Interobserver Agreement in the Trial of Org 10172 in Acute Stroke Treatment Classification of Stroke Based on Retrospective Medical Record Review

Background: The reliability of the Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification system of stroke in siblings of stroke-affected probands has not been tested. Similarly, the reliability of using clinical medical records to classify ischemic stroke has not been assessed. The purpose of this study was to establish the interrater reliability of sibling stroke subtyping by applying the TOAST criteria to retrospectively obtained medical records. Methods: Thirty medical records were randomly sampled from among the records of all siblings previously classified as stroke affected by the Siblings with Ischemic Stroke Study (SWISS) Stroke Verification Committee. Blinded medical records for these individuals were sent to 6 physician reviewers who independently classified TOAST stroke subtype on the basis of record review. Results: Using the κ statistic to assess interrater reliability, the overall reliability (SE) for assigning a TOAST subtype of stroke was 0.54 (0.03). Pair-wise comparisons between the original SWISS Stroke Verification Committee diagnoses and the diagnoses made by other reviewers exhibited moderate reliability (κ range 0.41-0.56). The κ statistics for common stroke subtypes were large vessel, 0.80 (0.06); cardioembolic, 0.80 (0.06); small vessel, 0.53 (0.06); and unknown cause, 0.40 (0.06). Conclusion We conclude that TOAST subtyping had moderate interrater reliability. Large-artery and cardioembolic subtype diagnoses seemed most reliable.