A 10-item Fugl-Meyer Motor Scale Based on Machine Learning

Abstract Objective The Fugl-Meyer motor scale (FM) is a well-validated measure for assessing upper extremity and lower functions in people with stroke. FM contains numerous items (50), which reduces its clinical usability. purpose of this study was to develop short form the stroke using machine-learning methodology (FM-ML) compare efficiency (ie, number items) psychometric properties FM-ML those other versions, including original FM, 37-item 12-item FM. Methods This observational follow-up used secondary data analysis. For developing FM-ML, random lasso method ML select 10 most informative (in terms index importance). Next, scores were calculated an artificial neural network. Finally, concurrent validity, predictive responsiveness, test–retest reliability all versions examined. Results fewer (80% than 73% 17% FM) achieve comparable (concurrent validity: Pearson r = 0.95–0.99 vs 0.91–0.97; responsiveness: 0.78–0.91 0.33–0.72; reliability: intraclass correlation coefficient 0.88–0.92 0.93–0.98). Conclusion findings preliminarily support 10-item FM-ML. Impact has potential substantially improve function assessments patients