Scholarlink

Background <p>Insomnia is a common complication in ischemic stroke convalescence (ISC) patients. While the interaction of clinical, psychological, and social factors remains unclear, developing a predictive model system is urgently needed. Currently, few studies have established insomnia risk prediction models.</p> Objectives <p>To construct a decision tree model for insomnia risk among ISC patients based on the classification and regression tree algorithm.</p> Design <p>Across-sectional study.</p> Setting <p>China.</p> Participants <p>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.</p> Methods <p>Following the TRIPOD+AI guidelines, we constructed a decision tree model utilizing data from the Pittsburgh Sleep Quality Index (PSQI), Fatigue Severity Scale (FSS), Social Support Scale (SSRS), and other assessment tools. Model validation encompassed 10-fold cross-validation, incorporating confusion matrix, ROC curves, calibration curve, and Brier scores. The model was trained on 623 patients and externally validated on an independent cohort of 200 cases.</p> Results <p>The study revealed an insomnia prevalence of 37.72%camong ISC patients. Univariate analysis identified BMI, SAS, SSRS, FSS, SDS, and NIHSS as significant factors. The decision tree model delineated 24 pathways (depth = 6), with predictive contributions ranked as follows: SAS &gt; SSRS &gt; FSS &gt; SDS &gt; BMI &gt; NIHSS, which were integrated into a nomogram. Internal validation exhibited robust predictive accuracy (90.4%), with a sensitivity of 0.96, specificity of 0.84, Youden index of 0.80, and F1 score of 0.89. The AUC was 0.96 (95% CI: 0.93–0.98; <i>p</i> &lt; 0.001), indicating well-calibrated predictions (χ² = 9.36, <i>p</i> = 0.404). Brier scores were 0.06 for the training set and 0.08 for the validation set. External validation demonstrated an accuracy of 82%. The decision curve analysis demonstrated acceptable clinical utility.</p> Conclusion <p>This model demonstrates promise in forecasting insomnia among ISC patients. Anxiety and social support emerged as the most influential predictors, with fatigue, depression, BMI, and stroke severity collectively offering a comprehensive outlook for anticipating post-stroke insomnia. These results have implications for informing future strategies in managing insomnia. The model's applicability is moderately robust, necessitating additional refinement to accurately pinpoint insomnia.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

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

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月19日

查看全文
  • Xuefeng Sun,
  • Zilin Wang,
  • Yuqing Song,
  • Deyu Cong,
  • Shu Sun,
  • Xinye Zhang,
  • Ye Zhang,
  • Hongshi Zhang

摘要

Background

Insomnia is a common complication in ischemic stroke convalescence (ISC) patients. While the interaction of clinical, psychological, and social factors remains unclear, developing a predictive model system is urgently needed. Currently, few studies have established insomnia risk prediction models.

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

Following the TRIPOD+AI guidelines, we constructed a decision tree model utilizing data from the Pittsburgh Sleep Quality Index (PSQI), Fatigue Severity Scale (FSS), Social Support Scale (SSRS), and other assessment tools. Model validation encompassed 10-fold cross-validation, incorporating confusion matrix, ROC curves, calibration curve, and Brier scores. The model was trained on 623 patients and externally validated on an independent cohort of 200 cases.

Results

The study revealed an insomnia prevalence of 37.72%camong ISC patients. Univariate analysis identified BMI, SAS, SSRS, FSS, SDS, and NIHSS as significant factors. The decision tree model delineated 24 pathways (depth = 6), with predictive contributions ranked as follows: SAS > SSRS > FSS > SDS > BMI > NIHSS, which were integrated into a nomogram. Internal validation exhibited robust predictive accuracy (90.4%), with a sensitivity of 0.96, specificity of 0.84, Youden index of 0.80, and F1 score of 0.89. The AUC was 0.96 (95% CI: 0.93–0.98; p < 0.001), indicating well-calibrated predictions (χ² = 9.36, p = 0.404). Brier scores were 0.06 for the training set and 0.08 for the validation set. External validation demonstrated an accuracy of 82%. The decision curve analysis demonstrated acceptable clinical utility.

Conclusion

This model demonstrates promise in forecasting insomnia among ISC patients. Anxiety and social support emerged as the most influential predictors, with fatigue, depression, BMI, and stroke severity collectively offering a comprehensive outlook for anticipating post-stroke insomnia. These results have implications for informing future strategies in managing insomnia. The model's applicability is moderately robust, necessitating additional refinement to accurately pinpoint insomnia.