a sociology essay

A E

Y a

b

c

a

A R R A A

J I I I

K A H I A Y M

1

a a u a y S o y t S

I U

a

h 0

Journal of Health Economics 39 (2015) 171–187

Contents lists available at ScienceDirect

Journal of Health Economics

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e c o n b a s e

ccess to health insurance and the use of inpatient medical care: vidence from the Affordable Care Act young adult mandate

aa Akosa Antwi a, Asako S. Moriya b, Kosali I. Simon b,c,∗

Department of Economics, Indiana University-Purdue University Indianapolis (IUPUI), United States The School of Public and Environmental Affairs (SPEA), Indiana University, United States National Bureau of Economic Research (NBER), United States

r t i c l e i n f o

rticle history: eceived 5 December 2013 eceived in revised form 5 June 2014 ccepted 22 November 2014 vailable online 28 November 2014

EL classification: 11 13 18

a b s t r a c t

The Affordable Care Act of 2010 expanded coverage to young adults by allowing them to remain on their parent’s private health insurance until they turn 26 years old. While there is evidence on insurance effects, we know very little about use of general or specific forms of medical care. We study the implications of the expansion on inpatient hospitalizations. Given the prevalence of mental health needs for young adults, we also specifically study mental health related inpatient care. We find evidence that compared to those aged 27–29 years, treated young adults aged 19–25 years increased their inpatient visits by 3.5 percent while mental illness visits increased 9.0 percent. The prevalence of uninsurance among hospitalized young adults decreased by 12.5 percent; however, it does not appear that the intensity of inpatient treatment changed despite the change in reimbursement composition of patients.

© 2014 Elsevier B.V. All rights reserved.

eywords: ffordable Care Act ealth insurance regulation

npatient medical care ccess to health insurance oung adults

y o

e b e i h ( a

ental health care

. Introduction

In the United States, the transition from adolescence to young dulthood is associated with the loss of health insurance cover- ge (Anderson et al., 2012). Prior to the Affordable Care Act (ACA), ninsurance among the non-elderly peaked at around ages 21–23 t close to 40 percent.1 The precarious health insurance status of oung adults motivated the early ACA provision that starting in eptember 2010 has allowed young adults to remain as dependents n their parents’ private health insurance plans until they turn 26

ears old. The mandate has substantially reduced uninsurance in his population (Cantor et al., 2012; Sommers and Kronick, 2012; ommers et al., 2013; Akosa Antwi et al., 2013). However, there is as

∗ Corresponding author at: The School of Public and Environmental Affairs (SPEA), ndiana University, Room 359, 1315 East Tenth Street, Bloomington, IN 47405-1701, nited States. Tel.: +1 812 856 3850.

E-mail addresses: [email protected] (Y. Akosa Antwi), [email protected] (A.S. Moriya), [email protected] (K.I. Simon). 1 Author’s calculation using 2008 Current Population Survey data.

m l o s fi t o

r e u

ttp://dx.doi.org/10.1016/j.jhealeco.2014.11.007 167-6296/© 2014 Elsevier B.V. All rights reserved.

et sparse evidence on the effect of this health insurance expansion n young adults’ use of health care.

Young adults are a key population targeted under the ACA ven beyond the specific mandate we study. On the one hand, ecause young adults are comparatively healthy, high young-adult nrollment is seen as an important goal for the success of health nsurance exchanges (Weaver and Radnofsky, 2013). On the other and, young adults tend to have high mental health care needs Grant and Potenza, 2010). Estimating the effect of this ACA young dult insurance expansion on health care use in general as well as ental health care use in particular is crucial for understanding the

aw fully, as well as for anticipating the effects of later expansions n this population. Empirical evidence of the impact of this provi- ion is of high interest to policymakers, who will likely continue to ne-tune the details of the ACA for some time; it also contributes to he growing academic literature on the effect of health insurance n medical care use.

We use the Nationwide Inpatient Sample (NIS), a nationally- epresentative database of inpatient admissions, to evaluate the arly effect of the ACA young adult insurance expansion on the se of inpatient medical care in general and mental healthcare

1 Health

s t a c A p c W a m b e t t

o f t i e m a c F d r c a a l o

g 3 i a t t a i t s a o i fi n

2

2

i i i 2 t s a

h e o a t s A b

2

t i 1 m t e t w i i t s t s

2

a e c u s n a

a S o a i i

t e t y h e i h n u t

72 Y. Akosa Antwi et al. / Journal of

pecifically, on treatment intensity, and on insurance status of inpa- ient visits. Inpatient visits are rare but expensive medical events nd are a vital component of any effort to “bend the health care cost urve”. For instance, for 19-to-25-year age group targeted by the CA dependent care mandate, inpatient visits represented about 31 ercent of total health care expenditures even though only 5.4 per- ent of individuals in this age group had an inpatient visit in 2008.2

e identify the effects of the policy on the targeted age group using differences-in-differences (DD) method that compares the treat- ent group of 19-to-25-year-olds to 27-to-29-year-olds, the latter

eing a comparison group that is close to but excluded from the xpansion. We also conduct extensive robustness checks regarding he assumptions underlying this identification strategy and find hem to be viable.

