Abstract
BACKGROUND: Sleep disturbance is a common physical and mental disease, which has a high incidence in breast cancer population and has a great impact on patients treatment outcomes. This study aimed to construct a predictive model for sleep disturbance in breast cancer using machine learning (ML) algorithms and identify the most vital risk predictors. METHODS: Through cross-sectional stratified random sampling, 644 breast cancer patients completed a face-to-face interview questionnaire. Sleep disturbances were assessed using the Patient-Reported Outcomes Measurement Information System Sleep Disturbance 8-item short form. Based on the Maximum Relevant and Minimum Redundancy (MRMR) method, the importance of 26 predictive factors was ranked. Four ML algorithms were used to develop models that predict sleep disturbances in breast cancer. RESULTS: The prevalence of sleep disturbance among breast cancer was 30.59%. The four ML models showed good prediction performances in the test set (area under the receiver operating characteristic curve: 0.74–0.83; accuracy: 0.73–0.82). The features strongly associated with sleep disturbance were loneliness, new possibilities of post-traumatic growth, anxiety, depression, and social support. CONCLUSIONS: ML can effectively predict sleep disturbances in breast cancer. Interventions targeting on the reduction of loneliness, anxiety and depression and enhancement of post-traumatic growth (new possibilities) and social support are needed to alleviate sleep disturbances. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12888-025-07424-9.