E

Dc

NT

a

Ub

c

d

RA

f

0h

pilepsy Research (2013) 107, 172—180

j ourna l h om epa ge: www.elsev ier .com/ locate /ep i lepsyres

isparities in access to specialized epilepsyare

icholas K. Schiltza, Siran M. Koroukiana, Mendel E. Singera,homas E. Lovea,b,c, Kitti Kaiboriboond,∗

Department of Epidemiology & Biostatistics, Case Western Reserve University, Cleveland, OH,nited StatesDepartment of Medicine, CWRU at MetroHealth Medical Center, Cleveland, OH, United StatesCenter for Health Care Research and Policy, CWRU at MetroHealth Medical Center, United StatesDepartment of Neurology, University Hospitals Case Medical Center, Cleveland, OH, United States

eceived 20 April 2013 ; received in revised form 15 July 2013; accepted 4 August 2013vailable online 16 August 2013

KEYWORDSEpilepsy;Healthcare access;Healthcare disparities

SummaryObjective: To examine the impact of individual and community characteristics on access tospecialized epilepsy care.Methods: This retrospective cross-sectional study analyzed data from the California State Inpa-tient Sample, the State Ambulatory Surgery Database, and the State Emergency DepartmentDatabase, that were linked with the 2009 Area Resource File and the location of the NationalAssociation of Epilepsy Center’s epilepsy centers. The receipt of video-EEG monitoring wasmeasured and used to indicate access to specialized epilepsy care in subjects with persistentseizures, identified as those who had frequent seizure-related hospital admissions and/or ERvisits. A hierarchical logistic regression model was employed to assess barriers to high qualitycare at both individual and contextual levels.Results: Among 115,632 persons with persistent seizures, individuals who routinely receivedcare in an area where epilepsy centers were located were more likely to have access to spe-

cialized epilepsy care (OR: 1.81, 95% CI: 1.20, 2.72). Interestingly, the availability of epilepsycenters did not influence access to specialized epilepsy care in people who had private insur-ance. In contrast, uninsured individuals and those with public insurance programs includingMedicaid and Medicare had significant gaps in access to specialized epilepsy care. Other indi-vidual characteristics such as age, race/ethnicity, and the presence of comorbid conditionswere also associated with disparities in access to specialized care in PWE.

∗ Corresponding author at: 11100 Euclid Avenue, Lakeside 3200, Cleveland, OH 44106, United States. Tel.: +1 216 844 8799;ax: +1 216 844 3160.

E-mail address: Kitti.Kaiboriboon@UHhospitals.org (K. Kaiboriboon).

920-1211/$ — see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.eplepsyres.2013.08.003

Specialized epilepsy care 173

Conclusion: Both individual and community characteristics play substantial roles in access tohigh quality epilepsy care. Policy interventions that incorporate strategies to address disparities

to improve access to specialized care for PWE.s reserved.

(DcAfSdvtdp

rppaQtrbi2iifwt

pifHtoFw

4tstmi

S

IClassification of Diseases, Ninth Revision, Clinical Modifica-tion (ICD-9-CM) code for epilepsy (ICD-9-CM: 345.xx), or 2 ormore occurrences, at least 30 days apart, of ICD-9-CM codes

at both levels are necessary

© 2013 Elsevier B.V. All right

Introduction

Disparities in access to specialized care pose significant chal-lenges to the treatment and quality of care improvement inpersons with epilepsy (PWE) (Institute of Medicine, 2012a).Previous studies provide evidence that gaps in access to com-prehensive high-quality epilepsy care exist in people withlow socioeconomic status, and in racial minority popula-tions (Kelvin et al., 2007; Begley et al., 2009; Burneo et al.,2009). In addition, insurance status has also been found tobe important factor that can influence access to specializedcare in PWE (McClelland et al., 2007, 2010; Begley et al.,2009; Halpern et al., 2011; Baca et al., 2013). So far, very lit-tle attention has been paid to the role of contextual factorsin the variation of access to epilepsy care. Several studies inother disease entities have shown that characteristics of thearea of residence such as socioeconomic status of the neigh-borhood, degree of urbanization, availability of resourcesand services including provider availability and managedcare penetration also play an important role in access tospecialty care (Gresenz et al., 2000; Diez Roux et al., 2001;Litaker et al., 2005). A recent study found that geographiclocation of residence influenced the timing of referral forepilepsy surgery evaluation (Hauptman et al., 2013). As theavailability of specialized epilepsy care is mostly limited tourban large medical centers (Schiltz et al., 2013), compre-hensive analysis of the influence of contextual and individualattributes is a vital step for planning effective policies toreduce disparities in access to care for PWE.

The objective of this study was to examine the natureof multilevel relationships in accessing specialized epilepsycare in PWE. We used the receipt of video-EEG (VEEG) moni-toring as an indicator for access to specialized epilepsy careand examined barriers to VEEG monitoring at both individualand contextual levels. VEEG monitoring is an important diag-nostic test that is commonly used to confirm a diagnosis ofa seizure disorder and to classify seizure type, especially inpersons with persistent seizures (Cascino, 2002). Moreover,determination of candidacy for epilepsy surgery requiresVEEG monitoring. It, therefore, is likely that most, if notall, of individuals with persistent seizures would have under-gone VEEG monitoring, had they had access to specializedepilepsy care.

