Frequent inaccuracies in ABCD² scoring in non-stroke specialists' referrals to a daily Rapid Access Stroke Prevention service

The 'accuracy' of age, blood pressure, clinical features, duration and diabetes (ABCD²) scoring by non-stroke specialists referring patients to a daily Rapid Access Stroke Prevention (RASP) service is unclear, as is the accuracy of ABCD² scoring by trainee residents. In this prospective study, referrals were classified as 'confirmed TIAs' if the stroke specialist confirmed a clinical diagnosis of possible, probable or definite TIA, and 'non-TIAs' if patients had a TIA mimic or completed stroke. ABCD² scores from referring physicians were compared with scores by experienced stroke specialists and neurology/geriatric medicine residents at a daily RASP clinic; inter-observer agreement was examined. Data from 101 referrals were analysed (mean age = 60.0 years, 58% male). The median interval between referral and clinic assessment was 1 day. Of 101 referrals, 52 (52%) were 'non-TIAs': 45 (86%) of 52 were 'TIA mimics' and 7 (14%) of 52 were completed strokes. There was only 'fair' agreement in total ABCD² scoring between referring physicians and stroke specialists (κ = 0.37). Agreement was 'excellent' between residents and stroke specialists (κ = 0.91). Twenty of 29 patients scored as 'moderate to high risk' (score 4-6) by stroke specialists were scored 'low risk' (score 0-3) by referring physicians. ABCD² scoring by referring doctors is frequently inaccurate, with a tendency to underestimate stroke risk. These findings emphasise the importance of urgent specialist assessment of suspected TIA patients, and that ABCD² scores by non-stroke specialists cannot be relied upon in isolation to risk-stratify patients. Inter-observer agreement in ABCD² scoring was 'excellent' between residents and stroke specialists, indicating short-term training may improve accuracy.