ABCD2 score guidelines: what NICE and the AHA say now

What NICE NG128 says about the ABCD2 score

NICE guideline NG128, Stroke and transient ischaemic attack in over 16s, was published in May 2019 and last updated in April 2022. Its recommendation 1.1.6 is direct: "Do not use scoring systems, such as ABCD2, to assess risk of subsequent stroke or to inform urgency of referral for people who have had a suspected or confirmed TIA." Recommendation 1.1.5 replaces score-based triage with one rule for everyone: "Refer immediately people who have had a suspected TIA for specialist assessment and investigation, to be seen within 24 hours of onset of symptoms." The committee explained that ABCD2 and ABCD3 used in isolation are poor at telling low from high stroke risk. Adding brain and carotid imaging, as in ABCD2-I and ABCD3-I, modestly improves discrimination, but that imaging is not available in general practice or to paramedics, where the tools would mostly be used. Arranging specialist assessment less urgently on the basis of such a tool could cause harm, so every suspected TIA is treated as potentially high risk.

NICE gave one more reason: no reliable diagnostic test for TIA exists, and the risk tools are not diagnostic tests. In practice, 30% to 50% of suspected TIAs turn out to have another diagnosis, so a specialist needs to confirm or refute the diagnosis quickly.

The evidence that changed the ABCD2 guidance

The ABCD2 guidance changed because studies after 2007 found the score separated risk less well than its derivation suggested. In 2015, Wardlaw and colleagues pooled 29 studies of 13,766 patients with TIA in a meta-analysis in Neurology, at a time when ABCD2 was widely recommended for triage. A score of 4 or more was sensitive, at 86.7%, but not specific, at 35.4%, for recurrent stroke within 7 days. One in five patients who scored under 4 (20%) had carotid stenosis over 50% or atrial fibrillation, findings that need urgent action. The score did not sort out mimics either: 35% to 41% of TIA mimics and 66% of true TIAs scored 4 or more. The authors concluded that ABCD2 does not reliably discriminate low from high risk, identify those patients, or streamline clinic workload, and that services need capacity for prompt specialist assessment of all suspected TIA patients.

A prospective study in eight Canadian emergency departments (Perry and colleagues, CMAJ, 2011) reached the same view. Among 2,056 patients, 1.8% had a stroke within 7 days. A score above 5 caught only 31.6% of them, and a score above 2 caught 94.7% but had a specificity of 12.5%. The area under the curve was 0.56 when calculated by the enrolling physician and 0.65 by the coordinating centre, which the authors called inaccurate at any cut-point.

What the American Heart Association says about ABCD2

The American Heart Association and American Stroke Association (AHA/ASA) have addressed the ABCD2 score in two scientific statements rather than ruling it out. The 2009 statement by Easton and colleagues, which set the tissue-based definition of TIA, found it reasonable to admit patients to hospital if they present within 72 hours of the event and have an ABCD2 score of 3 or more, or if their evaluation cannot be completed rapidly as an outpatient. It also advised neuroimaging within 24 hours of symptom onset, preferably MRI with diffusion sequences, and imaging of the neck vessels. Perry's Canadian study later found that this cut-off was sensitive but nonspecific. The 2023 AHA scientific statement on TIA in the emergency department (Amin and colleagues) describes identifying high-risk patients through comprehensive protocols that combine early imaging of the brain and blood vessels, thoughtful use of risk stratification scales and other tests, with the goal of deciding who can safely go home and who should be admitted.

So the two bodies differ in emphasis. NICE removes the score from triage altogether and sends everyone to a specialist within 24 hours. The AHA/ASA statements keep risk scores as one input, alongside imaging, in deciding on admission from the emergency department.

ABCD2 vs ABCD3-I

ABCD2 vs ABCD3-I comes down to whether imaging results are available. ABCD3-I, published by Merwick and colleagues in Lancet Neurology in 2010, keeps the five ABCD2 items and adds three more worth 2 points each: dual TIA (an earlier TIA within 7 days of the index event), stenosis of at least 50% on carotid imaging, and an abnormal diffusion-weighted MRI. Its range is 0 to 13. In the derivation sample of 2,654 patients assessed in secondary care, its c statistic for stroke at 2 days was 0.90. A pooled analysis of 2,176 specialist-confirmed TIA patients from 16 cohorts in Asia, Europe and the USA (Kelly and colleagues, 2016) found 2-day c statistics of 0.64 for ABCD2, 0.74 for ABCD2-I and 0.84 for ABCD3-I. The authors suggested considering ABCD3-I-guided care with immediate specialist assessment, urgent MRI and vascular imaging. NICE accepted that imaging modestly improves discrimination, but not that it justifies score-based triage.

How to read an ABCD2 score today

An ABCD2 score today is best read as a description of risk features, not as a decision about urgency. Under NICE NG128, the score should not change when a person with a suspected TIA is seen: everyone goes to a specialist within 24 hours of symptom onset. Where a local pathway or an emergency department still records ABCD2, as the AHA/ASA statements allow, two points from the evidence help. A low score is not reassurance: the low group in Johnston's 2007 paper still had a 1.0% stroke risk at 2 days, and in Wardlaw's meta-analysis about one in five patients scoring under 4 had carotid stenosis over 50% or atrial fibrillation. A high score does not confirm a TIA, since many mimics also score 4 or more. Take a 55-year-old with a blood pressure of 128/78 mmHg, 5 minutes of tingling in one hand and no diabetes: the ABCD2 score is 0, yet NICE still asks for specialist assessment within 24 hours.

Enter the features in the ABCD2 score calculator to see the score, its risk group and the 2-day stroke risk from Johnston 2007, shown with the NICE caveat.