Long-term risk of stroke after transient ischaemic attack: a hospital-based validation of the ABCD2 rule
- Rose Galvin1Email author,
- Penka A Atanassova2,
- Nicola Motterlini^1,
- Tom Fahey1 and
- Borislav D Dimitrov1, 3
© Galvin et al.; licensee BioMed Central Ltd. 2014
Received: 15 April 2014
Accepted: 22 April 2014
Published: 4 May 2014
The ABCD2 clinical prediction rule is a seven point summation of clinical factors independently predictive of stroke risk. The purpose of this cohort study is to validate the ABCD2 rule in a Bulgarian hospital up to three years after TIA.
All consecutive admissions to an emergency department with symptoms of a first TIA were included. Baseline data and clinical examinations including the ABCD2 scores were documented by neurologists. Discrimination and calibration performance was examined using ABCD2 cut-off scores of ≥3, ≥4 and ≥5 points, consistent with the international guidelines. The Hosmer-Lemeshow test was used to examine calibration between the observed and expected outcomes as predicted by ABCD2 score within the logistic regression analysis.
Eighty-nine patients were enrolled to the study with a mean age of 63 years (+/- 12 years). Fifty-nine percent (n = 53) of the study population was male. Seven strokes (7 · 8%) occurred within the first year and six further strokes within the three-year follow-up period. There was no incident of stroke within the first 90 days after TIA. The rule demonstrated good predictive (OR = 1 · 58, 95% CI 1 · 09-2 · 29) and discriminative performance (AUCROC = 0 · 72, 95% CI 0 · 58-0 · 86), as well as a moderate calibration performance at three years.
This validation of the ABCD2 rule in a Bulgarian hospital demonstrates that the rule has good predictive and discriminative performance at three years. The ABCD2 is quick to administer and may serve as a useful tool to assist clinicians in the long-term management of individuals with TIA.
The ABCD2 score was derived and validated to predict risk of stroke up to 90 days after TIA in specialist clinic cohorts and emergency departments (ED) from Oxfordshire in the UK and from California, USA [5, 10]. A number of studies have validated the rule for predicting short term risk of stroke after TIA and reported inconsistent results, ranging from excellent predictive value to little better than chance, largely due to the varied methodological approaches. However, the evidence from two recent systematic reviews demonstrates that the rule has good predictive and discriminative accuracy up to 90 days after stroke [2, 10]. More recently, studies have attempted to validate the rule to predict long term risk of stroke after TIA [11–13]. While these studies have reported that the rule has moderate predictive and discriminative ability, the widespread applicability of the ABCD2 score to predict long-term risk of stroke after TIA depends on its consistency of performance in different studies and settings . Therefore, further broad validation studies by investigators from different specialties and healthcare settings are necessary. This is the first broad validation study that examines the long term predictive value of the ABCD2 rule in a Bulgarian cohort of patients admitted to ED after a TIA.
Study design and setting
The Plovdiv project is a hospital-based cohort study from 2002–2005, designed to determine predictors of incidence and prognosis of major vascular events including the recurrence of cerebrovascular or cardiovascular events. The protocol for this project is described in detail elsewhere [14, 15]. Our study represents a secondary analysis of data collected from a prospective cohort of consecutive patients, presenting with symptoms of TIA to the emergency department at the University Hospital, Medical University of Plovdiv, Bulgaria. The STROBE standardized reporting guidelines were followed to report the findings of the study. Ethical approval was obtained from the Medical University of Plovdiv ethics committee.
Patients were included in this study if they presented to the emergency department with symptoms of a first TIA, were aged >40 years and resident in Plovdiv. Consecutive TIA patients were registered prospectively. Baseline data and the clinical examinations, including those contained in the ABCD2 score, were documented by the study neurologists at the recruitment of patients. Determination of the ABCD2 risk score was performed in a manner identical to that reported by its developers. All patients underwent a CT or MRI scan to determine the index event. Patients with a clinical diagnosis other than TIA, or who were unwilling to give informed consent were excluded from the study.
