Hematoma expansion in intracerebral hemorrhage: Development of a preoperative risk prediction model.

Intracerebral hemorrhage (ICH) is one of the most lethal and disabling forms of acute cerebrovascular disease. Hematoma expansion is a critical predictor of early deterioration and poor prognosis in ICH patients, making accurate preoperative risk prediction essential for clinical decision-making. This study aims to develop a preoperative risk prediction model for hematoma expansion in ICH patients to provide robust decision support for clinicians. A total of 156 patients with spontaneous ICH admitted to our hospital between January 2022 and July 2024 were included. The cohort was randomly divided into a training set (104 cases, 66.67%) and a validation set (52 cases, 33.33%). Baseline data were recorded for all patients, including age, sex, body mass index, admission systolic and diastolic blood pressure, Glasgow Coma Scale score at admission, initial hematoma volume, hematoma density, hematoma morphology, mean arterial pressure at onset, history of diabetes, use of anticoagulants, use of antiplatelet agents, prothrombin time, ICH grading score, computed tomography angiography spot sign score, Glasgow Coma Scale score, and National Institutes of Health Stroke Scale score. Univariate logistic regression analysis was used to preliminarily identify potential risk factors for hematoma expansion. Significant variables were further analyzed using multivariate logistic regression to determine independent predictors. A dynamic risk prediction model based on DynNom was constructed. The model was validated internally using the Bootstrap method to evaluate its calibration and discrimination. Diagnostic thresholds and predictive performance were assessed using receiver operating characteristic curves. The final model included initial hematoma volume, hematoma density, prothrombin time, and computed tomography angiography spot sign score as independent predictors. The nomogram model demonstrated good concordance between predicted and observed outcomes, with a concordance index of 0.798 (95% confidence interval: 0.771-0.825) and an area under the curve of 0.798 in the internal validation (split-sample) cohort. These metrics have been updated to match those reported in the main results section, ensuring consistency across the manuscript and accurately reflecting model performance in the selected dataset. This preoperative risk prediction model showed strong performance for early detection of hematoma expansion in ICH patients, with great potential for clinical application.