Variables included in the multivariate model were age, gender, race, SES, and MOI.
For multivariate model, each SNP together with all available relevant covariates (multivariate model) were included using the Enter algorithm (Forced Entry).
Variables tested in the multivariate model were: age at admission, delay from ICU admission to tumour imaging, delay from ICU admission to first immunomodulating treatment, type of first treatment and white blood cell (WBC) count in the first CSF examination.
All patient characteristics retained in the final multivariate model were significant at the 0.05 level.
Two-way interactions between biological plausible and significant variables in the multivariate model were investigated.
The variables introduced into the multivariate model were selected considering the clinical context.
The variables which remained in the final multivariate model were checked for interaction.
Only survey variables which remained significant in the multivariate model were retained.
The results from the final multivariate model were reported as the adjusted HRs.
For brevity, the non-zero coefficients in the multivariate model were denoted significant effects.
Parameters included in the multivariate model were selected according to the statistical significance in univariate models.
