Poverty, insurance, and region as predictors of epilepsy treatment among US adults

Abstract

Disparities in epilepsy treatment have previously been reported. In the current study, we examine the role of socioeconomic status, health insurance, place of residence, and sociodemographic characteristics in past-year visit to a neurology or epilepsy provider and current use of antiseizure medications. Multiple years of data were compiled from the National Health Interview Surveys, Sample Adult Epilepsy Modules. The sample (n = 1655) included individuals 18 years and older who have been told by a doctor to have epilepsy or seizures. Independent variables included number of seizures in the past year, health insurance, poverty status, education, region, race/ethnicity, foreign-born status, age, and sex/gender. Two sets of weighted hierarchical logistic regression models were estimated predicting past-year epilepsy visit and current medication use. Accounting for recent seizure activity and other factors, uninsured and people residing outside of the Northeast were less likely to see an epilepsy provider, and people living in poverty were less likely to use medications, relative to their comparison groups. However, no racial/ethnic and nativity-based differences in specialty service or medication use were observed. Further research, including longitudinal studies of care trajectories and outcomes, are warranted to better understand healthcare needs of people with epilepsy, in particular treatment-resistant seizures, and to develop appropriate interventions at the policy, public health, and health system levels.

Keywords: Epilepsy treatment; Health insurance; Healthcare services; National Health Interview Survey; Region; Specialist care.

Figure 1.

Differences in the predicted probabilities of epilepsy care between selected values of model covariates Notes: Predicted probabilities were calculated in postestimation following the logistic regression models for taking medicine or seeing a specialist (see Model 5 in Tables 1 and 2 for model coefficients). If the difference in probability is negative, then the reference group has a larger probability of the outcome. A difference in probabilities is not significant at .05 if its confidence interval is overlapping 0.