Materials and methods

The study protocol was approved by Institutional ReviewBoard at Case Western Reserve University.

Data sources

We performed a retrospective cross-sectional study using2005—2009 data from the California State Inpatient Sam-ple (SID), the California State Ambulatory Surgery Database

fcir

SASD), and the California State Emergency Departmentatabase (SEDD). These datasets are part of the Health-are Cost and Utilization Project (HCUP) sponsored by thegency for Healthcare Research and Quality (AHRQ) (Agencyor Healthcare Research and Quality, 2011). The HCUP SID,ASD, and SEDD provide complete information on all hospitalischarges, ambulatory surgeries, and emergency room (ER)isits. Each record contains patient demographic informa-ion, hospital and county identifier, diagnoses, procedures,ischarge status, length of stay, as well as total charges andayment source.

Since an individual record in the HCUP State databasesepresents one discharge abstract rather than an individualatient, we used the HCUP revisit variables to link multi-le records belonging to the same individual across facilitiesnd hospital settings (Agency for Healthcare Research anduality, 2012), thus making it possible to analyze the data athe individual level, rather than at the encounter level. Theevisit variables are states and year specific, and can onlye created for states that provide unique encrypted patientdentifier (Agency for Healthcare Research and Quality,012). In 2005, the revisit variables for all HCUP state datancluding SID, SASD, and SEDD were available in 4 statesncluding California, Florida, Nebraska, and Utah (Agencyor Healthcare Research and Quality, 2013). California dataere readily available to us, and therefore were used for

he analysis of this study.In addition, we used the Area Resource File (ARF), which

rovided a comprehensive collection of contextual data,ncluding socioeconomic and environmental characteristicsor each county within the US (US Department of Health anduman Services, 2011). ARF 2009 data elements were linkedo the California SID, SASD, and SEDD based on the locationf hospital and/or ambulatory centers by a common 5-digitederal Information Processing Standards (FIPS) code thatas unique to each US county.

We also obtained the address of all Level 3 and Level epilepsy centers (ECs) in the State of California fromhe National Association of Epilepsy Center (NAEC) web-ite (http://www.naec-epilepsy.org/find.htm) and linkedhe location of ECs to other socioeconomic and environ-ental variables in the ARF by matching the county of each

ndividual center to a corresponding FIPS code.

tudy population

ndividuals with at least 1 occurrence of the International

or convulsion (ICD-9-CM: 780.39) were included. Theseriteria have been shown to have high degree of accuracy fordentifying epilepsy cases from facility-based administrativeecords (Jette et al., 2010; Faught et al., 2012). We further

1

lsrpttf2b(o

O

Tb

I

Dt

aabpaMipiaodeMespn

Fc

74

imited our study population to individuals with persistenteizures, identified as those who had frequent (≥5) seizure-elated hospital admissions and/or ER visits during the studyeriod. This operational definition was designed based onhe Institute of Medicine report that patients with persis-ent seizures require a referral to more advance level of careor further diagnosis and treatment (Institute of Medicine,012a). The sensitivity analysis in subjects with higher num-er of seizure-related hospital admissions and/or ER visits≥10 and ≥15) was also performed to test the robustness ofur findings. Individuals over age 65 were excluded.

utcomes

he outcome was the receipt of VEEG monitoring, identifiedy the presence of primary ICD-9-CM procedure codes 89.19.

ndividual level characteristics

ata on all individual level variables were obtained fromhe California SID, SASD, and SEDD. Age was categorized

hplK

igure 1 Location of Level 3 and Level 4 epilepsy centers in the Staenters by counties.

N.K. Schiltz et al.

s children (0—17 years), young adults (18—34 years), anddults (35—64 years). Race/ethnicity was classified as white,lack, Hispanic, and other races. Based on the source ofayment documented in hospital discharge records, insur-nce status was categorized as private (or commercial),edicaid, Medicare, uninsured, and others. In the US, Med-

caid and Medicare programs are public health insurancerograms funded by state and federal governments. Med-caid, the largest health insurance program in the US, is

program for low-income and medically vulnerable groupsf people including children, pregnant women, adults withependent children, people with severe disabilities, and thelderly (Schneider and Elias, 2002). In order to qualify foredicaid, a person must belong to one of the Medicaid’s cat-gorically eligible groups and meet financial criteria. Eachtate determines its eligibility criteria for Medicaid program,articularly the minimum income threshold and the defi-ition of disability (Schneider et al., 2000). Medicare is a

ealth insurance program for individuals older than 65, peo-le suffering from end-stage renal disease, or amyotrophicateral sclerosis, regardless of their income (The Henry J.aiser Family Foundation, 2010). Persons younger than 65

te of California and classification of the availability of epilepsy

Ncd

S

Wiaaus(upBv(watI

Specialized epilepsy care

can also be qualified for Medicare benefits if they havepermanent disabilities and receive Social Security Disabil-ity Income for at least 2 years (The Henry J. Kaiser FamilyFoundation, 2010). Some low-income seniors and personswith disabilities receive both Medicaid and Medicare ben-efits. Medicare is the primary source of coverage for thesedual eligible beneficiaries (Jacobson et al., 2012). We usedElixhauser comorbid measurement (Elixhauser et al., 1998)to identify the patient’s comorbid conditions.

