KMAN Publication Inc. Quality of Life and Health Sciences 2 Serial Number 4 2026 04 01 LightGBM-Based Prediction of Poor Quality of Life Among Patients With Heart Failure Using Clinical, Behavioral, and Psychological Features LightGBM-Based Prediction of Poor Quality of Life Among Patients With Heart Failure Using Clinical, Behavioral, and Psychological Features 1 14 10.61838/eqrm4q87 EN Journal Article 2026 01 04 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> This study aimed to develop and interpret a LightGBM-based machine learning model for predicting poor quality of life among patients with heart failure using clinical, behavioral, and psychological features.</p> <p><strong>Methods and Materials: </strong>This multicenter cross-sectional predictive modeling study was conducted among 428 patients with heart failure recruited from cardiology outpatient clinics and heart failure units in Sweden. Clinical data included left ventricular ejection fraction, NYHA functional class, duration of heart failure, comorbidities, hospitalization history, blood pressure, heart rate, and medication-related indicators. Behavioral variables included physical activity, sleep quality, and heart failure self-care behaviors. Psychological variables included depressive symptoms, anxiety symptoms, perceived stress, and perceived social support. Quality of life was assessed using the Kansas City Cardiomyopathy Questionnaire, and poor quality of life was defined as an overall summary score below 50. The dataset was divided into training and testing subsets, and an optimized LightGBM model was developed using cross-validation and hyperparameter tuning. Model performance was evaluated using AUC, accuracy, sensitivity, specificity, precision, F1-score, balanced accuracy, and Brier score. SHAP analysis was used for model interpretation.</p> <p><strong>Findings: </strong>Poor quality of life was identified in 174 patients, representing 40.7% of the sample. Compared with patients without poor quality of life, patients with poor quality of life had significantly lower left ventricular ejection fraction, higher NYHA class, longer disease duration, higher comorbidity burden, more previous heart failure hospitalization, lower physical activity, poorer sleep quality, weaker self-care, higher depression, higher anxiety, higher perceived stress, and lower perceived social support. The optimized LightGBM model achieved strong predictive performance in the test set, with an AUC of 0.859, accuracy of 0.806, sensitivity of 0.788, specificity of 0.818, F1-score of 0.764, balanced accuracy of 0.803, and Brier score of 0.137. SHAP analysis identified depressive symptoms, NYHA functional class, sleep quality, anxiety symptoms, previous hospitalization, self-care maintenance, physical activity, ejection fraction, perceived stress, and social support as the most influential predictors.</p> <p><strong>Conclusion: </strong>The LightGBM model demonstrated strong ability to predict poor quality of life among patients with heart failure and showed that quality-of-life impairment is influenced by a combination of clinical severity, behavioral limitations, sleep disturbance, psychological distress, and reduced social support.</p> </td> </tr> </tbody> </table> https://www.journals.kmanpub.com/index.php/qlhs/article/download/5800/10976