A decision tree model for hematoma expansion prediction in women after spontaneous intracerebral hemorrhage.
Authors
Affiliations (3)
Affiliations (3)
- Medical Quality Control Department, The First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China.
- Department of Radiology, The First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China.
- Information Department, The First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China.
Abstract
Women are at a higher risk of poor outcomes following spontaneous intracerebral hemorrhage (ICH) compared to men, necessitating closer clinical monitoring. Preventing hematoma expansion (HE) represents a promising therapeutic target in the management of spontaneous ICH. This study aimed to develop a clinically practical decision tree model to predict HE in women. We retrospectively reviewed women with spontaneous ICH. All patients underwent initial and follow-up non-contrast CT scans within 6 h and 72 h after symptom onset, respectively. Univariate and multivariate logistic regression analyses were used to identify independent predictors of HE. A decision tree model was developed for HE prediction. A total of 417 patients were included, with 64 (15.3%) exhibiting HE on follow-up imaging. Multivariate analysis revealed that midline shift (odds ratio [OR], 1.18; 95% confidence interval [CI], 1.07-1.30; <i>p</i> = 0.001), time to initial CT scan (OR, 0.71; 95% CI, 0.57-0.88; <i>p</i> = 0.002), and presence of blend sign (OR, 2.81; 95% CI, 1.33-5.97; <i>p</i> = 0.007) were independently associated with HE. Our decision tree model achieved an AUC of 0.803 (95% CI, 0.736-0.856), a sensitivity of 81.1% and specificity of 67.1% in the training set, and 0.748 (95% CI, 0.581-0.880), 81.8 and 65.8% in the test set, respectively. It outperformed the HEP model. Although the BRAIN model had a higher AUC, our model achieved a higher sensitivity (81.8% vs. 72.7%), a key advantage for identifying patients needing timely intervention. We developed a simple, interpretable decision tree model to predict HE in women. This tool may support clinicians in identifying high-risk patients and guiding timely interventions.