Objectives: To systematically assess the discrimination and calibration of the Intracerebral Hemorrhage score for prediction of short-term mortality in intracerebral hemorrhage patients and to study its determinants using heterogeneity analysis.
Data sources: PubMed, ISI Web of Knowledge, Scopus, and CENTRAL from inception to September 15, 2018.
Study selection: Adult studies validating the Intracerebral Hemorrhage score for mortality prediction in nontraumatic intracerebral hemorrhage at 1 month/discharge or sooner.
Data extraction: Data were collected on the following aspects of study design: population studied, level of care, timing of outcome measurement, mean study year, and mean cohort Intracerebral Hemorrhage score. The summary measures of interest were discrimination as assessed by the C-statistic and calibration as assessed by the standardized mortality ratio (observed:expected mortality ratio). Random effect models were used to pool both measures. Heterogeneity was measured using the I statistic and explored using subgroup analysis and meta-regression.
Data synthesis: Fifty-five studies provided data on discrimination, and 35 studies provided data on calibration. Overall, the Intracerebral Hemorrhage score discriminated well (pooled C-statistic 0.84; 95% CI, 0.82-0.85) but overestimated mortality (pooled observed:expected mortality ratio = 0.87; 95% CI, 0.78-0.97), with high heterogeneity for both estimates (I 80% and 84%, respectively). Discrimination was affected by study mean Intracerebral Hemorrhage score (β = -0.05), and calibration was affected by disease severity, with the score overestimating mortality for patients with an Intracerebral Hemorrhage score greater than 3 (observed:expected mortality ratio = 0.84; 95% CI, 0.78-0.91). Mortality rates were reproducible across cohorts for patients with an Intracerebral Hemorrhage score 0-1 (I = 15%).
Conclusions: The Intracerebral Hemorrhage score is a valid clinical prediction rule for short-term mortality in intracerebral hemorrhage patients but discriminated mortality worse in more severe cohorts. It also overestimated mortality in the highest Intracerebral Hemorrhage score patients, with significant inconsistency between cohorts. These results suggest that mortality for these patients is dependent on factors not included in the score. Further studies are needed to determine these factors.