What question did this study set out to answer?

This study aims to identify key digital engagement features that correlate with significant reductions in blood glucose levels in a real-world diabetes cohort.

June 7, 2026

1048-OR: Digital Engagement Features Associated with Blood Glucose Improvement in Real-World Diabetes

Key Points

This study aims to identify key digital engagement features that correlate with significant reductions in blood glucose levels in a real-world diabetes cohort.
Retrospective analysis of 8,056 members using the Dario platform with baseline average BG >154 mg/dL.
Engagement features were analyzed for correlations with changes in average BG using Pearson correlations.
Random forest modeling predicted BG reduction based on engagement metrics.
44.8% of members achieved a clinically meaningful eA1c reduction of ≥1%.
Members with high-risk baseline BG levels showed a significant average BG decrease of 35 mg/dL at month 12 (p<0.001).
Total engagement over time had the strongest correlation to glucose improvement (r=0.09, p<0.001).

Abstract

Introduction and Objective: Digital health platforms generate rich engagement data, yet real-world evidence remains limited on which features are most strongly associated with clinically meaningful glycemic improvement. Understanding these relationships is critical for optimizing personalization and program design. This study used the Dario digital health platform to identify engagement features most strongly associated with reductions in average blood glucose (BG) through enhanced feature engineering and correlation analysis in a real-world diabetes population. Methods: A retrospective analysis of members who joined the Dario digital health platform in 2025 with baseline average BG above target (154 mg/dL; estimated HbA1c (eA1c) 7.0%) examined engagement across integrated app features. A feature-engineering pipeline derived metrics capturing longitudinal trends, timing (early vs late engagement), and frequency across domains. Pearson correlations were calculated between each engagement feature and the decrease in average BG. A random forest model (500 trees) predicted BG reduction, with feature importance assessed. Results: In a cohort of 8,056 members, 44.8% achieved a clinically meaningful eA1c reduction of ≥1%. Among those with high-risk baseline BG levels (183 mg/dL; eA1c 8.0), average BG decreased significantly by 35 mg/dL at month12 (p0.001). Increasing total engagement over time showed the strongest correlation with glucose improvement (r=0.09, p0.001). Food tracking trend, physical activity trend, and statistics viewing trend were significantly associated with BG reduction (r= 0.05-0.07, p0.01). Random forest models identified total engagement consistency as the strongest predictor of glucose improvement. Conclusion: In this real-world digital diabetes cohort, consistent, longitudinal engagement rather than a single interaction was strongly linked to glycemic improvement, supporting the use of sustained engagement metrics to guide personalization and optimize digital diabetes care. Disclosure Y.F. Hershcovitz: Employee; Current; DarioHealth Corp. I. Breuer Asher: Employee; Current; DarioHealth Corp. O. Manejwala: Employee; Current; DarioHealth Corp. Stock/Shareholder; Current; DarioHealth Corp.

Ask AI

Helpful

Bookmark