Multidimensional Health Phenotyping and Metabolic Syndrome Prediction in Chinese Community-Dwelling Elderly: An Integrated Data-Driven Approach

Background Population aging in China poses significant healthcare challenges, with chronic diseases accounting for over 80% of disability-adjusted life years among older adults. Traditional single-disease assessments fail to capture the complex interactions between multiple health domains. Methods This retrospective cross-sectional study utilized health examination data from 10, 639 community-dwelling elderly residents (≥ 60 years) from Jiangling Sub-district, Suzhou City, China, between April 1 and August 31, 2025. Health indicators included anthropometric measurements, blood pressure, fasting blood glucose, lipid profiles, and self-reported chronic diseases. K-means clustering was used to identify distinct health patterns. A machine learning model was developed to predict metabolic syndrome using random forest, as a key application of the health phenotyping framework. Performance metrics included area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), calibration (Hosmer-Lemeshow test), and clinical utility (decision curve analysis). Results Five distinct health patterns were identified: (1) Severe Metabolic Disorder Type (5.6% ), (2) Healthy Lean Hypercholesterolemia Type (21.7% ), (3) Elderly Hypertension Type (21.3% ), (4) Middle-aged Obese Multi-risk Type (25.4% ), and (5) Relatively Healthy Type (26.0% ). Validation via principal component analysis showed clear separation, with PC1 (38.2% variance) driven by metabolic load and PC2 (18.5% ) by age and cholesterol. The metabolic syndrome prediction model demonstrated high discriminative ability (AUC = 0.964, 95% CI: 0.958–0.970), substantially outperforming a baseline model using age and sex alone (AUC = 0.712) and showing good calibration (Hosmer-Lemeshow P = 0.38). Triglycerides (Gini importance = 0.318), fasting blood glucose (0.199), and BMI (0.195) were the most important predictors. Decision curve analysis showed positive net benefit across threshold probabilities of 0.2–0.5. Conclusion This study established a comprehensive multidimensional health assessment framework for Chinese community-dwelling elderly, identifying five clinically meaningful health patterns. The high-performance metabolic syndrome prediction model has practical applications for preventive care, while the identified patterns provide a roadmap for stratified interventions in geriatric care. This integrated data-driven approach represents an important advancement in geriatric medicine, with the potential to transform community-based preventive care and resource allocation for older adults.