Adequate sleep is crucial for mood regulation and injury prevention among high school female athletes. However, many in this demographic experience insufficient sleep, which can negatively impact their performance and overall well-being. This study investigates the feasibility and preliminary effects of an AI-supported mobile health (mHealth) intervention that provides personalized sleep feedback to high school female soccer players.
Study Overview
The pilot randomized controlled trial involved 25 high school female soccer players from a single team. The study commenced with a briefing and system setup on August 18, 2025, and was retrospectively registered on April 7, 2026. Participants were stratified by median baseline total sleep time (TST) and Pittsburgh Sleep Quality Index (PSQI) scores, and randomly allocated to either an intervention group (n=12) or a control group (n=13).
During the intervention phase, which lasted from weeks 6 to 9, the intervention group received personalized sleep feedback generated by GPT-4o via the LINE messaging platform on days with valid prior-night Fitbit data. In contrast, the control group received daily non-personalized sleep information. The primary outcome measured was sleep efficiency, while secondary and exploratory outcomes included TST, wake after sleep onset (WASO), subjective sleep quality, mood state, and sports injury severity.
Results
All participants completed the trial, demonstrating a 100% success rate in message delivery and questionnaire completion across both groups. The median valid Fitbit wear nights were 26 in the intervention group compared to 22 in the control group. Importantly, no major technical issues, adverse events, or privacy breaches were reported.
The primary efficacy outcome, sleep efficiency, showed a favorable change in the intervention group, with a Hodges-Lehmann between-group difference of 2% (95% CI 0.50%-4%; P=.046; r=0.35). Additionally, nominal between-group differences were noted in the PSQI-J total score and sleep difficulty score, as well as exploratory outcomes such as REM sleep duration.
What this study does not show
This study does not establish long-term effects or health outcomes resulting from the AI-generated sleep feedback intervention, as it primarily assesses feasibility and preliminary efficacy.
Evidence scorecard
- Analytic validity: Established
- Clinical validity: Established
- Clinical utility: Partial
The evidence is derived from a human randomized controlled trial.