Akademik Veri Yönetim Sistemi
Diagnostics, cilt.15, sa.19, 2025 (SCI-Expanded)
Objective: The objective was to evaluate the ability of Comprehensive Geriatric Assessment (CGA) parameters to predict all-cause mortality in older adults using both traditional statistical methods and machine learning (ML) approaches. Methods: A total of 1.974 older adults from a university hospital outpatient clinic were included in this study. Ninety-six CGA-related variables encompassing functional and nutritional status, frailty, mobility, cognition, mood, chronic conditions, and laboratory findings were assessed. Cox proportional hazards regression and six ML algorithms (logistic regression, support vector machine, decision tree, random forest, extreme gradient boosting, and artificial neural networks) were employed to identify mortality predictors. Model performance was evaluated using area under the curve (AUC), sensitivity, and F1-score. Results: During a median follow-up of 617 days (interquartile range [IQR]: 297–1015), 430 participants (21.7%) died. Lower Lawton instrumental activities of daily living scores, unintentional weight loss, slower gait speed, and elevated C-reactive protein levels were consistent mortality predictors across all models. The artificial neural network demonstrated the highest predictive performance (AUC = 0.970), followed by logistic regression (AUC = 0.851). SHapley Additive explanations (SHAP) analysis confirmed the relevance of these key features. Conclusions: CGA parameters provide robust prognostic information regarding mortality risk in older adults. Functional decline and inflammation markers offer greater predictive power than chronological age alone in assessing overall health and survival probability.