UCLA UCLA Previously Published Works Title Reducing Young Adults' Health Care Spending through the ACA Expansion of Dependent Coverage. Permalink https://escholarship.org/uc/item/0jz5q6xg Journal Health services research, 52(5) ISSN 0017-9124 Authors Chen, Jie Vargas-Bustamante, Arturo Novak, Priscilla Publication Date 2017-10-01 DOI 10.1111/1475-6773.12555 Peer reviewed eScholarship.org Powered by the California Digital Library University of California Health Services Research ©HealthResearchandEducationalTrust DOI:10.1111/1475-6773.12555 RESEARCHARTICLE ’ Reducing Young Adults Health Care Spending through the ACA Expansion of Dependent Coverage JieChen,ArturoVargas-Bustamante,andPriscillaNovak Objective. Toestimatehealthcareexpendituretrendsamongyoungadultsages19– 25beforeandafterthe2010implementationoftheAffordableCareAct(ACA)provi- sionthatextendedeligibilityfordependentprivatehealthinsurancecoverage. Data Sources. Nationally representative Medical Expenditure Panel Survey data from2008to2012. StudyDesign. Weconductedrepeatedcross-sectionalanalysesandemployedadif- ference-in-differences quantile regression model to estimate health care expenditure trendsamongyoungadultsages19–25(thetreatmentgroup)andages27–29(thecon- trolgroup). PrincipalFindings. Ourresultsshowthatthetreatmentgrouphad14percentlower overall health care expenditures and 21 percent lower out-of-pocket payments com- paredwiththecontrolgroupin2011–2012.Theoverallreductioninhealthcareexpen- dituresamongyoungadultsages19–25inyears2011–2012wasmoresignificantatthe higherendofthehealthcareexpendituredistribution.Youngadultsages19–25had significantlyhigheremergencydepartmentcostsatthe10thpercentilein2011–2012. Differencesinthetrendsofcostsofprivatehealthinsuranceanddoctorvisitsarenot statisticallysignificant. Conclusions. Increased health insurance enrollment as a consequence of the ACA provisionfordependentcoveragehassuccessfullyreducedspendingandcatastrophic expenditures,providingfinancialprotectionsforyoungadults. Key Words. Healthcareexpenditures,youngadults,theAffordableCareAct TheAffordableCareAct(ACA)extendedeligibilityfordependentcoverage underprivatefamilyhealthinsuranceuptoage26(KaiserFamilyFoundation 2013). This provision allowed young adults to remain insured under their parents’ private health insurance plan until age 26. Extended eligibility for dependent coverage was one of the earliest ACA provisions to be imple- mented,comingintoforceonSeptember23,2010.Recentestimatesshowthat this provision has reduced the number of uninsured young adults by an 1 2 HSR:HealthServicesResearch absolutenumberofatleast3millionindividuals(KaiserCommissiononMed- icaid and the Uninsured 2014). In percentage terms, in 2013, among young adults ages 19–25, 22.9 percent were uninsured for the entire year. This is lowerthanforthesamepopulationin2009,when27.5percentwereuninsured fortheentireyear(Rhoades2015). The population between ages 19 and 30 was labeled by the media and governmentinstitutionsas“younginvincibles”(Smith 2009),becauseoftheir belief that young adults were healthy enough to be uninsured without major consequences.BeforetheimplementationoftheACA,approximately30per- cent of young adults were uninsured, representing approximately one in five uninsured individualsintheUnitedStates (Centers for Medicare & Medicaid Services2010).Youngadultsexperiencedthehighestuninsuredratecompared with older adults and children, primarily because young adults experienced the lowest amount of employer-provided coverage (Centers for Medicare & MedicaidServices2010).Youngadultswerefrequentlyemployedinentryand part-timepositions,whichwerelesslikelytoofferemployersponsoredhealth insurance(MerluzziandNairn1999;CallahanandCooper2005). Different ACA-related strategies aimed to increase coverage among young adults, by expanding Medicaid coverage, creating health insurance exchanges,andextendingprivatehealthinsurancedependentcoverageeligibil- ity from age 19 to age 26 (Claxton, Rae, and Panchal 2013; Mulcahy et al. 2013; Sommers et al. 