Novel Prediction Model for ICU Admission in Acute Intracerebral Hemorrhage

Accurate prediction of intensive care unit (ICU) admission in intracerebral hemorrhage (ICH) patients could improve triage, resource use, and outcomes, yet admission criteria remain unclear. Thus, we aimed to develop and externally validate a predictive model for ICU admission in patients with acute ICH.

The derivation cohort included 419 patients with ICH admitted within 2 weeks of onset to Peking Union Medical College Hospital (November 2014-October 2022). External validation used 642 prospectively enrolled patients with ICH from Tongji Hospital, Wuhan (September 2020-August 2023) with identical enrollment criteria. Using multivariable logistic regression, we developed a simplified 6-predictor model of intracerebral hemorrhage (HEM-6), incorporating tachycardia, leukocytosis, Glasgow Coma Scale ≤8, modified Rankin Scale ≥4, hemorrhage volume ≥25 mL, and combined vascular events. Model performance was assessed by discrimination and calibration.

Among 419 patients in the derivation cohort, 169 (40.3%) were female, and the median age was 58 years. Overall, 166 patients (39.6%) were admitted to the ICU. The HEM-6 score demonstrated strong discriminative power for ICU admission (AUC=0.84, 95% CI: 0.80-0.88) and was successfully validated in the validation (ICU admission rate=42.7% [274 of 642], AUC=0.88, 95% CI: 0.86-0.91). In external validation cohort, HEM-6 showed excellent calibration (Hosmer-Lemeshow test P=0.8783). The ICU admission rates were 6.4%, 28.4%, 63.4% and 94.7% for HEM-6 scores of 0-1, 2-3, 4-6, and ≥7, respectively.

The HEM-6 model provided a practical and reliable tool for predicting ICU admission in patients with acute ICH. With its robust discriminative power and successful external validation, HEM-6 has the potential to optimize ICU triage decisions, improve resource allocation, and ultimately enhance patient care in emergency and critical care settings.