Machine learning techniques for mortality prediction in emergency departments: a systematic review

This systematic review aimed to assess the performance and clinical feasibility of ML algorithms in prediction of in-hospital mortality for medical patients using vital signs at emergency departments.

Design: A systematic review was performed.

Setting: The databases including Medline (PubMed), Scopus, and Embase (Ovid) were searched between 2010 and 2021, to extract published articles in English, describing ML-based models utilizing vital signs variables to predict in-hospital mortality for patients admitted at emergency departments. CHARMS checklist was used for study planning and data extraction. The risk of bias for included papers was assessed using the PROBAST tool.

Participants: Admitted patients to the ED

Main outcome measure: In-hospital mortality.

Results: Fifteen articles were included in the final review. We found that eight models including logistic regression, decision tree, K-nearest neighbors, support vector machine, gradient boosting, random forest, artificial neural networks, and deep neural networks have been applied in this domain. Most studies failed to report essential main analysis steps such as data preprocessing and handling missing values. Fourteen included studies had a high risk of bias in the statistical analysis part, which could lead to poor performance in practice. Although the main aim of all studies was developing a predictive model for mortality, nine articles did not provide a time horizon for the prediction.

Conclusion: This review provided an updated overview of the state-of-the-art and revealed research gaps; based on these, we provide eight recommendations for future studies to make the use of ML more feasible in practice. By following these recommendations, we expect to see more robust ML models applied in the future to help clinicians identify patient deterioration earlier.

Methods
Three databases including Medline (PubMed), Scopus, and Embase (Ovid) were targeted. Relevant articles were extracted using a broad range of relevant keywords. We stratified keywords into five groups namely, ML keywords, medical keywords, document type, publication year, and language. The keywords in a group were paired using OR operators and all groups were paired using AND operator.

本系统综述旨在评估基于急诊科（Emergency Department, ED）生命体征构建的机器学习（Machine Learning, ML）算法在预测内科患者住院死亡结局方面的性能与临床可行性。

研究设计：本研究为系统综述。

研究场景：本研究于2010年至2021年间检索Medline（PubMed）、Scopus及Embase（Ovid）三大数据库，提取以英文发表的、描述基于生命体征变量构建机器学习模型以预测急诊科收治患者住院死亡结局的相关文献。研究规划与数据提取环节采用CHARMS检查表（CHARMS checklist）进行规范；纳入文献的偏倚风险则通过PROBAST工具（PROBAST tool）进行评估。

研究对象：急诊科收治的患者。

主要结局指标：住院死亡结局。

研究结果：最终纳入15篇文献进行综述。本综述发现，该领域已应用的机器学习模型共8类，包括逻辑回归、决策树、K近邻（K-nearest neighbors）、支持向量机、梯度提升、随机森林、人工神经网络及深度神经网络。多数研究未报告数据分析的核心关键步骤，如数据预处理与缺失值处理。纳入的14项研究在统计分析环节存在较高偏倚风险，这可能导致模型在实际应用中表现不佳。尽管所有研究的核心目标均为构建死亡预测模型，但其中9篇文献未明确提及预测的时间窗。

研究结论：本综述对当前该领域的前沿进展进行了系统性梳理，并明确了现有研究的不足；基于此，我们针对未来研究提出8项建议，以提升机器学习技术在临床实践中的应用可行性。我们期望，遵循上述建议后，未来将有更稳健的机器学习模型问世，助力临床医师更早识别患者病情恶化情况。

研究方法：本研究以Medline（PubMed）、Scopus及Embase（Ovid）三大数据库为检索目标。通过多组相关关键词组合提取符合要求的文献：将关键词划分为5大类，即机器学习关键词、医学关键词、文献类型、出版年份及语言；每一类内的关键词以“或（OR）”逻辑运算符连接，各类别间则以“与（AND）”逻辑运算符组合。