Abstracts

Rationale: The pace of technological innovation is rapidly changing the nature of epilepsy surgery. Although many centers are reporting decreases in traditional open resective surgery, some recent studies have disputed whether or not this is truly occurring. In addition, significant effort has been directed at increasing access to surgical epilepsy care to underrepresented minority groups and to increase access to healthcare more generally through implementation of the Affordable Care Act in 2014. As a result, we decided to analyze data from the Nationwide Inpatient Sample to critically examine trends in the volume and type of epilepsy surgery practiced in the United States over the last decade of available data (2006-2016). In addition, we aimed to use this data to examine what factors affected access to epilepsy surgery, specifically with regards to race and insurance payer status. Methods: The Nationwide Inpatient Sample (NIS) is the largest publicly available inpatient hospital database in the United States. It catalogues a stratified sample of 20% of the discharges from every hospital participating in the Health Care Utilization Project (HCUP), over 4,500 hospitals representing more than 97% of the population. We utilized data from 2006 to 2016, the most recent year available. Because this dataset spanned the transition from ICD-9 to ICD-10 coding, we used both sets of codes to identify discharges with intractable epilepsy (ICD-9 code 345.41, 345.51, ICD-10 codes G40.111, G40.119, G40.211, G40.219). Once the dataset was narrowed, we then identified discharges of various types of epilepsy surgery using ICD-9 and ICD-10 codes. This allowed us to initially classify a discharge as surgical or non-surgical. Once classified as surgical, we were then able to identify and classify surgeries as open resection, laser ablation, intracranial monitoring, vagal nerve stimulation and neurostimulation using ICD-9 and ICD-10 codes. Next, we used univariate analysis to examine the effects of a number of patient and hospital-based variables on access to epilepsy surgery. These included age, sex, race, insurance payer type, median income quartile of patient’s zip code, county type based on population, hospital bed size, hospital location, and hospital region. Because we assumed that many of these variables may be correlated, we ran a random forest machine learning algorithm to identify the variables that were most important in classifying a discharge as surgical. Once we identified these variables, we included them in a hierarchical mixed effects logistic regression model to examine the effect of these variables on surgical admissions. Odds ratios and 95% confidence intervals were calculated for each variable. We also split the discharges into three approximately equal time frames based on number of discharges (2006-2010, 2011-2014, 2014-2016) to examine changes in the effects of these variables over time. Results: Total annual admissions for intractable epilepsy have steadily increased (7,224 to 18,605/year) over the last ten years while annual surgical admissions increased from 2006 to 2010 (1,084 to 2,013/year) and have remained essentially stable since then (between 1,800 and 2,070/year). As a result, the percentage of all epilepsy discharges that included surgery has decreased from 15.0 to 11.1% over the same time period. This appears to be mainly due to a reduction in the number of open resections as the percentage of all epilepsy discharges with open resections has decreased from 13% to 6.6%. A variable importance plot using a random forest machine learning algorithm demonstrated that insurance payer, race, age, hospital location, county population and median income quartile of the patient’s zip code were the most important variables linked to surgical epilepsy admissions. On multivariate analysis, variables favoring epilepsy surgery were private insurance (OR 2.03, CI 1.73-2.39), living in a metropolitan county of 250-999,000 people (OR 1.44, CI 1.24-1.67 ), and urban teaching hospital location (OR 4.92, CI 2.21-14) while black race was associated with a lower likelihood of epilepsy surgery (OR 0.54, CI 0.45-0.65). When looking at trends in variables over time, more patients with Medicaid were being admitted for both non-surgical and surgical admissions but their likelihood of undergoing surgery did not change. On the other hand, black patients were more likely to undergo epilepsy surgery over time when split into three time cohorts (OR 0.38->0.44->0.60). Overall, 90.5% of all epilepsy admissions occur at urban, teaching hospitals but 96.3% of surgeries are performed there. Interestingly, an increasing percentage of epilepsy surgeries are occurring at urban, teaching hospitals such that in 2016, 99.5% of all epilepsy surgeries occurred there. 0.5% occurred at urban, non-teaching hospitals and there have been zero cases at rural hospitals for both 2015 and 2016. Conclusions: The number of inpatient epilepsy admissions continues to increase in the United States while the number of surgical epilepsy admissions has stabilized as the percentage of surgical admissions has decreased. This appears to be due to a decrease in the number of open resections performed over time. Although changes in healthcare law have likely improved patients’ access to care (increase in number of Medicaid patients), this has not improved their access to epilepsy surgery. Black patients continue to have the lowest likelihood of undergoing epilepsy surgery although race is the only variable to demonstrate improvement in access to epilepsy surgery over time. Further efforts need to be towards establishing epilepsy surgery registries to examine the impact of these changing variables on seizure freedom rates. Funding: MnDRIVE neuromodulation scholar