Uso da inteligência artificial na classificação de risco nos serviços de emergência: revisão de escopo

Abstract

Objective: to map the evidence available in the literature on the use of artificial intelligence in risk classification in Emergency services. Method: This is a Scoping Review conducted based on the nine-step methodological framework adopted by the Joanna Briggs Institute and the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA ScR). The search was carried out in thirteen data sources: CINAHL, Cochrane Library, PubMed Central, Scielo, Web of Science, SCOPUS, Science Direct, Virtual Health Library, Embase, CAPES, Brazilian Digital Library of Theses and Dissertations, Open Access Scientific Repository of Portugal, and Theses Canada. Research available in open access, without time frame and language, was selected. Results: In total, 17 studies were selected, of which 6 were retrospective and 7 were cross-sectional studies, the others were systematic reviews and observational studies. Published between 2013 and 2024, with the years 2021 and 2023 being the years with the highest number of productions. As for the countries of origin, Asian countries stand out (52.94%), followed by European countries (29.41%). Regarding the techniques used, machine learning stands out with 41.18%, followed by its combination with natural language processing, with 17.65%. In general, the analyses performed show that models based on machine learning outperform traditional models in terms of their accuracy. The techniques were applied through automated classification, acting to predict the level of clinical severity of patients, as well as in defining their priority, according to the objective and subjective data collected in the evaluation. Final Considerations: The findings in this review demonstrated the satisfactory use of artificial intelligence in the risk classification of patients in emergency services, in addition to improving the flow of the service and being a good tool for supporting clinical decision-making.