Prognostic Risk Stratification Instruments for TIA

February 2011

Prognostic Risk Stratification Instruments for TIA

Your friendly neighborhood medical librarian (Susan Fowler, MLIS provides you with the following search strategy. First I did a google search on “Transient Ischemic Attack ABCD” and located a page from the National Stroke Association that describes how the tool was validated and the study that published in Lancet:

You are working a moonlight Friday 3-11p shift in your busy community emergency department. It is the start of a three day weekend so patients are taking every opportunity to receive emergency care before leaving town on vacation. EMS just notified you that they’re transporting a 63 year-old female, known history of NIDDM, hypertension and high cholesterol, with witnessed onset of aphasia and right arm/leg paralysis, last known well 20-minutes ago. Her fasting blood sugar is 184 so your stroke response team is appropriately activated. Upon arrival, you find a very relieved EMS crew and patient stating ‘almost complete’ resolution of symptoms over the last few minutes. A quick physical exam reveals normal vital signs, and an NIH Stroke scale = 2 (1-right arm drift +1-slight loss in speech fluency). Lab work is quickly sent, the patient is rushed off to CT scan and you go see the new shoulder dislocation in Bed 4. Upon return from CT 10-minutes later, the patient’s symptoms have resolved completely. A repeat NIH Stroke scale is 0. The laboratory tests, including a Troponin, CBC, BMP and coagulation studies, EKG and CT brain are all normal. You are concerned that this was a transient ischemic attack (TIA). The patient feels better and asks to go home since she is hosting a large family gathering at her lake house this weekend. She promises to follow-up with her primary care physician after the long weekend. Her on-call physician promises to pass on the information and will have her seen no later than next Wednesday. When you recall a lecture discussing the short-term risk of stroke following TIA and a clinical decision tool to quantify stroke risk in TIA patients, you become uncomfortable with this plan. What are the short-term risks of TIA and the diagnostic accuracy of the ABCD methods for predicting short-term risk of stroke following an ED evaluated TIA?

Then I looked up the study in Lancet in PubMed to see how it was indexed…
MeSH Terms:

  • California
  • Female
  • Great Britain/epidemiology
  • Humans
  • Ischemic Attack, Transient/classification
  • Ischemic Attack, Transient/complications
  • Ischemic Attack, Transient/epidemiology
  • Logistic Models
  • Male
  • Middle Aged
  • Prognosis
  • Risk Factors
  • Stroke/diagnosis
  • Stroke/etiology
  • Time Factors
  • United States/epidemiology

Since we wanted something more specific to the ED, the key terms were broadened from “Risk Factors” to “Risk,” removed subheadings from “Ischemic Attack, Transient” and added plan keywords like “Risks,” “TIA” and “ED” to capture some that may not have been exactly indexed. Added “ABCD” or else there was no way to determine if search results would be about that specific tool. There is no standardized term for ABCD so did that as a plan keyword but also noticed that it can be referred to ABCD2 as well so added that. Double checked the search results against a search done on narrow/specificity in clinical queries for “clinical predictors” and “etiology” and did not get any more valid results. Here is the final search strategy which brought up 15 results…

Limits: English
(“Risk”[Mesh] OR “risk” OR “risks”) AND (“Ischemic Attack, Transient”[Mesh] OR “TIA”) AND (“Emergency Service, Hospital”[Mesh] OR “ER” OR “ED”) AND (“ABCD” OR “ABCD2”)

You can see the results here:

You can narrow those results to Comparative Effectiveness for which there are 6, here:

PICO Question

Population: Adult emergency department patients with a TIA

Intervention: Clinical decision rule aided assessment

Comparison: Unaided clinical gestalt

Outcome: Prognostic and diagnostic accuracy (sensitivity, specificity, likelihood ratio) for short-term stroke risk and patient-oriented functional outcomes


First years: Short-term Prognosis After Emergency Department Diagnosis of TIA, JAMA 2000;284: 2901-2906. (

Second years: Higher ABCD2 score predicts patients most likely to have true transient ischemic attack, Stroke 2008; 39 (11): 3096-8. (

Third years: Clinical Prediction Rules to stratify short-term risk of stroke among patients diagnosed in the emergency department with a transient ischemic attack Ann Emerg Med 2009;53(5):662-73. (

Fourth years Stratified, urgent care for transient ischemic attack results in low stroke rates, Stroke 2010; 41(11): 2601-5. (

*Note: that the Clinical Decision Rule critical appraisal forms inquire about the decision rule derivation and validation trials. A useful reference for these questions may be Carpenter CR, Keim SM, Crossley J, Perry JJ; Post-transient ischemic attack early stroke stratification: the ABCD2 prognostic aid, J Emerg Med 2009; 36: 194-200. Also, be aware that the PGY IV paper is an “impact analysis” of the ABCD2 score and could have been evaluated with a “Therapy” critical appraisal form using application of the ABCD2 score as the therapy. For simplicity, and to emphasize the topic of “impact analysis” the CDR form was selected.


