Machine learning models screen for osteosarcopenia in older Iranians

Machine learning models using routine clinical data can screen for osteosarcopenia and related musculoskeletal conditions with high sensitivity, according to a study of over 4,100 older Iranian adults. In the Bushehr Elderly Health cohort (median age 66, 54.2% female), Random Forest detected osteoporosis with ROC-AUC 0.821 and sensitivity 0.946; XGBoost detected sarcopenia with ROC-AUC 0.865 and sensitivity 0.862. Osteosarcopenia models achieved ROC-AUC about 0.85. Key risk predictors included age, weight/BMI, forearm circumference, and functional limitations. Authors say external validation across diverse populations is needed before deployment as screening tools.