Acute neurological illnesses such as traumatic brain injury (TBI) and stroke are leading causes of death and disability, with a disproportionately high burden in LMICs. As advanced biomarkers are often inaccessible, this study evaluates routine low-cost laboratories (e.g., neutrophil percentage, albumin, HbA1c/EAG) alongside imaging which can refine early risk assessment, particularly where baseline under-nutrition is prevalent and may influence outcome.
This retrospective, single-center proof-of-concept study included consecutive neuro-ICU admissions with mixed etiologies. Predictors comprised a four-level composite imaging score (very low/low/moderate/high mortality risk) and admission laboratories (neutrophil %, albumin, total leukocyte count, HbA1c, estimated average glucose, NLR, PLR, CONUT and SHR). Outcomes were ICU death, poor disposition (death or DAMA), and early death (≤7 days). Penalized logistic regression with 5-fold cross-validation compared Imaging-only versus Imaging+lab models, summarized by mean AUROC and ΔAUC.
Ninety-five patients contributed to at least one endpoint. Imaging-only AUCs were 0.446 for early death, 0.589 for death, and 0.572 for poor disposition. Adding neutrophil percentage markedly improved early-death discrimination to AUC 0.837 (ΔAUC +0.391, N=94); albumin provided a moderate gain to AUC 0.621 (ΔAUC +0.175, N=93). For death, HbA1c/EAG modestly increased AUC to 0.648 (ΔAUC +0.058, N=94). No laboratory improved the poor-disposition endpoint. Exploratory analyses showed SHR, NLR, PLR, and CONUT had limited discrimination for these outcomes in this cohort.