Cost-effectiveness of the 3E model in diabetes management: a machine learning approach to assess long-term economic impact

Abstract

Background: This study investigated the cost-effectiveness and clinical impact of the 3E model (education, empowerment, and economy) in diabetes management using advanced machine learning techniques. Methods: We conducted an observational longitudinal descriptive analysis involving 320 patients, who were grouped into intervention and control groups over a 24-month period. Results: The 3E model demonstrated significant cost reductions, with the intervention group achieving a 74.3% decrease in total costs compared to 41.8% in the control group while maintaining the same level of glycemic control. Machine learning models, including random forest and K-means clustering, were used to identify key factors influencing treatment costs and to segment patient subgroups that were most responsive to the intervention. Natural language processing techniques revealed medication patterns associated with greater cost reductions. Long-term projections using ensemble methods (such as XG Boost, Exponential Smoothing, and Prophet) predicted that, on average, each year contributes approximately 20% to the total cumulative savings over 5 years. No significant correlations were observed between cost reduction and socioeconomic factors, gender, or age, suggesting the broad applicability of the 3E model. Conclusion: These findings demonstrate the potential of the 3E model to achieve significant reductions in diabetes management costs without compromising care quality, highlighting its value for healthcare policy and resource allocation in chronic disease management.