Interobserver variability in the interpretation of computed tomography following aneurysmal subarachnoid hemorrhage

Abstract

Object Numerous abnormal findings may be evident on CT scans after aneurysmal subarachnoid hemorrhage (SAH). Here, the authors assess the interobserver variability in the radiological interpretation of the initial CT scan following SAH. Methods Two experienced reviewers, a neurosurgeon and a neuroradiologist, independently prospectively reviewed the initial CT scans of 413 patients enrolled in the CONSCIOUS-1 trial. Measured variables included SAH, intraventricular hemorrhage, intracerebral hemorrhage, subdural hematoma, chronic infarction, midline shift, and hydrocephalus. To assess interobserver variability, weighted kappa values and intraclass correlation coefficients (ICCs) were calculated and Bland-Altman analysis was performed. Results Moderate to substantial agreement was found for most of the CT scanning findings. There was fair to moderate interobserver agreement between reviewers when determining the extent of SAH based on a descriptive categorical classification (kappa 0.41; 95% CI 0.33–0.49), and better agreement when a semiquantitative scale was used (ICC 0.56; 95% CI 0.49–0.62). There was poor agreement between reviewers for the presence of hydrocephalus (kappa 0.34; 95% CI 0.20–0.48), but substantial to near perfect agreement on ventriculocranial ratio measurements (ICC 0.77; 95% CI 0.72–0.81). Conclusions The authors' findings suggest that there is considerable interobserver variability in the interpretation of CT scans after SAH. Quantitative measures may reduce interobserver variability in comparison with qualitative or categorical scales. Variability in interpretation of CT scans has implications for patient care and conduct of clinical trials. It may be beneficial to develop standardized assessments to ensure consistent evaluation of measured variables.