Development and validation of a decision tree model for prediction of insomnia risk among ischemic stroke convalescence patients

Background: Insomnia is a prevalent complication among ischemic stroke convalescence (ISC) patients. Although the interplay of clinical, psychological, and social factors remains unclear in ISC patients, a model prediction system was necessary. Limited research developed a prediction model for insomnia risk. Objectives: To construct a decision tree model for insomnia risk among ISC patients based on the classification and regression tree algorithm. Design: Across-sectional study. Setting: China. Participants: The study enrolled 823 adult ISC patients between February 2023 and October 2024. Participants were recruited from stroke units in two tertiary hospitals in Jilin Province. Methods: A decision tree model was guided by the TRIPOD+AI report. The Pittsburgh Sleep Quality Index (PSQI), Fatigue Severity Scale (FSS), Social Support Scale (SSRS) and other scales were used to collect data. The confusion matrix, ROC curves, H-L test, and calibration curve were employed for internal and external validation by using a bootstrap resampling method. 623 patients were used to construct the decision tree model, while the remaining 200 non-homologous cases were used for external validation. Results: This study showed that the prevalence of insomnia among ISC patients was 37.72 %. Univariate analysis revealed that factors such as BMI, SAS, SSRS, FSS, SDS, and NIHSS were critical. The decision tree model yielded 24 paths with a depth of 6. The predictive contribution was ranked as follows: SAS > SSRS > FSS > SDS > BMI > NIHSS, which were identified to create the nomogram. Internal validation indicated that the model had strong predictive accuracy at 88.2%, with a sensitivity of 0.96, specificity of 0.84, and a Youden index of 0.80. The area under the curve was 0.96 (95% CI: 0.93~0.98; p < 0.001); Additionally, the H-L test showed that the model was well-calibrated (χ2 = 9.36, p = 0.404). External validation proved that the model had stability across different data. Conclusion: This decision tree model demonstrates potential for predicting insomnia in ISC patients, and these predictors can inform the development of future insomnia management strategies. The ultimate objective is to alleviate the distress caused by insomnia and to facilitate the recovery process in stroke patients.