Risk Assessment of Sarcopenia in Patients With Type 2 Diabetes Mellitus Using Data Mining Methods.

Purpose: Sarcopenia is a geriatric syndrome, and it is closely related to the prevalence of type 2 diabetes mellitus (T2DM). Until now, the diagnosis of sarcopenia requires Dual Energy X-ray Absorptiometry (DXA) scanning. This study aims to make risk assessment of sarcopenia with support vector machine (SVM) and random forest (RF) when DXA is not available. Methods: Firstly, we recruited 132 patients aged over 65 and diagnosed with T2DM in Changchun, China. Clinical data were collected for predicting sarcopenia. Secondly, we selected 3, 5, and 7 features out of over 40 features of patient's data with backward selection, respectively, to train SVM and RF classification models and regression models. Finally, to evaluate the performance of the models, we performed leave one out and 5-fold cross validation. Results: When training the model with 5 features, the sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) were favorable, and it was better than the models trained with 3 features and 7 features. Area under the receiver operating characteristic (ROC) curve (AUC) were over 0.7, and the mean AUC of SVM models was higher than that of RF. Conclusions: Using SVM and RF to make risk assessment of sarcopenia in the elderly is an option in clinical setting. Only 5 features are needed to input into the software to run the algorithm for a primary assessment. It cannot replace DXA to diagnose sarcopenia, but is a good tool to evaluate sarcopenia.