We first examine the impact of the ACA young adult mandate n the total number of non-birth hospitalizations. The transition rom adolescence to adulthood often involves many life changes hat can trigger mental distress, making services related to mental llness an important component of young adult medical care (Patel t al., 2007; Yu et al., 2008). Not surprisingly, mental disorder is the ost frequent reason why young adults seek hospital-based care,

side from visits related to childbirth.3 Thus, the second aspect we onsider is the effect of the mandate on mental illness admissions. or both types of utilization, we analyze the impact of the man- ate separately on admissions that originate from the emergency oom (ER) and those that are direct admissions. To shed light on the hanges that hospitals will likely face in reimbursement for young dult care as a result of reform, we estimate the change in insur- nce composition as well. We end by evaluating the impact of the aw on treatment intensity as measured by length of stay, number f procedures, and hospital charges.

We find that compared to slightly older young adults, those tar- eted by the law increased their overall non-birth inpatient visits by .5 percent after the law’s implementation. This is driven by direct

npatient admissions which tend to be more discretionary than dmissions that result from the ER, and might be more responsive o changes in health insurance coverage. Consistent with evidence hat mental illness treatment is more responsive to health insur- nce coverage than general medical care, mental-illness-related npatient visits increase by 9.0 percent, driven mainly by visits hat originate through the ER. Corresponding to these changes in ervice utilization, we find that the fraction of hospitalized young dults without insurance decreased by 12.5 percent compared to ther adults who did not benefit from the expansion. In examin- ng whether treatment intensity of hospital usage is affected, we nd no robust evidence of the impact of the law on length of stay, umber of procedures, and total charges.

. Prior literature

.1. The effect of health insurance expansion on inpatient care

A large body of empirical research studies the effect of health nsurance expansions on the use of inpatient care. Studies on Med- caid and Medicare, for instance, find that they lead to an increase n the consumption of inpatient medical care (Dafny and Gruber, 005; Card et al., 2008; Finkelstein et al., 2012). But research on

he impact of a near-universal health insurance coverage expan- ion in Massachusetts finds no change in hospitalizations (Kolstad nd Kowalski, 2012).

2 Author calculations from 2008 Medical Expenditure Panel Survey. 3 This is based on author’s calculations from the Nationwide Inpatient Sample.

t i p

r

Economics 39 (2015) 171–187

Research especially relevant for our work studies the impact of ealth insurance on medical-care use by young adults. Anderson t al. (2012, 2014) estimate the effect of health insurance coverage n inpatient and ER visits by exploiting the sharp change in insur- nce coverage rates that result from young adults “aging out” of heir parent’s health insurance plans at ages 19 and 23. They find ignificant reductions in inpatient and ER visits at both age cutoffs. ging out of health insurance at 19 and 23 reduces inpatient visit y 1.7 and 0.8 percent, respectively.

.2. Effect of insurance expansion on inpatient mental health care

Evidence from the RAND health insurance experiment shows hat mental health care is almost three times as responsive to nsurance generosity as other forms of health care (Manning et al., 989). While this evidence pertains to outpatient care only, it ight have implications for inpatient care as well. Research from

he 2007 Massachusetts health insurance expansion in Meara t al. (2014) shows that there was little to no change in inpa- ient admissions for mental health among young adults (consistent ith Kolstad and Kowalski (2012)’s result for inpatient care

n general), and a small but statistically significant reduction n admissions for substance use disorders. The authors caution hat the generalizability of evidence from Massachusetts is not traightforward since the availability of behavioral health outpa- ient providers is much higher in Massachusetts than in other tates.4

.3. Effect of ACA dependent coverage on use of medical care

Although there is no prior work on the effect of the young dult mandate on inpatient care, Sommers et al. (2013) find vidence that the mandate increased self-reported access to are but had no statistically significant effect on self-reported sual source of care. Mulcahy et al. (2013) find that the expan- ion led to a 3.1-percentage-point increase in the share of on-discretionary emergency care that is paid by private insur- nce.

Taken together, prior literature on the impact of health insur- nce on all and mental health inpatient admissions is rather mixed. ome studies find a sizeable increase in use while others find small r no effects. There is no literature on the impact of the ACA young dult mandate on inpatient admissions, but evidence suggests an ncrease in access to healthcare in general, but not necessarily an ncrease in a usual source of care.

Our research makes several distinct contributions. In addition o providing the first evidence on the impact of the ACA depend- nt coverage expansion on inpatient care use, we contribute to he literature on the effect of insurance on use of care among oung adults by evaluating the effect of gaining rather than losing ealth insurance on medical use. Evidence provided by Anderson t al. (2012, 2014) measure the effect of anticipated loss of health nsurance on medical-care use. The effect of gaining and losing ealth insurance may not be symmetric. Third, we examine the ational impact of providing coverage to young adults on their se of inpatient mental health care, a particularly high need in his age group. Fourth, we examine the effect of the law on

he prevalence of insurance coverage among young adults using npatient services, and its implications for intensity of treatment rovided.

