Paper Type
Complete
Paper Number
PACIS2026-1547
Description
User attrition in mobile health behavior change support systems (mHBCSS) is typically operationalized as a discrete dropout event, obscuring the temporal dynamics that determine whether inactivity is reversible. This study characterizes re-engagement as a longitudinal process using behavioral log data from 605 users across a 52-week clinically deployed mHBCSS. Analyzing consecutive inactivity spells, we model how recovery probability evolves as uninterrupted non-use accumulates and derive an empirically grounded engagement typology. Results reveal a continuous recovery gradient in which re-engagement likelihood declines monotonically with each additional inactive week, from 89% at one week to below 10% after 20 weeks. Sensitivity analyses identify 17 consecutive inactive weeks as the optimal threshold for classifying prolonged disengagement. Four engagement states are derived: sustained engagement, episodic engagement, prolonged inactivity with recovery, and prolonged inactivity without recovery. Findings have direct implications for the timing of adaptive re-engagement strategies in mHBCSS design.
Recommended Citation
Kulasinghe, Kavinda; Savolainen, Markku; and Oinas-Kukkonen, Harri, "Re-engagement Dynamics in a Mobile Health Behavior Change Support System: Inactivity, Recovery, and Engagement Typologies" (2026). PACIS 2026 Proceedings. 9.
https://aisel.aisnet.org/pacis2026/ishealthcare/ishealthcare/9
Re-engagement Dynamics in a Mobile Health Behavior Change Support System: Inactivity, Recovery, and Engagement Typologies
User attrition in mobile health behavior change support systems (mHBCSS) is typically operationalized as a discrete dropout event, obscuring the temporal dynamics that determine whether inactivity is reversible. This study characterizes re-engagement as a longitudinal process using behavioral log data from 605 users across a 52-week clinically deployed mHBCSS. Analyzing consecutive inactivity spells, we model how recovery probability evolves as uninterrupted non-use accumulates and derive an empirically grounded engagement typology. Results reveal a continuous recovery gradient in which re-engagement likelihood declines monotonically with each additional inactive week, from 89% at one week to below 10% after 20 weeks. Sensitivity analyses identify 17 consecutive inactive weeks as the optimal threshold for classifying prolonged disengagement. Four engagement states are derived: sustained engagement, episodic engagement, prolonged inactivity with recovery, and prolonged inactivity without recovery. Findings have direct implications for the timing of adaptive re-engagement strategies in mHBCSS design.
Comments
14-Healthcare