Abstract WP434: The SAH score: An Outcome prediction model

Abstract

Objective: To create a reliable model for predicting mortality following aneurysmal subarachnoid hemorrhage (aSAH) based on admission variables. Background: Hunt & Hess grade is commonly used as a grading scale to predict mortality following aSAH. The scale relies only on clinical presentation and does not incorporate other admission factors making it suboptimal for outcome prediction. Methods: Prospectively collected data of aSAH patients admitted to our institution between 1991-2009 were reviewed. We analyzed factors that impacted in-hospital SAH mortality following multiple logistic regression analysis. Scores were ‘weighted’ based on relative risk of mortality following stratification of each of these variables. Hunt & Hess grade was subdivided into grades I/II, III, IV and V; age was split into 4 subgroups: 18-49, 50-69, 70-79 and >80. Medical co-morbidities were subdivided into none, 1 or >/=2 based on co-morbidities derived either from Charlson index or other factors (hypertension, cocaine) historically known to impact SAH outcomes, only if they were associated with increased mortality on univariate analysis. Results: 1134 patients were included; all-cause SAH hospital mortality was 18.3%. Hunt & Hess Grade, age and medical co-morbidities significantly impacted mortality following multivariate analysis (P< 0.05). Association with mortality based on Hunt & Hess Grade was 7%(I/II; score 0), 16%(III; score 1), 31%(IV; score 2) and 65%(V; score 4). Mortality based on age was 13%(18-49; score 0), 18%(50-69; score 1), 34% (70-79; score 2) and 46% (>80; score 3). Relationship of co-morbidities and mortality was 9%(none; score 0), 17%(one; score 1) and 32%(two/more; score 2). Summated Scores ranged from 0-9 with progressively increasing mortality at higher scores (0=1%/ 1=4%/ 2=9%/ 3=13%/ 4=22%/ 5=52%/ 6=77%/ 7=88%/ 8=100%/ 9=100%). PPV for scores in the range 7-9 was 90%; 6-9 was 83%. NPV for range 0-3 was 93% and 0-4 was 91%. The area under the curve (AUC) was 0.825 (good accuracy), which was superior to Hunt & Hess Grade (AUC 0.775, fair accuracy). Conclusions: The SAH score is a more accurate prediction model than the Hunt & Hess grade in estimating likelihood of hospital mortality following SAH.