Validated clinical prediction model for mortality from COVID-19 in hospitalized patients. What is truly important?

Objective: To develop and validate a clinical prediction model aimed at improving resource management and determining the prognosis of patients hospitalized with COVID-19. Materials and methods: A retrospective, single-center cohort study conducted at the University Hospital of Guadalajara, including 1,043 patients hospitalized with COVID-19 between March and May 2020. Data were extracted from hospital records and anonymized. Demographic, clinical, laboratory, radiological, and therapeutic variables were collected, and statistical analysis was performed to identify factors associated with mortality. Logistic regression and Cox models were employed to evaluate mortality predictors. Validation was conducted by comparing ROC curves. Results: The median age of the patients was 70.4 years (P25-P75: 59-84), with 59.2% being male, and a mortality rate of 23.2%. The most common comorbidities were hypertension (54.8%), dyslipidemia (36.3%), and diabetes (27.1%). Independent predictors of mortality included age over 80years (OR: 6.18), chronic obstructive pulmonary disease (OR: 2.35), oxygen saturation < 90% (OR: 1.7), multilobar pneumonia (OR: 2.4), and elevated LDH levels (OR: 1.2). The area under the curve (AUC) for the derivation model was 0.805 (P < .001), and for the validation model, the AUC was 0.78 (P < .001). Conclusions: Advanced age, chronic obstructive pulmonary disease, low oxygen saturation, multilobar pneumonia, and elevated LDH levels are significantly associated with increased mortality risk. The validated predictive model enables classification of patients into high- or low-risk groups, thereby facilitating improved clinical decision-making and resource management. (c) 2025 Sociedad Espanola de Medicos de Atencion Primaria (SEMERGEN). Published by Elsevier Espana, S.L.U. All rights are reserved, including those for text and data mining, AI training, and similar technologies.