Abstract 49: Diagnostic Accuracy of Telestroke Consultation

Abstract

Introduction: Given the rapid expansion of telestroke systems, an increasing number of patients with acute stroke-like presentations are receiving tele-evaluation for guidance of diagnostic and therapeutic decisions. There is limited data, however, regarding the accuracy of telestroke diagnosis. We examined the accuracy of telestroke diagnosis and evaluated predictors of misdiagnosis. Methods: We conducted a retrospective study of all telestroke consultations at our center between April 2015 and April 2016. Consultations were classified into one of 3 diagnostic categories: stroke/TIA, mimic, and uncertain. Telestroke diagnosis was compared with final diagnosis determined after review of additional emergency department testing or hospital admission. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for diagnosis of stroke/TIA vs mimic were calculated. Area under receiver-operating characteristic curve (AUC) analysis to predict true stroke was performed. Bivariate analysis based on the diagnostic categories examined association with sex, age, NIH Stroke Scale (NIHSS), stroke risk factors (CAD, DM, HLD, HTN, history of stroke/TIA, smoking), TPA given, bleeding after TPA, symptom onset to last known normal, symptom onset to consult, time of day of consult by 6 hour epoch, and consult duration. Logistic regression (LR) was performed as indicated by bivariate analysis. Results: 874 telestroke evaluations were included in our analysis. Sensitivity, specificity, PPV, NPV were 97.8%, 82.5%, 93.7% and 93.4%, respectively. AUC (95% CI) was 0.9061 (0.8749-0.9283). LR revealed OR (95% CI) of misdiagnosis for female gender of 1.9 (1.3-2.9). Conclusions: Our telestroke system is highly accurate in discriminating between stroke/TIA and stroke mimics with a tendency towards overdiagnosis of stroke. Female gender was associated with misdiagnosis.