Contextual level characteristics

The contextual characteristics were obtained at the countylevel from the ARF including the proportion of residents liv-ing in poverty, percentage with a 4-year college degree,urban/rural designation, and whether the county was des-ignated as a high unemployment area as defined by theNational Bureau of Economic Research. We matched eachindividual patient to a county based on the location of thehospital and/or ambulatory centers (s)he attended mostoften. For patients with no regular source of care (1.7%),the location of the first hospital or ambulatory centers vis-

ited was used. We classified counties where the patientsreceived their medical care into 3 mutually exclusive groupsbased on the availability of Level 3 or Level 4 ECs (Fig. 1).EC counties were home to at least 1 Level 3 or Level 4 ECs.

mftd

Table 1 Individual level characteristics of all adults with persistecare in California between 2005 and 2009.

Study population Sub

n (%) n (%

Overall 115,632 (100%) 3692

GenderMale 58,728 (51%) 153Female 56,686 (49%) 206

Age0—17 15,934 (14%) 8418—34 25,758 (22%) 11535—64 73,940 (64%) 169

RaceWhite 63,841 (55%) 209Black 18,852 (16%) 3Hispanic 27,183 (24%) 90Other 5347 (5%) 2Missing 409 (0%) 1

Insurance statusPrivate 28,126 (24%) 163Medicare 27,439 (24%) 71Medicaid 42,276 (37%) 112Uninsured 11,828 (10%)

Others 5963 (5%) 1

Comorbid conditionsNone 60,584 (52%) 237One 25,082 (22%) 82Two or more 29,966 (26%) 49

175

earby counties shared a geographic border with an ECounty, while the remaining counties were categorized asistant.

tatistical analysis

e used hierarchical logistic regression to examine thempact of individual and contextual characteristics onccess to specialized epilepsy care. This multilevel analyticpproach accounts for shared characteristics of individ-als within the same clusters or units, and provides robusttandard errors and reduces the potential for type I errorBarcikowski, 1981). Hierarchical models, therefore, areseful when the associations of interest arise from multi-le sources or within organizational units (Raudenbush andryk, 2002). We first analyzed the association between indi-idual level factors and the receipt of VEEG monitoringTable 3; Individual model). All individual characteristicsere retained for the second analysis, which assessed thessociation between the presence of epilepsy centers andhe receipt of VEEG monitoring (Table 3; Hierarchical model). Our third model kept the covariates from the first 2

odels and then added social and economic community

actors (Table 3; Hierarchical model II). In addition, weested several a priori cross-level interactions to exploreifferential relationships between individual and contextual

nt seizures, and those who had access to specialized epilepsy

jects with access to specialized epilepsy care

) Row percent

(100%) 3.2%

5 (42%) 2.6%4 (56%) 3.6%

6 (23%) 5.3%5 (31%) 4.5%1 (46%) 2.3%

1 (57%) 3.3%38 (9%) 1.8%3 (24%) 3.3%33 (6%) 4.4%27 (3%) 31.1%

7 (44%) 5.8%7 (19%) 2.6%1 (30%) 2.7%87 (2%) 0.7%30 (4%) 2.2%

6 (64%) 3.9%6 (22%) 3.3%0 (13%) 1.6%

1

cspe(an

R

OsipuulVaa

sLooTau

dami

pmab(tnct

ao9fli(ts

ovn

D

Tip

76

haracteristics. Specifically, interactions between insurancetatus and all contextual level variables, and betweenroximity to ECs and all individual level variables werexamined. We used version 9.2 of the SAS system for UNIXSAS Institute Inc., Cary, NC) for all descriptive statisticsnd HLM version 7 for Windows (Scientific Software Inter-ational, Lincolnwood, IL) for hierarchical analyses.

esults

ur study included 115,632 individuals with persistenteizures, among whom 3692 subjects underwent VEEG mon-toring during the study period. Table 1 describes our studyopulation by various individual characteristics. In partic-lar, the rates of VEEG monitoring in patients who wereninsured or who had public insurance were significantlyower than the rates for patients who had private insurance.EEG monitoring was least common among blacks, males,nd those with multiple comorbid conditions. In addition,dults (age 35—64) had the lowest rates of VEEG monitoring.

Table 2 summarizes the contextual characteristics of thetudy population. There were 6 counties that had Level 3 orevel 4 ECs in the State of California (Fig. 1). The majorityf persons with persistent seizures (87%) in our study lived inr adjacent to the counties where these ECs were located.hey also lived in the metropolitan area and in counties with

high proportion of people with college education and lownemployment.

Multiple logistic regression on the individual level pre-

ictors revealed that several factors were associated withccess to specialized epilepsy care (Table 3, Individualodel). Patients who were uninsured or publicly insured

ncluding Medicaid and Medicare were less likely than

imlt

Table 2 Contextual characteristics of all adults with persistent sein California between 2005 and 2009.

