Risk Score for Predicting In-Hospital Mortality in COVID-19 (RIM Score)

Abstract

A multicenter retrospective cohort study including 1955 consecutive patients admitted to HM Hospitales (Spain) between March and June 2020 developed and validated a prognostic model for in-hospital mortality due to COVID-19. Patients were randomly divided into training (n = 1310) and validation (n = 645) cohorts. Logistic regression models were constructed using routinely available variables at admission (age, sex, oxygen saturation, C-reactive protein, neutrophil-to-lymphocyte ratio (NLR), neutrophil-to-platelet ratio (NPR)) and incorporating the rate of change of hemogram-derived ratios during the first week (VNLR and VNPR). The models demonstrated high discriminative capacity with AUC values ranging from 0.853 to 0.861 in baseline validation models and up to 0.896 when including velocity parameters. Calibration curves and decision curve analysis confirmed robustness. The resulting RIM Score is a simple, reproducible and widely accessible tool for early risk stratification of hospitalized COVID-19 patients.