Article 1: Short Term TIA Prognosis_JAMA 2000

Article 2: ABCD2 Associated Higher Stroke Risk_Stroke 2008

Article 3: CDRs for TIA Risk Stratification_Ann EM 2010

Article 4: Urgent Care for TIA_Stroke 2010

Bottom Line

In the United States 240,000 TIA’s are reported annually. TIA’s precede 23% of strokes. Most TIA’s resolve within 24-hours, but 20%-50% will have evidence of acute tissue infarction on MRI. TIA may be the result of large artery atherosclerosis (thrombosis-generated distal embolism from plaque rupture or vessel occlusion), cardioembolism, or small deep cerebral vessel occlusion/stenosis in addition to less common causes such as infectious vasculopathy, pharmagological vasospasm, or hypercoagulable states. The differential diagnosis of TIA includes stroke, seizure, hypo or hyperglycemia, migraine, CNS mass lesions, cervical or lumbar spine disease, benign positional vertigo, and psychogenic etiologies. Here are the theoretical pros/cons of admitting TIA patients:


  • Prompt recognition of CVA in-hospital within 4.5-hour thrombolytic window.
  • Significant emergency physician (Prabhakaran 2008) and neurologist (Kraaijeveld 1984, Koudstaal 1989, Castle 2010) variability in the diagnosis of TIA.
  • Expeditious diagnosis of culprit lesion (carotid stenosis, paroxysmal atrial fibrillation, valvular heart disease) allowing prompt palliative/curative therapy and (?) reduction in post-TIA stroke risk at 2-days (unlikely), 7-days, and/or 30-days.


  • Expense.
  • 95% of TIA patients will NOT have a stroke within 2 days.
  • Neuro/PCP reticence about admitting asymptomatic patients.

Far more (56.7% versus 6.6%) TIA patients are admitted in the United States than in Canada. Since society probably cannot afford to admit every TIA patient presenting to the ED, identifying a subset most likely to benefit from admission (i.e. with the highest short-term risk of stroke) is worth exploring. Recent research suggests that emergency physicians would accept a TIA clinical decision rule (CDR) to modify admission decisions with a sensitivity of 97%. Neurologists, on the other hand, would accept a TIA CDR with a median sensitivity of 92%. Emergency department physician impression of TIA is usually accurate (94% of cases) as judged by a reviewing neurologist. The Johnston-derived “California Rule” (below) was subsequently merged with the ABCD rule to form the ABCD2 rule (Shah 2009, Carpenter 2009). Among those with true TIA, the ABCD2 score is generally higher (85% in the 3-6 range) and the score correlates with 90-day stroke risk. In non-TIA patients ABCD2 scores are generally <4 (59%) and not associated with any 90-day stroke risk.

At the time of publication 11-years ago, the California Rule and ED-based post-TIA stroke risk provided startling contrast to the previous conceptions about short-term stroke risk. In 2011, though, this prognostic data is well accepted after being replicated in other settings and the result has been multiple interventional trials (SOS-TIA, Detroit, EXPRESS) to reduce 48-hour post-TIA stroke risk.

The ABCD rule is the only TIA-prognostic CDR that has been validated in multiple settings and can be reliably applied to heterogeneous populations, but the ABCD2 is very similar and probably will exhibit similar properties when ultimately tested. An ABCD ≤ 3 is associated with a 0% 7-day stroke risk and might be used to identify a subset of TIA patients appropriate for outpatient work-up. The Ottawa trial (PGY IV article) demonstrated that TIA-risk stratification instruments like the ABCD2 can be used to prioritize outpatient work-up priorities between patients. Reductions in stroke risk can be achieved without hospital admission. Subsequent trials are needed to assess physician acceptance and accuracy in applying the ABCD2 at the bedside in less “clinical decision rule friendly environments” while assessing for cause-effect mechanisms of post-TIA stroke incidence.

Future trials are needed to:

  • Validate the ABCD2, particularly in non-Caucasian populations;
  • Evaluate heterogeneous patient populations’ (differing levels of health literacy, various ethnic and language groups’) ability to comprehend post-TIA stroke risk as communicated to them by physicians in making healthcare decisions real-time in the emergency department.
  • Assess the reliability and accuracy of the ABCD or ABCD2 when used prospectively at the bedside by busy EM clinicians;
  • Assess the impact of ABCD/ABCD2 use on resource utilization and patient-centric outcomes.

California Rule:

  • Age > 60 years
  • Symptoms > 10 – minutes
  • Weakness
  • Speech Deficit


Risk FactorPoints
Age > 60 years1
BP > 140/901
Clinical Features
Unilateral weakness
Language disturbance without weakness
Symptom duration
> 60 minutes 2
10-59 minutes 1
< 10 minutes 0


Risk FactorPoints
Age > 60 years1
BP > 140/901
Clinical Features
Unilateral weakness
Language Disturbance without weakness
Symptom duration
> 60 minutes 2
10-59 minutes 1
< 10 minutes 0