ABCD2 Score Calculator
ABCD2 Score Calculator
Score the five ABCD2 items for early stroke risk after a transient ischaemic attack — and read what happened next. Later validation found it discriminates poorly, and NICE now recommends against using it to stratify risk or set referral urgency.
ABCD2 score
Five items, 0 to 7Age 71, blood pressure 164/92 mmHg, 25 minutes of right arm and leg weakness, no diabetes
Scoring
Range 0 to 7 · 2-day stroke risk 1.0% (0–3), 4.1% (4–5), 8.1% (6–7)
- the derivation
- Johnston and colleagues combined two earlier scores on the 1,916 patients those had been derived in, then validated the result in four independent cohorts totalling 2,893 patients from California and the United Kingdom. Validation c statistics ranged from 0.62 to 0.83
- what the validation found
- Wardlaw and colleagues pooled 29 studies and 13,766 patients. At 4 or more the score was sensitive (86.7 per cent, 95% CI 81.4–90.7) but not specific (35.4 per cent, 33.3–37.6) for recurrent stroke within 7 days, and they concluded it “does not reliably discriminate those at low and high risk of early recurrent stroke, identify patients with carotid stenosis or AF needing urgent intervention, or streamline clinic workload”. Perry and colleagues, prospectively in 2,056 patients, found sensitivity of 31.6 per cent at a score above 5 and 94.7 per cent with specificity 12.5 at a score above 2, concluding the score is “inaccurate, at any cutpoint”
- what replaced it
- specialist assessment of everybody. NICE NG128 recommendation 1.1.7 (2019): “Do not use scoring systems, such as ABCD2, to assess risk of subsequent stroke or to inform urgency of referral.” Wardlaw’s reasoning was economic as well as clinical: triaging low-scoring patients to slower investigation both prevented fewer strokes and cost more
Worked example
Age 71, blood pressure 164/92 mmHg, 25 minutes of right arm and leg weakness, no diabetes
Age 71 = 1 · systolic 164 = 1 · unilateral weakness = 2 · 25 minutes = 1 · no diabetes = 0
1 + 1 + 2 + 1 + 0 = 5 points, in the 4-or-5 stratum with a 2-day stroke risk of 4.1 per cent in the derivation and validation cohorts
Now read the same patient through the validation. Take 1,000 similar patients at Perry’s measured 1.8 per cent 7-day stroke rate: 18 will stroke, of whom about 16 score 4 or more — and about 635 of the 982 who do not stroke also score 4 or more. The positive predictive value at the cut-off is under 3 per cent
Lengthen the history to 70 minutes and the score becomes 6, the top stratum. Shorten it to 8 minutes and it becomes 4. The same patient, the same arteries, two strata apart, on a recollection of how long an arm was weak
The five items
| Item | 0 points | 1 point | 2 points |
|---|---|---|---|
| Age | Under 60 years | 60 years or over | — |
| Blood pressure | Systolic under 140 and diastolic under 90 mmHg | Systolic 140 or over, or diastolic 90 or over | — |
| Clinical features | Neither of the below | Speech disturbance without weakness | Unilateral weakness |
| Duration | Under 10 minutes | 10 to 59 minutes | 60 minutes or more |
| Diabetes | Absent | Present | — |
What the derivation said, and what the validation found
| Study | Population | Finding |
|---|---|---|
| Johnston 2007 | 1,916 derivation, 2,893 validation, specialist-diagnosed TIA | 2-day stroke risk 1.0% (score 0–3), 4.1% (4–5), 8.1% (6–7); validation c statistics 0.62–0.83 |
| Perry 2011, prospective | 2,056 emergency-department patients with TIA | 7-day stroke 1.8%, 90-day 3.2%. Score above 5: sensitivity 31.6%, specificity 86.9%. Score above 2: sensitivity 94.7%, specificity 12.5%. “Inaccurate, at any cutpoint” |
| Wardlaw 2015, meta-analysis | 29 studies, 13,766 patients with TIA | Score 4 or more for 7-day stroke: sensitivity 86.7% (81.4–90.7), specificity 35.4% (33.3–37.6). 20% of patients scoring under 4 had carotid stenosis above 50% or atrial fibrillation |
| NICE NG128, 2019 | Suspected or confirmed TIA, over 16s | Recommendation 1.1.7: “Do not use scoring systems, such as ABCD2, to assess risk of subsequent stroke or to inform urgency of referral” |
The score its own validation took apart
ABCD2 was assembled in 2007 from two earlier transient ischaemic attack scores, on the 1,916 patients those had been derived in, and validated in four further cohorts totalling 2,893. Five items: age 60 or over, blood pressure of 140/90 or more, unilateral weakness or speech disturbance, symptom duration, and diabetes. The reported 2-day stroke risks were 1.0 per cent for 0 to 3, 4.1 per cent for 4 or 5 and 8.1 per cent for 6 or 7, with validation c statistics between 0.62 and 0.83. It spread quickly because it is memorable, needs no investigations, and appeared to answer the practical question of who has to be seen tonight.