County characteristics Number of counties Study pop

Total 58 115,632 (

Location of epilepsy centersWithin the county 6 59,993Adjacent to the county 21 38,512Far from the county 31 17,127

Poverty level of the communityUnder state median 29 55,520Over state median 29 60,112

Proportion of individuals with 4 year college educationLowest quartile 15 11,076Middle quartile 28 67,172Highest quartile 15 37,384

EmploymentHigh unemployment 18 14,523Low unemployment 40 101,109

Urban/Rural ClassMetro 37 111,959Non-metro or Rural 21 367

N.K. Schiltz et al.

atients who were privately insured to have access to VEEGonitoring. Gaps in access to specialized epilepsy care were

lso noted among blacks, Hispanics, and those with comor-id conditions. Age also played a role. Younger patientschildren and young adults) were more likely to have accesso specialized epilepsy care compared to adults. We foundo change in significance or direction of individual leveloefficients after the addition of contextual level charac-eristics.

Moving to our contextual covariates, receiving care inn EC county was associated with an increased likelihoodf the receipt of sophisticated epilepsy care (OR: 1.81,5% CI: 1.20, 2.72) (Table 3, Hierarchical model I). We alsoound that none of the other county level covariates such asevel of poverty, education, and unemployment had signif-cant impact on the access to specialized care among PWETable 3; Hierarchical model II). Tests of pairwise combina-ions of individual and contextual levels did not show anyignificant cross-level interactions.

The sensitivity analysis in individuals with higher numberf seizure-related hospital admissions and/or ER visits pro-ided essentially similar results as the primary analysis (dataot shown).

iscussion

his is the first population-based study that explores thempact of individual and community characteristics on dis-arities in access to specialized care in PWE. We showed that

ndividuals with persistent seizures who routinely receivededical services in the area where the epilepsy centers were

ocated tended to have access to high quality epilepsy carehan those who received care elsewhere. The availability of

izures and those who had access to specialized epilepsy care

ulation (%) Subjects with access to specialized care (%)

100%) 3692 (100%)

(52%) 2418 (65%) (33%) 831 (23%) (15%) 443 (12%)

(48%) 1778 (48%) (52%) 1914 (52%)

(10%) 246 (7%) (58%) 2101 (57%) (32%) 1345 (36%)

(13%) 328 (9%) (87%) 3364 (91%)

(97%) 3606 (98%)3 (3%) 86 (2%)

Specialized epilepsy care 177

Table 3 Adjusted odds ratio of individual and contextual characteristics on the access to specialized epilepsy care based onthe Andersen behavioral model of health service utilization (Andersen, 2008).

Individual model OR (95% CI) Hierarchical model I OR (95% CI) Hierarchical model II OR(95% CI)

Predisposing factorsGender

Female 1.34 (1.25—1.43)* 1.34 (1.25—1.43)* 1.34 (1.25—1.43)*

Age0—17 2.42 (2.20—2.66)* 2.42 (2.20—2.66)* 2.42 (2.20—2.66)*

18—34 2.04 (1.88—2.22)* 2.04 (1.88—2.22)* 2.04 (1.88—2.22)*

35—64 Reference Reference Reference

RaceBlack 0.54 (0.48—0.61)* 0.54 (0.47—0.60)* 0.54 (0.47—0.60)*

Hispanic 0.87 (0.80—0.95)** 0.87 (0.80—0.95)** 0.87 (0.80—0.95)**

Other race 1.03 (0.89—1.19) 1.02 (0.89—1.18) 1.02 (0.89—1.18)White Reference Reference Reference

Enabling factorsInsurance status

Medicaid 0.49 (0.45—0.53)* 0.49 (0.45—0.53)* 0.49 (0.45—0.53)*

Medicare 0.69 (0.63—0.76)* 0.69 (0.63—0.76)* 0.69 (0.63—0.76)*

Uninsured 0.14 (0.11—0.18)* 0.14 (0.11—0.18)* 0.14 (0.11—0.18)*

Private Reference Reference Reference

Need factorsComorbid conditions

One 0.90 (0.83—0.98)** 0.90 (0.83—0.98)** 0.90 (0.83—0.98)**

Two or more 0.50 (0.45—0.55)* 0.50 (0.45—0.55)* 0.50 (0.45—0.55)*

None Reference Reference Reference

Health system factorsLocation of epilepsy centers

In the county 1.81 (1.20—2.72)** 1.49 (0.97—2.29)In nearby county 1.05 (0.77—1.43) 0.95 (0.70—1.29)Far from the county Reference Reference

Social and economic factorsPoverty level of the community

Higher than mediana 0.78 (0.54—1.14)

Proportion of individuals with 4 year college educationHighest quartile 1.36 (0.69—2.68)Middle quartiles 1.14 (0.65—2.00)Lowest quartile Reference

EmploymentHigh unemployment 1.16 (0.68—1.97)

Abbreviations: CI: Confidence interval; OR: Odds ratio.a Reference categories: Poverty level of the community below the state median.* Level of statistical significance: P ≤ 0.001.

c2aa

b

** 0.001 ≤ P ≤ 0.05.

ECs, therefore, plays an important role in access to special-ized care in PWE. Simply providing health insurance to theuninsured without expansion of ECs is unlikely to eliminateor reduce disparities. Actions in the form of policy changes,such as an expansion of the Medicaid program (Baicker et al.,2013), an implementation of health insurance exchange

program that subsidizes the purchase of private insurancefor low-income people (Rosenbaum and Sommers, 2013),an improvement of reimbursement policies (Cunninghamand Nichols, 2005), and an enactment of effective care

apta

oordination particularly specialty referral (Bodenheimer,008) are needed to ensure that PWE particularly those whore uninsured and those on Medicaid and/or Medicare haveccess to the highly effective epilepsy management.

