JMIR Publications

ABSTRACT

Background:

Mobile health (mHealth) apps are increasingly used to support self-care, but poor user engagement limits their effectiveness. This is particularly true for fully automated mHealth apps without any human support. Human support in mHealth apps is associated with better engagement but at the cost of reduced scalability.

Objective:

This paper explores engagement over 10 weeks with a fully automated, theory- and evidence-based self-care app with a text-based conversational agent as a social component designed to reduce distress in people with cancer (CanRelax app 2.0). Contrary to most studies, we examine engagement on both a “micro” level (ie, app usage and user experience) and a “macro” level (ie, engagement with the target behavior).

Methods:

We developed the CanRelax app 2.0 in iterative processes involving input from people with cancer and relevant experts. The app includes evidence-based relaxation exercises, personalized weekly coaching sessions with a rule-based conversational agent, 39 self-enactable behavior change techniques, a self-monitoring dashboard with gamification elements, highly tailored reminder notifications, an educational video clip, and personalized in-app letters. We applied multiple measures of microlevel engagement (completed coaching sessions, relaxation exercises practiced with the app, feedback on the app) and macrolevel engagement (relaxation exercises practiced without the app and self-efficacy toward self-set weekly relaxation goals). We used exploratory analysis to investigate data from 100 participants in a larger randomized controlled trial. German-speaking adults diagnosed with cancer within the last five years were recruited online in Switzerland, Austria and Germany.

Results:

In week 10, 62% (62/100) of participants were still actively using the CanRelax app 2.0. We did not identify associations between engagement and level of distress at baseline, sex assigned at birth, educational attainment, or age. On the micro level, 3520 relaxation exercises (72% of all relaxation exercises) and 714 coaching sessions were completed in the app, and all participants who provided feedback (52/100) expressed positive app experiences. On the macro level, 28% of relaxation exercises (1377/4897) were completed without the app, and participants’ self-efficacy remained stable at a high level. At the same time, participants raised their weekly relaxation goals, which indicates a relative increase in self-efficacy.

Conclusions:

The CanRelax app 2.0 achieved promising engagement even though we provided no human support. Fully automated social components might have compensated for the lack of human involvement and should be investigated further. More than a quarter of all relaxation exercises were practiced without the app, highlighting the importance of assessing engagement on multiple levels. Clinical Trial: German Clinical Trials Register DRKS00027546; https://drks.de/search/en/trial/DRKS00027546