Challenges in measurement of linkage following HIV self‐testing: examples from the STAR Project

Knowledge of HIV status through HIV testing constitutes the 
first step towards HIV treatment and prevention services. HIV 
self-testing (HIVST), whereby individuals collect their own 
specimen, conduct their own test and interpret the results, 
allows individuals to learn their HIV status conveniently and 
privately, as well as to decide when and where to attend posttest 
services. Accurate estimation of the proportion of those 
tested who link to additional HIV care, treatment and prevention 
services is critical in quantifying the health impact of HIV 
testing. As HIVST becomes integrated into testing programmes 
worldwide, implementers in diverse settings will 
need to measure the effectiveness of their programmes to 
ensure self-testers link to onward care and services. This can 
be challenging, and community health programmes in many 
contexts find it difficult to track referral uptake and equity [1]. 
We draw upon experience from the Self-Testing in AfRica 
(STAR) Initiative in 2015 to 2017 to identify three lessons for 
measurement of linkage following HIVST. In STAR, two pragmatic 
cluster-randomized trials evaluated the effectiveness of 
continuous HIVST distribution over 12 months in increasing 
testing coverage and linkage to care in Malawi and Zambia. A 
third trial, in Zimbabwe, evaluated the effectiveness of an 
incentive to promote linkage following a short, campaign-style 
HIVST distribution programme, and included a non-randomized 
component assessing the association between HIVST distribution 
and antiretroviral therapy (ART) initiations in nearby clinics. 
Details are provided elsewhere [2,3]. Each trial incorporated a 
household survey and data collection from health facilities to 
evaluate changes in HIV testing coverage and linkage to confirmatory 
testing, care and prevention (Table 1).

Knowledge of HIV status through HIV testing constitutes the first step towards HIV treatment and prevention services. HIV self-testing (HIVST), whereby individuals collect their own specimen, conduct their own test and interpret the results, allows individuals to learn their HIV status conveniently and privately, as well as to decide when and where to attend posttest services. Accurate estimation of the proportion of those tested who link to additional HIV care, treatment and prevention services is critical in quantifying the health impact of HIV testing. As HIVST becomes integrated into testing programmes worldwide, implementers in diverse settings will need to measure the effectiveness of their programmes to ensure self-testers link to onward care and services. This can be challenging, and community health programmes in many contexts find it difficult to track referral uptake and equity [1].
We draw upon experience from the Self-Testing in AfRica (STAR) Initiative in 2015 to 2017 to identify three lessons for measurement of linkage following HIVST. In STAR, two pragmatic cluster-randomized trials evaluated the effectiveness of continuous HIVST distribution over 12 months in increasing testing coverage and linkage to care in Malawi and Zambia. A third trial, in Zimbabwe, evaluated the effectiveness of an incentive to promote linkage following a short, campaign-style HIVST distribution programme, and included a non-randomized component assessing the association between HIVST distribution and antiretroviral therapy (ART) initiations in nearby clinics. Details are provided elsewhere [2,3]. Each trial incorporated a household survey and data collection from health facilities to evaluate changes in HIV testing coverage and linkage to confirmatory testing, care and prevention (Table 1).

LESSON 1. COLLECT USAGE INFORMATION FROM USERS, NOT JUST FROM DISTRIBUTORS AND CLINICS
User household surveys and qualitative research are typically more time-consuming to collect than routine data from distributors and clinics. However, this type of data is vital for understanding how self-tests are being used. For example, in Zimbabwe, 289 survey respondents reported a reactive HIVST, of whom 216 (75%) were already on ART [2]. Formative qualitative work indicated that individuals already known to be positive self-tested in trial communities to avoid inadvertent disclosure, to confirm their status or check for cure. Without accounting for unintended use of HIVST by persons already diagnosed and linked into HIV care, the number of HIVSTs used by undiagnosed individuals will be overestimated, and both linkage and cost-effectiveness measures are likely to be overestimated.

LESSON 2. THE QUALITY OF LINKAGE MEASURES USING CLINIC DATA DEPEND NOT ONLY ON THE QUALITY OF CLINIC DATA, BUT ALSO ON THE SELF-TESTING DISTRIBUTION MODEL AND LOCAL CONTEXT
In two STAR trials, we had contrasting findings about the effectiveness of self-test distribution on increasing ART initiations at local clinics, measured by routine clinic data. In Malawi, clinics in areas with self-test distribution had a 14% increase in ART initiations compared with clinics in areas without HIVST distribution (adjusted risk ratio (aRR) = 1.14; 95% CI: 0.75 to 1.75) [4], but the wide confidence intervals suggest limited statistical evidence for effectiveness. In contrast, the study in Zimbabwe showed a somewhat larger increase in local ART initiations in areas with HIVST distribution with stronger evidence for effectiveness (aRR=1.27; 95% CI: 1.14 to 1.43) [5].
These studies used different designs, and the trial in Zimbabwe had more clusters receiving HIVST distribution and more power to detect differences. However, we believe these estimates may also have been driven by the context and the duration of distribution. Universal test-and-treat interventions, including the HPTN 071 (PopART) and ANRS 12249 trials, have found that linkage to care remains low even after substantial efforts supporting testing and linkage. This suggests that it is necessary to consider how contextual factors apart from the availability of HIV testing services impact linkage to care [6,7]. The Zimbabwe study used a shorter, campaign approach to distributing HIVST and the implementing organization (Population Services International) had initiatives facilitating linkage-to-care across the study area. In contrast, distribution in Malawi was continuous over 12 months, with linkage more diffuse across an extended period. It is likely easier to detect increased linkage at facilities if HIVST kits are distributed in short concentrated periods and in contexts where linkage after testing is relatively easy for users. For programmes working in areas or in populations where linkage is difficult, or for those distributing HIVST over an extended period, alternative methods of measuring linkage may provide more power to detect effects.

LESSON 3. MEASURE BOTH LINKAGE TO TREATMENT AND TO PREVENTION
Increasing the availability and uptake of HIV prevention services is increasingly recognized as necessary for reducing HIV incidence rates [8]. As testing and treatment expand, the number of new diagnoses identified by testing programmes is likely to decrease, and consequently, the denominators of linkage-to-care measures will shrink. Expanding to include linkageto-prevention of persons not infected with HIV will increase power to detect impacts of HIVST on population health. In the STAR trials described above, linkage to prevention was a consideration, but not the focus. For example, in Zambia, voluntary medical male circumcision (VMMC) mobilizers affiliated with HIVST intervention clinics received HIVST kits to distribute to clients. However, only four of the six intervention clinics had affiliated VMMC mobilizers and the proportion of test kits distributed within each area by VMMC mobilizers was low overall (<1% to 11% across areas). There was a positive association between the intervention and self-reported VMMC uptake, but this was not statistically significant (aRR: 1.36; 95% CI: 0.49 to 3.78) [5]. With collaborators, STAR is now undertaking trials to assess whether HIVST can be used to generate demand for prevention services, including pre-exposure prophylaxis (PrEP) among adolescent girls and young women in KwaZulu-Natal, South Africa [9] and VMMC among men in urban Zimbabwe.
To summarize, defining and estimating linkage following HIVST is complex, and there is no single measurement strategy that will fit the needs of all HIVST implementers. Researchers and implementers must account for contextual factors that may affect users' likelihood of linking and the amount of time required to link when developing strategies to measure linkage following HIVST. In STAR studies, it was difficult to anticipate who would use HIVST, and where and when self-testers would link to care. Where possible, collecting data on users' self-testing and linkage experiences can provide valuable insight.