Seizure risk prediction models after intracerebral hemorrhage: a systematic review and meta-analysis

Current models for predicting seizure risk after intracerebral hemorrhage (ICH) frequently demonstrate suboptimal accuracy. Despite the increasing number of these predictive tools, their utility in clinical practice and research remains poorly defined. A search of eight databases from their inception through September 18, 2025, was undertaken to identify studies of predictive models related to post-ICH seizures. Risk of bias and applicability were evaluated using the Prediction model Risk of Bias Assessment Tool (PROBAST). From 2,578 retrieved studies, eight prediction models from nine studies were included. The observed incidence of post-hemorrhagic stroke seizures (PHSS) ranged from 3.07% to 12.75%. All studies were determined to have a high risk of bias, primarily due to poor reporting of the analysis domain. Meta-analysis indicated that a hematoma volume ≥ 10 mL, early seizures, cortical involvement, and surgical intervention served as independent predictors of PHSS. The pooled area under the curve for the eight models was 0.81 (95% CI: 0.76-0.86), showing that the models have moderate to good discriminative capacity. Future research must prioritize rigorous model validation, adhering to PROBAST standards to ensure methodological quality.