2013; Busch, Golberstein, and Meara 2014; Chua and Sommers2014).Collectively,thesestrategies offered youngadults asmoother transition to their own health insurance coverage. According to Cantor et al. (2012a,b), the “ACA-dependent coverage expansion represents a rare public policysuccessintheefforttocovertheuninsured.” Ourstudyfocusesonhealthcareexpendituretrendsamongyoungadults before and after the extension of dependent coverage eligibility. Chua and Sommers(2014)examinedtotalhealthcarecostsandout-of-pocket(OOP)pay- mentsamongyoungadults.TheirresultsshowedsignificantreductioninOOP expendituresin2011,butfoundnochangesinoverallhealthcareexpenditures from2002to2011.Theseresultsmaybesensitivetotheshorttimeperiodafter the ACAexpansion. The association between the ACAexpansion and health AddresscorrespondencetoJieChen,Ph.D.,DepartmentofHealthServicesandAdministration, SchoolofPublicHealth,UniversityofMaryland-CollegePark,3310E,CollegePark,MD20742; e-mail:[email protected],Ph.D.,iswiththeDepartmentofHealthPol- icyandManagement,FieldingSchoolofPublicHealth,UniversityofCalifornia,LosAngeles, CA. Priscilla Novak, M.P.H., is with the Department of Health Services and Administration, SchoolofPublicHealth,UniversityofMaryland-CollegePark,CollegePark,MD. ReducingYoungAdults’SpendingunderACA 3 careexpendituresmightalsodifferalongthedistributionofhealthcareexpen- ditures(CookandManning2009;Chenet al.2014).Higherhealthcareexpen- diture may indicate higher intensity of care, such as cancer treatment, and lower health care expenditures may indicate demand for primary health care services (e.g., prescription drug use and physician visits) (Cook and Manning 2009;Chenet al.2014).SincethisACAprovisionexpandedinsurancecover- age,beneficiariesofthisprovision,thoseyoungadultsages19–25,wouldhave better access to primary care, and potentially reduce their emergency depart- ment (ED) utilization. In other words, we may observe the ACA provision is positivelyassociatedwiththelowerendofthehealthexpendituresdistribution (e.g.,spendingonprimarycare)andnegativelyassociatedwiththehigherend ofthisdistribution(e.g.,spendingonEDvisits). In this study, we use quantile regression with difference-in-differences estimates to identify changes along the health expenditures distribution (Chen et al. 2014). We implement before and after comparisons of health care services used by young adults to identify any changes in health care spending patterns as a consequence of the dependent coverage expansion up to age 26. In addition, it is likely that the cost shifting among different types of health insurance plans might have happened during this transition. Hence, we also examine the pattern of health care spending across payers. Young adults’ health spending patterns are expected to impact aggregate U.S. healthspending inthe long term. Lackof health care access and health insurance may result in delaying or forgoing necessary treatment, poten- tially leading to health problems and higher health expenditures in mid- adulthood (Merluzzi and Nairn 1999; Callahan and Cooper 2005). Our study provides first evidence on the changes of health spending patterns under this ACAprovision. METHOD Data Weusenationallyrepresentativedatafromthe2008to2012MedicalExpen- diturePanelSurvey(MEPS)(Cohen,Cohen,andBanthin2009).MEPSisa nationally representative survey of the civilian, noninstitutionalized popula- tionintheUnitedStates.Itprovidesrespondents’detailedhealthcarespend- ing during the survey year, as well as their demographics, socioeconomic characteristics,healthstatus,andhealthinsurancestatus. 