| Sumario: | • Machine learning has a total of 5 models in clinical triage. • Neural networks are suitable for large datasets but not for smaller, single-center experimental clinical trial data. • Gradient Boosting Decision Trees are suitable for using both categorical and continuous variables. • The random forest algorithm can improve the performance of a model and generate more effective models. • Support Vector Machines are developed based on small sample statistical theory. To investigate the application status of machine learning model in the prediction of clinical outcomes in emergency pre-examination and triage, and to analyze its characteristics, advantages and disadvantages, so as to add an objective tool for medical staff to predict the clinical outcome of patients in the process of pre-examination and triage. The literature review method was used to search PubMed, Web of Science, Embase, Cochrane Library, China Biomedical Literature Database, CNKI, Wanfang, VIP and other databases, and the literature that met the inclusion criteria was screened and the specific information of the machine learning model in the literature was extracted. A total of 12 articles that met the criteria were included, including 5 machine learning models, which were mainly used in clinical outcomes such as hospital admission, death, intensive care unit admission, hospital transfer, and home. The overall sensitivity of the machine learning model is high, but there are few literature studies on the prediction of clinical outcomes for pre-test triage, so relevant large-sample studies should be carried out in clinical practice to achieve the combination of subjective and objective evaluation tools to improve the accuracy of prediction and ensure patient safety.
|
|---|