Impact of insurance status on access to care in PWE haseen studied very rarely. A survey of health care utilization

t 4 epilepsy clinics that serve sociodemographically diverseopulations shows that individuals with public insurance andhose who are uninsured had lower rates of specialty carend higher rates of hospitalizations than privately insured

1

iiniiawftw(pmYaectNcpkatretcce

ctatslvtisob

iotccthacirteabf

f

rIosea

tspaMpHhpnMsqgc

silgcttlipw(pse

uatcsttpoctpiVeTia

78

ndividuals (Begley et al., 2009). An analysis of the Med-cal Expenditure Panel Survey (MEPS), which provides aationally representative sample of the US civilian non-nstitutionalized population, also suggests that uninsuredndividuals with epilepsy have significantly fewer physiciannd ER visits but greater costs for medications than thoseho are privately insured (Halpern et al., 2011). Moreover,

ew studies that examine the receipt of epilepsy surgery findhat privately insured individuals are more likely than thoseho are uninsured or publicly insured to undergo surgery

McClelland et al., 2007, 2010). Public insurance programs,articularly Medicaid, are designed to provide access toedical care for those who otherwise would be uninsured.

et, people who carry this most basic form of health insur-nce continue to face barriers to specialized care (Goldt al., 2006; Bisgaier and Rhodes, 2011). The reason for inac-essibility to high quality care in Medicaid enrollees appearso be multifactorial. Low reimbursement (Cunningham andichols, 2005), delay in payment, difficulty of payment pro-edures (Berman et al., 2002), size of community whereractices are located, underlying providers’ attitudes, andnowledge about Medicaid programs (Margolis et al., 1992)re important determinants of physicians’ decisions to par-icipate in Medicaid. As a result, Medicaid programs oftenely on a network of a small number of specialists (Landont al., 2007; Bisgaier and Rhodes, 2011). There is little doubthat Medicaid coverage improves access to general medicalare for PWE (Halpern et al., 2011). However, access to spe-ialized high quality care in this indigent population remainsxtremely limited.

Studies in other disease entities usually show that Medi-are as well as privately insured patients are more likelyo receive better care than Medicaid patients or those whore uninsured (Liu et al., 2006; LaPar et al., 2012). In con-rast, Medicare beneficiaries with epilepsy are found to haveignificantly lower likelihood of specialist visits and higherikelihood of visits with general physicians, compared to pri-ately insured individuals (Begley et al., 2009). We confirmhat having Medicare does not guarantee access to high qual-ty epilepsy management. However, since we limited ourtudy population to individuals younger than 65, the resultsf this study can be applied only to the non-elderly Medicareeneficiaries.

Our findings highlight the significance of racial disparitiesn accessing specialized epilepsy care. We showed that notnly blacks but also Hispanics with epilepsy were less likelyhan their white counterparts to have access to specializedare. The reasons for racial disparities in access to care areomplex and likely influenced by several factors related tohe patients themselves (e.g., attitudes toward risk, theealth care system, and care-seeking behavior) (Zuvekasnd Taliaferro, 2003), as well as those related to physi-ians (e.g., communication skills, quantity and quality ofnformation provided) (Griggs and Engel, 2005). An inverseelationship between age and comorbid conditions in accesso epilepsy care has been shown previously (Navaneethant al., 2007; Mukherjee et al., 2010). As a large number ofging population increases, the impact of age and comor-

idities on access to high quality care in PWE deservesurther study.

Measuring access to care is complex. In this study, weocused mainly on the receipt of VEEG monitoring, which

astr

N.K. Schiltz et al.

epresented merely one aspect of specialized epilepsy care.n addition, very few individuals may undergo surgery with-ut VEEG monitoring. We, therefore, could have missed amall group of people who in fact had access to specializedpilepsy care. Exclusion of these subjects was unlikely toffect our findings.

Whether our findings can be generalized to other states orhe entire US is arguable. California is a large state that haseveral unique characteristics, including a substantial pro-ortion of ethnic minorities, especially Asian and Hispanic,nd a high number of uninsured individuals (Fronstin, 2012).oreover, a large presence of integrated delivery systems,articularly Kaiser Permanente, and a high penetration ofealth Maintenance Organization (HMO) in California mightave augmented access to specialized epilepsy care amongrivately insured individuals through better care coordi-ation and referral. Since each state operates its ownedicaid program, the variation of Medicaid policies at a

tate level, leading to significant differences in access anduality of care among Medicaid enrollees in different geo-raphic regions (Subramanian and Chen, 2013), must bearefully considered.

