Introduction: Neutrophil-to-lymphocyte (NLR) and lymphocyte-to-C-reactive protein (LCR) ratios are used to predict severity and mortality in various infections. Objective: To establish the best NLR and LCR cutoff point to predict mortality in patients hospitalized for COVID-19 in Mexico. Method: Analytical cross-sectional study of patients hospitalized for severe COVID-19 in a specialty hospital. Results: Out of 242 analyzed patients, 34 % died. The deceased subjects were older (62 vs. 51 years; p < 0.001), had a higher prevalence of > 10 years with systemic arterial hypertension (59.4 vs. 45.1 %, p = 0.022), as well as a higher NLR (17.66 vs. 8.31, p < 0.001) and lower LCR (0.03 vs. 0.06, p < 0.002] with regard to those who survived. The cutoff points to predict mortality were NLR > 12 and LCR < 0.03. The combination of NLR/LCR had a sensitivity of 80 %, specificity of 74 %, positive predictive value of 46.15 %, negative predictive value of 93.02 % and an odds ratio of 11.429 to predict mortality. Conclusion: NLR > 12 and LCR < 0.03 are useful biomarkers to evaluate the risk of mortality in Mexican patients with severe COVID- 19. Keywords: COVID-19. In-hospital mortality. Neutrophil-to-lymphocyte ratio. DOI: 10.24875/GMM.20000525