4 Meara et al. (2014) report that Massachusetts has 32.4 psychiatrists per 100,000 esidents compared with 14.5 for the United States.

Health

3

e l M

t y T e c a c a n c

c 1 e e w d c w d c i c h t s m p t 2 ( b s 2 o c t m B s t t

t h a p c t 2 s m h p e i a s

4

I ( c h o t

w t g w s g f O s p 1 b s s s a a t l w h

5

s f d 1 a s t m w 2 w e t have led the treatment group to experience different rates of med- ical care access and use after reform. We start our empirical work by testing the validity of the DD estimator assumptions including

5 Data from Alabama, Delaware, the District of Columbia, and Idaho are not in our sample because they do not contribute data to the NIS in any year of our sample. We exclude data from California, Maine, and Texas because precise information on age, which is crucial for our study design, is not available and thus prevents us from separating the control and treatment groups. However, we note that results are qualitatively the same when we include these states and estimate models with less than ideal control and treatment group age definitions.

Y. Akosa Antwi et al. / Journal of

. Mechanisms

The mechanisms by which the availability of health insurance is xpected to affect medical-care use have been well covered in prior iterature (Finkelstein et al., 2012; Kolstad and Kowalski, 2012;

iller, 2012), thus our discussion of the hypotheses remains brief. Access to health insurance reduces the cost of medical care for

he newly insured, and moral hazard suggests that newly insured oung adults would increase their consumption of medical care. he effect of health insurance on how newly insured young adults nter the medical system is not clear. If inpatient and outpatient are are substitutes for some conditions, then having health insur- nce could reduce hospital admissions through better outpatient are. Several studies reviewed above find inpatient care increases fter insurance expansions, suggesting hospital care could be on et a complement rather than a substitute for other forms of are.

The moral hazard of insurance is stronger for mental health- are than other types of healthcare (Horgan, 1986; Taube et al., 986; Manning et al., 1989; Frank and McGuire, 2000), leading to an xpectation that the ACA will likely increase mental health care use specially in light of recent reductions in social stigma associated ith seeking treatment (Mojtabai, 2007). Garfield et al. (2011) pre- ict that full implementation of the ACA will increase mental health are use (including inpatient forms) by 4.5 percent. This increase ill most likely not be uniform across all US states given evi- ence from Massachusetts suggesting that inpatient mental health are use may not increase in response to an expansion of health nsurance coverage if there is capacity for improved outpatient are (Meara et al., 2014). In addition, in response to the perceived igher moral hazard of mental health care, insurers have tradi- ionally placed greater limitations on coverage for mental health ervices than other forms of care. These restrictions gave rise to a ovement of state mandates aimed at parity. Although a federal

arity statute passed in 1996, that law is relatively weak compared o state and federal laws passed since 2000 (Buchmueller et al., 007). The 2008 Mental Health Parity and Addiction Equity Act MHPAEA) was effective January 2010, and required closer parity etween mental and medical care coverage. It is unclear whether trong parity laws will lead to an increase in use of services (HCCI, 014), partly because insurers have been adopting aggressive forms f management for mental health care delivery. Providers could onstrain the use of expensive forms of care which is usually inpa- ient based by contracting with specialized managers to carve out

ental health care benefits (Goldman et al., 1998; Sturm, 1997). arry and Ridgely (2008) find evidence that insurance plans inten- ified their use of utilization management techniques in response o the increased likelihood of moral hazard due to parity legisla- ion.

There are competing hypotheses about the likely effect of he mandate on overall treatment intensity, conditional on ospitalization. The law is expected to change the health insur- nce composition of hospitalized young adults by reducing the roportion of uninsured visits and increasing the proportion overed by private insurance. If private health insurance sta- us leads to more treatment by health care providers (Doyle, 005), then we would expect an increase in treatment inten- ity after the law. Intensity of care may decrease, if the arginal young adults seeking care after the expansion are

ealthier than others, especially since young adults whose arents have private insurance tend to be from higher socio-

conomic status families. Thus, the direction of predicted change n intensity is ambiguous, and we explore this point using nalyses that do and do not control for the case-mix of admis- ions.

d

( h

Economics 39 (2015) 171–187 173

. Data

The main data we use for our analysis is from the Nationwide npatient Sample (NIS), Healthcare Cost and Utilization Project HCUP), Agency for Healthcare Research and Quality. The NIS data ontains all discharges from a 20% stratified sample of community ospitals in the United States. On average the NIS has information n about eight million hospital stays a year from about 40 states in he U.S.5