Our contextual-level findings should be interpreted withome caution. Defining proximity to an EC at the county levels somewhat crude, especially for areas where counties arearge such as the southern part of California. More detailedeographic data at the patient level such as address or zipode would allow for a better measure of proximity. In addi-ion, smaller geographic units would also yield more powero detect effects at the contextual level, as our power wasimited by the number of counties. Nonetheless, our find-ngs were in line with a recent study that assessed a referralattern among children with refractory epilepsy who under-ent surgery at the University of California, Los Angeles

Hauptman et al., 2013). Hauptman et al. (2013) found thatatients from Los Angeles County had shorter interval fromeizure onset to referral, compared to those from the West-rn US.

Limitations, especially those that are inherent to studiessing claims data, should also be considered. The SID, SASD,nd SEDD datasets do not contain detailed clinical informa-ion or prescribed medications. The accuracy of epilepsyase ascertainment, therefore, is mainly based on diagno-is codes without other information. Recent studies showedhat the use of diagnosis codes to identify PWE in adminis-rative data generally had high sensitivity and good positiveredictive values (Kee et al., 2012; Reid et al., 2012). More-ver, our operational definition of epilepsy was designed toapture subgroup of people with persistent seizures, ratherhan people with refractory epilepsy. This particular group ofeople with frequent seizures should be referred to special-zed epilepsy centers for further investigation, particularlyEEG monitoring. Some of them might turn out to have non-pileptic or psychogenic spells after a thorough work-up.he robustness of our findings was confirmed by sensitiv-

ty analysis in subjects with a higher number of admissionsnd/or ER visits.

Besides the accuracy of case definition, coding errors are

lso important and could potentially affect the results of thistudy. Nonetheless, such errors are likely diluted (owing tohe large sample) and not likely to strongly bias the overallesults. It is also important to note that we identified the

B

B

B

B

B

C

C

D

E

F

F

G

G

G

H

H

I

I

J

Specialized epilepsy care

location of specialized epilepsy centers based on the mem-bership of the NAEC. The NAEC membership is voluntaryand the classification of Level 3 and Level 4 ECs are self-designated without external evaluation to assess whetherthe NAEC’s ECs actually provide the level of care that theyreport (Institute of Medicine, 2012b). However, the NAECis the only organization that defines the criteria for andcollects the data on comprehensive epilepsy managementin the US. Many hospitals that provide specialized epilepsycare, such as the Department of Veteran Affairs (VA) EpilepsyCenters of Excellence and the Kaiser Foundation Hospitalsmay not join the NAEC and, therefore, are not included inthe contextual level analysis of this study. Our recent reporton trends in pre-surgical evaluations and epilepsy surgery inthe US showed that >90% of VEEG monitoring are still per-formed in large academic medical centers (Schiltz et al.,2013), most of which are likely to be Level 4 ECs of theNAEC. Nevertheless, an inclusion of hospitals that providecomprehensive epilepsy care, like the VA Epilepsy Center ofExcellence and the Kaiser Foundation Hospitals, in our con-textual level analysis would likely strengthen the associationbetween the availability of ECs and the access to specializedepilepsy management.

Acknowledgements

Mr. Schiltz is supported by the Agency for HealthcareResearch and Quality (AHRQ) T32 Institutional TrainingGrant, #5T32HS000059-18. Dr. Kaiboriboon is supported bythe Epilepsy Foundation. This study was also supportedin part by the Case Western Reserve University/ClevelandClinic CTSA Grant Number UL1 RR024989 from the NationalCenter for Research Resources (NCRR), a component of theNational Institutes of Health and NIH roadmap for MedicalResearch. Its contents are solely the responsibility of theauthors and do not necessarily represent the official view ofNCRR or NIH.

References

Agency for Healthcare Research and Quality, 2013. HCUP Supple-mental Variables for Revisit Analyses, Retrieved (4.7.2013) fromhttp://www.hcup-us.ahrq.gov/toolssoftware/revisit/revisit.jsp#user

Agency for Healthcare Research and Quality, 2012. HCUP Supple-mental Files for Revisit Analyses, Retrieved (4.7.2013) fromwww.hcup-us.ahrq.gov/toolssoftware/revisit/revisit.jsp

Agency for Healthcare Research and Quality, 2011. HCUPDatabases, Retrieved July (4.7.2013) from http://www.hcup-us.ahrq.gov/databases.jsp

Andersen, R.M., 2008. National health surveys and the behavioralmodel of health services use. Med. Care 46, 647—653.

Baca, C.B., Vickrey, B.G., Vassar, S., Hauptman, J.S., Dadour, A.,Oh, T., Salamon, N., Vinters, H.V., Sankar, R., Mathern, G.W.,2013. Time to pediatric epilepsy surgery is related to diseaseseverity and nonclinical factors. Neurology 80, 1231—1239.