The TIA, as index event, was defined as a focal neurological deficit, clearly related to one of the main vascular territories, lasting from few minutes up to 24 hours and resolving completely within 24 hours, without evidence of a recent infarction on a CT scan or MRI scan of the brain . Our primary outcome was the occurrence of stroke at three years after TIA. However, in-keeping with other studies of that have reported the short-term predictive value of the rule, we also examined the incidence of stroke at seven and 90 days also. A subsequent stroke was defined using the World Health Organization (WHO) definition of a clinical syndrome consisting of ‘rapidly developing clinical signs of focal (at times global) disturbance of cerebral function lasting more than 24 hours or leading to death with no apparent cause other than that of vascular origin’ and this definition was used for our study . All strokes were assessed by MRI or CT scan and confirmed by a study neurologist. The determination of the outcome endpoint was done when blinded to the ABCD2 score and its variables. In terms of patient follow-up, evaluations were conducted by the study physicians during hospital visits every three months for the first year and then annually for 36 months. Every evaluation was carried out by contact with the patient and where relevant with the family member or caregiver.
Descriptive statistics including means and standard deviations were computed for baseline data. We examined two aspects of validity of our results, discrimination and calibration. Discrimination refers to the ability of ABCD2 score to distinguish correctly the patients with different outcomes (stroke/no stroke). Logistic regression, Cox regression with cumulative hazard function, and receiver operation characteristics (ROC) curves with 95% confidence intervals (95% CI) were used for the analysis. The dependent variable was binary (stroke/no stroke) and the predictor variables examined were the range of scores on the ABCD2 rule. In addition, 2×2 cross-tables were used to calculate discriminative accuracy (sensitivity and specificity) at ≥3, ≥4 and ≥5 points on the ABCD2 rule. These different cut-points are used to identify people at low and high risk of stroke following TIA in several international guidelines on the management of patients with TIA. The c statistic, or area under the curve (AUC), with 95% CI were also estimated to describe model discrimination. The c statistic ranges from 0 · 5 (no discrimination) to a theoretical maximum of 1, values between 0 · 7 and 0 · 9 represent moderate accuracy and greater than 0 · 9 represents high accuracy . A c statistic of 1 represents perfect discrimination, whereby scores for all cases with stroke are higher than those for all the non-cases with no overlap.
Calibration (or reliability) reflects how closely predicted outcomes agree with the actual outcomes. For this purpose, the Hosmer-Lemeshow test (HLT) was used for comparison between the observed outcomes and those expected as predicted by ABCD2 score within the logistic regression analysis. We also used the frequency distribution patterns in the initial derivation study of the ABCD2 rule as a predictive model against which our validation study was compared. Therefore the predicted number of patients with stroke at seven and 90 days (based on the probability calculated in the derivation study) was compared with the observed number of patients with stroke from our validation study. All tests were two sided. An association was considered significant at p < 0 · 05. All statistical analyses were completed using STATA (version 12, Stata Corp, College Station, Texas, USA) and IBM SPSS Statistics (Ver.21, IBM Corporation, USA).
Baseline characteristics of patients with TIA
Systolic BP (mmHg)
Speech disturbance without weakness
Duration of symptoms (minutes)
Time from onset to enrollment (hours)
The incidence of stroke up to three years stratified according to the ABCD2 score
All patients (n = 89)
Stroke within 7 days
Stroke within 90 days
Stroke within 1 year
Stroke within 3 years
Discriminative value of the ABCD 2 rule to identify patients at high risk of stroke at one and three years
Patients at high risk of stroke (as per international guidelines)
Discrimination (95% confidence interval)
ABCD2 ≥ 3 (US guidelines)
Area under curve
1.00 (0.59 - 1.00)
0.34 (0.24 - 0.45)
0.67 (0.62 – 0.72)
0.92 (0.64 - 1.00)
0.36 (0.25 - 0.47)
0.64 (0.55 – 0.73)
ABCD 2 ≥ 4 (UK & New Zealand guidelines)
0.86 (0.42 - 1.00)
0.48 (0.36 - 0.59)
0.67 (0.55 – 0.76)
0.77 (0.46 - 0.95)
0.49 (0.37 - 0.60)
0.63 (0.50 – 0.76)
ABCD 2 ≥ 5 (Australian guidelines)
0.71 (0.29 - 0.96)
0.67 (0.56 - 0.77)
0.69 (0.51 – 0.88)
0.62 (0.32 - 0.86)
0.68 (0.57 - 0.79)
0.65 (0.50 – 0.80)
At three years, the analysis of the overall calibration performance indicates that there is no statistically significant difference between the observed and expected outcomes as predicted by the ABCD2 rule (χ2HLT = 4.11, p = 0 · 533). When further using the original derivation study as a short-term predictive model, the ABCD2 rule tends to over-predict the risk of stroke at seven days across all three original risk strata: low risk, [RR 3 · 00, 95% CI (0 · 13-71 · 51)]; intermediate risk, [(RR 5 · 00, 95% CI (0 · 25-99 · 82)]; and high risk, [(RR 4 · 33, 95% CI (0 · 75-24 · 90)]. The results are broadly similar at 90 days across all three risk strata: low risk, [RR 3 · 00, 95% CI (0 · 13-71 · 51)]; intermediate risk, [(RR 5 · 00, 95% CI (0 · 25-99 · 82)]; and high risk, [(RR 9 · 00, 95% CI (0 · 52-156 · 91)]. There are seven strokes predicted at 90 days when using the original derivation study as a predictive model but no incident of stroke was observed in our dataset within this time across all three risk strata of risk.