We use the NIS data from 2007 to 2011 (the latest year for hich data is available) to allow a sufficient look-back period

o test for differences in trends between treatment and control roups. Each observation in the NIS is a patient discharge abstract, hich includes detailed clinical information, such as primary and

econdary diagnoses, and demographic information, such as age, ender, and race/ethnicity. We restrict our attention to discharges or ages 19–29. We follow prior literature by using non-birth visits.6

ur final sample is composed of 794,392 hospital visits. In our econdary analysis of mental illness visits, we restrict our sam- le 128,310 visits that belong to Major Diagnostic Category (MDC) 9: Mental Diseases and Disorders.7 Henceforth we refer to non- irth admissions as “all” admissions to distinguish them from our econdary analysis which singles out mental health related admis- ions. Our data also includes information on health insurance status uch as private insurance, Medicaid, Medicare, other insurance, nd no insurance. We cannot separately identify treated young dults who are dependents on their parent’s health insurance from hose who have their own ESI or non-group insurance. To calcu- ate our dependent variables measuring number of inpatient visits,

e aggregate hospital discharges by age and year-quarter for each ospital.

. Method

To estimate the effect of the federal young adult coverage expan- ion on inpatient medical care, one must isolate the policy’s impact rom contemporaneous national changes. We use a differences -in- ifferences (DD) method that compares all targeted young adults, 9–25-year-olds and a control group of 27–29-year-olds, who are s similar as possible in age but excluded from the insurance expan- ion. We also conduct sensitivity checks comparing separately hose who are 19–22 years and 23–25 years as alternate treat-

ent subgroups to those 27–29 years (control group). In addition, e compare young adults on either side of the age cutoff, that is,

5 year-olds vs. age 27 year-olds. We exclude 26-year-olds since e cannot accurately classify them as part of either group. Our

mpirical strategy rests on the strong assumption that the con- rol group will account for other time-varying factors that would

6 Non-birth visits here denote all inpatient visits except those classified as major iagnostic category 14: pregnancy, childbirth and puerperium. 7 The majority of mental health hospitalizations occur in community hospitals

Meara et al., 2014), but we note here that specialty psychiatric hospitals and prison ospitals, which do not report data to AHRQ, are also a major source of care.

174 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Table 1 Summary statistics of treatment and control group.

19–25 years old 27–29 years old

Before ACA enactment

After ACA implementation

Before ACA enactment

After ACA implementation

Demographic characteristics Age 22.1 22.1 28.0 28.0 Indicator: male 0.500 0.501 0.474 0.473 Indicator: white 0.622 0.602 0.617 0.607 Indicator: African-American 0.207 0.235 0.210 0.233 Indicator: Hispanic 0.103 0.097 0.104 0.097

Clinical characteristics Indicator: mental illness 0.177 0.192 0.152 0.157 Number of diagnosis codes 4.74 5.22 5.15 5.66 Indicator: admitted on weekend 0.238 0.221 0.237 0.223 Indicator: admitted through ER 0.654 0.652 0.628 0.638

Utilization measures Length of stay (LOS) 4.38 4.36 4.33 4.35 Log of LOS 1.39 1.39 1.40 1.40 Number of procedure codes 1.08 1.08 1.15 1.15 Total charges 24,298 28,972 24,333 29,648 Log of total charges 9.55 9.71 9.58 9.75

Health insurance status Indicator: covered by private insurance 0.389 0.395 0.379 0.324 Indicator: uninsured 0.214 0.193 0.195 0.203 Indicator: covered by Medicaid 0.275 0.291 0.263 0.298 Indicator: covered by Medicare 0.045 0.048 0.095 0.103 Indicator: covered by other insurance 0.073 0.065 0.064 0.065

Number of observations 523,487 213,482 270,630 106,715

Note: Sample estimates from the NIS data, 2007–2011, using data of non-birth related admissions of young adults aged from 19 to 29, except for the removal of 26 year olds who are in neither control nor treatment. Means of the variables are obtained for treatment and control groups before ACA enactment (2007 Q1–2010 Q1) and after ACA implementation (2010 Q4 and onward). The data from California, Maine and Texas are excluded because precise information on age was not available. Observations in which length of stay exceeds 90 days are excluded. The mean for the race categories is calculated using observations in which race/ethnicity variable is available in the data. Not a

c t

i m u e

Y

w h f 2 t t 2 t i i a h o g

v i i

r a n u g t s a m d a c f s p r w f o t w f n l t w

ll states provide information on race/ethnicity for all years in the NIS dataset.

ommon trends and also conduct placebo tests using data prior to he ACA.

Our main analysis examines how the reform affected the use of npatient medical care. Using aggregated discharge data, we esti-

ate a model similar to that of Kolstad and Kowalski (2012), who se the NIS to evaluate health care reform in Massachusetts. Our stimating equation is:

hsgt = ̨ + �Ageg + ıImplementt + ϑEnactt + �(Treatg × Implementt ) + �(Treatg × Enactt ) + Xst ̌ + �t + �hs + εhsgt , (1)

here Yhsgt represents our outcome variables of interest for Ageg in ospital h, state s, and quarter t. Implementt represents a dummy

or the period after the law was implemented in September 010. Since our analysis is at the quarterly level, our implemen- ation phase starts from the fourth quarter of 2010 and runs hrough the latest period of available data, the fourth quarter of 011. The variable Treatg is a binary variable for membership in he 19–25 age range (relative to the 27–29 range); in the non- nteracted term Ageg we include a full set of age indicators. The nteraction of Implementt and Treatg captures the average impact fter the law was implemented in September 2010 by comparing ospital visits during this period to visits before the enactment f the law among the treatment group relative to the control roup.

