Prevalence and a LASSO-derived prediction model for screening-positive mild cognitive impairment among older adults...

This study aimed to investigate the prevalence of screening-positive mild cognitive impairment (s-MCI) and to develop a parsimonious prediction model using machine learning methods to identify high-risk older adults in Wuhan, China. A total of 2,190 community-dwelling adults aged ≥ 60 years were recruited through multistage cluster sampling from 30 residential committees in 13 districts. MCI screening was conducted with the Community Screening Instrument for Dementia (CSI-D). The Least Absolute Shrinkage and Selection Operator (LASSO) was employed to select predictive variables, and multiple machine learning classifiers were compared. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) was used to interpret the best-performing model. The screening-positive rate of MCI was 35.3%. The final prediction model retained four predictors: age, occupation, sleep disorders, and literacy level. XGBoost showed slightly better discrimination than logistic regression in the validation set (AUC: 0.693 vs. 0.671). Based on SHAP values, sleep disorders, advanced age, lower education level, and farmer occupation were the most important contributors to the prediction. The burden of screening-positive MCI is considerable among older adults in Wuhan. The LASSO-derived prediction model, comprising easily obtainable variables, can serve as a practical risk-stratification tool to support targeted early screening, without implying a unique etiological role for the selected predictors.