Statement of principal findings
This broad validation of the ABCD2 clinical prediction rule in Bulgarian patients demonstrates that the rule has a good predictive and discriminative performance, as well as a moderate calibration performance at three years. The results also indicate that the rule over-predicts the risk of stroke at seven and 90 days in comparison with the frequency distributions patterns from the original derivation study.
Results in the context of the current literature
Our findings are broadly in-keeping with two other retrospective studies that have examined the long term predictive value of the ABCD2 rule [11, 12]. In our study, 7 · 8% of patients had a stroke within the first year and six further strokes occurred within the three year follow-up period (14 · 94%). These results are similar to a German multi-centre hospital based validation of the ABCD2 rule where the overall risk of stroke at one year was estimated at 6 · 5% in a cohort of 1448 patients . Our findings are also in keeping with a retrospective study by Harrison and colleagues who, although not showing results at three years, had reported that the overall absolute risk of stroke in patients who attended an outpatient clinic with a TIA was 7 · 3% within one year, 16 · 2% within five years and 28 · 0% within ten years . Similar to our study, patients with higher ABCD2 scores (≥3 points) experienced higher stroke risk within one year and this risk persisted at five and ten years .
A recent systematic review of studies that validated the ABCD2 rule reported a low overall rate of stroke at seven days (1 · 72%) and 90 days (2 · 63%) in patients identified as low risk using an ABCD2 score of 0–3 points . Furthermore, these patients represented one third of the overall cohort of patients in the meta-analysis . Almost 45% of individuals in our dataset were classified as low risk using the ABCD2 cut-off score and there was no incident of stroke in this patient group within seven or 90 days. This finding in not surprising given the low stroke rate reported in the pooled analysis of studies. In addition, our results are similar to three other hospital based validation studies where no event of stroke was recorded in the low risk group at seven days [19–21]. While we reported no incident of stroke across all three strata of risk at seven or 90 days, a study by Coutts and colleagues also reported no incident of stroke across all three risk groups at seven or 90 days in a cohort of 87 patients . However, our results need to be interpreted with caution due to the large number of patients in the low risk category and limited sampled population included in the validation study.
The overall discriminative value of the ABCD2 score at three years is significantly better than chance, as measured by ROC area under the curve and associated 95% CIs being above the null value of 0 · 5. While no other study has examined the discriminative performance of the rule at three years, the accuracy of the rule has been examined up to 13 · 8 years and an acceptable level of accuracy by AUC of 0 · 630 (95% CI 0 · 58-0 · 67) was reported . Our results are also similar to those reported in a systematic review of the discriminative accuracy of the ABCD2 at seven days [AUC 0 · 72 (95% CI 0 · 63-0 · 82)] , suggesting that the rule has reasonable ability to correctly distinguish patients with and without stroke at three years.
Strengths and weaknesses of the study
This is the first broad hospital based validation study to examine the predictive and discriminative value of the ABCD2 score in a Bulgarian cohort up to three years after TIA. The risk of stroke is presented in risk strata consistent with those applied in the international guidelines on the management of stroke so that the value of the ABCD2 score across these strata can be interpreted in a clinically meaningful way. Furthermore, the ABCD2 rule was applied to our cohort of patients according to the methods described by original developers . Our method of calibration analysis further examines the specific performance ability of the rule by using the ratio of predicted stroke (from the original derivation study) to observed stroke in the present validation study. However, the generalisability of our results may be influenced to some extent by the relatively limited number of the patients included in our study. The comparability of our findings to the current literature is restricted by the limited number of studies that examine the long-term discriminative performance of the ABCD2. Therefore there is a need for future large multi-centre cohort studies to examine the long term predictive value of the rule in different clinical settings. The impact of the rule in the ED setting also warrants further investigation, particularly in terms of patient outcome, clinician behaviour, cost effectiveness and resource use, or any combination of these.
