作者：杨贵分，张少华，刘卫珍，陈敏霞，姚兰，廖旭

单位：

430070湖北省武汉市，中部战区总医院心血管内科

Units：

Department of Cardiology, General Hospital of Central Theater Command, Wuhan 430070, China

关键词：

急性冠脉综合征；猝死，心脏；院内心搏骤停；机器学习；预测

Keywords：

Acute coronary syndrome; Death, sudden, cardiac; In-hospital cardiac arrest; Machine learning; Forecasting

CLC：

R 542.2

DOI：

10.12114/j.issn.1008-5971.2024.00.070

Funds：

湖北省卫生健康委科研项目（WJ2023M090）

摘要：

　目的　比较两种机器学习模型〔决策树早期预警得分（DTEWS）模型、决策树模型〕对急性冠状动 脉综合征（ACS）患者发生院内心搏骤停（IHCA）的预测价值。方法　采用便利抽样法回顾性选取2018—2022年入 住某三甲医院心血管内科的发生IHCA的ACS患者53例为IHCA组，选取同期入住某三甲医院心血管内科的未发生IHCA 的ACS患者706例为非IHCA组。基于两种机器学习模型中的预测因子，收集患者一般资料，采用ROC曲线评估两种机 器学习模型对ACS患者发生IHCA的预测价值。结果　两组年龄、糖尿病发生率、致命性心律失常发生率、吸氧者占 比、入院方式和入院时Killip分级、收缩压、舒张压、血尿素氮、心肌肌钙蛋白T、意识状态及住院时间比较，差异 有统计学意义（P＜0.05）。ROC曲线分析结果显示，DTEWS模型、决策树模型预测ACS患者发生IHCA的AUC分别为 0.815〔95%CI（0.785～0.842）〕、0.824〔95%CI（0.795～0.851）〕。DTEWS模型、决策树模型预测ACS患者发生 IHCA的AUC比较，差异无统计学意义（P＞0.05）。结论　DTEWS模型、决策树模型均对ACS患者发生IHCA有中等预 测价值，其中DTEWS模型纳入变量较少、易获取且计算方式简单，更适宜临床推广。

Abstract：

Objective To compare the predictive value of two machine learning models [decision-tree early warning score (DTEWS) model and decision tree model] for in-hospital cardiac arrest (IHCA) in patients with acute coronary syndrome (ACS) . Methods Fifty-three ACS patients with IHCA who were admitted to the Cardiovascular Department of a tertiary hospital from 2018 to 2022 were retrospectively selected as the IHCA group by convenience sampling method, and 706 ACS patients without IHCA who were admitted to the Cardiovascular Department of a tertiary hospital during the same period were selected as the non-IHCA group. Based on the predictors in the two machine learning models, general information of patients was collected, and ROC curve was used to assess the predictive value of the two machine learning models for IHCA in ACS patients. Results There were significant differences in age, incidence of diabetes, incidence of fatal arrhythmia, percentage of oxygen users, mode of admission and Killip classification, systolic blood pressure, diastolic blood pressure, blood urea nitrogen, cardiac troponin T and state of consciousness at admission, and length of hospital stay between the two groups (P < 0.05) . ROC curve analysis showed that the AUC of DTEWS model and decision tree model for predicting IHCA in ACS patients was 0.815 [95%CI (0.785-0.842) ] and 0.824 [95%CI (0.795-0.851) ] , respectively. There was no statistically significant difference in AUC of DTEWS model and decision tree model for predicting IHCA in ACS patients (P < 0.05) . Conclusion Both DTEWS model and decision tree model have moderate predictive value for IHCA in ACS patients. Among them, DTEWS model has fewer variables, easy access and simple calculation method, and is more suitable for clinical promotion.

ReferenceList：