Brain–heart interaction-driven prognostic model in intracerebral hemorrhage based on cardiac features

Brain–heart interactions modulate both cardiac and cerebral functions through neural, humoral, and hemodynamic pathways. However, they have rarely been systematically evaluated in patients with intracerebral hemorrhage (ICH). This study aimed to develop a cardiac injury-based model to assess brain–heart interactions after ICH. This retrospective study enrolled 574 patients with spontaneous ICH who underwent baseline cardiac assessments, including medical history, electrocardiographic abnormalities, and cardiac injury biomarkers. The primary outcome was 90-day mortality. A two-stage modeling framework was used. First, the Heart Enzymes, Arrhythmia, and Risk Traits for Intracerebral Hemorrhage (ICH–HEART) score was developed using cardiac biomarkers and electrocardiographic variables. Second, the ICH–HEART score was incorporated as a composite cardiac predictor, together with clinical and imaging variables, to construct the Heart–Brain Integration for ICH mortality prediction (HBI–ICH) model. Statistical significance was defined as a two-sided p < 0.05. The 90-day mortality rate was 23.2%. Nonsurvivors exhibited elevated cardiac biomarkers and more frequent ECG abnormalities ( p < 0.05). The ICH–HEART model achieved area under the curve (AUC) values of 0.732 and 0.767 in the training and validation cohorts, respectively. After incorporation of the ICH–HEART score, the HBI–ICH model achieved AUC values of 0.843 in the training cohort and 0.836 in the validation cohort. However, in the internal validation cohort, the HBI–ICH model did not demonstrate statistically significant superiority over the original ICH score ( p = 0.057) or the modified ICH score ( p = 0.070). This study developed and internally validated the ICH–HEART and HBI–ICH models, which showed promising risk stratification capabilities for 90-day mortality, an outcome partially shaped by treatment-limitation decisions in approximately one-third of deaths. The extent to which the observed prognostic associations reflect biological severity versus clinician-driven care decisions requires further evaluation in cohorts with standardized treatment protocols.