The Role of Emotional Intelligence in Diabetes Management: Evaluating Predictive Power for Self-Care Behaviors Using Artificial Intelligence Models

Objective: This study examined the effect of emotional intelligence on self-care behaviors in individuals with type 2 diabetes and their prediction using artificial intelligence based models. Material and Methods: This cross-sectional, quantitative study included 200 individuals diagnosed with type 2 diabetes mellitus (DM). “Personal Information Form’’, “Diabetes Self-Care Activities Questionnaire”, and “Trait Emotional Intelligence Questionnaire-Short Form (TEIQue-SF)” were used. Random Forest, Support Vector Machine, Artificial Neural Networks, and Linear Regression algorithms were applied in the analyses. Results: The mean age of the participants was 54.88 ± 6.41 years, 76.0% were women, and 94.0% were married. The overall emotional intelligence level was moderate (71.48 ±1.51). Among sub-dimensions, emotionality had the highest mean score (4.55 ± 0.47), while sociability had the lowest (3.14±0.19). Emotionality was positively associated with diet and blood glucose monitoring but negatively associated with exercise. In contrast, sociability showed a negative relationship with diet and a positive relationship with exercise. Among the applied models, Random Forest demonstrated the best performance, explaining approximately 14% of the variance in the total self-care score (R2 = .146) while Support Vector Machine (R2 = -.077), Linear Regression (R2 = -.023), and Artificial Neural Network (R2 = -3.538) models yielded negative R2 values, indicating that emotional intelligence sub-dimensions alone were insufficient to generate meaningful predictions. Feature importance analysis identified “Well-being” and “Self-control” as the primary contributors, while “Emotionality” ranked lowest despite its strong bivariate correlations reflecting methodological differences between the two analyses. Conclusion: These exploratory findings indicate that emotional intelligence sub-dimensions alone have limited predictive value (R2 = .146), suggesting that self-care is a multidimensional construct extending beyond emotional capacity.