Development and Validation of a 7-Year Risk Prediction Model for Progression from Subjective Cognitive Decline to Mild Cognitive Impairment: A Prospective Cohort Study Using CHARLS Data

Background Subjective cognitive decline (SCD) frequently precedes mild cognitive impairment (MCI), yet few prospective tools quantify 7-year progression risk in Chinese community-dwelling older adults. Methods Using nationally representative CHARLS 2011–2018 data, we included 4,981 participants aged ≥50 y with baseline SCD and intact cognition. After unified multiple-imputation, 36 candidate variables were screened by seven complementary feature-selection algorithms. Nine consensus predictors entered nine machine-learning algorithms; the best-performing logistic-regression model was internally validated (five random 60/40 splits; n = 2,989/1,992). Discrimination, calibration (Brier score), decision-curve analysis and SHAP explainability were evaluated. Results Incident MCI occurred in 1,194 (24.0 %) individuals. Logistic regression achieved mean AUC 0.783 (95 % CI 0.765–0.806), superior Brier score 0.148, and positive net benefit across 10–40 % threshold probabilities. SHAP analysis identified education (mean |SHAP| = 0.35) and baseline cognitive score (0.28) as dominant protective factors, whereas visual impairment (0.09) and arthritis (0.12) increased risk. A clinician-friendly nomogram translated coefficients into a 0–100 points scale for bedside use. Limitations S CD/MCI ascertainment relied on psychometric rather than clinical diagnosis; external validation is pending. Conclusions The first CHARLS-derived, pragmatic 7-year SCD-to-MCI risk calculator demonstrates robust discrimination, excellent calibration and actionable interpretability, facilitating early detection and preventive intervention in primary-care and mental-health settings.