Inter-rater Agreement for the Diagnosis of Stroke Versus Stroke Mimic

Abstract

Background and purpose: Patients who present emergently with acute neurological signs and symptoms represent unique diagnostic challenges for clinicians. We sought to characterize the reliability of physician diagnosis in differentiating aborted or imaging-negative acute ischemic stroke from stroke mimic.

Methods: We constructed 10 case-vignettes of patients treated with thrombolysis with subsequent clinical improvement who lacked radiographic evidence of infarction. Using an online survey, we asked physicians to select a most likely final diagnosis after reading each case-vignette. Inter-rater agreement was evaluated using percent agreement and κ statistic for multiple raters with 95% confidence intervals reported.

Results: Sixty-five physicians participated in the survey. Most participants were in practice for ≥5 years and over half were vascular neurologists. Physicians agreed on the most likely final diagnosis 71% of the time, κ of 0.21 (95% confidence interval, 0.06-0.54). Percent agreement was similar across participant practice locations, years of experience, subspecialty training, and personal experience with thrombolysis.

Conclusions: We found modest agreement among surveyed physicians in distinguishing ischemic stroke syndromes from stroke mimics in patients without radiographic evidence of infarction and clinical improvement after thrombolysis. Methods to improve diagnostic consensus after thrombolysis are needed to assure acute ischemic stroke patients and stroke mimics are treated safely and accurately.