Rationale:
Simultaneous recording of video-EEG (v-EEG) monitoring in an inpatient epilepsy monitoring unit (EMU) setting has been the gold standard for distinguishing epileptic seizures (ES) from psychogenic non-epileptic seizures (PNES). However, capturing events during inpatient monitoring is difficult in patients with infrequent events. Patient recordings of habitual events may bridge this diagnostic gap if they are sufficiently predictive of the diagnosis. Using v-EEG diagnosis as a gold standard, we assessed the predictive diagnostic value of expert evaluation of smartphone videos of habitual events with or without additional limited demographic data in US Veterans referred for evaluation of epilepsy.
Method:
This was a prospective, blinded diagnostic accuracy study conducted at the Michael E. DeBakey VA Medical Center in Houston TX between 12/2015 and 06/2019, involving adult patients who had a definitive diagnosis of paroxysmal episodes from EMU v-EEG recording. Three epileptologists with ABPN boards in Epilepsy and Clinical Neurophysiology, were blinded to the diagnosis but formulated a diagnostic impression based upon the review of the smartphone video alone and smartphone video plus limited demographic data.
Results:
Fifty patients (30 men [60%]; mean [range] age, 47.7 [22-72] years) submitted smartphone videos. Of these, 14 had ES and 33 had PNES diagnosed by v-EEG. Diagnostic accuracy was calculated for each reviewer and combined across all the ratings (see Table 1 for results). Providing raters with basic patient demographic information in addition to the smartphone videos did not significantly improve diagnostic accuracy above the smartphone videos alone (p >0.05); however this increased subjective confidence ratings by 26%, 36% and 32% for the three epileptologists. Inter-rater reliability between three raters based on smartphone video using Cronbach’s alpha was 0.75, suggesting good correlation. Significant predictors of PNES diagnosis (using binary logistic regression) included greater seizure frequency, greater number of anti-seizure medications, and more non-epileptic features. Specific non-epileptic features that were most consistently seen in the videos among those with PNES compared with ES included complex non-physiologic motor (59.6% vs. 11.9%), generalized rhythmic tremor (22.2% vs. 4.8%), staring and eye blinking (17.2% vs. 4.8%), and violent flailing movements (15.2% vs. 2.4%).
Conclusion:
This study demonstrates that smartphone videos of habitual patient events could be critical tools in reliably diagnosing ES vs PNES in US Veterans referred for evaluation of epilepsy when v-EEG is not feasible or unsuccessful in capturing habitual events. Although demographic information did not significantly contribute to the predictive value of smartphone video, it overall increased the confidence level in reviewers about one third of the time.
Funding:
:None
FIGURES
Figure 1