To examine possible anticipatory changes, we add a dummy ariable, Enactt, to capture the period between enactment and mplementation of the law, from April 2010 to September 2010, and ts interaction with the treatment dummy variable. The Xst vector

f m l r

epresents quarterly linear state-specific time trends. The recession nd the subsequent slow recovery span our sample period. These ational macroeconomic conditions could differentially affect the se of hospital-based medical care by our treatment and control roups. Although we conduct tests of differential time trends prior o reform to validate our study design, we account for the pos- ibility that the macroeconomy could also influence age trends fter reform by including two variables: the quarterly unemploy- ent rate, and its interaction with age dummies. We also include

ummy variables for year and quarter in �t, to control for season- lity and year fixed effects that are common to the treatment and ontrol groups. We include hospital fixed effects in �hs to account or time-invariant hospital characteristics. We use ordinary least quares to estimate all of our continuous outcomes and the linear robability models for our binary outcomes. To account for cor- elation of hospital-level errors over time due to common forces, e cluster standard errors by hospital. However, inference drawn

rom this model may be incorrect as the policy variation occurs nly at the age by time level. We are not able to block-bootstrap he standard errors as suggested by Bertrand et al. (2004) because e have an extremely small number of “groups” (two). Thus, we

ollow Cameron et al. (2008) by estimating our main model alter- atively by aggregating data to the national age by year-quarter

evel, and implement the wild cluster bootstrap-t method using ime as the grouping variable. Our results from this estimation, hich we report in the appendix, reinforce the results stemming

rom our models that cluster at the hospital level. We also esti- ate all reported models by aggregating data to the age by time

evel and clustering at the age level, and find qualitatively similar esults [results available upon request].

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 175

Fig. 1. Number of admissions. Notes: (1) Sample estimates from the NIS data, using data from 2007 to 2011. (2) The solid line indicates number of admissions among 19–25-year-olds per hospital, and the dashed line indicates number of admissions among 27–29-year-olds per hospital. (3) The first vertical line indicates the first quarter of 2010 when the ACA was passed, the second vertical line indicates the third quarter of 2010 when the dependent coverage mandate was implemented, and the third vertical l ter th

6

6

b e t i h p t t t T d s

s t l v t

c g

6

n a m b a i c s v m during the past recession. A volatile macroeconomy is linked to greater indications of depressive symptoms (Tefft, 2011), increased suicide (Luo et al., 2011), and loss of health insurance (Cawley et al.,

ine indicates the first quarter of 2011 when most new insurance plan years start af

. Results

.1. Summary statistics

We present sample means of our treatment and control groups, efore and after ACA implementation, in Table 1. Other than for the xpected age differences between the older and younger cohorts, he means of most demographic and clinical variables appear sim- lar across the groups and over time. Targeted young adults have a igher likelihood of a mental health admission (17.7 percent) com- ared to slightly older adults (15.2 percent) in the period prior o ACA enactment. The spread increases in the post implemen- ation period, with the likelihood rising to 19.2 percent for the argeted group relative to 15.7 percent for the slightly older group. here is very little other evidence in the descriptive statistics of ifferential changes in clinical characteristics or utilization mea- ures.

When we consider health insurance status as our outcome, we ee some stark differences in the mean changes experienced by the

wo groups. Conditional on seeking inpatient care, older adults are ess likely to have private health insurance over time (37.9 percent s. 32.4 percent) while private coverage increases slightly among argeted young adults (38.9 percent vs. 39.5 percent). There is i

e implementation of the mandate.

orresponding evidence of reductions in uninsurance among tar- eted young adults relative to slightly older adults.8

.2. Validity of study design

Estimating the impact of policies by comparisons between ational treatment and control groups involve especially strong ssumptions. For example, when we identify the impact of the andate on hospitalizations by comparing the count of admissions

y our treatment and control groups before and after the law, we ssume the control group and treatment group would have sim- lar trends absent the law. Given the fact that our sample period oincides with the most recent recession, this assumption is even tronger if there is a differential impact of the recession on younger s. older adults. Roberts and Terrell (2014) find evidence that job arket prospects deteriorated furthest for new job market entrants

8 Means of aggregated variables such as group totals of admissions are reported n regressions tables.

176 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

F the pe l rance F

2 ( n ( i c e a

i t h o t r s p

I c t 2 W n i

t f o l f t

fi t i a f b a l v s

s W h

ig. 2. Percent of admissions insured. Notes: (1) The solid line indicates the mean of evel, and the dashed line indicates mean of the percent of admissions for each insu ig. 1.

