Analysis plan.

Values were expressed in numbers and proportions or means ± SD where appropriate. Continuous variables were compared using Mann-Whitney U test for nonnormal data distributions, and by independent t test for normal data distributions. Dichotomous variables were compared using Pearson’s χ² test or Fisher’s exact test.

Correlations between HMGB and IL-6 concentrations at ICU admission with peak laboratory data were estimated by Spearman’s correlation. The ICU fatality prediction of HMGB levels, IL-6 levels, neutrophil/lymphocyte ratio, D-dimer level, SOFA scores, and COVID-GRAM risk scores at ICU admission were further evaluated by receiver operating characteristic curve (ROC) analysis, and areas under the curve (AUCs) were calculated. Youden Index was calculated on the basis of the ROC to help set the appropriate cutoff value as the best obtainable balance of sensitivity and specificity. Comparison of survival curves used Kaplan-Meier survival analysis, and differences were compared using the log-rank test. Multivariate analysis was performed to adjust for disparate baseline characteristics contributing to mortality.

Data were analyzed using IBM SPSS Statistics for Windows version 18.0 (IBM Corp., Armonk, NY) and GraphPad Prism version 7.0 (GraphPad Software Inc., San Diego, CA).