Abstracts

Rationale: Inpatient long-term video-EEG monitoring (LTM) is an important diagnostic tool for patients with seizures and other paroxysmal behavioral events. The yield of video-EEG monitoring is highly dependent on the recording of events. A subset of patients admitted to an inpatient epilepsy monitoring unit will not have their typical events recorded, which makes diagnosis more problematic. Identifying the variables that predict the diagnostic value of long term video EEG monitoring will help in reducing LTM failure rates. The purpose of this study is to determine if specific indicators can help identify patients who are more likely to have non-diagnostic studies. Methods: This was an observational, cross-sectional study involving retrospective chart review. All patients over18 years of age admitted to the Epilepsy Monitoring Unit at Spectrum Health for LTM between December 2014 and December 2015 were included in the study. Patients were divided into two groups (non-diagnostic and diagnostic) based on findings. Each patient's electronic chart was reviewed to determine demographics (age, gender, education), and historical information (duration of disorder, event semiology, event frequency, diurnal event pattern variation, history of abnormal EEG, past psychiatric history, past history of abuse, medications at admission, review of systems). Data pertinent to the LTM study was also reviewed (study duration, activations employed, medication taper). Results: Ninety-six patients (34.2%) without events (non-diagnostic) and 184 (65.7%) with events (diagnostic) were included. 72.9% of patients in the non-diagnostic group and 66.8% in the diagnostic group were females (p=0.298). Patients with a median monthly event frequency of 4 or less were more likely to have a non-diagnostic study whereas a median event frequency of 8 or more per month was more likely to yield a diagnostic study (p=0.016). In terms of semiology, having convulsive events was more likely to be associated with a diagnostic study (46.8% diagnostic vs 28.6% non-diagnostic, p=0.027) whereas nonconvulsive events with altered awareness, or events with preserved awareness did not affect the yield. Other semiological findings noted to be non-predictive of yield included patient-reported tongue bite or urinary incontinence, and diurnal variation to events. In our group, report of a previous abnormal EEG did not affect the likelihood of a diagnostic study (49.3% diagnostic vs 50.7% non-diagnostic). In female patients, a history of abuse (physical, emotional or sexual) or more than 5 positive systems on a 14-systems review was more likely to be associated with a non-diagnostic study (p=0.025 and 0.024 respectively). There was no statistically significant difference between the non-diagnostic and diagnostic groups with respect to the age (p=0.305), duration of disorder (p=0.748), and a past psychiatric history (p=0.221). Conclusions: Our results suggest that patients with non-convulsive events, or an event frequency of 4 or less per month are more likely to have non-diagnostic studies. Female patients with a history of abuse, or multiple positive systems on comprehensive review of systems were more likely to have non diagnostic studies. Funding: N/A