JMIR Publications

ABSTRACT

Background:

Mobile health (mHealth) applications have become important tools for HIV/AIDS management. While the Mobile App Rating Scale (MARS) and its user version (uMARS) are widely used for quality assessment, how different stakeholder groups combine quality dimensions to form overall judgments remains unclear. Discrepancies between researcher evaluations and patient perceptions may critically affect app adoption and clinical effectiveness.

Objective:

This exploratory study aimed to compare the configurational pathways associated with high subjective quality ratings of HIV/AIDS apps between researchers and patients using fuzzy-set qualitative comparative analysis (fsQCA).

Methods:

Twenty-one HIV/AIDS mobile applications were evaluated by three researchers using the original MARS and two HIV-positive patients using the uMARS. Six conditions were examined: engagement, functionality, aesthetics, information quality, privacy protection, and developer type. Necessary and sufficient configurations were identified for both evaluator groups. Group differences in raw ratings were tested with Wilcoxon signed-rank tests. Robustness was examined through sensitivity analyses across four calibration anchor sets, four raw-consistency cutoffs, four proportional reduction in inconsistency (PRI) cutoffs, and an alternative multi-value operationalization of privacy protection.

Results:

Researcher and patient ratings differed significantly for functionality (median difference 0.50; P=.003) and engagement (median difference −0.20; P=.04), but not for aesthetics (P=.16), information (P=.07), or subjective quality (P=.06). Five sufficient configurations were identified for researchers (solution consistency=0.981, coverage=0.649) and four for patients (solution consistency=0.998, coverage=0.796). Configurations that included privacy protection appeared only among researchers, whereas configurations that included the absence of privacy protection were common among patients. For patient evaluations, functionality showed the highest necessity consistency (0.907); information quality was also high (0.915) and strongly correlated with functionality (r=0.816), so neither was interpreted as an independent necessary condition. Illustrative cases included Wu Ai Fang Hua and Ai Yi Jian for the researcher-oriented pathway and Chabei for the patient-oriented pathway; Chabei was the only application appearing as an illustrative case in both groups. Sensitivity analyses left the configurational structure unchanged.

Conclusions:

Researcher and patient evaluations of HIV/AIDS mHealth applications appear to follow different configurational logics, with several equifinal pathways to high subjective quality in each group and no single configuration common to all pathways. Because the patient panel comprised only two evaluators and 21 applications were analyzed with six conditions, these findings should be read as hypothesis-generating rather than as representative of people living with HIV. They nonetheless suggest that both expert and end-user perspectives are needed during mHealth development, and that the gap between expert and user attention to privacy deserves regulatory and educational attention in a population that is highly vulnerable to disclosure-related harm.