A Machine Learning and Traditional Chinese Medicine Constitution-Based Prediction Model for Mild Cognitive Impairment in Community-Dwelling Older Adults.

Abstract

Objective: To develop and validate a nomogram screening model for mild cognitive impairment (MCI) in community-dwelling older adults by integrating Traditional Chinese Medicine (TCM) constitution classification with machine learning-based feature selection, aiming to provide a practical tool for early identification in primary care.

Methods: A cross-sectional study was conducted among 1,503 older adults (aged ≥60 years) at a community health service center in Guangzhou, China. Data were prospectively collected during standardized community health examinations between January and December 2025. Participants were randomly divided into training (n = 1,052) and validation (n = 451) sets. Four machine learning algorithms-LASSO regression, random forest, decision tree, and XGBoost-were applied to identify stable predictors. Variables selected by all four methods were entered into multivariable logistic regression, and a nomogram was constructed. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA).

Results: The prevalence of MCI, defined by education-adjusted Chinese Mini-Mental Status (CMMS) cutoffs based on Petersen criteria, was 24.1%. Seven independent correlates were identified: increasing age, female sex, Qi-deficiency constitution, Yin-deficiency constitution, elevated serum creatinine, regular physical exercise, and Balanced constitution. The nomogram achieved AUCs of 0.813 (training) and 0.747 (validation), with satisfactory calibration. Adding TCM constitution to a clinical reference model significantly improved predictive performance (NRI > 0, p < 0.05).

Conclusion: The nomogram incorporating TCM constitution types demonstrated good discrimination, calibration, and clinical utility for community-based MCI screening, providing a practical tool for early identification and risk stratification in primary care settings.