Development of a Model for Predicting Mortality Among Patients Hospitalized with COVID-19 During Their Stay in a Clinical Centre

Background: Various tools have been proposed for predicting mortality among patients hospitalized with COVID-19 to improve clinical decision-making, the predictive capacities of which vary in different populations. The objective of this study was to develop a model for predicting mortality among patients hospitalized with COVID-19 during their time in a clinical centre. Methods: This was a retrospective study that included 201 patients hospitalized with COVID-19. Mortality was evaluated with the Kaplan-Meier curve and Cox proportional hazards models. Six models were generated for predicting mortality from laboratory markers and patients' epidemiological data during their stay in a clinical centre. Results: The model that presented the best predictive power used D-dimer adjusted for C-reactive protein (CRP) and oxygen saturation. The sensitivity (Sn) and specificity (Sp) at 15 days were 75% and 71.9%, respectively. At 30 days, Sn was 75% and Sp was 75.4%. Conclusions: These results allowed us to establish a model for predicting mortality among patients hospitalized with COVID-19 based on D-dimer laboratory biomarkers adjusted for CRP and oxygen saturation. This mortality predictor will allow patients to be identified who require more continuous monitoring and health care.