Baicker, K., Taubman, S.L., Allen, H.L., Bernstein, M., Gruber, J.H.,Newhouse, J.P., Schneider, E.C., Wright, B.J., Zaslavsky, A.M.,Finkelstein, A.N., Carlson, M., Edlund, T., Gallia, C., Smith, J.,

2013. The Oregon experiment — effects of Medicaid on clinicaloutcomes. N. Engl. J. Med. 368, 1713—1722.

Barcikowski, R.S., 1981. Statistical power with group mean as theunit of analysis. J. Educ. Behav. Stat. 6, 267—285.

J

179

egley, C.E., Basu, R., Reynolds, T., Lairson, D.R., Dubinsky, S.,Newmark, M., Barnwell, F., Hauser, A., Hesdorffer, D., Her-nandez, N., Karceski, S.C., Shih, T., 2009. Sociodemographicdisparities in epilepsy care: results from the Houston/NewYork City health care use and outcomes study. Epilepsia 50,1040—1050.

erman, S., Dolins, J., Tang, S.F., Yudkowsky, B., 2002. Factors thatinfluence the willingness of private primary care pediatriciansto accept more Medicaid patients. Pediatrics 110, 239—248.

isgaier, J., Rhodes, K.V., 2011. Auditing access to specialty care forchildren with public insurance. N. Engl. J. Med. 364, 2324—2333.

odenheimer, T., 2008. Coordinating care — a perilous jour-ney through the health care system. N. Engl. J. Med. 358,1064—1071.

urneo, J.G., Jette, N., Theodore, W., Begley, C., Parko, K., Thur-man, D.J., Wiebe, S., 2009. Disparities in epilepsy: report ofa systematic review by the North American Commission of theInternational League Against Epilepsy. Epilepsia 50, 2285—2295.

ascino, G.D., 2002. Clinical indications and diagnostic yieldof video-electroencephalographic monitoring in patients withseizures and spells. Mayo Clin. Proc. 77, 1111—1120.

unningham, P.J., Nichols, L.M., 2005. The effects of medicaidreimbursement on the access to care of medicaid enrollees: acommunity perspective. Med. Care Res. Rev. 62, 676—696.

iez Roux, A.V., Merkin, S.S., Arnett, D., Chambless, L., Mass-ing, M., Nieto, F.J., Sorlie, P., Szklo, M., Tyroler, H.A., Watson,R.L., 2001. Neighborhood of residence and incidence of coronaryheart disease. N. Engl. J. Med. 345, 99—106.

lixhauser, A., Steiner, C., Harris, D.R., Coffey, R.M., 1998. Comor-bidity measures for use with administrative data. Med. Care 36,8—27.

aught, E., Richman, J., Martin, R., Funkhouser, E., Foushee, R.,Kratt, P., Kim, Y., Clements, K., Cohen, N., Adoboe, D., Knowl-ton, R., Pisu, M., 2012. Incidence and prevalence of epilepsyamong older US Medicare beneficiaries. Neurology 78, 448—453.

ronstin, P., 2012. California’s uninsured: treading water, Retrieved(12.7.2013) from http://www.chcf.org/∼/media/MEDIA%20LIBRARY%20Files/PDF/C/PDF%20CaliforniaUninsured2012.pdf

old, M., Kuo, S., Taylor, E.F., 2006. Translating research to action:improving physician access in public insurance. J. Ambul. CareManage. 29, 36—50.

resenz, C.R., Stockdale, S.E., Wells, K.B., 2000. Communityeffects on access to behavioral health care. Health Serv. Res.35, 293—306.

riggs, J.J., Engel, J.J., 2005. Epilepsy surgery and the racialdivide. Neurology 64, 8—9.

alpern, M.T., Renaud, J.M., Vickrey, B.G., 2011. Impact of insur-ance status on access to care and out-of-pocket costs for U.S.individuals with epilepsy. Epilepsy Behav. 22, 483—489.

auptman, J.S., Dadour, A., Oh, T., Baca, C.B., Vickrey, B.G.,Vassar, S.D., Sankar, R., Salamon, N., Vinters, H.V., Math-ern, G.W., 2013. Sociodemographic changes over 25 years ofpediatric epilepsy surgery at UCLA. J. Neurosurg. Pediatr. 11,250—255.

nstitute of Medicine (IOM), 2012a. Epilepsy Across the Spectrum:Promoting Health and Understanding. The National AcademicPress, Washington, DC.

nstitute of Medicine (IOM), 2012b. Health care: quality, access, andvalue. In: England, M.J., Liverman, C.T., Schultz, A.M., Straw-bridge, L.M. (Eds.), Epilepsy across the spectrum: Promotinghealth and understanding. The National Academy Press, Wash-ington DC, pp. 113—161.

acobson, G., Neuman, T., Damico, A., 2012. Medicare’s rolefor dual eligible beneficiaries, Retrieved (July 4.7.2013) fromhttp://kaiserfamilyfoundation.files.wordpress.com/2013/01/

8138-02.pdf

ette, N., Reid, A.Y., Quan, H., Hill, M.D., Wiebe, S., 2010. Howaccurate is ICD coding for epilepsy? Epilepsia 51, 62—69.

