Development and Validation of Interpretable Machine Learning Models for Triage Patients Admitted to the Intensive Care Unit

Objectives: Developing and validating interpretable machine learning (ML) models for predicting whether triaged patients need to be admitted to the intensive care unit (ICU).Measures: This was a single-center, retrospective study. Emergency Severity Index (ESI), vital signs, demographic characteristics, history, and chief complaints of triaged patients were extracted from the Medical Information Mart for Intensive Care IV database, and the predicted outcome was admission to the ICU.Three models were compared: Model 1 based on ESI, Model 2 on vital signs, and Model 3 on vital signs, demographic characteristics, medical history, and chief complaints. Nine ML algorithms were employed. The area under the receiver operating characteristic curve (AUC), Matthews correlation coefficient (MCC), learning curves, recall curves, calibration curves, and decision curves analysis were used to evaluate the performance of the models.

研究目标：开发并验证可解释机器学习（Machine Learning, ML）模型，用于预测经分诊的患者是否需要收入重症监护病房（Intensive Care Unit, ICU）。研究方法：本研究为单中心回顾性研究。从重症监护医学信息数据库IV（Medical Information Mart for Intensive Care IV, MIMIC-IV）中提取经分诊患者的急诊严重指数（Emergency Severity Index, ESI）、生命体征、人口统计学特征、病史及主诉信息，模型的预测结局为患者收入ICU。共对比3种模型：模型1仅基于急诊严重指数，模型2仅基于生命体征，模型3则整合生命体征、人口统计学特征、病史与主诉信息。本次研究共采用9种机器学习算法，并通过受试者工作特征曲线下面积（Area Under the Receiver Operating Characteristic Curve, AUC）、马修斯相关系数（Matthews Correlation Coefficient, MCC）、学习曲线、召回率曲线、校准曲线以及决策曲线分析，对各模型的性能进行评估。