- 95% inter-quantile ranges are presented for each estimate. Specificity is obtained by evaluating at a set sensitivity of either 90% or 95%. All reduced variable models include age, and a stated number of biomarkers. The reduced variable standard GLM uses age and 16 biomarkers that had non-zero coefficients on \(>=\)50% LASSO Cross-validation trials. If at least one categorical level for a particular biomarker (e.g. severe) met this requirement, all levels for that biomarker were included in the model. The 4 biomarker projective prediction model uses all categorical levels for Urea, PT, CRP, and NLR. Pairwise AUC difference is presented in comparison to the Bayesian (Horse shoe prior) model
