Abstracts

Rationale: Despite the prevalence and heterogeneity of epilepsy, there is limited national-level information on the quality of epilepsy care. This scarcity is largely due to the limited availability of population-wide databases and the uncertain validity of claims-based definitions of epilepsy. We examined the accuracy of a claims-based definition for generalized seizures and active epilepsy. Methods: We conducted a cross-sectional validation analysis using two sources of linked data: a) administrative claims (i.e., primary, secondary, tertiary ICD-10 diagnosis codes); b) electronic health records (EHR). In a tertiary academic epilepsy clinic six providers used an identical method of documentation (reference-standard) for seizure frequency and type (documented based on all relevant clinical, EEG, and imaging data) during the routine epilepsy care of all patients between January and April of 2019. We excluded patients that were employees (n<11), patients seen by providers that did not use the standardized documentation (n=44), and patients who had incomplete standardized documentation (n=48), resulting in a cohort of 123 patients.From their standardized seizure documentation, we classified patients based on: 1) seizure type (focal, generalized, unknown); 2) epilepsy diagnosis (yes, no, unsure); and 3) date of last seizure. We defined reference-standard generalized seizures as patients who had exclusively generalized onset seizures (i.e., patients with documentation of both focal and generalized onset seizure were classified with focal seizures). We defined reference-standard active epilepsy as patients who had at least one seizure within the past two years (from the date of the encounter). We first evaluated the accuracy of a claims-based definition for seizure type (i.e., generalized=G40.0X, G40.1X, or G40.2X; focal=G40.AX, G40.BX, G40.3X or G40.4X). We then verified the accuracy of a claims-based definition for active epilepsy (i.e., any of the following: G40.019; G40.119; G40.211; G40.219; G40.411; G40.419; G40.804). For each analysis, we compared the predictive performance of the aforementioned claims-based definitions against our reference-standard (i.e. EHR-based classification) using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUROC). Results: In this cohort of 123 patients, 118 (95.9%) had true epilepsy. The claims-based definition of seizure type correctly classified 81% of patients with true generalized epilepsy and 81.7% of patients with true focal epilepsy (AUROC=0.77, 95% CI 0.67-0.87; sensitivity=60.7%, 95% CI 40.6%-78.5%; PPV=81%, 95% CI 58.1%-94.6%; specificity=92.5%, 95% CI 81.8%-97.9%; NPV=81.7%, 95% CI 69.6%-90.5%). The claims-based definition of active epilepsy correctly identified 90.9% of patients with seizures within the past two years (AUROC=0.64, 95% CI 0.56-0.71; sensitivity=37.0%, 95% CI 26%-48.5%; PPV=90.9%, 95% CI 75.7%-98.1%; specificity=91.2%, 95% CI 76.3%-98.1%; NPV=37.8%, 95% CI 27.3%-49.2%). Conclusions: Claims-based definitions for generalized seizures and active epilepsy that used ICD-10 codes provided by epilepsy specialists differentiate generalized versus focal seizures and identify patients with active epilepsy with reasonable accuracy. Further validation of these definitions in different care settings is necessary. Funding: No funding