UTILIZAÇÃO DE MACHINE LEARNING NA SAÚDE: PREVENÇÃO E DETECÇÃO DE RISCO DE AGRAVAMENTO EM PACIENTES COM COVID-19 DEVIDO À DEFICIÊNCIA DE ZINCO, FERRO E VITAMINA D

Abstract

Introduction: COVID-19 is a viral disease caused by SARS-CoV-2, a member of the Coronaviridae family, which results in a severe acute respiratory syndrome. Mathematical and simulation models, combined with artificial intelligence, offer an advanced approach to improving management in health decision-making, including epidemiological surveillance of diseases such as COVID-19. Objective: To develop a predictive model for the detection and prevention of deaths in COVID-19 patients at risk of deterioration due to deficiencies in vitamin D, zinc, and iron in the blood serum. Methods: This study was conducted as an experimental, observational, and quantitative research focusing on a group of 75 patients who tested positive for COVID-19 and presented severe symptoms resulting in death. The k-means clustering algorithm was used based on this data. Subsequently, four predictive algorithms were employed: Support Vector Machines (SVM), Decision Trees (DT), Naive Bayes (NB), and K-Nearest Neighbours (KNN). These algorithms included epidemiological, clinical, laboratory, and imaging variables to predict patterns of adverse outcomes leading to deaths. Results: The obtained results revealed the existence of 8.208 symptomatic pairs and 7.318 asymptomatic cases. Only 75 symptomatic patients with a positive result for vitamin D, zinc, and iron deficiency were included based on classification and selection by Bloom Filters using Machine Learning and Deep Learning techniques to achieve the expected outcomes. Two patterns of adverse outcomes leading to death were derived from the sampling. The results showed statistically significant associations between nutritional deficiencies (vitamin D, zinc, and iron) and adverse outcomes of COVID-19 in a specific population in Campo Grande, MS. Conclusion: The studies demonstrate a significant association between vitamin D deficiency and the severity of COVID-19, while zinc and iron are associated with higher mortality and an exacerbated inflammatory response. An AI-based algorithm showed efficacy in screening, contributing to the early identification of critical cases.