Personalizing mHealth Interventions for Occupational Stress: Protocol for a Randomized Pilot Study

Occupational stress is associated with detrimental consequences that are addressed by mobile health (mHealth) solutions. Previous developments of apps for occupational stress have not yet fully exploited the potential of multilevel diagnostics through the integration of wearable sensors for interventions. Personalizing mHealth approaches in terms of intervention time and content, which requires the use of artificial intelligence, is the next logical developmental step. The "Relax" approach developed a corresponding prototype of an app-wearable system, which will be evaluated for effectiveness in terms of stress reduction and usability.

This study protocol describes an evaluation study used to test the effectiveness and usability of the Relax approach.

The evaluation study was designed as a 2-arm randomized trial with 2 phases, each with a 3-week intervention period. In both phases, employees were required to use the app to record daily stress and to wear a wearable sensor to measure heart rate variability. The app offered interventions based on algorithms, which altered the probability of their selection after learning from the data, thereby personalizing the user experience. In the second phase of the study, the sample was divided into 2 groups, varying the degree of personalization of the app. To analyze effectiveness, a 2-factorial mixed within-between design will be applied to compare the outcomes between both groups as well as in a pre-post comparison. In addition, exploratory analyses of the usability of the approach are planned.

The study was conducted during the spring and summer of 2024, with a total of 46 participants enrolled, and is ready for data analysis.

The Relax approach, including a number of factors related to personalization that have not yet been incorporated into mHealth in current research, will provide new insights into the next steps of advanced mHealth solutions. Limitations of the study design, such as the lack of a control group and the sample representativity, have to be addressed.

Open Science Foundation 10.17605/OSF.IO/MYRD9; https://osf.io/myrd9.

DERR1-10.2196/68012.