Scholarlink

This study aimed to predict the Intensive Care Unit (ICU) Length of Stay for individuals with Inflammatory Bowel Disease (IBD) using the XGBoost model. Electronic Health Record (EHR) data from The Medical Information Mart for Intensive Care (MIMIC) - IV (version 2.2) dataset was used in the study. Variables included patients’ demographic information, first measurements of basic physical and laboratory tests after ICU admission, and some important comorbidities. After data preprocessing, 652 IBD patients and their first ICU records were selected. With ten-fold cross-validation, the XGBoost model demonstrated strong performances with area under the curve (AUC) (0.787), accuracy (0.768), and sensitivity (0.943). The most significant variables include age, mean corpuscular volume (MCV), red blood cell distribution width (RDW), acute respiratory failure, among others. Our study could serve as a predictive model for the ICU Length of Stay and offer guidance and references for medical organizations in resource allocation, staff assignments, and bed preparation. This, in turn, can improve efficiency and reduce patient mortality rates.

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

Predicting Intensive Care Unit Length of Stay for Inflammatory Bowel Diseases Patients Using Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ke Xu,
  • Jiayi Nie,
  • Yifan Chen,
  • Ziqi Ban,
  • Ling Liu,
  • Kang Li,
  • Di Liu,
  • Rong Yin

摘要

This study aimed to predict the Intensive Care Unit (ICU) Length of Stay for individuals with Inflammatory Bowel Disease (IBD) using the XGBoost model. Electronic Health Record (EHR) data from The Medical Information Mart for Intensive Care (MIMIC) - IV (version 2.2) dataset was used in the study. Variables included patients’ demographic information, first measurements of basic physical and laboratory tests after ICU admission, and some important comorbidities. After data preprocessing, 652 IBD patients and their first ICU records were selected. With ten-fold cross-validation, the XGBoost model demonstrated strong performances with area under the curve (AUC) (0.787), accuracy (0.768), and sensitivity (0.943). The most significant variables include age, mean corpuscular volume (MCV), red blood cell distribution width (RDW), acute respiratory failure, among others. Our study could serve as a predictive model for the ICU Length of Stay and offer guidance and references for medical organizations in resource allocation, staff assignments, and bed preparation. This, in turn, can improve efficiency and reduce patient mortality rates.