KMAN Publication Inc. Quality of Life and Health Sciences 1 Serial Number 2 2025 10 01 Predicting Health-Related Quality of Life Among Patients With Type 2 Diabetes Using Random Forest and SHAP-Based Feature Importance Analysis Predicting Health-Related Quality of Life Among Patients With Type 2 Diabetes Using Random Forest and SHAP-Based Feature Importance Analysis 1 16 10.61838/sxnr2h27 EN Journal Article 2025 04 10 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> This study aimed to predict health-related quality of life among patients with type 2 diabetes in Mexico using a Random Forest model and to identify the most influential clinical, behavioral, psychological, and sociodemographic predictors through SHAP-based feature importance analysis.</p> <p><strong>Methods and Materials:</strong> This cross-sectional predictive modeling study was conducted among 462 adult patients with type 2 diabetes receiving outpatient care in Mexico City, Guadalajara, and Monterrey. Health-related quality of life was assessed using the 36-Item Short Form Health Survey. Clinical, behavioral, psychological, and sociodemographic data were collected using medical records and standardized self-report instruments, including measures of diabetes self-care and psychological distress. The dataset was divided into training and testing subsets using an 80:20 split. A Random Forest regression model was developed to predict total health-related quality of life scores. Model performance was evaluated using the coefficient of determination, root mean square error, mean absolute error, and mean squared error. SHAP analysis was applied to interpret global feature importance and the direction of predictor effects.</p> <p><strong>Findings:</strong> The Random Forest model demonstrated strong predictive performance, explaining 73.5% of the variance in health-related quality of life in the independent testing set. The model achieved an RMSE of 8.34, MAE of 6.49, and MSE of 69.56. Cross-validation and out-of-bag estimation showed comparable performance, with R² values of 0.721 and 0.704, respectively. SHAP analysis identified depression score as the strongest predictor of lower health-related quality of life, followed by HbA1c, number of diabetes-related complications, physical activity self-care, duration of diabetes, body mass index, stress score, medication adherence, sleep duration, age, hypertension, anxiety score, foot care behavior, household income, and insulin use.</p> <p><strong>Conclusion:</strong> The findings indicate that health-related quality of life among patients with type 2 diabetes can be predicted with acceptable accuracy using Random Forest modeling, while SHAP analysis provides clinically interpretable evidence that psychological distress, glycemic control, complication burden, and self-care behaviors are central determinants of patient-perceived health.</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/qlhs/article/download/5784/10945