Annual Meeting Abstracts: View

<< Back to Search Results

(Abst. 1.205), 2019

Video Quality Using Outpatient Smartphone Videos in Epilepsy: Results from the OsmartViE Study
Authors: William O. Tatum IV, Mayo Clinic; Larry Hirsch, Yale University; Michael Gelfand, University of Pennsylvania; Emily K. Acton, University of Pennsylvania; Curt LaFrance, Brown University; Robert Duckrow, Yale University; David K. Chen, Baylor University; Andrew S. Blum, Brown University; John D. Hixson, University of California San Francisco; Joseph Drazkowski, Mayo Clinic Arizona; Selim R. Benbadis, University of South Florida; Gregory D. Cascino, Mayo Clinic Rochester
Content: Rationale: Misdiagnosis of epilepsy may occur when incomplete or inaccurate reporting occurs. Outpatient video of a typical event may serve as a surrogate but depends upon video quality and the skill of the interpreter. We sought to evaluate video quality of outpatient smartphone videos in adults undergoing diagnostic evaluation for epilepsy.  Methods: A prospective, multicenter, blinded trial involving outpatient-generated smartphone video of patients with seizures was performed. Diagnosis was confirmed for patients with 1) epileptic seizures (ES), 2) psychogenic nonepileptic attacks (PNEA), and (3) physiologic nonepileptic events (PhysNEE) after video-EEG monitoring (VEM). Experts and senior neurology residents viewed smartphone videos via HIPAA-protected web-based transfer for diagnosis without clinical information or EEG. Quality was assessed by survey response for technical and operational measures. P-value of < 0.05 was significant. Results: Forty-four patients [31 F, age 45.1 yrs. (range 20-82)] had 530 smartphone video views by a mean of 6.6 experts and 5.5 residents per video averaging 133.8 sec. (range: 9-543). 30 patients had a final diagnosis of PNEA, 11 with ES, and 3 had PhysNEE following admission for a mean of 3.1 days (range: 1-9) of video-EEG monitoring. Overall accuracy by all reviewers for ES was 82.7% (95% CI: 78.7 - 86.2%), and PNEA was 81.0% (95% CI: 76.8 - 84.7%). The positive predictive value for a smartphone video was 70.7% (60.2 - 79.7%) for ES and 84.1% (79.4 - 88.1%) for PNEA. Overall quality in pooled reviews was 70.8% (81.8% by the majority) without distinction between raters. 120 views were non-diagnostic due to; difficult diagnosis (42.5%), poor video (24.2%), both (31.7%) or neither (1.7%). Videos with 100% accuracy noted good audio in 86.2% vs 75.4% for others (p=0.01). A trend for level of light to yield correct diagnosis was present (p=0.06), though absent for clarity (p=0.59). Limited interactivity, atypical semiology, restricted fields and short video duration hindered diagnosis, and was facilitated by the presence of motor features. Inter-rater reliability was moderate for ES (k= 0.44) when judged by experts and fair (k= 0.3) for residents. Conclusions: Overall, smartphone video quality is adequate for clinical interpretation in majority of patients. Technical features are less important than operational ones. Enhancing interactivity, prolonging duration and large fields of view optimize recording. We speculate higher yield with smartphone video are attainable with patient education. Funding: $5000.00 support for coordinator fees from Mayo Clinic