Development and validation of an explainable machine learning model for predicting the risk of sleep disorders in older adults with multimorbidity: a cross-sectional study

Xia Wang¹Dan Zhang²Liu Lu³Shujie Meng¹Yong Li⁴Rong Zhang⁴Jingjie Zhou⁵Qian Yu³Li Zeng³Jiang Zhao⁴Yu Zen⁴Ru Gao^2*

• ¹Chengdu University, Chengdu, China
• ²The People's Hospital of Wenjiang Chengdu, Chengdu, China
• ³Yibin Fourth People's Hospital, Yibin, China
• ⁴Sichuan Health Rehabilitation Vocational College, Zigong, Sichuan Province, China
• ⁵Tellyes Scientific Inc, Tianjin, China

Objective To develop and validate an explainable machine learning model for predicting the risk of sleep disorders in older adults with multimorbidity.Methods A total of 471 older adults with multimorbidity were recruited between October and November 2024. We employed six machine learning (ML) methods, namely logistic regression (LR), neural network (NN), support vector machine (SVM), gradient boosting machine (GBM), K-Nearest Neighbours (KNN), and light gradient boosting machine (LightGBM), to predict the risk of sleep disorders based on their sociodemographic data, health behaviour factors, mental health, and disease-related data. The optimal model was identified through the evaluation of the area under the curve (AUC). This study also employed explainable machine learning techniques to provide insights into the model's predictions and outcomes using the SHAP (Shapley Additive Explanations) approach.The prevalence of sleep disorders was 28.7%. Among the six models developed, the GBM model achieved the best performance with an AUC of 0.881.The analysis of feature importance revealed that the top seven predictors of sleep disorders were frailty, cognitive status, nutritional status, living alone, depression, smoking status, and anxiety.This study is the first to predict sleep disorders in Chinese older adults with multimorbidity using explainable machine learning methods and to identify seven significant risk factors. The SHAP method enhances the interpretability of machine learning models and helps medical staff better understand the rationale behind the predicted outcomes more effectively.