013) which likely further exacerbate health conditions. Maclean 2013) finds that leaving school in a bad economy has a persistent egative impact on the health outcomes of men and Kuhn et al. 2009) find that job loss increases the use of mental health care, ncluding inpatient care. Our treatment group of 19–25 year-olds ontain ages when young adults leave school and are new work- rs, thus we turn next to testing the time trends of outcomes by ge prior to the law.

Following Akosa Antwi et al. (2013), one approach to increas- ng confidence in our ability to use those older than 26 to capture rends in hospitalizations that would have affected younger adults ad the policy not occurred would be to examine the trends of our utcomes of interest prior to the law’s implementation. If we were o find that hospitalization rates increased for the treatment group elative to the control group even prior to the policy, this would uggest that the impact we estimate could be a continuation of rior trends.

We first present visual pre-trends of main analysis in Figs. 1–4. n Fig. 1, we show trends for number of visits for our treatment and ontrol groups. The vertical lines represent the passage of the law, he implementation of the law in September 2010 and the start of

011, when most new health insurance plans start, respectively. e plot graphs for overall visits, visits through an ER, and visits

ot through an ER for our main sample and sample with mental llness diagnosis. The plots are quarterly unconditional means of

v v a i

rcent of admissions for each insurance type among 19–25-year-olds at the hospital type among 27–29-year-olds at the hospital level. (2) See Notes (1) and (3) under

he aforementioned variables. For all visits, visits that originated rom the ER, and visits not originating from the ER (top row), we bserve differences between the treatment and control groups in evels, but no visible differences in trends. There is evidence of dif- erential trends after the implementation of the law, especially in he bottom row graph for mental illness admissions through an ER.

In Fig. 2, we examine trends by health insurance status. In the rst row of plots, visual inspection of the first graph shows that he two groups follow a similar trend before the law, with a sharp ncrease in the share of treatment group visits paid by private insur- nce after the implementation. The second plot shows the trend or the uninsured. Again, the two groups follow a similar trend efore the law, with a sharp post-law decrease in the share of visits ccounted for by uninsured young adults who are targeted by the aw. In Figs. 3 and 4, we show trends for our treatment intensity ariables for all visits and mental illness visits. These plots show no trong pre-trends and no sharp changes post-law either.

We formalize our trend tests in Table 2 by estimating regres- ions with our outcomes of interest as left-hand-side variables.

e only use data prior to the enactment of the law. The right- and-side variables for this regression include the same control

ariables as our main model, described above, except that the key ariable of interest is an interaction between the linear time trend nd the treatment group dummy instead of the usual difference- n-difference variables. In the first panel of Table 2, which covers

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 177

F ates th c r-olds

a o t t f c g t t a t a e t s t a s t

6 c

d

h o v o m v d d d o

o e i o a b a

6 v

ig. 3. Treatment intensity of non-birth admissions. Notes: (1) The solid line indic onditions, and the dashed line indicates mean of each outcome among 27–29-yea

ll admissions and mental illness admissions, we show that none f the trends are statistically significantly different between the reatment and control groups in the period prior to the ACA. In he second and third panels, we report our pre-reform trend test or insurance results and intensity of treatment for both types of are. For all health insurance outcomes, our treatment and control roups follow similar pre-trends. For the intensity of treatment rends, we see one marginally statistically significantly different rend for log length of stay (LOS) when risk adjustment variables re included, and four statistically significantly different trends in erms of charges (both levels and logs), when risk adjustment vari- bles are excluded. Overall, these pre-trend tests are reassuring, specially for admissions volume, and health insurance composi- ion of the admissions. As additional checks on the validity of our tudy design, we also conduct placebo tests, and estimate models hat compare between 25 and 27 year olds, who are very close in ge, and thus less likely to be differentially affected by other forces uch as recession. All these analyses provide reassuring evidence hat the assumptions behind the DD study design are valid.

.3. Impact of the ACA young adult mandate on use of medical

are and source of admission

Table 3 contains regression results from Eq. (1), where the ependent variable is the number of admissions at the quarter,

m t

e mean of each outcome among 19–25-year-olds admitted for non-birth related admitted for non-birth related conditions. (2) See Notes (1) and (3) under Fig. 1.

ospital, and age level. The results indicate that the implementation f the law had a statistically significant impact on overall inpatient isits. Relative to a mean average quarterly visit per hospital per age f 7.27 prior to the passage of the law, the estimated ACA imple- entation coefficient of 0.253 indicates that the overall number of

isits increased by 3.5 percent. Our result on the impact of the ACA ependent coverage expansion on overall visits is comparable in irection with Anderson et al. (2012), who find that young adults ecrease their visits for non-birth inpatient visits when they age ut of insurance.

In considering the possible substitution away from ER care to utpatient care, we investigate the effect of the health insurance xpansion on the source of inpatient admission by estimating the mpact of the law on ER and non-ER admissions. In columns 2 and 3 f Table 3, we find that our overall result is driven by direct hospital dmissions that bypass the ER. Direct hospital admissions tend to e scheduled visits and are likely more sensitive to health insurance ccess, hence the larger percent increase is not surprising.