1

K

K

L

L

L

L

M

M

M

M

N

R

R

R

S

S

S

S

T

U

80

ee, V.R., Gilchrist, B., Granner, M.A., Sarrazin, N.R., Carnahan,R.M., 2012. A systematic review of validated methods for iden-tifying seizures, convulsions, or epilepsy using administrativeand claims data. Pharmacoepidemiol. Drug Saf. 21 (Suppl. (1)),183—193.

elvin, E.A., Hesdorffer, D.C., Bagiella, E., Andrews, H., Pedley,T.A., Shih, T.T., Leary, L., Thurman, D.J., Hauser, W.A., 2007.Prevalence of self-reported epilepsy in a multiracial and multi-ethnic community in New York City. Epilepsy Res. 77, 141—150.

andon, B.E., Schneider, E.C., Normand, S.L., Scholle, S.H.,Pawlson, L.G., Epstein, A.M., 2007. Quality of care in Med-icaid managed care and commercial health plans. JAMA 298,1674—1681.

aPar, D.J., Stukenborg, G.J., Guyer, R.A., Stone, M.L., Bhamidi-pati, C.M., Lau, C.L., Kron, I.L., Ailawadi, G., 2012. Primarypayer status is associated with mortality and resource utilizationfor coronary artery bypass grafting. Circulation 126, S132—S139.

itaker, D., Koroukian, S.M., Love, T.E., 2005. Context and health-care access: looking beyond the individual. Med. Care 43,531—540.

iu, J.H., Zingmond, D.S., McGory, M.L., SooHoo, N.F., Ettner, S.L.,Brook, R.H., Ko, C.Y., 2006. Disparities in the utilization of high-volume hospitals for complex surgery. JAMA 296, 1973—1980.

argolis, P.A., Cook, R.L., Earp, J.A., Lannon, C.M., Keyes, L.L.,Klein, J.D., 1992. Factors associated with pediatricians’ partic-ipation in Medicaid in North Carolina. JAMA 267, 1942—1946.

cClelland, S., Curran, C.C., Davey, C.S., Okuyemi, K.S., 2007.Intractable pediatric temporal lobe epilepsy in the UnitedStates: examination of race, age, sex, and insurance status asfactors predicting receipt of resective treatment. J. Neurosurg.107, 469—473.

cClelland, S., Guo, H., Okuyemi, K.S., 2010. Racial disparities inthe surgical management of intractable temporal lobe epilepsy

in the United States: a population-based analysis. Arch. Neurol.67, 577—583.

ukherjee, D., Zaidi, H.A., Kosztowski, T., Chaichana, K.L., Brem,H., Chang, D.C., Quinones-Hinojosa, A., 2010. Disparities in

Z

N.K. Schiltz et al.

access to neuro-oncologic care in the United States. Arch. Surg.145, 247—253.

avaneethan, S.D., Nigwekar, S., Sengodan, M., Anand, E., Kadam,S., Jeevanantham, V., Grieff, M., Choudhry, W., 2007. Referralto nephrologists for chronic kidney disease care: is non-diabetickidney disease ignored? Nephron. Clin. Pract. 106, c113—c118.

audenbush, S.W., Bryk, A.S., 2002. Hierarchical Linear Models:Applications and Data Analysis Methods. Sage Publications, Inc.,Thousand Oaks, California.

eid, A.Y., St Germaine-Smith, C., Liu, M., Sadiq, S., Quan, H.,Wiebe, S., Faris, P., Dean, S., Jette, N., 2012. Development andvalidation of a case definition for epilepsy for use with adminis-trative health data. Epilepsy Res. 102 (3), 173—179.

osenbaum, S., Sommers, B.D., 2013. Using medicaid to buy privatehealth insurance — the great new experiment? N. Engl. J. Med.369, 7—9.

chiltz, N.K., Koroukian, S.M., Lhatoo, S.D., Kaiboriboon, K., 2013.Temporal trends in pre-surgical evaluations and epilepsy surgeryin the U.S. from 1998 to 2009. Epilepsy Res. 103, 270—278.

chneider, A., Elias, R., 2002. The Medicaid Resource Book. KaiserCommission on Medicaid and the Uninsured, Washington, DC.

chneider, A., Strohmeyer, V., Ellberger, R., Retrieved 4.7.2013 fromhttp://www.kff.org/medicaid/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=13323 2000. Medicaids Eligibilityfor Individuals with Disabilities.

ubramanian, S., Chen, A., 2013. Treatment patterns and survivalamong low-income medicaid patients with head and neck cancer.JAMA Otolaryngol. Head Neck Surg. 139, 489—495.

he Henry J. Kaiser Family Foundation, Retrieved July 4, 2013 fromhttp://kaiserfamilyfoundation.files.wordpress.com/2013/01/7615-03.pdf 2010. Medicare: A Primer.

S Department of Health and Human Services, 2011.Area Resource File (ARF), Retrieved (4.7.2013) from

http://arf.hrsa.gov/index.htm

uvekas, S.H., Taliaferro, G.S., 2003. Pathways to access: healthinsurance, the health care delivery system, and racial/ethnicdisparities, 1996—1999. Health Aff. (Millwood) 22, 139—153.