, at smaller as well as larger sites; and sites with higher viral suppression at baseline (Table 3).

26This site-randomized, community-based study included a large number of sites that provide HIV testing and HIV care in the Bronx and Washington, DC, and for a substantial proportion of the HIV-positive patients in both communities. This study, the largest to evaluate the effectiveness of FI on HIV care-related behaviors, demonstrated that FI significantly increased by 3.8% the proportion of patients with viral suppression at FI compared with SOC sites. The effects observed were stronger in patients not consistently suppressed prior to the intervention. Financial incentives also substantially increased the proportion of patients reporting regularly for quarterly clinic visits. However, FI did not have a significant effect on linking HIV-positive individuals to care when compared with linkage at SOC sites.

27Few studies have assessed the efficacy of FI on HIV care-related behaviors. In 1 study conducted among people who inject drugs (PWID) in India, the use of vouchers for food or household goods was associated with enhanced linkage to care. In 2 US studies of PWID, contingency management combined with voucher prize drawings increased adherence to antiretroviral therapy. In 2 other US studies, 1 found that cue training with monetary reinforcement was associated with a transient increase in adherence but had no effect on viral suppression, whereas the other did not demonstrate benefit of FI on viral suppression among substance users.

28In this study, while FI had a statistically significant overall effect on the proportion of patients virally suppressed, this effect was larger in Washington, DC. This may be owing to the fact that New York State was an earlier adopter of initiating antiretroviral therapy, regardless of CD4+ cell count, which may have made it more difficult to detect the effect of FI at the Bronx sites. In addition, FI may have had a more pronounced influence in getting patients to make regular clinic visits and achieve viral suppression in Washington, DC, because a smaller proportion of patients made regular clinic visits at baseline compared with the Bronx. We also found a significant effect of FI at hospital-based care sites, possibly because they provided care for less financially secure patients for whom FI may be more meaningful.

29It should also be noted that, in this study, all patients on treatment at a site randomized to FI were eligible to receive them if virally suppressed, including those already virally suppressed. Thus, the larger effect of FI observed in the subset of patients not consistently suppressed at baseline, albeit an outcome not defined a priori, may reflect the potential added value of FI among such patients. In a study that assessed the effect of FI on adherence to warfarin, a significant effect was only noted in individuals with characteristics associated with nonadherence rather than in all patients.

30In this study, FI also significantly increased the proportion of patients regularly attending clinic visits. Poor retention in care has been associated with higher mortality. Regular clinical and laboratory monitoring is important to enable provision of adherence support, prevention counseling, or screening for other health conditions. However, care should be taken to safeguard against overutilization of clinic services motivated by FI.

31Use of FI to motivate behaviors remains controversial. We took measures to prevent untoward consequences. Prior to study initiation, we consulted with the study’s community advisory group to determine the appropriate value of the FI. In addition, we used a site- rather than individual-randomized design to avoid the potential disruption of services and perceived inequities if only some patients within a site received FI. Last, to discourage patients from transferring care to FI sites, we required that patients be in care at a site for at least 3 months before receiving a gift card.

32The study had several important strengths. It was a community-wide effort implemented at hospital and community-based sites with a wide range of patient volumes. The outcome measure was laboratory-reported viral suppression, a biological indicator rather than self-reported adherence. A unique feature was the use of data from the US HIV Surveillance System to measure outcomes, an approach particularly well-suited for assessing a community-level intervention and one that avoided the selection bias inherent in recruitment and follow-up of individual participants.

33The study also had several limitations. The effect of FI may have been diluted because, while only patients receiving antiretroviral therapy were eligible for FI, all patients in care at an HIV care site (including those not on treatment) were included in the assessment of the viral suppression outcome because treatment status is not available in the surveillance system. Overall data completeness was superior in the Bronx compared with Washington, DC; moreover, for some nonresident patients receiving care at sites in Washington, DC, CD4+ and viral load data were reported to their jurisdiction of residence rather than to Washington DC, possibly hindering complete capture of data. We also attempted to mitigate the potential of assessment bias caused by increased frequency of viral load assessment owing to FI by defining the viral load suppression outcome over a 6-month period consistent with recommendations in prevailing guidelines; however, we are unable to assess whether this approach was fully successful. Finally, the study had limited power to asses changes in linkage to care because linkage was high at baseline, and the number of HIV-positive individuals identified per site was substantially lower than anticipated.

34HPTN 065, the largest study to date to our knowledge to evaluate the effect of FI on HIV-related care outcomes, demonstrated the overall effectiveness of FI for achieving viral suppression and regular clinic attendance. While seemingly modest, an increase of 4% in viral suppression with FI may potentially have considerable clinical and preventive implications on a population level, particularly in settings and among patients with less robust viral suppression. A recent study indicated that one-third of HIV-positive patients in care in the United States had detectable viral load for substantial durations of time, limiting their individual benefit and increasing their risk for transmitting HIV to others. Our study also demonstrated both the feasibility of using FI in a large community-wide effort and the successful linkage of a large-scale research endeavor with the established HIV surveillance system. Further analyses are ongoing to determine the cost-effectiveness of FI on viral suppression based on our study findings. In our study, we found that FI had no significant effect on linkage to care, an area ripe for further research.

35In conclusion, while our findings offer an innovative intervention for achieving the treatment and prevention potential of antiretroviral therapy, a strategy that offers great promise for control of HIV in the United States and globally, more research is needed to determine how such an intervention can be implemented in programs and at scale.