.4. Impact of the ACA young adult mandate on mental illness isits

We study mental health admissions separately because they ay be especially impacted by the ACA young adult mandate given

he particular health needs of this population. We present our

178 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Fig. 4. Treatment intensity of mental illness admissions. Notes: (1) The solid line indicates the mean of each outcome among 19–25-year-olds admitted for mental illness, a tted fo

a t a a ( i b t r p n

6 i

h t f t t o e T t

i a a q a o e e t h 1 c a c f

t m p a h

nd the dashed line indicates mean of each outcome among 27–29-year-olds admi

nalysis of the impact of the mandate on the total number of men- al health admissions in the last three columns of Table 3. We find

statistically significant 9 percent increase in the number of visits fter implementation. These results differ from the Meara et al. 2014) finding for Massachusetts that shows no or small decreases n mental health inpatient visits after reform. Parsing our results y the source of admission, we find that visits originating from he ER comprise a significant portion of our overall mental health esults. Our estimate for non-ER admissions after the law, while ositive and marginally statistically significant, is smaller in mag- itude.

.5. Impact of the ACA young adult mandate on the health nsurance status of inpatient visits

In Table 4, we evaluate the impact of the mandate on the ealth insurance composition of young adults who sought inpa- ient care after the reform using the same sample and regression ramework we describe above. The dependent variable here is he fraction of the hospital-age-quarter-level inpatient popula- ion with private insurance, no insurance, Medicaid, Medicare, and

ther insurance. These health insurance categories are mutually xclusive so the horizontal sum of all the coefficients presented in able 4 is zero. Our DD estimate presented in the first column of he top panel of Table 4 shows that the proportion of young adults

d r p b

r mental illness. (2) See Notes (1) and (3) under Fig. 1.

n our treatment group with private insurance increased by 2.1 nd 6 percentage points, respectively, as a result of the enactment nd the implementation of the law. Relative to the mean baseline uarterly fraction of privately insured visits, these represent 5.3 nd 15.3 percent increases in private coverage of all visits among ur treatment group. This result is not surprising as the mandate xpanded health insurance coverage to young adults whose par- nts have private health insurance. Our next set of results shows hat the fraction of young adults in our treatment group without ealth insurance decreased by a marginally statistically significant

percentage point (4.4 percent) and statistically significant 2.9 per- entage points (12.5 percent) as a result of the law’s enactment nd implementation respectively. There is a statistically signifi- ant negative coefficient associated with reform implementation or Medicaid and other insurance coverage.

The bottom panel of Table 4 shows the impact of the law on he health insurance status of hospitalized young adults with a

ental illness diagnosis. In the first column, we estimate a 5.8 ercentage-point (17.7 percent) increase in the fraction of young dult mental-health-related inpatient visits paid through private ealth insurance. There is also suggestive evidence of a small

ecrease in the prevalence of uninsurance among mental-illness- elated inpatient care. We also find evidence of a large decline, 3.1 ercentage points (9.1 percent), in mental illness visits reimbursed y the Medicaid program. We find no meaningful impact of the

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 179

Table 2 Test for equality of pre-reform trends.

Number of admissions and admissions by source

Non-birth related admissions Mental illness admissions

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Interaction of time trend and a dummy variable for treatment group

−0.047 (0.046)

−0.050 (0.031)

0.003 (0.025)

−0.006 (0.031)

−0.008 (0.022)

0.002 (0.019)

Fraction of admissions by health insurance

Private insurance

Uninsured Medicaid Medicare Other insurance

Non-birth admissions Interaction of time trend and a dummy

variable for treatment group 0.000 (0.002)

−0.001 (0.002)

−0.002 (0.002)

0.001 (0.001)

0.002 (0.001)

Mental illness admissions Interaction of time trend and a dummy

variable for treatment group 0.005 (0.004)

−0.001 (0.003)

−0.001 (0.004)

0.000 (0.003)

−0.002 (0.002)

Intensity of treatment

LOS Log of LOS Number of procedures

Total charges Log of total charges

Non-birth admissions Risk-adjustment variables (including individuals’ demographic characteristics) are excluded Interaction of time trend and a dummy

variable for treatment group −0.022 (0.016)

−0.003 (0.002)

−0.004 (0.004)

−261.9*** (97.2)

−0.005** (0.002)

Risk-adjustment variables are included Interaction of time trend and a dummy

variable for treatment group −0.022 (0.014)

−0.003* (0.002)

−0.002 (0.004)

−216.4** (87.4)

−0.004** (0.002)

Mental illness admissions Risk-adjustment variables (including individuals’ demographic characteristics) are excluded Interaction of time trend and a dummy

variable for treatment group 0.040 (0.041)

0.000 (0.004)

0.001 (0.003)

85.1 (83.0)

0.003 (0.004)

Risk-adjustment variables are included Interaction of time trend and a dummy

variable for treatment group 0.054 (0.041)

0.002 (0.004)

0.001 (0.003)

115.8 (83.2)

0.004 (0.004)

