Clinical signatures of genetic epilepsy precede diagnosis in electronic medical records of 32,000 individuals

Abstract

An early genetic diagnosis can guide the time-sensitive treatment and care of individuals with genetic epilepsies. However, identification of a genetic cause often occurs long after onset of these disorders. Here, we aimed to identify early clinical features suggestive of genetic diagnoses in individuals with epilepsy by systematic large-scale analysis of clinical information from full-text patient notes in the electronic medical records (EMR). From the EMR of 32,112 individuals with childhood epilepsy, we retrieved 4,572,783 clinical notes spanning 203,369 total patient-years. A subcohort of 1,925 individuals had a known or presumed genetic epilepsy with 738 genetic diagnoses spanning 271 genes. We employed a customized natural language processing (NLP) pipeline to extract 89 million time-stamped standardized clinical annotations from free text of the retrieved clinical notes. Our analyses identified 47,641 clinical associations with a genetic cause at distinct ages prior to diagnosis. Notable among these associations were: SCN1A with status epilepticus between 9 and 12 months of age (P<0.0001, 95% CI=8.10-133); STXBP1 with muscular hypotonia between 6 and 9 months (P=3.4x10-4, 95% CI=3.08-102); SCN2A with autism between 1.5 and 1.75 years (P<0.0001, 95% CI=11.1-Inf); DEPDC5 with focal-onset seizure between 5.75 and 6 years (P<0.0001, 95% CI=12.8-Inf); and IQSEC2 with myoclonic seizure between 2.75 and 3 years (P=2.5x10 4, 95% CI=11.3-1.15x104). We also identified associations between clinical terms and gene groups. Variants in ion channel gating mechanisms were associated with myoclonus between 3 and 6 months of age (P<0.0001, 95% CI=5.23-24.2), and variants in calcium channel genes were associated with neurodevelopmental delay between 1.75 and 2 years (P<0.0001, 95% CI=4.8-Inf). Cumulative longitudinal analysis revealed further associations, including KCNT1 with migrating focal seizures from at 0 to 1.75 years (P<0.0001, 95% CI=96.8-4.50x1015). A neurodevelopmental abnormality presenting between 6 and 9 months of age was strongly associated with an individual having any genetic diagnosis (P<0.0001, 95% CI=3.55-7.42). The earliest features associated with genetic diagnosis occurred a median of 3.6 years prior to the median age of diagnosis. Latency to diagnosis was greater in older individuals (P<0.0001) and those who initially underwent less comprehensive genetic testing (P=5.5x10-3, 95% CI=1.23-3.35). In summary, we identified key clinical features that precede genetic diagnosis, leveraging EMR data at scale from a large cohort of individuals with genetic epilepsies. Our findings demonstrate that automated EMR analysis may assist clinical decision making, leading to earlier diagnosis and more precise prognostication and treatment of genetic epilepsies in the precision medicine era.