Predicting Long-Term Engagement in mHealth Apps:ComparativeStudy of Engagement Indices

Background: Digital health care apps, including digital therapeutics, have the potential to increase accessibility and improve patient engagement by overcoming the limitations of traditional facility-based medical treatments. However, there are no established tools capable of quantitatively measuring long-term engagement at present. Objective: This study aimed to evaluate an existing engagement index (EI) in a commercial health management app for long-term use and compare it with a newly developed EI. Methods: Participants were recruited from cancer survivors enrolled in a randomized controlled trial that evaluated the impact of mobile health apps on recovery. Of these patients, 240 were included in the study and randomly assigned to the Noom app (Noom Inc). The newly developed EI was compared with the existing EI, and a long-term use analysis was conducted. Furthermore, the new EI was evaluated based on adapted measurements from the Web Matrix Visitor Index, focusing on click depth, recency, and loyalty indices. Results: The newly developed EI model outperformed the existing EI model in terms of predicting EI of a 6- to 9-month period based on the EI of a 3- to 6-month period. The existing model had a mean squared error of 0.096, a root mean squared error of 0.310, and an R-2 of 0.053. Meanwhile, the newly developed EI models showed improved performance, with the best one achieving a mean squared error of 0.025, root mean squared error of 0.157, and R-2 of 0.610. The existing EI exhibited significant associations: the click depth index (hazard ratio [HR] 0.49, 95% CI 0.29-0.84; P<.001) and loyalty index (HR 0.17, 95% CI 0.09-0.31; P<.001) were significantly associated with improved survival, whereas the recency index exhibited no significant association (HR 1.30, 95% CI 1.70-2.42; P=.41). Among the new EI models, the EI with a menu combination of menus available in the app's free version yielded the most promising result. Furthermore, it exhibited significant associations with the loyalty index (HR 0.32, 95% CI 0.16-0.62; P<.001) and the recency index (HR 0.47, 95% CI 0.30-0.75; P<.001). Conclusions: The newly developed EI model outperformed the existing model in terms of the prediction of long-term user engagement and compliance in a mobile health app context. We emphasized the importance of log data and suggested avenues for future research to address the subjectivity of the EI and incorporate a broader range of indices for comprehensive evaluation.