Authors :
Presenting Author: John Mytinger, MD – Nationwide Children's Hospital, The Ohio State University
Dara Albert, DO, MS – Nationwide Children's Hospital, The Ohio State University; Shawn Aylward, MD – Nationwide Children's Hospital, The Ohio State University; Christopher Beatty, MD – Nationwide Children's Hospital, The Ohio State University; Sonam Bhalla, MBBS, MD – Emory University/Children's Healthcare of Atlanta; Sonal Bhatia, MD – Medical University of South Carolina; Guy Brock, PhD – Biomedical Informatics and Center for Biostatistics, Biostatistics Resource at Nationwide Children's Hospital (BRANCH) – The Ohio State University; Purva Choudhari, MD – University of Texas Southwestern Medical Center; Micheal Ciliberto, MD – University of Iowa; Daniel Clark, MD, PhD – Nationwide Children's Hospital; Theresa Czech, MD – University of Iowa; Megan Fredwall, MD – Nationwide Children's Hospital; Ernesto Gonzalez-Giraldo, MD – University of California San Francisco; Chellamani Harini, MD – Boston Children’s hospital; Senyene Hunter, MD, PhD – University of North Carolina at Chapel Hill; Akshat Katyayan, MD – Texas Children's Hospital/ Baylor College of Medicine; Isaac Kistler, MS – Biomedical Informatics and Center for Biostatistics, Biostatistics Resource at Nationwide Children's Hospital (BRANCH) – The Ohio State University; Neil Kulkarni, MD – Nationwide Children's Hospital, The Ohio State University; Virginia Liu, MD, PhD – Children's Hospital of Orange County | University of California Irvine; Jennifer Madan Cohen, MD – Connecticut Children's/University of Connecticut; Corinne McCabe, MD – Nationwide Children's Hospital; Thomas Murray, DO – Nationwide Children's Hospital; Kerri Neville, MD – Mott Children's Hospital, University of Michigan; Shital Patel, MD – Duke University Medical Center; Spriha Pavuluri, MD – University of Nebraska Medical Center; Donald Phillips, MD, MPH – Children's Hospital of Orange County | University of California Irvine; Debopam Samanta, MD – University of Arkansas for Medical Sciences/Arkansas Children's Hospital; Amanda Sandoval, MD – University of Utah | Primary Children's Hospital; Deepa Sirsi, MD – University of Texas Southwestern Medical Center; Emily Spelbrink, MD, PhD – Stanford University; Carl Stafstrom, MD, PhD – Johns Hopkins University; Maija Steenari, MD – Children's Hospital of Orange County | University of California Irvine; Danielle Takacs, MD – Texas Children's Hospital/ Baylor College of Medicine; Tyler Terrill, MD – University of Texas Southwestern Medical Center; Jorge Vidaurre, MD – Nationwide Children's Hospital, The Ohio State University; Daniel Shrey, MD – Children's Hospital of Orange County | University of California Irvine
Rationale:
The best possible outcomes for children with infantile epileptic spasms syndrome (IESS) requires the cessation of infantile spasms as well as the resolution of any EEG features consistent with an epileptic encephalopathy (EE). While the presence of hypsarrhythmia is an important EEG marker of EE, many children with IESS do not have hypsarrhythmia and its determination has poor interrater reliability (IRR). The
Burden of
Amplitude
S and
Epileptiform
Discharges (BASED) score has shown promise for the interictal assessment of children with IESS with high levels of IRR between three reviewers at a single center using five minute EEG clips. Utilizing a multicenter network of EEG readers, we developed a formal BASED training program and utilized a post-training test to assess IRR among learners. The primary outcomes of this study were to evaluate the IRR of learners for determining the BASED score and the presence (BASED 4-5) or absence (BASED 0-3) of EE (+/-EE) in children with IESS. Levels of IRR for the final BASED score, +/-EE, and all BASED score elements were assessed by reader demographics.
Methods:
After standardized training (2021 BASED article review plus < 2 hours of discussion/EEG tracing review), 31 learners (seven trainees, 24 faculty physicians) assessed 12 studies (six pretreatment, six post-treatment) using a web-based application (Persyst Mobile version 1.2.15, Solana Beach, CA). The length of unmarked EEG epochs ranged from one to six hours. Gold standard scores (as determined by three expert reviewers, using a best 2/3 method) were used to determine the accuracy of learners (expert scores were not included in the IRR outcomes). All learner and expert reviewers were blind to clinical data for EEG review.
IRR values for various subsets were calculated using two-way random, single measures, absolute agreement using intraclass correlation coefficient (ICC) (ICC command of psych package in R, version 4.3.0). Reliability was designated as follows: < 0.5 = poor, 0.5-0.75 = moderate, 0.75-0.9 = good, and > 0.9 = excellent. Results:
The IRR among different EEG reader demographics are listed in the Table. For all readers, the IRR was
good for determining the final BASED score (ICC 0.86) and
moderate for +/-EE (ICC 0.60). For all readers, the IRR was
good for determining
> 50% 3 spike foci,
moderate for grouped multifocal spikes and paroxysmal voltage attenuations, and
poor for < 50% 3 spike foci, < 3 spike foci and
> 1 channel with abnormal high amplitude waves. This latter finding contrasts with the complete agreement (ICC 1.0) among three expert reviewers in the assessment of background amplitude, suggesting that experience is important for this element. Among all learners, 87% correctly identified +/-EE in
> 9/12 studies.
Conclusions:
Following a standardized training program, the IRR of all learners was
good for the final BASED score and
moderate for +/-EE. These findings support the use our training program to learn the BASED scoring method. The
poor IRR of some BASED elements, including the assessment of background amplitude, may improve with additional experience.
Funding:
Pediatric Epilepsy Research Foundation