Risk Score for Early Prediction of In-Hospital Mortality After Aneurysmal Subarachnoid Hemorrhage: Pooled Analysis With Score Construction and Validation

Abstract

Objective

Aneurysmal subarachnoid hemorrhage (aSAH) has a high complications burden, with in-hospital mortality as the most devastating outcome. We aimed to develop and validate a risk score for early prediction of in-hospital mortality after aSAH.

Methods

Data from 2 university hospitals were pooled (n = 1070), with cohorts for score construction (n = 886) and external validation (n = 184). Several parameters assessable at admission were collected. Independent predictors of in-hospital mortality were used as mortality score components. Diagnostic accuracy of the novel score was compared to the Hunt and Hess Age, Intraventricular Hemorrhage, Rebleed and World Federation of Neurosurgical Societies (WFNS) scores.

Results

Overall rate of in-hospital mortality was 19% and 14.7% in construction and validation cohorts, respectively. The novel risk score (aSAH mortality score: 0–12 points) included patients' age (≤55 years: 0 points, 56–70 years: 1 point, >70 years: 2 points), aneurysm rebleeding (2 points), WFNS grade (grade I–II: 0 points, grade III–IV: 2 points, grade V: 5 points), and Hijdra sum score (≤10: 0 points, 11–20: 1 point, 21–30: 2 points, >30: 3 points). In-hospital mortality rates ranged 0.6% (0 points) to 100% (12 points) in the construction cohort, and 0% (0 points) to 60% (10 points) in the validation cohort. In the receiver operating characteristic analysis, the aneurysmal subarachnoid hemorrhage mortality score (area under the curve [AUC]: 0.829 and 0.824 in the construction and validation cohorts, respectively) was superior to the Hunt and Hess Age, Intraventricular Hemorrhage, Rebleed (AUC: 0.811 and 0.813) and WFNS scores (AUC: 0.768 and 0.795).

Conclusions

Risk of in-hospital mortality after aSAH can be predicted with high accuracy using baseline characteristics. The novel risk score showed best diagnostic performance in the construction and validation cohorts and can aid in early prognostication and treatment decisions.