A cross-sectional analysis of severe SARS cases evolution in a brazilian municipality using data mining techniques

Abstract

Abstract The first Severe Acute Respiratory Syndrome (SARS) outbreak occurred in China in November 2002. Since then, other coronavirus variants have emerged worldwide, such as Middle East Respiratory Syndrome (MERS) in 2012, 2019-nCOV in 2019, and Omicron in 2020. Several studies have been published, demonstrating the use of data mining (DM) to create relevant classification and decision systems for analyzing clinical data of patients with SARS. However, most of these studies lack the depth of consideration for the socioeconomic factors of the patients, such as income, education levels, race, among others, which could be relevant for classification algorithms. This study demonstrates the application of the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework and the use of DM techniques and tools, employing binary classification and decision algorithms to predict the progression of severe cases in SARS patients residing in the municipality of Recife, Pernambuco, Brazil. It is a cross-sectional study conducted using open datasets, considering various attributes related to symptoms, pre-existing conditions, and socioeconomic factors, including income, literacy rate, and patient domicile location. The analysis involved three healthcare experts (physicians). The results highlighted that the apriori algorithm performed better in rule induction, and the decision tree showed slightly better performance compared to logistic regression. Furthermore, the analysis brought to light interesting correlations between the progression of severe cases and the socioeconomic data of patients.