4 HSR:HealthServicesResearch Ouroutcomevariablesareannualhealthcareexpendituresperperson, includingtotalhealthcarespendingandspendingonspecifictypesofhealth care services (e.g., costs of physician visits). We also examine expenditures bypayers,suchaspatients(i.e.,OOPpayments),privateinsurers,andpublic payers. Health care expenditures are self-reported and validated by respon- dents’ physicians and pharmacists. All health expenditures are adjusted to constant U.S. dollars using the 2012 Consumer Price Index Medical Component. We parse the U.S. young adult population into two groups: the ACA- dependentcoverageexpansiontargetedgroup(adultsages19–25)andnontar- getedgroup(thosewhowere27–29 yearsold).Tobeconsistentwithprevious studies, we also use young adults 27–29 years old as the reference group (Busch, Golberstein, and Meara 2014; Barbaresco, Courtemanche, and Qi 2015;Golbersteinet al.2015).Weexclude26-year-oldindividualsduetothe lackofaccurateinformationtodeterminetheireligibilitywhentheACAwas implemented in 2010. Under the ACA, since September 23, 2010, young adults were able to join or remain on their parents’ private health insurance plan,regardlessofmaritalstatus,schoolattendance,financialdependenceon parents,oreligibilityforemployer-providedhealthinsurancecoverage(U.S. DepartmentofLabor2014).Somelargeemployersclaimthattheyexpanded coverageinearly2010.Tocaptureaprecisesnapshotofhealthcareexpendi- tures due to the ACA expansion, we use 2008–2009 and 2011–2012 as the pre-andpost-implementationperiod. The sample size of young adults ages 19–25 and 27–29 in 2008–2009 and2011–2012is12,564.Atotalof11,154observationsremainedafterdrop- ping the observations with missing variables. We further exclude 15 outliers withexpenditurehigherthan3standarddeviationsabovetheaverageexpen- diture. Our final sample has 7,623 young adults 19–25 years old and 3,516 youngadults27–29 yearsold. Analysis We first summarize overall and specific health care expenditures for young adultsages19–25andages27–29in2008–2009(beforetheACAexpansion) and 2011–2012 (after the ACAexpansion of young adults’ health insurance coverage was fully implemented). We then compare population characteris- tics between these two groups, with individuals ages 27–29 as the reference group.Thesurveyweightswereemployedtoadjustsamplecharacteristicsto be nationally representative. To capture this natural experiment of health ReducingYoungAdults’SpendingunderACA 5 insurance eligibility among young adults, we use a difference-in-differences approachtoidentify anychangesinhealthcareexpendituresamongthetar- geted population (young adults ages 19–25 years old) in the pre- (2008– 2009)andthepost-(2011–2012)implementationperiods,relativetochanges inhealthcareexpendituresofthenontargetedcomparisongroup(adultsages 27–29 yearsold). Difference-in-differences method assumes “parallel trends” (Wool- dridge 2002; Dimick and Ryan 2014) in outcomes between the treatment and comparison groups prior to the intervention, which is the ACAexpan- sion in our study. To test whether the trends of health care expenditures are parallelbeforetheintervention,weusedatafromearlieryearsoftheMEPS and construct the interaction terms between the treatment group “young adults ages 19–25” with each survey year. We test the significance of these interaction terms for all outcome variables, which are total health care expenditures, OOP payment, private health insurance expenditures, Medi- caid expenditures; and costs of doctor visits, prescription drugs, ED visits and inpatient visits. Results (Appendices SA2 and SA3) show that most of the interaction terms are statistically insignificant which indicate that health expenditure trends between young adults ages 19–25 and young adults ages 27–29areparallelpre-ACA.Thesefindingsjustifytheuseofadultsages27– 29asourtreatmentgroup. Importantly, the interaction terms in the regressions of Medicaid expenditure (year 2004 and year 2008), prescription drug cost (year 2006), and inpatient visit costs (in multiple years) were statistically significant (p < .05). This finding may suggest that young adults ages 27–29 are not the comparable treatment group to examine these outcomes. Thus, we exclude the analyses on Medicaid expenditures, prescription drug costs, and inpatient visit costs. Hence,ouroutcomevariablesaretotalhealthcareexpenditures,OOP, cost paid by private health insurers, and costs of ED and doctor visits. We use a repeated cross-sectional study design