Work

Development and Evaluation of an Ecological Momentary Assessment–Based mHealth Platform for Monitoring HIV PrEP Adherence

Year: 2026

Type: preprint

Abstract: Background Ecological momentary assessment (EMA) enables real-time behavioral monitoring and may reduce recall bias associated with conventional retrospective questionnaires. However, few studies have validated EMA-based adherence monitoring against pharmacologic biomarkers. This study evaluated the accuracy of an EMA-enabled mHealth platform for assessing adherence to oral HIV pre-exposure prophylaxis (PrEP), using dried blood spot (DBS) drug concentrations as the gold standard and traditional questionnaires as a comparator. Method A total of 342 real-world online PrEP users in China were enrolled in the Real-world E-consumers Cohort II of PrEP (RECOPE II) and followed at weeks 6, 12, and 18. The platform used event-triggered sampling for event-driven (ED) PrEP and fixed-interval sampling for daily PrEP to monitor medication adherence. Platform- and questionnaire-based adherence measures were compared with DBS drug concentrations by estimating sensitivity and specificity. Participants who never checked in on the platform, failed to return a DBS sample, or did not complete the corresponding questionnaire were excluded from the accuracy analysis. Result Among the ED users, the prevalence of optimal adherence identified by the platform at week 6, 12, and 18 was 30.0%, 36.4%, and 28.0%, respectively, compared with 42.5%, 43.0%, and 43.2% identified by the questionnaire. Among the daily users, the platform-measured prevalence was 55.9%, 69.6%, and 68.3%, while the corresponding questionnaire-based measures were 84.3%, 81.5%, and 83.3%.When validated against the DBS results among ED users, the platform demonstrated consistently higher specificity than questionnaires at all time points (0.64 vs. 0.63, 0.66 vs. 0.61, and 0.77 vs. 0.63), but lower sensitivity (0.19 vs. 0.53, 0.40 vs. 0.50, and 0.38 vs. 0.58). Similarly, the platform demonstrated higher specificity (0.69 vs. 0.25, 0.59 vs. 0.11, and 0.50 vs. 0.21), but comparable sensitivity (0.65 vs. 0.79, 0.86 vs. 0.78, and 0.86 vs. 0.86) than the questionnaire among the daily users. Conclusion The EMA-based mHealth platform demonstrated consistently higher specificity and comparable sensitivity for daily PrEP users than conventional questionnaires, supporting its validity for real-time adherence monitoring. These findings provide empirical evidence that EMA is a practical and scalable approach for monitoring medication adherence and may have broader applications in assessing health-related behaviors.

Source: Research Square

DOI: 10.21203/rs.3.rs-11044881/v1