We see a lot of patients in the ED with a history of what sounds very much like a transient ischaemic event (TIA). By definition the patient will have had resolution of their symptoms/signs and so there is no apparent need for immediate intervention, but we also know that a TIA may be precursor to a larger and thus more serious stroke. In some settings the ABCD2 score is used to risk stratify patients into those who are safe to go home and to those requiring urgent investigation or even admission. However, the ABCD2 score is not perfect and in 2014 a Canadian group prospectively derived a risk score that hoped to improve on current models. This Canadian TIA score has now been validated in a paper published in the BMJ this week.
As with most BMJ papers the paper copy that landed on my door only has summary information so if you want to read the full article you will need to go online and read the open access article. The abstract is below but as always we do strongly recommend you read the full article.
What kind of paper is this?
This is a prospective cohort paper. The authors are using this cohort to validate a previous paper that derived a risk score for patients following a TIA. It is vital that risk scores and other diagnostic tools are validated in differing cohorts from the one that they were derived in, but that does not always happen. Independent validation is important as a risk score will always perform best in the cohort from which it is derived. What we really want to know, and what this study aims to do is to find out whether the tool works in other cohorts.
Who was studied?
Patients were recruited from 13 Canadian emergency departments with symptoms of TIA or a minor stroke, although the definition of a minor stroke is a little unclear. They do say that they excluded patients who had symptoms for more than 24 hours, so perhaps these two terms are synonymous? Patients who had received thrombolysis or clot retrieval were not included. I was impressed that they were able to recruit patients sequentially, 24 hours a day and seven days a week. This is so important when validating risk scores, but is something that we rarely see in the literature. Patients who present at night/weekends/holidays are often different to normal working hours and so this is an important and welcome feature of this validation study (though in practice this meant that they managed to enrol roughly 80% of patients).
What did they do?
All patients had three TIA scores calculated. The ABCD2 score (which we use in Virchester), the ABCD2i score (which incorporates neuroimaging results) and the Canadian TIA score. The Canadian score is shown below, and is clearly more complicated than the ABCD scores. The authors acknowledge this and suggest calculating using an app. I suspect we will see it on sites such as MedCalc soon (though not currently on there).
Patients were then followed up for seven days post discharge to see if they developed a subsequent stroke or received carotid endarterectomy or stenting. They looked for this outcome through healthcare records and by contacting patients for telephone follow up.
What did they find?
It’s worth taking a little time to read through the methods section to understand how they approached validation, but in simple terms they did three things. Firstly they calculated the absolute risk for each individual score than the Canadian TIA score produces and compared this to what the original derivation study would have predicted. Secondly, the assessed the risk in three distinct groups of scores representing low, medium and high risk groups, thirdly they compared the performance of the score against ABCD2 and ABCD2i.
7607 patients were recruited of whom 182 had the outcome of stroke or revascularisation within seven days.
The Canadian TIA score showed an increasing risk with a higher score and when gouped into three risk bands did differentiate into risk groups.
- Score -3-3. Low risk. 16.3% of patients with 6/1236 adverse outcomes which equates to 0.5%
- Score 4-8. Medium risk. 72.1 of patients with 124/5360 adverse outcome which equates to 2.3%.
- Score 9-14. High risk. 11.6% of patients with 52/829 adverse outcomes which equates to 6.3%
These bandings seem reasonable in terms of risk assessments that might link to clinical strategies relating to discharge or rapid interventions for patients. Low risk being suitable for discharge, but high risk requiring a very rapid intervention to prevent an early stroke.
The ABCD scores were unable to identify a low (<1%) risk group and thus the Canadian TIA score appears to have a significant advantage. Additionally the bast majority of patients on the ABCD scores are classified as medium risk which is not as useful to clinicians making important decision about disposition and investigation prioritisation. More patients were classified as high risk by the Canadian TIA score as compared to the ABCD scores.
The results were not significantly different if the authors excluded revascularisation as an outcome (important as this may be a health economy effect).
So should we use the Canadian TIA score rather than the ABCD scores?