and apply a generalized linear model with log link and gamma distribution (GLM) to estimate health care expenditures.Themodelspecificationforthedifference-in-differencesanaly- sisispresentedinthefollowingequation. Health care expenditures¼b þb ðAge19(cid:2)25Þþb ðYears2011(cid:2)2012Þ 0 1 2 þb ðAge19(cid:2)25(cid:3)Years2011(cid:2)2012Þ 3 þb ðcovariatesÞþe 4 6 HSR:HealthServicesResearch WeusetheconceptualframeworkofAndersensocialbehavioralmodel (Aday and Andersen 1974; Andersen 1995) to select the covariates that are associatedwithhealthcareexpenditures.Thecovariatesincludedinourstudy canbecategorizedintothreedomains:thepredisposingfactors(race/ethnic- ity,gender,maritalstatus,U.S.bornvs.foreignborn,andinterviewlanguage) (Chen, Vargas-Bustamante, and Ortega 2013); enabling factors (education, familyincome,urban/rural,andU.S.CensusRegion);andclinicalneedsfac- tors(self-reportedphysicalandmentalhealth,SF12-physicalcomponentsum- mary (PCS), and mental component summary (MCS) (Ware, Kosinski, and Keller1996).Thesevariableshavebeenwidelyusedintheliteraturetoexam- ine health care expenditures and utilization (Cook and Manning 2009; Ku 2009;Chenet al.2014). Prior research suggests that health care expenditures are highly con- centratedandnotevenlydistributed,withonly5percentoftheU.S.popula- tion accounting for half of health care expenditures (Staton 2006), while approximately 50 percent of the population has minimal expenditure on health care. GLM estimation takes into account the skewed health care expenditures distribution,addressesheterogeneities,andprovidesconsistent andefficientestimates(Goldberger1964;Wooldridge2002;Manning,Basu, and Mullahy 2005). All our results are nationally representative using the survey weights provided by the MEPS. Stata 12 MP was used to conduct theanalysis. Health care spending may be indicative of consumers’ different health needs(CookandManning2009;Chenet al.2014).Thus,inthisstudy,weuse the quantile regression model to examine whether the effects of the ACA’s expansiononyoungadults’coverageeligibilityvaryalongthehealthexpendi- turedistributions(KoenkerandHallock2001;Koenker2005).Thecoefficients at the lower percentiles of the health expenditure distributions (e.g., the 10th, 25th, and 50th percentiles) reflect the associations between the ACA expan- sion and health expenditures on primary or routine health care. The coeffi- cientsatthehigherpercentilesofhealthexpenditures(e.g.,the75th,90th,and 95th percentiles) indicate the associations between the ACA expansion and theuseofmoreintenseandcostlyhealthcareservices(Chenetal.2014). RESULTS Figure 1presentsthetotalamountofhealthcareexpendituresbypayersand servicesforyoungadultsages19–25andages27–29.Comparedtoadultsages ReducingYoungAdults’SpendingunderACA 7 Figure 1: HealthCareExpendituresforYoungAdultsAges19–25andAges 27–29beforeandaftertheACAExpansion Notes.Dataset:MedicalExpenditurePanelSurvey2008–2009,2011–2012.Resultsarenationally representative. 27–29 years old, the younger cohort had relatively lower total health care expenditures, OOP, private health insurance expenditures, doctor visit expenditures,buthigherEDcosts.Theseexpenditureswererelativelylower or similar post-ACA expansion. However, among adults 27–29 years old, totalhealthcareexpendituresandEDvisitcostsincreasedpost-ACAexpan- sionofdependentcoverage. Table 1showspopulationcharacteristicsforourtreatmentandcompar- isongroupsbeforeandaftertheACAchangeinprivatehealthinsuranceeligi- bility. Adults 19–25 years old were less likely to be married, to report over 12 yearsofschooling,tobeemployed,ortoreportahighfamilyincome.The treatmentandcomparisongroupshadsimilarratesofchronicconditions. Table 2 presents the results of GLM regressions of overall health care expendituresanddisaggregatesexpendituresbypayersandservices.Results show that after controlling for socioeconomic and demographic factors, the associations between the year indictors (2011–2012) and total health care expendituresandexpendituresbypayerandserviceswerenotstatisticallysig- nificant. Young