Motor and Gait-Based Screening for Low MoCA-BJ Scores in Adults Aged ≥ 50 Years: A Cross-Sectional Study

Background/Objectives: Motor measurements may identify people who merit cognitive assessment, but the added value of detailed gait features is uncertain. We compared simple physical and detailed gait models for classifying Montreal Cognitive Assessment (MoCA) scores at two community sites. Methods: After excluding unsuccessful assessments, 371 adults aged ≥50 years with valid MoCA and gait data were included. Five fixed ridge logistic models were evaluated with stratified tenfold cross-validation and fold-specific preprocessing. Paired comparisons, calibration, site validation and threshold analyses were performed. Results: MoCA was <26 in 295 participants (79.5%). The parsimonious model had an AUC of 0.707 (0.643 to 0.763); the physical-plus-speed and full-gait models had AUCs of 0.710 and 0.693. Adding detailed gait features to physical measures and speed changed AUC by −0.017 (conditional 95% interval −0.034 to −0.002) and Brier score by +0.0034. Within-site parsimonious AUCs were 0.830 (Pinggu) and 0.565 (Changping). Targeting 90% sensitivity yielded 89.8% sensitivity but 22.4% specificity. Conclusions: Simple motor measures moderately discriminated MoCA score groups, and detailed gait features showed no advantage. Performance varied by site, and high sensitivity came with low specificity; harmonized measurement and external validation are the next steps toward screening use.