A Clinically Applicable Nomogram for Predicting In-Hospital Mortality in Intracerebral Hemorrhage: Development and External Validation

Objective Intracerebral hemorrhage (ICH) is a subtype of stroke associated with high mortality. However, effective clinical models for predicting in-hospital mortality in ICH patients are currently lacking. Therefore, this study aims to develop a predictive model for in-hospital mortality following ICH, which may facilitate early identification of patients at risk of poor outcomes. Materials and methods A retrospective analysis was conducted on 1,513 patients with ICH. A total of 1,026 patients from the MIMIC-IV database were included in the training and test sets, while 487 patients from the First Affiliated Hospital of Harbin Medical University comprised the external validation cohort. Least Absolute Shrinkage and Selection Operator (LASSO) regression and multivariate Logistic regression analyses were employed to identify optimal independent risk factors and to construct a logistic regression model. The model's discrimination, goodness-of-fit, and clinical utility were comprehensively evaluated using multiple indicators. Results Among patients from the MIMIC-IV database, 221 (21.5%) experienced in-hospital mortality, while 85 (17.5%) patients from the First Affiliated Hospital of Harbin Medical University died during hospitalization. Multivariate logistic regression identified age (OR 1.02, 95%CI 1.01-1.04), Glasgow Coma Scale (GCS) score (OR 0.76, 95%CI 0.71-0.80), minimum international normalized ratio (INR) (OR 5.00, 95%CI 1.35-19.79), length of hospital stay (OR 0.93, 95%CI 0.90-0.96), intraventricular hemorrhage (IVH) (OR 2.19, 95%CI 1.35-3.61), and maximum hematoma area (OR 1.05, 95%CI 1.04-1.07) as independent risk factors for in-hospital mortality in ICH patients. Based on these variables, a predictive model was developed, demonstrating excellent performance with the area under the curve (AUC) of 0.887in the training set, 0.913in the test set, and 0.786in the external validation cohort. The model exhibited good sensitivity and specificity, indicating strong clinical utility for early risk stratification in ICH patients. Conclusion The nomogram developed in this study more accurately predicts in-hospital mortality following ICH and has the potential to improve patient prognosis. Clinical trial number Not applicable