A clinically interpretable nomogram for predicting discharge outcomes in cerebral amyloid angiopathy-related lobar intracerebral hemorrhage: development and internal validation.

BACKGROUND

Prognostic models for cerebral amyloid angiopathy (CAA)-related lobar intracerebral hemorrhage (ICH) remain limited. We aimed to develop and internally validate an interpretable nomogram for predicting unfavorable functional outcome at discharge.

METHODS

We retrospectively included 124 patients aged ≥50 years with lobar ICH who were diagnosed with CAA according to the Boston criteria version 2.0. Discharge outcome was classified as favorable (modified Rankin Scale score 0-2) or unfavorable (3-6). The model was developed using Boruta feature screening, AIC-guided logistic regression, and prespecified CAA-specific markers. Internal validation was performed using full-pipeline bootstrap resampling with 500 resamples.

RESULTS

Seventy patients had unfavorable discharge outcomes. The final model included admission NIHSS score, hematoma volume, glucose level, prior cognitive decline, and cortical superficial siderosis grade. The model showed good apparent discrimination, with an AUC of 0.905. Full-pipeline bootstrap validation yielded an optimism-corrected AUC of 0.852. Calibration was acceptable, with an apparent Brier score of 0.127 and an optimism-corrected Brier score of 0.173. Decision curve analysis showed favorable clinical utility.

CONCLUSIONS

We developed and internally validated an interpretable nomogram for predicting unfavorable discharge outcome after CAA-related lobar ICH. The model combines acute neurological severity, hematoma burden, glycemic status, and CAA-specific markers, and may support early individualized risk assessment in this population.