In-Hospital Complications Associated with Mortality in Patients Hospitalized for COVID-19: A Prospective Single-Center Cohort Study

Background and Objectives: We assessed associations of variables recorded within 24 h of admission and in-hospital complications with mortality in adults hospitalized for COVID-19. Materials and Methods: This prospective single-center cohort included 249 adults hospitalized in 2021. Four mortality classifiers underwent repeated cross-validation and 0.632+ bootstrap validation, a resampling method that combines apparent and out-of-bag performance while adjusting for overfitting. Separate logistic regressions evaluated mortality and acute kidney injury (AKI), with AKI specifications including or excluding baseline renal laboratory predictors. Results: In-hospital mortality was 22.5%. The 0.632+ bootstrap estimates of the area under the receiver operating characteristic curve were 0.862 for random forest, 0.769 for ridge logistic regression, 0.757 for linear discriminant analysis, and 0.552 for quadratic discriminant analysis. Higher lymphocyte count was associated with lower odds of death (odds ratio 0.237; 95% confidence interval 0.089–0.632). AKI, shock, and renal replacement therapy were associated with high conditional mortality. Conclusions: Admission-assessment variables provided exploratory prognostic information on in-hospital mortality in adults hospitalized for COVID-19. Sparse events, incomplete covariate information, and the single-center setting limit inference. Independent validation, calibration, and assessment of clinical benefit are required before implementation.