BACKGROUND COVID-19 has generated an unprecedented impact. Patients in critical states present a reported mortality of up to 61%, and aggressive treatment does not seem to be enough. Different studies have described associated factors to mortality; however, in Latin America, the information is scarce. Thus, we aimed to determine the associated factors with mortality due to COVID-19 in a high-complexity hospital in Bogotá, Colombia. METHODS This case-control study included 282 individuals who died due to COVID-19, compared with 282 individuals who survived. Individuals were matched by age, sex, and month of admission to determine if there were associated variables with the primary outcome. Multiple imputations by chained equation (MICE) were implemented to account for missing variables. Classification and regression trees (CART) were estimated to evaluate the interaction of factors on admission and their role in predicting mortality during hospitalization. RESULTS Most of the patients included were males in the seventh decade. Recovered patients reported heterogeneous symptomatology, whereas deceased patients were most likely to present respiratory distress, dyspnea, and seizures on admission. In addition, the latter group exhibited a higher burden of comorbidities and alterations in laboratory biomarkers. After the imputation of datasets, CART analysis estimated 14 clinical profiles. The accuracy model for prediction was 85.6% (P < 0.0001). CONCLUSION Multivariate analysis yielded a reliable model to predict mortality in COVID-19. This analysis revealed new interactions between clinical and paraclinical features. Furthermore, this predictive model could offer new clues for the personalized management of this condition in clinical settings.
