Abstract
BACKGROUND: Viral suppression is key to ending the HIV epidemic, yet only 58% of people living with HIV (PLHIV) in sub-Saharan Africa are suppressed. Cash transfers are an effective strategy to improve retention in care, but little is known about optimization of implementation; for example, designing effective programs that integrate into existing clinic workflows. We studied implementation of an mHealth system to deliver cash transfers to support retention. METHODS: We conducted a mixed-methods study assessing implementation of an mHealth cash transfer study. This was part of a larger, hybrid implementation-effectiveness randomized controlled trial evaluating cash transfers conditional on visit attendance for viral suppression among Tanzanian PLHIV initiating ART. An mHealth system using fingerprint identification and mobile payments was used to automatically disburse mobile money to eligible PLHIV. We used Proctor's framework, assessing implementation of the mHealth system from the perspectives of PLHIV and clinicians. We analyzed mHealth system data and conducted surveys (n = 530) and in-depth interviews (n = 25) with PLHIV, clinic and pharmacy staff (n = 10), and structured clinic observations (n = 2293 visits). RESULTS: One thousand six hundred fifty-one cash transfers were delivered to 346 PLHIV in the cash arms, 78% through mobile money. Among those in the cash arms, 81% registered their mobile money account with the mHealth system by study end, signaling high adoption. While acceptability for fingerprinting and mobile payments was high among PLHIV, interviews revealed mixed views: some had privacy concerns while others felt the system was secure and accurate, and provided some legitimacy to the clinical visits. Pharmacists praised system efficiency, but concerns about duplicative recordkeeping and added work arose. Clinic staff voiced excitement for the system's potential to bring the cash program to all patients and simplify workflows; yet concerns about multiple systems, staffing, and intermittent connectivity tempered enthusiasm, highlighting structural issues beyond program scope. Structured observations revealed a steep learning curve; repeat fingerprint scans and manual entry declined as the system improved. CONCLUSIONS: Biometric identification and mobile payments were acceptable to most patients and staff. Fingerprinting encountered some feasibility limitations in the first months of testing; however, mobile payments were highly successful. Biometric identification and mobile payments may provide a scalable mechanism to improve patient tracking and efficiently implement financial incentives in low-resource settings. TRIAL REGISTRATION: Name of the registry: clinicaltrials.gov Trial registration number: NCT03351556 Date of registration: 11/24/2017 Checklists: StaRI (included with submission). Note CONSORT for cluster-randomized trials was used for the main trial but is not directly applicable to this manuscript.