Inter-Rater Reliability of the CASCADE Criteria

Abstract

Background and Purpose— There are limited data about the reliability of subtype classification in childhood arterial ischemic stroke, an issue that prompted the IPSS (International Pediatric Stroke Study) to develop the CASCADE criteria (Childhood AIS Standardized Classification and Diagnostic Evaluation). Our purpose was to determine the CASCADE criteria’s reliability in a population of children with stroke. Methods— Eight raters from the IPSS reviewed neuroimaging and clinical records of 64 cases (16 cases each) randomly selected from a prospectively collected cohort of 113 children with arterial ischemic stroke and classified them using the CASCADE criteria. Clinical data abstracted included history of present illness, risk factors, and acute imaging. Agreement among raters was measured by unweighted κ statistic. Results— The CASCADE criteria demonstrated a moderate inter-rater reliability, with an overall κ statistic of 0.53 (95% confidence interval [CI]=0.39–0.67). Cardioembolic and bilateral cerebral arteriopathy subtypes had much higher agreement (κ=0.84; 95% CI=0.70–0.99; and κ=0.90; 95% CI=0.71–1.00, respectively) than cases of aortic/cervical arteriopathy (κ=0.36; 95% CI=0.01–0.71), unilateral focal cerebral arteriopathy of childhood (FCA; κ=0.49; 95% CI=0.23–0.76), and small vessel arteriopathy of childhood (κ=−0.012; 95% CI=−0.04 to 0.01). Conclusions— The CASCADE criteria have moderate reliability when used by trained and experienced raters, which suggests that it can be used for classification in multicenter pediatric stroke studies. However, the moderate reliability of the arteriopathic subtypes suggests that further refinement is needed for defining subtypes. Such revisions may reduce the variability in the literature describing risk factors, recurrence, and outcomes associated with childhood arteriopathy.