OBJECTIVES: To measure health care costs related to epilepsy severity using real-life claims data (Pharmetrics®, IMS, USA) in a representative sample from a U.S. commercially-insured population.  METHODS: The observation period ranged from January 2006-December 2007. Patients with at least two diagnoses of epilepsy before the observation period and at least two claims for anti-epileptic drugs (AEDs) during the observation period were included.  As Pharmetrics® does not report data on disease severity, the number of epilepsy-related emergency room visits over two years (0, 1, 2, ≥3) was used as a proxy.  Covariates included age, gender, region, epilepsy-type, number of co-morbidities, concomitant AEDs, and treatment duration.  Annualized costs were split into AED medications and non-AED costs.  Non-AED costs included non-AED medications and ‘other’ costs including emergency room visits, hospitalizations, and physician visits.  RESULTS: A total fo 9163 patients were included, with 14% in the most severe category (≥3 ER visits).  Total annualized costs ranged from US$6,000 to $33,000 depending on disease severity.  AED costs were not linked to severity; however ‘other costs’ increased disproportionately with disease severity.  In the unadjusted analysis, mean annualized AED, non-AED medication and ‘other’ costs were $2,513, $1,276 and $2,522, respectively, for those with no ER visits and $3,279, $3,457 and $26,270, respectively, for those with ≥3 ER visits.  The rise in ‘other’ costs with severity was mainly attributable to hospitalization costs.  In the adjusted analysis, the difference between AED and ‘other’ costs increased significantly with epilepsy severity, number of co-morbidities, and age, whereas it decreased with improved AED compliance. CONCLUSIONS: Non-AED treatment costs increased disproportionately with epilepsy severity, driven mainly by hospitalization. AED medication costs were not related to disease severity.  This analysis suggests cost savings may be achieved through targeted strategies and improvement of patient compliance in cases of severe epilepsy.