What most influences severity and death of COVID-19 patients in Brazil? Is it clinical, social, or demographic factors? An observational study.

ObjectiveThis study aimed to assess the space distribution and factors associated with the risk of severe acute respiratory syndrome (SARS) and death in COVID-19 patients, based on routine register data; and to develop and validate a predictive model of the risk of death from COVID-19. MethodsA cross-sectional, epidemiological study of positive SARS-CoV-2 cases, reported in the south region of the city of Sao Paulo, SP, Brazil, from March 2020 to February 2021. Data were obtained from the official reporting databases of the Brazilian Ministry of Health for influenza-like illness (ILI) (esus-VE, in Portuguese) and for patients hospitalized for SARS (SIVEP-Gripe). The space distribution of cases is described by 2D kernel density. To assess potential factors associated with the outcomes of interest, generalized linear and additive logistic models were adjusted. To evaluate the discriminatory power of each variable studied as well as the final model, C-statistic was used (area under the receiver operating characteristics curve). Moreover, a predictive model for risk of death was developed and validated with accuracy measurements in the development, internal and temporal (March and April 2021) validation samples. ResultsA total of 16,061 patients with confirmed COVID-19 were enrolled. Morbidities associated with a higher risk of SARS were obesity (OR=25.32) and immunodepression (OR=12.15). Morbidities associated with a higher risk of death were renal disease (OR=11.8) and obesity (OR=8.49), and clinical and demographic data were more important than the territory per se. Based on the data, a calculator was developed to predict the risk of death from COVID-19, with 92.2% accuracy in the development sample, 92.3% in the internal validation sample, and 80.0% in the temporal validation sample. ConclusionsThe risk factors for SARS and death in COVID-19 patients seeking health care, in order of relevance, were age, comorbidities, and socioeconomic factors, considering each discriminatory power.