Computed tomography interobserver agreement in the assessment of aneurysmal subarachnoid hemorrhage and predictors for clinical outcome.

BACKGROUND

The severity of aneurysmal subarachnoid hemorrhage (SAH) is often assessed by the clinical state of the patient on presentation, but radiological evaluation of the extent of hemorrhage has rarely been examined in the literature. Several CT scan based grading systems exist yet only a few studies have investigated interobserver agreement. We evaluated five radiological grading systems and assessed their clinical value for early prognostication.

METHODOLOGY

This was a retrospective study of patients diagnosed with aneurysmal SAH with a CT scan performed within 72 hours of symptom onset. Four independent observers, blinded to patient outcome, evaluated each scan using the five grading systems. A separate assessor determined 6 month outcome from clinical records. The primary outcome was interobserver agreement for each grading system using the Fleiss κ statistic. The secondary endpoint was the 6 month modified Rankin Scale score, with poor outcome defined as a score of 4-6.

RESULTS

165 patients with a mean age of 59 years were assessed. Interobserver agreement for the Fisher, modified Fisher, Claassen, Barrow Neurological Institute, and Hijdra grading systems were as follows: k=0.53 (moderate), k=0.42 (moderate), k=0.38 (mild), k=0.20 (poor), and k=0.66 (good), respectively. The only independent clinical risk factor for poor outcome was a World Federation of Neurological Surgeons (WFNS) grade of 4 or 5 (adjusted OR 6.55; p<0.05). After adjusting for confounders, Fisher grade 4 (adjusted OR 17.84), modified Fisher grade 4 (adjusted OR 5.65), and Hijdra grade 3 (adjusted OR 3.34) were associated with poor outcome. Receiver operator characteristic analysis revealed that the Hijdra grading system (area under the curve=0.76) was more predictive of outcome compared with the Fisher and modified Fisher systems. A Hijdra cut-off score of 22 was associated with poor outcome (adjusted OR 5.92).

CONCLUSIONS

The Hijdra grading system had the best interobserver agreement and was a better independent early predictor for 6 month clinical outcome than the other systems. A Hijdra score ≥22 was associated with poor outcome.