Digital Engagement Phenotypes and 6-Month Weight Loss in a Tirzepatide-Supported Digital Weight-Loss Program: A Retrospective Cohort Study
Digital Engagement Phenotypes and 6-Month Weight Loss in a Tirzepatide-Supported Digital Weight-Loss Program: A Retrospective Cohort Study. Background and Objectives: Real-world persistence with medication-supported weight management programs is often low. While digital weight loss services (DWLS) provide multi-modal digital supports to improve engagement and counter attrition, the existing literature frequently relies on unidimensional or binary classifications of user engagement. This study used unsupervised machine learning to identify distinct digital engagement phenotypes and evaluated their independent associations with 6-month weight loss outcomes in patients prescribed tirzepatide. Materials and Methods: This retrospective cohort study analyzed deidentified data from 9470 medication-adherent, complete-case adult patients (out of 39,220 tirzepatide initiators) within a British DWLS who initiated tirzepatide between 20 May 20 and 2 December 2025. K-means clustering was performed on four continuous, longitudinal usage metrics: weekly app logins, health coach messaging, automated assistant (JuneBot) messaging, and weight tracking. To evaluate the primary clinical endpoint-6-month percentage weight loss-unadjusted pairwise comparisons (Tukey HSD) and a fully adjusted ordinary least squares multivariate linear regression model were executed to control for baseline demographic, clinical, and interim behavioral covariates. Results: Four stable engagement phenotypes emerged: non-engaged (n = 1772), high health coach engagement (n = 1141), high JuneBot engagement (n = 1472), and passive self-trackers (n = 5085). In a baseline-adjusted multivariate regression model, all active phenotypes were independently associated with 6-month weight loss relative to the low engagement baseline. Compared with this reference group, adjusted mean differences in percentage weight loss were 4.60% (SE = 0.26, p < 0.001) in the high health coach cluster, 4.69% (SE = 0.24, p < 0.001) in th…
0 Comments
The summary above is machine-written and the abstract is the authors' own pitch. This is where people who read the paper say what it actually found, what the summary missed, and which part is worth your time.
Log in to join the discussion.