Comparative Prognostic Performance of 18 Routine Laboratory-Derived Indices for In-Hospital Mortality in Patients with COVID-19

Background/Objectives: Numerous hematologic, inflammatory, biochemical, and nutritional indices have been proposed as prognostic markers in coronavirus disease 2019 (COVID-19), but direct comparison of a broad range of these indices within the same cohort remains limited. This study aimed to evaluate and compare the prognostic performance of multiple routinely available indices for in-hospital mortality. Methods: This retrospective observational cohort included 1952 adults hospitalized with COVID-19. Demographic, clinical, comorbidity, and laboratory data were obtained from medical records. Eighteen derived indices were evaluated, including NLR, PLR, SII, CAR, CLR, LAR, UAR, CPR, DAR, and PNI. Receiver operating characteristic analysis was performed to assess discrimination for in-hospital mortality, and optimal cut-offs were determined using the Youden index. Selected indices were further evaluated in separate multivariable logistic regression models adjusted for age, sex, diabetes mellitus, CAD/IHD, and eGFR. Results: In-hospital mortality occurred in 80 patients (4.1%). Non-survivors were older and had higher inflammatory and biochemical markers and lower lymphocyte counts and albumin levels. CAR and CLR showed the highest numerical AUCs for in-hospital mortality (0.870 for both), followed by PNI (0.857), CPR (0.852), DAR (0.829), UAR (0.823), LAR (0.815), and NLR (0.807). In separate multivariable models adjusted for age, sex, diabetes mellitus, CAD/IHD, and eGFR, ln(CLR), ln(CAR), ln(CPR), ln(LAR), ln(UAR), ln(NLR), ln(dNLR), and ln(DAR) remained independently associated with mortality (all p < 0.001). Among the adjusted models, the CLR model showed the highest explained variance (Nagelkerke R2 = 0.511) with good calibration (Hosmer–Lemeshow p = 0.997); ln(CLR) remained independently associated with mortality (aOR 3.590, 95% CI 2.517–5.120; p < 0.001). Conclusions: Several routinely available composite indices were associated with in-hospital mortality in COVID-19. Among the evaluated indices, CAR and CLR showed the highest individual discrimination, while the CLR-based adjusted model showed the most favorable model-fit profile among the multivariable models evaluated. Multiple selected indices remained independently associated with mortality after adjustment for clinical covariates. These findings suggest that such markers may have potential value as adjunctive tools for mortality risk stratification.