adults 19–25 years old had comparable total expenditures compared with the control group. In 2011–2012, however, total costs were reducedby14percent(=0.06(thecoefficientof“Age19–25”)–0.20(thecoeffi- cient of the interaction term)), and OOP costs were reduced by 21 percent (=0.08(thecoefficientof “Age19–25”)–0.29(thecoefficientoftheinteraction 8 HSR:HealthServicesResearch Table 1: PopulationCharacteristicsofYoungAdultsAges19–25YearsOld andAges27–29YearsOld Age19–25 Age19–25 Age27–29 Age27–29 (2008–2009), (2010–2012), (2008–2009), (2010–2012), mean mean mean mean Race/ethnicity White 0.67 0.65 0.69 0.65 Latinos 0.16 0.17 0.15 0.18 AfricanAmericans 0.12 0.13 0.11 0.11* Otherraces 0.05 0.05 0.05 0.07 Gender Female 0.54 0.56 0.58* 0.56 Maritalstatus Married 0.10 0.09 0.42*** 0.37*** InterviewedEnglish English 0.93 0.94 0.93 0.94 Education Schoolingyears<12years 0.56 0.49 0.37*** 0.31*** Schoolingyears12–16years 0.42 0.49 0.53*** 0.57*** Schoolingyears>16years 0.02 0.01 0.10*** 0.12*** Familyincome Under100%FPL 0.19 0.21 0.13*** 0.14*** 200–400%FPL 0.20 0.22 0.20 0.19*** Over400%FPL 0.60 0.56 0.67*** 0.67*** Unemployed 0.19 0.22 0.13*** 0.12*** Urban 0.84 0.85 0.88* 0.84 U.S.CensusRegion Northeast 0.16 0.18 0.18 0.16 Midwest 0.25 0.23 0.22 0.27 South 0.36 0.36 0.35 0.33 West 0.24 0.23 0.25 0.25* Self-reportedhealthstatus Healthpoororfair 0.08 0.06 0.08 0.08 Healthgood 0.21 0.21 0.25* 0.25 Healthverygoodor 0.70 0.72 0.68 0.68* excellent Self-reportedmentalhealthstatus Mentalhealthpoororfair 0.06 0.05 0.05 0.05 Mentalhealthgood 0.17 0.19 0.20 0.18 Mentalhealthverygood 0.77 0.76 0.75 0.77 orexcellent SF12-PCS 0.54 0.54 0.54 0.53* SF12-MCS 0.50 0.51 0.49 0.50* Diabetes 0.01 0.01 0.02 0.02* Asthma 0.07 0.07 0.06 0.06 continued ReducingYoungAdults’SpendingunderACA 9 Table1: Continued Age19–25 Age19–25 Age27–29 Age27–29 (2008–2009), (2010–2012), (2008–2009), (2010–2012), mean mean mean mean Heartdisease 0.01 0.01 0.00 0.00 Depress 0.07 0.07 0.08 0.09 Anxiety 0.06 0.07 0.07 0.09* Notes.***p<.001;**p<.01;*p<.05,thereferencegroupisyoungadultsages19–25yearsold inthesametimeperiod.Authors’analysisusingtheMedicalExpenditurePanelSurvey2008– 2009,2011–2012.Resultsarenationallyrepresentative. term))amongyoungadults19–25 yearsold,comparedwiththecostofyoung adults27–29 yearsold. Inaddition,Table 2alsoshowsthatamongyoungadults,beingaracial or ethnic minority (compared with whites) or male (compared with female) wasassociatedwithlowertotalhealthcareexpenditures,OOP,privatehealth insurance expenditures, and lower spending on doctor visits. Individuals speaking English were more likely to have higher total cost, greater private healthinsuranceexpenditures,andhigherexpendituresondoctorvisits.Indi- vidualswholivedintheSouthweremorelikelytohavehigherOOP,lower spendingondoctorvisits,buthigherEDspending. Table 3presentsthequantileregressionanalyses(10th,25th,50th,75th, 90th,and95thpercentiles)ofhealthcareexpenditures bypayer(wepresent thekeyfindingsinTable 3;fullsetsofresultsareincludedintheAppendices). The quantile regressions show that the coefficients for the year indicator 2011–2012aresignificantlynegativeatthelowerend(coef = (cid:2)0.3,p < .05at 25thpercentile;andcoef = (cid:2)0.16,p < .05,at50thpercentile)ofthedistribu- tion of total health care expenditures. The interaction terms between treat- mentgroupandyearindicatorshowthatatthe90thpercentile,thetreatment groupreported30percentlowertotalcostcomparedwiththecontrolgroup in 2011–2012 (30 percent = 0.01 (the coefficient of “age 19–25”)–0.31 (the coefficientoftheinteractionterm)). Quantile regressions in Table 3 show that OOP costs were signifi- cantly lower in year 2011–2012 at the 25th and 75th percentiles. OOP costs of the treatment group were reduced significantly at the 90th (the reduction reached 16 percent, p < .05) and 95th percentiles (the reduction reached 31 percent), compared with the OOP of the control group. Overall costs for private health insurance were similar between the treat- ment and comparison groups.