Evaluating effect modification by HIV testing history to understand the mechanisms behind the impact of announcing HIV self-testing availability in a clinic system in Kenya

Bibliographic Data

ID22072791
AuthorsElizabeth A Kelvin (0000-0002-0566-8697, City University of New York, corresponding author), Matthew L Romo (0000-0002-3038-0644, City University of New York), Gavin George (0000-0001-7258-8470, Lund University), Joanne E Mantell (0000-0002-0748-5974, Columbia University), Eva Mwai, Samuel Kinyanjui, Eston Nyaga, Eston N Nyaga, Jacob Odhiambo (0000-0002-4234-9740), Jacob O Odhiambo, Kaymarlin Govender (0000-0002-9586-1510, University of KwaZulu-Natal)
Year2023
Volume10
Pages880070-880070
Publication date2023-01-06
Open AccessYes
TypeARTICLE
VenueFrontiers in Public Health (JOURNAL)
Journal identifiersISSN: 2296-2565 • E-ISSN: 2296-2565
PublisherFrontiers Media SA (PUBLISHER • CH)
DOI10.3389/fpubh.2022.880070
PMID36684866
OpenAlexW4313644663
LanguageEN
Citations received1
References cited13

Background In sub-Saharan Africa, truckers and female sex workers (FSWs) have high HIV risk and face challenges accessing HIV testing. Adding HIV self-testing (HIVST) to standard of care (SOC) programs increases testing rates. However, the underlying mechanisms are not fully understood. HIVST may decrease barriers (inconvenient clinic hours, confidentiality concerns) and thus we would expect a greater impact among those not accessing SOC testing (barriers prevented previous testing). As a new biomedical technology, HIVST may also be a cue to action (the novelty of a new product motivates people to try it), in which case we might expect the impact to be similar by testing history. Methods We used data from two randomized controlled trials evaluating the announcement of HIVST availability via text-message to male truckers ( n = 2,260) and FSWs ( n = 2,196) in Kenya. Log binomial regression was used to estimate the risk ratio (RR) for testing ≤ 2 months post-announcement in the intervention vs. SOC overall and by having tested in the previous 12-months (12m-tested); and we assessed interaction between the intervention and 12m-tested. We also estimated risk differences (RD) per 100 and tested additive interaction using linear binomial regression. Results We found no evidence that 12m-tested modified the HIVST impact. Among truckers, those in the intervention were 3.1 times more likely to test than the SOC ( p < 0.001). Although testing was slightly higher among those not 12m-tested (RR = 3.5, p = 0.001 vs. RR = 2.7, p = 0.020), the interaction was not significant ( p = 0.683). Among FSWs, results were similar (unstratified RR = 2.6, p < 0.001; 12m-tested: RR = 2.7, p < 0.001; not 12m-tested: RR = 2.5, p < 0.001; interaction p = 0.795). We also did not find significant interaction on the additive scale (truckers: unstratified RD = 2.8, p < 0.001; 12m-tested RD = 3.8, p = 0.037; not 12m-tested RD = 2.5, p = 0.003; interaction p = 0.496. FSWs: unstratified RD = 9.7, p < 0.001; 12m-tested RD = 10.7, p < 0.001, not 12m-tested RD = 9.1, p < 0.001; interaction p = 0.615). Conclusion The impact of HIVST was not significantly modified by 12m-tested among truckers and FSWs on the multiplicative or additive scales. Announcing the availability of HIVST likely served primarily as a cue to action and testing clinics might maximize the HIVST benefits by holding periodic HIVST events to maintain the cue to action impact rather than making HIVST continually available

Computer security · Confidentiality · Environmental health · Family medicine · Poisson regression · Population · Computer Science · HIV Research and Treatment · HIV/AIDS drug development and treatment · HIV/AIDS Research and Interventions · Medicine · Nursing

Tools

Open DOIOpen Access