JMIR Publications

ABSTRACT

Background:

Running-related injuries are common among recreational runners, and access to sports medicine consultation in Indonesia remains limited, particularly outside large cities. Artificial intelligence (AI)–based mobile health apps that use large language models could provide personalized, low-cost injury prevention and management guidance, but their usability and market acceptance have rarely been evaluated in non-English-speaking populations.

Objective:

This study aimed to evaluate the usability, perceived usefulness, perceived ease of use, product-market fit, and willingness to recommend and pay for Roopin, an AI-based running injury prevention and management app, among recreational runners in Indonesia, and to explore users’ experiences and suggestions for improvement.

Methods:

We conducted a sequential explanatory mixed methods study. Adult runners using iOS devices were recruited online and used the Bahasa Indonesia version of Roopin for 4 weeks, then completed the System Usability Scale (SUS), perceived usefulness (PU) and perceived ease of use (PEOU) scales, the Sean Ellis product-market fit question, the Net Promoter Score (NPS), and willingness-to-pay items; engagement was recorded through usage logs. Prespecified hypotheses were tested using one-sample tests, rank-based tests, and logistic regression. A purposive subsample took part in focus group interviews, analyzed using reflexive thematic analysis and integrated with the quantitative findings in a joint display.

Results:

Of 478 registrations, 300 participants entered the cohort (mean age 30.9, SD 8.2 years; 93/300, 31.0% women) and 171 (57.0%) completed the end-of-study questionnaire. The median SUS score was 75.0, above the benchmark of 68 (Wilcoxon signed rank test, P=.01; mean 71.2, SD 18.3). Mean PU was 4.19 (SD 0.74) and mean PEOU was 4.16 (SD 0.74), both above the 3.5 benchmark (both P<.001), and each independently predicted intention to continue use (PU: odds ratio 2.37; PEOU: odds ratio 2.96). Market indicators were weaker: 17 of 171 participants (9.9%) would be very disappointed without the app (vs the 40% benchmark; P<.001), the NPS was +10.5, and 38 (22.2%) were willing to pay (vs 30%; P=.03). Overall, 194 of 300 participants (64.7%) opened the app at least once, with a median of 11 days of use (IQR 7-17); 106 (35.3%) never opened it. Ten focus groups (33 participants) highlighted the need for automatic data synchronization for non-Apple Watch devices, more concise AI responses, and stronger engagement features.

Conclusions:

Among participants who completed the study, Roopin showed above-average usability and high perceived usefulness and ease of use, and its AI injury guidance was rated among its strongest features. Early market indicators, however, did not reach prespecified thresholds: product-market fit, the Net Promoter Score, and willingness to pay all fell short, and engagement was low. Reliable data synchronization for non-Apple Watch devices, more concise AI responses, and stronger reasons to return are priorities for refinement before a larger effectiveness evaluation.