Lopez-Escobar, AlejandroMadurga, RodrigoCastellano, Jose MariaVelazquez, SaraSuárez Del Villar, RafaelMenendez, JustoPeixoto, AlejandroJimeno, SaraVentura, Paula SolRuiz de Aguiar, Santiago2021-09-152021-09-152021-03-26Diagnostics (Basel). 2021; 11(4):5962075-4418http://hdl.handle.net/20.500.12105/13398Infection by SARS-CoV2 has devastating consequences on health care systems. It is a global health priority to identify patients at risk of fatal outcomes. 1955 patients admitted to HM-Hospitales from 1 March to 10 June 2020 due to COVID-19, were were divided into two groups, 1310 belonged to the training cohort and 645 to validation cohort. Four different models were generated to predict in-hospital mortality. Following variables were included: age, sex, oxygen saturation, level of C-reactive-protein, neutrophil-to-platelet-ratio (NPR), neutrophil-to-lymphocyte-ratio (NLR) and the rate of changes of both hemogram ratios (VNLR and VNPR) during the first week after admission. The accuracy of the models in predicting in-hospital mortality were evaluated using the area under the receiver-operator-characteristic curve (AUC). AUC for models including NLR and NPR performed similarly in both cohorts: NLR 0.873 (95% CI: 0.849-0.898), NPR 0.875 (95% CI: 0.851-0.899) in training cohort and NLR 0.856 (95% CI: 0.818-0.895), NPR 0.863 (95% CI: 0.826-0.901) in validation cohort. AUC was 0.885 (95% CI: 0.885-0.919) for VNLR and 0.891 (95% CI: 0.861-0.922) for VNPR in the validation cohort. According to our results, models are useful in predicting in-hospital mortality risk due to COVID-19. The RIM Score proposed is a simple, widely available tool that can help identify patients at risk of fatal outcomes.engVoRhttp://creativecommons.org/licenses/by/4.0/Atribución 4.0 Internacional3381053411459610.3390/diagnostics11040596Diagnostics (Basel, Switzerland)open access