Classification of Cognitive Frailty in Elderly People from Blood Samples using Machine Learning

Cognitive Frailty (CF) is a prevalent age-related disease that is affecting many individuals worldwide. Medical intervention needs to be timely, as the late stages of CF prove to be challenging for both clinicians and caretakers. While the existing clinical diagnosis and screening tools for CF are capable of detecting the syndrome, a means of prediction is needed in order to identify CF in older adults before its onset. This paper proposes a machine learning model to classify patients into different levels of CF, using parameters from blood samples. A total of 7 different classification algorithms were used to predict between 6 levels of CF, the Robust and Non-Robust groups, as well as the Robust and Frail with MCI groups. The binary classification for Robust and Frail with MCI achieved the highest accuracy, with Gaussian Naïve Bayes showing the highest holdout method accuracy of 70.5%, as well as the highest cross validation accuracy of 74%.