Clinical and policy implications
Current international guidelines recommend that individuals with low ABCD2 scores should be triaged for specialist assessment within one week of onset of symptoms. While we observed no incident of stroke in our patients with ABCD2 score of 0–3 points at seven days, it is worth noting that the overall stroke rate in these patients was 7 · 5% within three years. Therefore, even at a lower risk initially, there is a need for a regular review of these specific patients in the primary care or specialist setting to monitor and manage their ongoing cerebrovascular risk and minimise the risk of subsequent stroke or other related cardiovascular events.
There is a lack of consistency in the international guidelines on what constitutes a ‘low’ and ‘high’ risk patient and we have reported our data accordingly. All guidelines relating to the management of individuals classified as ‘high risk’ include early neurology consultation to confirm the diagnosis of TIA, rapid diagnostic assessment, and implementation of aetiology-specific precautionary measures, including carotid endarterectomy and anticoagulation . However, there is a need for consensus in relation to the identification of individuals at high risk of subsequent stroke as the various cut-points of ABCD2 score used to categorise stroke risk have considerable economic and management implications.
This validation of the ABCD2 rule in a Bulgarian hospital setting demonstrates that the rule has a good predictive and discriminative performance at three years. The results also indicate that the rule tends to over-predict the risk of stroke at seven and 90 days in comparison with the frequency distributions patterns from the original derivation study. The ABCD2 is easy and quick to administer and it is a useful tool to assist clinicians in the long term management of individuals with TIA.
The authors would like to thank Prof. Zahari Zahariev, DMSc and Assoc. Prof. Nedka Chalakova, PhD at the Medical University of Plovdiv, Bulgaria for their helpful comments, support with the data retrieval and facilitating the data analysis and results interpretation.
This work was supported by the Noel Hickey Research Bursary 2010 in conjunction with the Irish Heart Foundation and also by the Health Research Board (HRB) of Ireland through the HRB Centre for Primary Care Research under Grant HRC/2007/1.
- Cumler E, Glasheen J: Risk stratification tools for transient ischemic attack: which patients require hospital admission?. J Hosp Med. 2009, 4 (4): 247-251. 10.1002/jhm.411.View ArticleGoogle Scholar
- Galvin R, Geraghty C, Motterlini N, Dimitrov BD, Fahey T: Prognostic value of the ABCD2 clinical prediction rule: a systematic review and meta-analysis. Fam Pract. 2011, 28: 366-376. 10.1093/fampra/cmr008.PubMedView ArticleGoogle Scholar
- Weimar C, Benemann J, Huber R, Mieck T, Kaendler S, Grieshammer S, Katsarava Z, Diener HC: Long-term mortality and risk of stroke after transient ischemic attack: a hospital-based cohort study. J Neurol. 2009, 256 (4): 639-644. 10.1007/s00415-009-0150-9.PubMedView ArticleGoogle Scholar
- Purroy F, Jimenez Caballero PE, Gorospe A, Torres MJ, Alvarez-Sabin J, Santamarina E, Martinez-Sanchez P, Canovas D, Freijo MM, Egido JA, Girón JM, Ramírez-Moreno JM, Alonso A, Rodríguez-Campello A, Casado I, Delgado-Medeiros R, Martí-Fàbregas J, Fuentes B, Silva Y, Quesada H, Cardona P, Morales A, de la Ossa N, García-Pastor A, Arenillas JF, Segura T, Jiménez C, Masjuán J, Stroke Project of the Spanish Cerebrovascular Diseases Study Group: Prediction of early stroke recurrence in transient ischemic attack patients from the PROMAPA study: a comparison of prognostic risk scores. Cerebrovasc Dis. 2012, 33 (2): 182-189. 10.1159/000334771.PubMedView ArticleGoogle Scholar
- Johnston SC, Rothwell PM, Nguyen-Huynh MN, Giles MF, Elkins JS, Bernstein AL, Sidney S: Validation and refinement of scores to predict very early stroke risk after transient ischaemic attack. Lancet. 2007, 369 (9558): 283-292. 10.1016/S0140-6736(07)60150-0.PubMedView ArticleGoogle Scholar
