Abstract 131: Interrater Reliability of the ABCD 2 Score

Abstract

Background: The ABCD 2 score is increasingly used for risk stratification of TIA patients, but little is known about the interrater reliability of the score. Methods: Patients presenting to the emergency department with suspected TIA (symptom duration <24 hours) were prospectively evaluated. Only those asymptomatic at the time of enrollment were eligible. Patients verbally consented to questioning by multiple raters. We compared ABCD 2 scores determined by raters of different medical specialties to the “gold standard” score of neurovascular attending physicians. Raters were given a worksheet with basic ABCD 2 scoring instructions to complete for each subject; all scores were based on history obtained directly from the patient by each rater. Estimated component, total scores, and ABCD 2 risk category (0-3, 4-5, 6-7) were compared both between raters and with the neurovascular attending score. Reliability was assessed using unweighted kappa statistics. Results: A total of 46 patients were assessed, with each scored by a mean of 3.3 raters. In addition to the neurovascular attending, scores were generated by neurology junior (n=37) and senior residents (n=9), and internal medicine (n=36) and emergency medicine residents (n=36). Based on neurovascular attending scores, 35% of patients were categorized as low-risk (ABCD 2 score 0-3), 50% as moderate risk (4-5), and 15% as high risk (6-7). Interrater reliability was limited for ABCD 2 total score (κ=0.28) and category (κ=0.39). Interrater reliability of the component scores was near perfect for age (κ=1) and diabetes (κ=0.94) and substantial for blood pressure (κ=0.66), but only moderate for clinical features (κ=0.54) and duration (κ=0.59). Agreement between ABCD 2 risk category scored by the neurovascular attending and that determined by other raters was as follows: 73% (95%CI:58-88%) for neurology junior residents, 89% (CI:63-100%) for neurology senior residents, 58% (CI:41-75%) for internal medicine residents, and 70% (CI:48-92%) for emergency medicine residents. Conclusions: The interrater reliability of the ABCD 2 score is poor. Clinical symptoms (C) and duration (D) displayed the greatest variability among ABCD 2 components.