What followed is the content of this page rather than a caveat at the end of it. Perry and colleagues ran the first large prospective validation in an emergency department: 2,056 patients, a 7-day stroke rate of 1.8 per cent, and at the original high-risk cut-point a sensitivity of 31.6 per cent — it missed around two thirds of the strokes. Pushed down to the cut-point of above 2 that the American Heart Association had recommended, sensitivity rose to 94.7 per cent and specificity collapsed to 12.5, which is not a triage tool but a way of writing “see everybody” in five letters. Their conclusion was that the score is inaccurate at any cut-point.
Wardlaw and colleagues then pooled 29 studies and 13,766 patients. At the usual threshold of 4 or more, sensitivity for recurrent stroke within 7 days was 86.7 per cent and specificity 35.4 — so about two thirds of the patients who were never going to have a stroke screened positive, and the positive predictive value at the cut-off is under 3 per cent. Worse for its intended purpose, 20 per cent of the patients scoring under 4 had carotid stenosis above 50 per cent or atrial fibrillation: the score is blind to precisely the two findings that change what happens next. They also found that 35 to 41 per cent of TIA mimics score 4 or more, because nothing in the five items is specific to cerebral ischaemia. NICE drew the obvious conclusion in 2019, in recommendation 1.1.7 of NG128: “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.” The replacement is prompt specialist assessment and imaging of everyone. A grade is not a diagnosis and a cohort risk is not this patient’s probability: a stratum in which 72 per cent died tells you about that cohort, not which 72 per cent. Every threshold here comes from a named cohort, and cohorts differ in case mix, era and treatment; where your unit’s protocol differs, it takes precedence.
Frequently asked questions
What are the ABCD2 score components?
Age 60 or over (1 point), blood pressure of 140/90 mmHg or more at first assessment (1), clinical features — unilateral weakness 2 points, speech disturbance without weakness 1 — duration (60 minutes or more 2, 10 to 59 minutes 1), and diabetes (1). The total runs 0 to 7.
What stroke risk does each ABCD2 stratum carry?
In the pooled derivation and validation cohorts the 2-day stroke risk was 1.0 per cent for 0 to 3 points, 4.1 per cent for 4 or 5, and 8.1 per cent for 6 or 7. Those are cohort figures from specialist-diagnosed TIA, and later prospective work found substantially poorer discrimination.
Why do guidelines no longer recommend the ABCD2 score?
Because its validation showed it does not do the job it was adopted for: a meta-analysis of 29 studies and 13,766 patients found sensitivity 86.7 per cent but specificity 35.4 at a cut-off of 4, and that 20 per cent of patients scoring under 4 had carotid stenosis above 50 per cent or atrial fibrillation. NICE NG128 recommendation 1.1.7 advises against using it.
Can a low ABCD2 score rule out a high-risk TIA?
No. One in five patients scoring under 4 in the meta-analysis had carotid stenosis above 50 per cent or atrial fibrillation, and the clinical item scores nothing for isolated sensory symptoms, isolated visual loss, vertigo or ataxia, so a posterior circulation event cannot score above 5 however high the risk.
Which cut-off should be used, 4 or 6?
Neither performs well. At a score above 5 the prospective sensitivity for 7-day stroke was 31.6 per cent; at 4 or more, pooled sensitivity 86.7 with specificity 35.4; at above 2, sensitivity 94.7 with specificity 12.5. The authors of the prospective cohort concluded the score is inaccurate at any cut-point.
Related calculators
References
- Johnston SC, Rothwell PM, Nguyen-Huynh MN, et al. Validation and refinement of scores to predict very early stroke risk after transient ischaemic attack. Lancet. 2007;369(9558):283–292.
- Wardlaw JM, Brazzelli M, Chappell FM, et al. ABCD2 score and secondary stroke prevention: meta-analysis and effect per 1,000 patients triaged. Neurology. 2015;85(4):373–380.
- Perry JJ, Sharma M, Sivilotti MLA, et al. Prospective validation of the ABCD2 score for patients in the emergency department with transient ischemic attack. CMAJ. 2011;183(10):1137–1145.
- National Institute for Health and Care Excellence. Stroke and transient ischaemic attack in over 16s: diagnosis and initial management. NICE guideline NG128, recommendation 1.1.7 (2019).
Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/