JMIR Publications

ABSTRACT

Background:

Diabetes management is complex, and user personalization in diabetes management programs has been identified to enhance engagement and clinical outcomes. However, with 50% of individuals with diabetes unable to achieve glycemic control, there appears to be a gap in delivery of self-management education and behavior change. Machine learning and recommender systems, which have been used within the healthcare setting, could be a feasible application to provide a personalized user experience and improve user engagement and outcomes.

Objective:

To evaluate machine learning models using member level engagements to predict improvement in estimated A1c (eA1c) and develop personalized action recommendations within a remote diabetes monitoring program to improve clinical outcomes.

Methods:

Member engagement was analyzed within five action categories (interacting with a coach, reading education content, monitoring blood glucose level, tracking physical activity, and monitoring nutrition) to build a member level model to predict if a specific type and level of engagement could lead to improved eA1c for members with type 2 diabetes. Engagement and improvement in eA1c can be correlated; therefore, Doubly Robust Learning method was used to model the heterogeneous treatment effect of action engagement on improvements in eA1c.

Results:

Treatment effect was successfully computed within the five action categories on eA1c reduction for each member. Results show interaction with coaches and monitoring glucose levels were the actions that resulted in the highest average decrease in eA1c (1.7%) and were the most recommended action for 52% of the population. However, these were found to be not the optimal interventions for all members; 48% members were predicted to have better outcomes with one of the other 3 interventions. Members who engaged within their recommended actions had a 1.1% higher change in eA1c than those who did not engage within recommended actions.

Conclusions:

Personalized action recommendations using heterogeneous treatment effects to compute the impact of member actions can reduce eA1c and be a valuable tool for diabetes management programs in encouraging members toward actions to improve clinical outcomes.