Emergency triage requires medical personnel to use a scientific and reasonable triage system to assess emergency patients in a short period of time, quickly identify their conditions, and classify them based on the severity of their illness. This ensures that critically ill patients receive timely treatment and that emergency resources are used effectively. This places high demands on the professional level of medical personnel and the triage system. Currently, various countries have gradually established relatively complete emergency triage systems and have also developed operational manuals for the clinical application of these systems. For example, there are the Chinese Emergency Triage Standards based on the principles of emergency patient condition grading [1], the American Emergency Severity Index (ESI) [2], the Australasian Triage Scale (ATS) in Australia, and the Canadian Emergency Department Triage and Acuity Scale (CTAS) [3,4]. These widely recognized emergency triage systems, although most standards are objective indicators and the operational manuals provide clear and detailed guidance for the clinical application of each standard, are still subject to the subjective judgment and experience of triage personnel due to the uncertainty of emergency patient conditions and the inherent complexity of triage. This can lead to triage errors and over-triage, as well as issues such as non-standard call systems and insufficiently intelligent triage and connection processes, which can easily lead to patient dissatisfaction [5]. Therefore, there is an urgent need for new technologies to assist triage personnel. Machine learning, as a branch of artificial intelligence, can extract effective objective indicators from clinical big data to achieve accurate triage, thereby reducing the incidence of medical adverse events and improving the quality of medical care. It can also predict emergency waiting times and optimize emergency processes. Therefore, machine learning models are currently being used in the research of emergency triage systems, such as pre-hospital emergency care, diagnostic decision-making, triage, and prognosis prediction. Machine learning includes various algorithms, such as decision trees, support vector machines, and artificial neural networks. As triage is the first step in patient care, choosing the appropriate machine learning model is crucial. Additionally, due to the generation of large and complex datasets using precision medical diagnostic methods, new technologies are needed to process and understand these complex data.

Therefore, this article summarizes the types of machine learning models currently used to predict clinical outcomes in triage, the current application status of machine learning models in clinical outcomes of triage, and the advantages and disadvantages of each model through a systematic literature review, in order to provide ideas and references for clinical application selection.