Evaluation of the CNS-LAND Score for ICU Mortality Prediction in Patients with Spontaneous Intracerebral Hemorrhage

Objectives: This study investigated the prognostic value of the CNS-LAND score in predicting Intensive Care Unit (ICU) mortality among patients admitted with spontaneous intracerebral hemorrhage (ICH). Methods: In this single-center, retrospective cohort study, 305 adult ICU patients with spontaneous ICH (September 2022-March 2026) were included. The primary endpoint was ICU mortality. Prognostic performance was evaluated using receiver operating characteristic (ROC) curve analysis and the bootstrap-based DeLong method. The independent contribution of CNS-LAND beyond the Glasgow Coma Scale (GCS) was assessed using partial correlation and logistic regression analyses. Results: ICU mortality occurred in 71 of 305 patients (23.3%; median age 66.0 years). CNS-LAND scores were significantly higher in non-survivors than survivors (median 6.0 vs. 3.0, p < 0.001). ROC analysis showed the highest discriminative performance for GCS (AUC: 0.925), followed by Sequential Organ Failure Assessment (SOFA) (AUC: 0.906), Acute Physiology and Chronic Health Evaluation II (APACHE II) (AUC: 0.840), and CNS-LAND (AUC: 0.816; p = 0.467 vs. APACHE II). At a cutoff of ≥5, CNS-LAND yielded 83.1% sensitivity, 70.9% specificity, and a negative predictive value of 93.3%. Spearman analysis revealed near-perfect collinearity between National Institutes of Health Stroke Scale (NIHSS) and GCS (rs = -0.964, p < 0.001). After controlling for GCS, CNS-LAND lost statistical significance (partial rs = -0.051, p = 0.373), and logistic regression confirmed no independent prognostic contribution beyond GCS (LRT: χ² = 0.121, p = 0.728). Conclusions: CNS-LAND demonstrated mortality prediction performance comparable to APACHE II in critically ill ICH patients. However, this apparent performance was almost entirely attributable to the score's NIHSS component: NIHSS and GCS were near-perfectly collinear (rs = -0.964), and after adjustment for GCS, CNS-LAND was no longer significantly associated with mortality (partial rs = -0.051, p = 0.373; likelihood-ratio test p = 0.728; ΔAUC ≈ 0.000). CNS-LAND therefore provides essentially no independent prognostic information beyond GCS in this population, indicating that its discriminative performance is redundant with, rather than complementary to, standard neurological assessment. This redundancy is rooted in the parallel progression of neurological deficit and consciousness characteristic of severe ICH-a relationship fundamentally distinct from the acute ischemic stroke population in which CNS-LAND was originally derived.