Probably in my opinion. The derivation and validation studies for this score appear robust and it appears to offer advantages over our current scores.
However, there are always caveats and in this paper my principle concern is that the score has been derived and validated in very similar patient populations. Although an additional number of recruiting centres were added in the validation paper, this is still Canadian emergency medicine centres, many of which recruited to both derivation and validation data sets. Although there are clear advantages in doing this, there are also benefits to validating in other health systems and with different patients. In general the less ‘alike’ patients are from the original derivation study then the less well a score will perform and thus I suspect that the score may not work as well in metropolitan Virchester as in Canada, though it’s worth noting that exactly the same argument can be made against the ABCD score we currently record (it’s just in the nature of all derivation/validation studies). In their defence the authors acknowledge this and call for more data from implementation studies.
We also need to consider how this score might be used on a local basis. Stratifying into three risk groups is only useful if it then leads to differential actions. How that might happen will require a shared decision with neurology/vascular/imaging/stroke colleagues and will also depend on how your local services are currently set up and how quickly they can respond. In past years in Virchester we used the ABCD score to decide who was admitted and who was discharged, but these days all are discharged with antiplatelet treatment and followed up as an out patient within 24 hours (as per NICE guidance).
High risk groups may also benefit from dual antiplatelet treatment (though at higher risk of bleeding if you do).
Global health perspective from Stevan Bruijns aka @codingbrown
From a high-income perspective, the Canadian TIA score may soon be redundant. As already mentioned, in the NHS there is little need for risk stratification as everyone can be followed up within 24 hours (often with a MRI). Other countries with the resources to do so are either already practicing the same, or will be reasonably soon – likely sooner in countries with a universal healthcare system, such as Canada, than those without.
From a low- to middle-income perspective, the Canadian TIA score provides a possible upgrade to the ABCD scoring systems. But external validation of the Canadian TIA score will be required to understand its usefulness in these settings – ironically lacking for ABCD scoring as well. It is unclear what the value of risk stratification is, when primary prevention without further investigation, is often all that’s on offer. Magnetic resonance imaging isn’t universally accessible (some countries only have access to MRI by sending patients to neighbouring countries), access to specialist care can be hit and miss, and the cost for either within these settings are highly prohibitive. This is depressing as around a quarter or more of all global stroke deaths and morbidity occur in LMICs and have been steadily rising (whilst the incidence of stroke has been declining in HICs). The ability to better predict risk can only translate into better outcomes if the resources exist to manage the risk – pointing to a developing problem is not quite the same as preventing the problem. These complex systems issues are likely to remain for the foreseeable future whether the Canadian TIA score replaces ABCD2 or not in LMIC EDs.
The bottom line
The Canadian TIA score offers advantages over current ABCD scoring systems. We should consider how this might affect patient management pathways with colleagues in related specialities.
- Perry JJ, Sharma M, Sivilotti ML et al . A prospective cohort study of patients with transient ischemic attack to identify high-risk clinical characteristics. Stroke2014;45:92-100. doi:10.1161/STROKEAHA.113.003085 pmid:24262323
- ABCD2 score on MedCalc https://www.mdcalc.com/abcd2-score-tia
- Giles M et al. Addition of brain infarction to the ABCD2 Score (ABCD2I): a collaborative analysis of unpublished data on 4574 patients https://pubmed.ncbi.nlm.nih.gov/20634480/
- MedCalc https://www.mdcalc.com/
- Just the Facts: Diagnosis and risk-stratification following transient ischemic attack https://www.cambridge.org/core/journals/canadian-journal-of-emergency-medicine/article/just-the-facts-diagnosis-and-riskstratification-following-transient-ischemic-attack/8FC148BD8A26DCF612920D0B884F4B45
- Carley S. Risky Business at St Emlyn’s https://www.stemlynsblog.org/?s=risky+business
- NICE guideline [NG128] Published date: 01 May 2019 Stroke and transient ischaemic attack in over 16s: diagnosis and initial management. https://www.nice.org.uk/guidance/ng128/chapter/recommendations#imaging-for-people-who-have-had-a-suspected-tia-or-acute-non-disabling-stroke