Annual Meeting Abstracts: View

<< Back to Search Results

(Abst. 2.012), 2019

Validating StatNet EEG in the ICU: Preliminary Results
Authors: Laura Gill, University of British Columbia; Farzad Moien Afshari, University of British Columbia; Christopher Uy, University of British Columbia; Shananjit Thiara, University of British Columbia; Manouchehr Javidan, University of British Columbia; Donald Griesdale, University of British Columbia; Mypinder Sekhon, University of British Columbia
Content: Rationale: Nonconvulsive status epilepticus (NCSE) is common in the ICU and requires EEG for diagnosis. At many instiutions, it is difficult to get an after-hours EEG. Using StatNet EEG could potentially reduce delays to getting EEGs and diagnosing NCSE as they can be applied by non-technologists. StatNet EEG has not been validated in the ICU setting.Our hypothesis:1. StatNet EEG is accurate compared to conventional EEG for detection of NCSE in the ICU.2. StatNet EEG will significantly reduce delay to EEG. Methods: Each StatNet electrode set is a one time peel-and-stick device. The F3, F4, P3, P4, Pz electrodes are not included, so the posterior electrodes are shifted to ensure adequate coverage of the posterior head. StatNet studies are performed by trained neurology residents after a 1-hour teaching session.Patients are identified at the time a conventional EEG is ordered using the following inclusion/ exclusion criteria:Inclusion criteria1.Patients ≥17 years of age in ICU2.EEG requested for:-Query seizures or encephalopathy3. Severely depressed mental status and one of:-Acute CNS insult-Toxic/ metabolic disturbance-Drug overdose-Hx of epilepsy-Witnessed seizure without return to baseline4. Any unexplained fluctuation in level of alertnessExclusion criteria1.Patients on doses of anesthetics known to cause burst suppression2.Patients with skull defectsWe plan to collect 30 studies total (15 of each StatNet and conventional.)The StatNet and conventional EEGs are deidentified, shortened to 20 mins and the montage is changed to the StatNet montage. Videos/ notes are deleted.  Readers are not allowed to change the montage. The studies are read in a random blinded fashion by two epileptologists.Intra rater agreement is then assessed with respect to presence of:-Epileptiform discharges-Electrographic seizures/ NCSEA kappa coefficient will be generated for assessment.Inter rater agreement will be assessed with respect to presence of:-Epileptiform discharges-Electrographic seizures/ NCSEA kappa coefficient will be generated for assessment.We will calculate if there is a significant difference between the start of the StatNet and conventional recordings using a wilcoxon signed-rank test. Results: We have collected 16 studies total (8 StatNet and 8 conventional)For Reader 1 there was good agreement with respect to the presence of epileptiform discharges and seizures/NCSE. For both: % agreement 100, Kappa 1.00, p=0.04. Reader 2 also had high % agreement with respect to the presence of epileptiform discharges and seizures/NCSE. However the kappa coefficients were zero given the low rate of positive studies.Epileptiform discharges: % agreement 87.5, Kappa 0Seizures/ NCSE: % Agreement 100, Kappa 0There was poor inter rater agreement with respect to both the presence of epileptiform discharges and seizures. Epileptiform discharges StatNet: % agreement 50.0, Kappa -0.23; p=0.053Epileptiform discharges conventional: % agreement 62.5, Kappa 0Seizures/ NCSE StatNet: % agreement 62.5, Kappa 0Seizures/ NCSE conventional:  % agreement 62.5, Kappa 0The difference between the start of the StatNet and conventional EEGs was significant: z-statistic 2.10, p=0.04.  Conclusions: StatNet is potentially reliable as the intra rater agreement was high between the StatNet and conventional EEGs. However, Reader 2 had kappa values of 0 given low number of studies felt to be positive for epileptiform discharges, seizures. StatNet EEG can be done faster than conventional EEG. The poor inter rater agreement was due to differences in distinguishing triphasic waves/ epileptiform discharges, which are known to be difficult to differentiate at times in ICU. Funding: Equipment will be purchased with support from UBC Internal Funding (G-Fund, Neurology Operational Fund.)
Figure 1
Figure 2