Optimizing machine learning-based identification of sexual health influencers for HIV self-testing distribution among men who have sex with men in China: a secondary analysis of a quasi-experimental trial.
Among sexual health influencers distributing HIV self-tests to peers, effective distribution was associated with stronger awareness of peers' HIV status, more active sharing of HIV information, and having been accompanied during one's own HIV test.
Key Findings
Results
Among 196 sexual health influencers (SHIs) included in the analysis, 46.9% achieved effective distribution, defined as motivating at least one peer to test.
The analysis was a secondary analysis of a parent quasi-experimental trial (ChiCTR2000039632) conducted in five provinces in southern China.
Mean age of SHIs was 28 years.
196 SHIs with complete follow-up were included.
Participants were randomized to either an empirical scale or machine-learning identification arm, and SHIs were then selected within each arm.
Results
A total of 286 unique peers returned verified HIV self-testing results, of whom 118 (41.3%) were first-time testers.
Secondary distribution of HIV self-testing was able to reach a substantial proportion of individuals who had never previously tested.
Verified HIV self-testing results were used to confirm peer uptake.
The trial targeted men who have sex with men (MSM) in China.
Results
Stronger awareness of peers' HIV status was associated with greater peer HIV testing uptake among SHIs.
Adjusted incidence rate ratio (aIRR) for awareness of peers' HIV status: 1.31 (95% CI 1.16–1.48).
This association was identified using multivariable Poisson LASSO models with multiple imputation and cross-validation.
Predictor variables captured sociodemographic, behavioral, and network characteristics.
Results
More active sharing of HIV information by SHIs was associated with greater peer HIV testing uptake.
aIRR for active sharing of HIV information: 1.28 (95% CI 1.10–1.50).
This was one of the key behavioral predictors identified in the multivariable Poisson LASSO model.
The finding suggests that SHIs who more proactively disseminate HIV-related information are more effective at motivating peers to test.
Results
SHIs who had been accompanied during their own HIV test recruited more newly tested peers.
aIRR for having been accompanied during one's own HIV test: 1.94 (95% CI 1.12–3.37).
This was the strongest predictor of recruiting newly tested (first-time tester) peers.
The finding implies that personal testing experience with social support may enhance an SHI's ability to engage untested peers.
Background
The parent quasi-experimental trial found that SHIs identified by a machine-learning model achieved greater peer uptake than those identified by an empirical scale.
The parent trial was registered as ChiCTR2000039632.
The current secondary analysis sought to identify factors driving distribution effectiveness beyond the comparison of identification methods.
The machine-learning arm outperformed the empirical scale arm in the parent trial, but specific factors explaining this difference were unclear prior to this analysis.
Discussion
The authors identified a need for capacity strengthening related to peer engagement in HIV self-testing based on the identified predictors of effective distribution.
Findings have implications for secondary distribution of HIV self-test kits.
Modifiable factors such as HIV information sharing and awareness of peers' HIV status suggest potential intervention targets.
The experience of being accompanied during HIV testing highlights the role of social support in enabling SHIs to engage their networks.
What This Means
This research studied how HIV self-testing kits can be distributed more widely through trusted community members—called sexual health influencers (SHIs)—among men who have sex with men in China. The study analyzed data from a trial in five Chinese provinces to understand what personal characteristics and behaviors made some SHIs more effective than others at encouraging their peers to get tested. Out of 196 SHIs studied, about half successfully motivated at least one peer to test, and over 40% of the peers who tested were doing so for the first time.
The study found three main factors linked to SHIs being more effective at distributing HIV tests. SHIs who were more aware of their peers' HIV testing status, those who more actively shared HIV-related information, and those who had themselves been accompanied by someone during their own HIV test were all significantly more likely to get peers tested. Notably, having had a supportive companion during their own HIV test nearly doubled the likelihood that an SHI would recruit a first-time tester.
This research suggests that training and supporting community members to distribute HIV self-tests should focus on building their knowledge of their social networks' testing status and their confidence in sharing HIV information. It also suggests that encouraging people to bring a friend or partner when getting tested themselves could have downstream benefits—those who had that supportive experience may be better equipped to bring others into HIV testing. These findings could help programs that use peer networks to expand HIV testing reach more people who have never been tested before.
Zou Q, Zhou Y, Zhang Y, Wu D, Zhang X, Jing F, et al.. (2026). Optimizing machine learning-based identification of sexual health influencers for HIV self-testing distribution among men who have sex with men in China: a secondary analysis of a quasi-experimental trial.. Sexual health. https://doi.org/10.1071/SH25214