Comparison of all 19 published prognostic scores for intracerebral hemorrhage

Background and aims: We evaluated the accuracy of 19 published prognostic scores to find the best tool for predicting mortality after intracerebral hemorrhage (ICH). Methods: A retrospective single-center analysis of consecutive patients with ICH (n = 1013). After excluding patients with missing data (n = 131), we analyzed 882 patients for 3-month (primary outcome), in-hospital, and 12-month mortality. We analyzed the strength of the individual score components and calculated the c-statistics, Youden index, sensitivity, specificity, negative and positive predictive value (NPV and PPV) for the scores. Finally, we included every score component in a multivariable model to analyze the maximum predictive value of the data elements combined. Results: Observed in-hospital mortality was 23.6%, 3-month mortality was 31.0%, and 12-month mortality was 35.3%. For in-hospital mortality, the National Institutes of Health Stroke Scale (NIHSS) performed equally good as the best score for the other outcomes, the ICH Functional Outcome Score (ICH-FOS). The c-statistics of the scores varied from 0.6293 (95% CI 0.587-0.672) to 0.8802 (0.855-0.906). With all variables from all the scores in a multivariable regression model, the c-statistics did not improve, being 0.89 (0.867-0.913). Using the Youden index cutoff for the ICH-FOS score, the sensitivity (73%), specificity (90%), PPV (76%), and NPV (88%) for the primary outcome were good. Conclusions: A plethora of scores exists to help clinicians estimate the prognosis of an acute ICH patient. The NIHSS can be used to quantify the risk of in-hospital death while the ICH-FOS performed best for the other outcomes. (C) 2017 Elsevier B.V. All rights reserved.