Development and validation of an interpretable machine learning model for predicting intrinsic capacity decline in...

The purpose of this study was to use machine learning algorithms to construct a risk prediction model for the decline of intrinsic capacity in older hospitalized patients.

In this study, the demographic and disease-related data of older hospitalized patients from three tertiary general hospitals were collected for follow-up. The older hospitalized patients from two of the hospitals formed the modeling cohort, while those from the third hospital formed the validation cohort. Four machine learning methods were used to construct prediction models, and the discrimination, calibration, and clinical application value of different models were evaluated. The optimal model was selected based on these indicators. The Shapley Additive exPlanations method was employed to explain the optimal model.

A total of 533 older hospitalized patients were included in this study. During the one-year follow-up period, the incidence of decline in intrinsic capacity was 50.8 %. Eight predictors were selected for model construction. The Support Vector Machine model had the best predictive performance, and the area under the receiver operating characteristic curve of the training set, test set and validation set were 0.923 (0.893-0.954), 0.932 (0.871-0.993) and 0.882 (0.831-0.933), respectively. The five most important characteristics for predicting decreased intrinsic capacity were identified as heart disease, hypertension, grip strength, age and triglycerides.

The Support Vector Machine model can effectively identify high-risk older hospitalized patients with decreased intrinsic capacity. Early risk identification and targeted intervention measures applied in clinical practice are helpful to improve the health level of older hospitalized patients and achieve healthy aging.