Clinical Scores for Predicting Recurrence After Transient Ischemic Attack or Stroke

Abstract

Risk scores are commonly used in the prediction of disease outcome. In the context of cerebrovascular diseases, risk scores have been created to identify stroke risk after transient ischemic attack (TIA) and (minor) stroke, to identify subgroups of patients with high risk of stroke (for instance, correlated with grade of carotid stenosis), or to predict functional outcome after stroke. Identifying high-risk patients after TIA is important because early assessment and management of these patients is pivotal. Confident detection of the low-risk patient, however, is of similar importance. Performing multiple acute diagnostic investigations for all suspected TIA and stroke patients might overwhelm the medical system and might not be feasible because of resource limitations. Simple and reliable risk estimation of recurrence might be beneficial to high-risk patients to be admitted and investigated early. Additionally, the medical health system might benefit as well, because low-risk patients can be seen in less expensive outpatient clinics. We performed a systematic review of published risk scores that predict recurrence risk after stroke or TIA. We checked the quality of the risk scores based on the characteristics of the various derivation and replication studies. One investigator (S.S.) performed a PubMed search with the search terms prognostic models stroke and prognostic scores stroke for the period 1992 to 2011, and additionally explored the reference lists of the identified articles. We excluded specific risk scores for stroke risk in atrial fibrillation, for instance, CHADS2,1 CHADS2VASC2, or global vascular estimates, such as the QRISK2 and SCORE.3 The other exclusion criteria, internal and external validity, statistical methodology, validation of the models, and clinical applicability were evaluated (for details, see Methods and Table I in the online-only Data Supplement) by 3 independent researchers (R.L., S.S., and V.T.). All results were compared between researchers and inconsistencies were resolved by …