JMIR Publications

ABSTRACT

Background:

Depression is a growing problem among youth, and providers need scalable solutions for augmenting cognitive behavioral therapy and treating unresolved sleep disturbance and fatigue. Just-in-time adaptive interventions (JITAIs) that leverage digital health technology such as smartphones and smartwatches can help address depression treatment challenges in adolescents, but existing JITAIs provide limited options for personalizing interventions.

Objective:

The primary objective of this pilot study was to design and user-test a novel JITAI approach that uses digital health data to (1) establish baseline physical activity and sleep behaviors among adolescents; (2) measure day-to-day changes in physical activity and sleep; and (3) apply personalized, algorithm-based decision rules for selecting intervention messages designed to engage treatment targets of physical activity and sleep.

Methods:

Adolescents (n=8) were enrolled in a 4-week user-testing protocol to test the JITAI. Baseline physical activity and sleep data were collected for ~2 weeks via smartwatch and ecological momentary assessments (EMA) before beginning the intervention phase. Baseline data were used to set sleep and activity goals and establish success benchmarks. Participants were randomized into receiving up to one physical activity and one sleep intervention message each day for ~2 weeks. Physical activity (n=84) and sleep (n=84) messages were designed based on existing evidence-based interventions, with message wording adjusted using 6 validated behavior change principles. Selection of a message for delivery was first based on equal priority for all messages. If smartwatch or survey data indicated that a participant’s sleep or activity goal was met following intervention delivery, the JITAI algorithm increased the likelihood that future messages would be drawn from the same behavior change category; if the goal was not met, the likelihood was reduced. Following the user-testing study, message selection patterns for each behavior change category were examined to refine JITAI algorithm before implementation in a pilot clinical trial.

Results:

141 participant days of baseline data were collected, and 114 participant days of intervention data were collected. Smartwatch wear compliance exceeded 90% for all participants, and average EMA compliance exceeded 70%. Physical activity and sleep varied both within and between participants, and weekday versus weekend differences were observed. A total of 55 physical activity and 42 sleep behavior change interventions were provided. Across the cohort, the algorithm increased preference for some categories and decreased preference for others. Message categories that were consistently less effective were removed from the final JITAI algorithm.

Conclusions:

Our novel JITAI demonstrated high smartwatch and moderate to high EMA compliance. The JITAI algorithm successfully evaluated intervention success, and selected personalized physical activity and sleep intervention messages across evidence-based behavior change categories. Future experiments should consider using a baseline period of 2-4 weeks to capture variation in behavior and should ensure an adequate window for completing surveys to increase compliance.