Construction of a prediction model for sarcopenic obesity based on machine learning

In the context of the rapidly aging global population, sarcopenic obesity (SO) in older adults is associated with significantly higher rates of disability and mortality. SO has become a serious and critical public health concern. This study aimed to develop and validate predictive models using machine learning (ML) to identify SO in patients.

Data from 386 participants collected at the Affiliated Hospital of Qingdao University were divided into an 8:2 ratio, with 80% used for training and 20% for testing. Univariate analysis was performed to identify the factors correlated with SO, and multivariate logistic regression analysis was performed to determine the independent factors influencing SO. The Shapley Additive exPlanations (SHAP) diagram was used to illustrate the importance of variables in the model. To develop a predictive model for SO, we used five models and applied internal five-fold cross-validation to determine the most suitable hyperparameters for the model.

Among 386 participants, 61 were diagnosed with sarcopenic obesity (15.8%). We identified four independent predictive factors, namely BMI, Barthel Index score, grip strength, and calf circumference. Notably, calf circumference plays an important role in assessing the risk of SO in older adults. The area under the curve (AUC) values of the test set for the Random forest (RF), naive Bayes (NB), Light Gradient Boosting Machine (LightGBM), k-nearest neighbor algorithm (KNN), and eXtreme Gradient Boosting (XGBoost) models were recorded as 0.839, 0.815, 0.808, 0.794, and 0.798, respectively. Among these models, the RF model exhibited the best average performance in the training set, with an AUC value of 0.839.

We constructed a predictive model based on the results of the RF model, combining four clinical predictors-BMI, Barthel Index score, grip strength, and calf circumference-to reliably predict SO.