Abstract 149: Evaluating the Risk of Cerebrovascular Events Among Those Seeking Information on Transient Ischemic Attack Symptoms on the Internet

Abstract

Background: Despite the high risk of stroke after transient ischemic attack (TIA), many people do not seek urgent medical attention because symptoms may be mild or have already resolved or be incorrectly attributed to a benign mimic. Since the Internet has become an increasingly important source of medical information, we sought to reach these people by developing and testing an instrument to identify cerebrovascular events among those seeking information about TIA online. Methods: We developed an instrument to evaluate possible TIA symptoms ( http://tia.ucsf.edu/ ) using items from the ABCD2 score excluding blood pressure since it may not be readily available by self-report. Using targeted Internet text advertisements for terms such as “transient ischemic attack symptoms” or “mini-stroke symptoms” we invited subjects to complete a 20-item demographic, medical history, and symptom inventory online. Neurologists that were blinded to these responses then contacted subjects by telephone to assess the likelihood that a true TIA or stroke had occurred. Logistic regression was used to assess univariate and multivariable associations and ROC curves were generated to assess model discrimination. Results: The study website received 26,602 hits over 122 days of study enrollment. A total of 289 visitors started the enrollment and eligibility screening process, 253 provided consent and recorded responses for the instrument, and 173 completed the follow-up telephone assessment. The mean age was 58.7 years (range 23-88) and 63% were women. A total of 36 (21%) subjects had had symptoms within 24 hours and 60 (37%) had not yet sought medical advice for their symptoms. The follow-up telephone assessment process determined that 43 (24.9%) had a TIA, and 67 (38.7%) had either a TIA or a stroke. The ABCD2 score (excluding blood pressure) demonstrated moderate discrimination overall (c-statistic 0.69, 95% CI 0.61-0.76), but 1 (8.3%) of 12 subjects with a score of 0 had a cerebrovascular event (isolated diplopia). Those who reported multiple prior episodes of similar symptoms in the past year were less likely to have had a cerebrovascular event (OR 0.18, 95% CI 0.10-0.37) and incorporating this predictor and the presence of diplopia into the model significantly improved discrimination (c-statistic 0.78, 95% CI 0.71-0.85, p=0.03) with no events among the 7 patients who had a score of 0. Conclusion: Patients with possible TIA symptoms turn to the Internet for medical information and prior to any attempt to seek formal medical advice. Although applying self-reported elements of the ABCD2 score was inadequate to rule-out cerebrovascular disease, a modified score with additional elements may provide an efficient, immediate, and individualized mechanism to reach patients with TIA who might not otherwise seek medical attention.