Authors :
Presenting Author: Aran Groves, MD, PhD – University of California, Los Angeles
Zachary Grinspan, MD, MS – Cornell Weill Medicine
Shaun A. Hussain, MD, MS – Division of Pediatric Neurology, Department of Pediatrics, UCLA Mattel Children's Hospital, David Geffen School of Medicine
Rationale:
Approximately one-third of children with Infantile Epileptic Spasms Syndrome (IESS) later meet diagnostic criteria for autism. It remains unclear whether this association reflects shared etiologic/genetic risk factors (e.g. tuberous sclerosis complex), or arises secondary to greater IESS disease burden (e.g., delayed diagnosis, longer duration of IESS, higher burden of seizure or epileptic encephalopathy, etc.). Prior efforts to explore these relationships have been limited by small cohort sizes. In this study, we leveraged electronic medical record (EMR) data to identify children with IESS and autism, and to extract standard treatment and response attributes. Methods:
Using single-center (UCLA) data housed within the Pediatric Epilepsy Learning Healthcare System (PELHS) infrastructure, we defined IESS as presence of an ICD diagnostic code and prescription of at least one standard therapy (prednisolone, ACTH, or vigabatrin). Autism was defined by ICD diagnostic code. Treatment response in children treated before 18 months was defined as an absence of (a) any second course of standard IESS therapy or (b) any new anti-seizure medication (ASM) prescribed within 3 months of the first IESS treatment. Children whose most recent follow-up occurred before age 3 were excluded to ensure sufficient observation time for autism assessment. Automated abstractions were validated by manual chart review. Results:
We identified 602 children with IESS; 220 were excluded for follow-up < 3 years, leaving 382 for analysis. Among these, 69 (18%) had an autism diagnosis. Validation in 86 IESS charts and 27 autism charts showed 98.9% and 96.2% concordance with ICD codes, respectively. The median latency from clinical IESS onset to standard therapy was 1.9 months (IQR 0.6–1.96 months). Of 82 children with verified standard therapy, response method 1 (i.e., second IESS therapy) identified 47 responders, method 2 (i.e., new ASM) identified 44, and both methods agreed in 25 cases. Chart review confirmed treatment outcomes in 60% of reviewed cases; 15% were discordant and 25% had incomplete data due to out-of-system therapies.Conclusions:
Automated EMR abstraction can reliably identify IESS and autism diagnoses and approximate key treatment parameters. However, precise dating of IESS onset and response still benefits from manual review or advanced NLP methods. Scaling this approach to multi-center datasets may clarify the mechanistic link between IESS severity and subsequent autism risk.Funding:
This study was supported by the John C. Hench Foundation.