Scholarlink

Background <p>Prolonged postoperative length of stay (PLOS) is associated with several clinical risks and increased medical costs. This study aimed to develop a prediction model for PLOS based on clinical features throughout pre-, intra-, and post-operative periods in patients undergoing laparoscopic gastrointestinal surgery.</p> Methods <p>This secondary analysis included patients who underwent laparoscopic gastrointestinal surgery in the FDP-PONV randomized controlled trial. This study defined PLOS as a postoperative length of stay longer than 7&#xa0;days. All clinical features prospectively collected in the FDP-PONV trial were used to generate the models. This study employed six machine learning algorithms including logistic regression, K-nearest neighbor, gradient boosting machine, random forest, support vector machine, and extreme gradient boosting (XGBoost). The model performance was evaluated by numerous metrics including area under the receiver operating characteristic curve (AUC) and interpreted using shapley additive explanations (SHAP).</p> Results <p>Cases on or before 2021–12-18 formed the training cohort (802 patients) and those after that date formed the validation cohort (339 patients). Among the models evaluated, the XGBoost model demonstrated the best performance in the validation cohort, achieving an AUC of 0.756. After five times repeated tenfold cross-validation, the XGBoost model demonstrated a mean AUC of 0.816. SHAP plots were generated based on the predictors: education level; preoperative hypoalbuminemia and insufficient sleep; surgery type and duration; blood loss; postoperative hypotension; postoperative albumin infusion; highest pain score during 73–120&#xa0;h after surgery; time to first flatus and defecation; postoperative infection; and postoperative acute kidney injury.</p> Conclusions <p>This study developed a prediction model using machine learning to identify patients at high risk of PLOS after laparoscopic gastrointestinal surgery.</p> Trial registration <p>The FDP-PONV trial was registered at clinicaltrials.gov (NCT04853147) on 2021–04-27.</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 machine learning-based prediction model for prolonged length of stay after laparoscopic gastrointestinal surgery: a secondary analysis of the FDP-PONV trial

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月8日

查看全文
  • Jiankun Shi,
  • Yabin Huang,
  • Jiaxin Han,
  • Shimin Zhang,
  • Xingshan Cheng,
  • Tongfeng Luo

摘要

Background

Prolonged postoperative length of stay (PLOS) is associated with several clinical risks and increased medical costs. This study aimed to develop a prediction model for PLOS based on clinical features throughout pre-, intra-, and post-operative periods in patients undergoing laparoscopic gastrointestinal surgery.

Methods

This secondary analysis included patients who underwent laparoscopic gastrointestinal surgery in the FDP-PONV randomized controlled trial. This study defined PLOS as a postoperative length of stay longer than 7 days. All clinical features prospectively collected in the FDP-PONV trial were used to generate the models. This study employed six machine learning algorithms including logistic regression, K-nearest neighbor, gradient boosting machine, random forest, support vector machine, and extreme gradient boosting (XGBoost). The model performance was evaluated by numerous metrics including area under the receiver operating characteristic curve (AUC) and interpreted using shapley additive explanations (SHAP).

Results

Cases on or before 2021–12-18 formed the training cohort (802 patients) and those after that date formed the validation cohort (339 patients). Among the models evaluated, the XGBoost model demonstrated the best performance in the validation cohort, achieving an AUC of 0.756. After five times repeated tenfold cross-validation, the XGBoost model demonstrated a mean AUC of 0.816. SHAP plots were generated based on the predictors: education level; preoperative hypoalbuminemia and insufficient sleep; surgery type and duration; blood loss; postoperative hypotension; postoperative albumin infusion; highest pain score during 73–120 h after surgery; time to first flatus and defecation; postoperative infection; and postoperative acute kidney injury.

Conclusions

This study developed a prediction model using machine learning to identify patients at high risk of PLOS after laparoscopic gastrointestinal surgery.

Trial registration

The FDP-PONV trial was registered at clinicaltrials.gov (NCT04853147) on 2021–04-27.