As common prognostication models in intracerebral hemorrhage (ICH) are developed variably including patients with early (<24 hours) care limitations (ECL), we investigated its interaction with prognostication in maximally treated patients and sought to provide a new unbiased severity assessment tool.Methods:
This observational cohort study analyzed consecutive ICH patients (n = 583) from a prospective registry over 5 years. We characterized the influence of ECL on overall outcome by propensity score matching and on conventional prognostication using receiver operating characteristic analyses. We established the max-ICH score based on independent predictors of 12-month functional outcome in maximally treated patients and compared it to existing models.Results:
Prevalence of ECL was 19.2% (n = 112/583) and all of these patients died. Yet propensity score matching displayed that 50.7% (n = 35/69) theoretically could have survived, with 18.8% (n = 13/69) possibly reaching favorable outcome (modified Rankin Scale score 0–3). Conventional prognostication seemed to be confounded by ECL, documented by a decreased predictive validity (area under the curve [AUC] 0.67, confidence interval [CI] 0.61–0.73 vs AUC 0.80, CI 0.76–0.83; p < 0.01), overestimating poor outcome (mortality by 44.8%, unfavorable outcome by 10.1%) in maximally treated patients. In these patients, the novel max-ICH score (0–10) integrates strength-adjusted predictors, i.e., NIH Stroke Scale score, age, intraventricular hemorrhage, anticoagulation, and ICH volume (lobar and nonlobar), demonstrating improved predictive accuracy for functional outcome (12 months: AUC 0.81, CI 0.77–0.85; p < 0.01). The max-ICH score may more accurately delineate potentials of aggressive care, showing favorable outcome in 45.4% (n = 214/471) and a long-term mortality rate of only 30.1% (n = 142/471).Conclusions:
Care limitations significantly influenced the validity of common prognostication models resulting in overestimation of poor outcome. The max-ICH score demonstrated increased predictive validity with minimized confounding by care limitations, making it a useful tool for severity assessment in ICH patients.