- Easton J, Saver J, Albers G, Alberts M, Chaturvedi S, Feldmann E, Hatsukami T, Higashida R, Johnston S, Kidwell C, Lutsep HL, Miller E, Sacco RL, American Heart Association; American Stroke Association Stroke Council: Definition and evaluation of transient ischemic attack. A scientific statement for healthcare professionals from the American heart association/American stroke association stroke council; council on cardiovascular surgery and anesthesia; council on cardiovascular radiology and intervention; council on cardiovascular nursing; and the interdisciplinary council on peripheral vascular disease. Stroke. 2009, 40: 2276-2293. 10.1161/STROKEAHA.108.192218.PubMedView ArticleGoogle Scholar
- National Institute for Health and Care Excellence: Diagnosis and initial management of acute stroke and transient ischaemic attack: CG68. 2008, London: National Institute for Health and Care ExcellenceGoogle Scholar
- National Stroke Foundation: Clinical Guidelines for Stroke Management. 2010, Melbourne: National Stroke FoundationGoogle Scholar
- Stroke Foundation of New Zealand: For the Assessment and Management of People With Recent Transient Ischaemic Attack. 2008, Wellington: Stroke Foundation of New ZealandGoogle Scholar
- Giles M, Rothwell P: Systematic review and pooled analysis of published and unpublished validations of the ABCD and ABCD2 transient ischaemic attack risk scores. Stroke. 2010, 41: 667-673. 10.1161/STROKEAHA.109.571174.PubMedView ArticleGoogle Scholar
- Fothergill A, Christianson TJH, Brown RD, Rabinstein AA: Validation and refinement of the ABCD2 score: a population-based analysis. Stroke. 2009, 40 (8): 2669-2673. 10.1161/STROKEAHA.109.553446.PubMedPubMed CentralView ArticleGoogle Scholar
- Harrison JK, Sloan B, Dawson J, Lees KR, Morrison DS: The ABCD and the ABCD2 as predictors of stroke in transient ischemic attack clinic outpatients: a retrospective cohort study over 14 years. QJM: Monthly J Assoc Phys. 2010, 103 (9): 679-685.View ArticleGoogle Scholar
- Holzer K, Feurer R, Sadikovic S, Esposito L, Bockelbrink A, Sander D, Hemmer B, Poppert H: Prognostic value of the ABCD2 score beyond short-term follow-up after transient ischemic attack (TIA)–a cohort study. BMC Neurol. 2010, 10: 50-10.1186/1471-2377-10-50.PubMedPubMed CentralView ArticleGoogle Scholar
- Atanassova PA, Semerdjieva MA, Naydenov VI, Dzhurkova AG, Traykova NI, Chalakova NT: Transient ischaemic attacks and minor strokes (analyses and clinically significant comparisons). Folia Med (Plovdiv). 2006, 48 (2): 30-36.Google Scholar
- Atanassova PA, Chalakova NT, Dimitrov BD: Major vascular events after transient ischaemic attack and minor ischaemic stroke: post hoc modelling of incidence dynamics. Cerebrovasc Dis. 2008, 25 (3): 225-233. 10.1159/000113860.PubMedView ArticleGoogle Scholar
- Advisory Council for the National Institute of Neurological and Communicative Disorders and Stroke: A classification and outline of cerebrovascular diseases II. Stroke. 1975, 6 (5): 564-616.View ArticleGoogle Scholar
- Hatano S: Experience from a multicentre stroke register: a preliminary report. Bull World Health Organ. 1976, 54 (5): 541-553.PubMedPubMed CentralGoogle Scholar
- Swets J: Measuring the accuracy of diagnostic systems. Science. 1988, 240 (4857): 1285-1293. 10.1126/science.3287615.PubMedView ArticleGoogle Scholar
- Asimos AW, Johnson AM, Rosamond WD, Price MF, Rose KM, Catellier D, Murphy CV, Singh S, Tegeler CH, Felix A: A multicenter evaluation of the ABCD(2) score’s accuracy for predicting early ischemic stroke in admitted patients with transient ischemic attack. Ann Emerg Med. 2010, 55 (2): 201-210. 10.1016/j.annemergmed.2009.05.002.PubMedView ArticleGoogle Scholar
- Coutts SB, Eliasziw M, Hill MD, Scott JN, Subramaniam S, Buchan AM, Demchuk AM, VISION study group: An improved scoring system for identifying patients at high early risk of stroke and functional impairment after an acute transient ischemic attack or minor stroke. Int J Stroke. 2008, 3 (1): 3-10. 10.1111/j.1747-4949.2008.00182.x.PubMedView ArticleGoogle Scholar
- Cucchiara BL, Messe SR, Sansing L, MacKenzie L, Taylor RA, Pacelli J, Shah Q, Kasner SE: Lipoprotein-associated phospholipase A2 and C-reactive protein for risk-stratification of patients with TIA. Stroke. 2009, 40 (7): 2332-2336. 10.1161/STROKEAHA.109.553545.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.