Persistent reminders, event-driven collection, and other strategies to improve active and passive mobile data collection for health studies (Preprint)

Abstract

Machine learning models often use passively recorded sensor data streams as inputs to train machine learning models that predict outcomes captured through EMA. Despite the growth of mobile data collection, challenges in obtaining proper authorization to send notifications, receive background events, and perform background tasks persist.

We aimed to explore these challenges in real-world settings to develop design guidelines for mobile sensing applications to ensure better data collection.

We developed mobile sensing applications for iOS and Android devices. We tested them through a mixed-method user study involving college students (n = 145) for 30 days to answer the following research questions. (1) How do contextual prompting and setup prompting affect scheduled notification delivery and the response rate of notification-initiated EMA? (2) Which authorization paradigm, setup or contextual prompting, is more successful in leading users to grant authorization to receive background events? (3) Which polling-based method, persistent reminders or scheduled background tasks, completes more background sessions?

For RQ1, setup and contextual prompting yielded no significant difference (ANOVA F(1,144) = 0.0227, P = .88) in EMA compliance, with an average of 23.4 out of 30 completed (SD 7.36). However, qualitative analysis showed that contextual prompting on iOS devices resulted in inconsistent notification delivery. For RQ2, contextual prompting for background events was 55.5% (X2=4.42, df=1, P < .035) more effective in gaining authorization for background events. For RQ3, users showed resistance to installing the persistent reminder, but when installed, the persistent reminder performed 226.5% more background sessions than traditional background tasks.

We developed the following guidelines from both quantitative and qualitative results to improve mobile sensing on consumer mobile devices. Qualitative results showed that contextual prompts on iOS devices resulted in inconsistent notification delivery, unlike setup prompting on Android devices. We recommend using setup prompting for notifications. Consistent with prior works, contextual prompting is more efficient for authorizing background events. Developing a persistent reminder and requiring participants to install it provides an additional way to poll for sensor and user data and could improve data consistency.