Predicting frail syndrome using adverse geriatric health outcomes: Comparison of different statistical classifiers

The present work seeks to evaluate the role of statistical classifiers based on sociodemographic data, anthropometric measurements and adverse health outcomes for identifying frail individuals. Three classifiers were considered: Logistic Regression, Random Forest and Support Vector Machines. The best results were obtained with Logistic Regression models based solely on 3 predictors, which reached an area under the curve [0.80, 0.85], specificity [75%, 81%] and sensitivity [80%, 86%]. These models display a promising role as a brief and easy screening tool for Frail Syndrome and should be further assessed, notably by external validation.