Note: (1) First panel: Number of observations is 95,040 for the first three columns and 34,730 for the last three columns. Second panel: Number of observations is 70,782 for the first set of rows and 24,536 for the second set of rows. Third panel: Number of observations in the first set of rows is 731,727 in the first four columns and 729,116 in the last two columns, and the number of observations in the second set of rows is 122,546 in the first four columns and 122,275 in the last two columns. (2) First and second panels: Quarterly hospital-age level variables are calculated, using young adults aged from 19 to 29, except for the removal of 26 year olds who are in neither control nor treatment. Third panel: Individual level variables are used. (3) Data: The NIS data for the period from the first quarter of 2007 to the fourth quarter of 2009, which is prior to the passage of the ACA in March 2010. The data from California, Maine and Texas are excluded because precise information on age was not available. Observations in which length of stay exceeds 90 days are excluded. (4) Dependent variables in the first panel - column 1: number of all non-birth related admissions; column 2: number of non-birth related admissions originated from the ER; column 3: number of non-birth-related admissions that did not originate from the ER; column 4: number of total mental illness admissions; column 5: number of mental illness admissions originated from the ER; and column 6: number of mental illness admissions that did not originate from the ER. Second panel—column 1: ratio of non-birth admissions covered by private health insurance; column 2: ratio of uninsured non-birth admissions; column 3: ratio of non-birth admissions covered by Medicaid; column 4: ratio of non-birth admissions covered by Medicare; and column 5: ratio of non-birth admissions covered by other insurance. Third panel—column 1: length of stay; column 2: log of the sum of length of stay and one; column 3: number of procedure codes (up to six codes); column 4: total charges; and column 5: log of the sum of total charges and one. (5) Cells of the table contain: coefficients, and standard errors in parentheses. Coefficients are from the interaction of a dummy variable for treatment group and a linear measure for time trend (number of quarters since the first quarter of 2007), which shows whether there was a different time trend for the control vs. the treatment group in the period prior to policy enactment. (6) Other regressors are a linear time trend, a dummy variable for the treatment group, and all other explanatory variables included in our main specification. (7) Standard errors are clustered at the hospital level.

l c

6 i

i i t i

o a r t c s m

aw on Medicare or on other forms of insurance payment (last two olumns).

.6. Impact of the ACA young adult mandate on the treatment ntensity of inpatient visits

We study treatment intensity using a DD regression model that

s similar to Eq. (1) except that the unit of observation for this model s at the individual level rather than the hospital-quarter level. In he Massachusetts context, Kolstad and Kowalski (2012) argue that f health insurance expansion alters the observable characteristics

p E c a

f patients who seek care, then including these characteristics in regression framework would blunt any estimated impact of the eform. As a result they estimate models with and without con- rolling for patient demographics and clinical characteristics and onsider the model without patient characteristics as the preferred pecification. We do the same in our evaluation of the impact of the andate on treatment intensity. For models in which we include

atient characteristics, we include a new vector Zighst not present in q. (1). This vector contains patient demographics and clinical indi- ators such as age, gender, race/ethnicity, whether the patient was dmitted on the weekend, Charlson Index, the number of diagnosis

180 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Table 3 Effect of mandate on the number and sources of admissions.

All non-birth admission Mental illness admissions

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of admissions

Number of admissions through ER

Number of admissions not through ER

ACA enactment effect (2010 Q2–Q3) 0.124 (0.117)

0.047 (0.078)

0.077 (0.064)

0.032 (0.091)

−0.007 (0.067)

0.039 (0.052)

ACA implementation effect (2010 Q4–) 0.253**

(0.113) 0.092 (0.076)

0.162***

(0.061) 0.315***

(0.086) 0.234***

(0.061) 0.081*

(0.048)

Dependent variable means Treatment, before ACA enactment 7.27 4.75 2.52 3.52 2.09 1.43 Control, before 8.76 5.50 3.26 3.64 2.15 1.49 Treatment, after ACA implementation 7.70 5.02 2.69 4.08 2.41 1.67 Control, after 8.99 5.73 3.26 3.89 2.25 1.65

Notes: (1) Number of observations is 158,740 in the first three columns and 57,930 in the last three columns. (2) Quarterly hospital-age level variables are calculated using the non-birth related admissions (for columns 1–3) and mental illness admissions (for columns 4–6) of young adults aged from 19 to 29, except for the removal of 26 year olds who are in neither control nor treatment. (3) Cells of the table contain: coefficients and standard errors in parentheses. Coefficients in the first row are from the interaction of a dummy variable for treatment group (19–25 years old) and a dummy variable for the period after ACA enactment but before implementation (the second and third quarters of 2010); coefficients in the second row are from the interaction of a dummy variable for treatment group and a dummy variable for the period after ACA implementation (the fourth quarter of 2010 and onwards). (4) Data: The NIS data from 2007 to 2011. The data from California, Maine and Texas are excluded because precise information on age was not available. Observations in which length of stay exceeds 90 days are excluded. (5) Dependent variables—column 1: number of all non-birth related admissions; column 2: number of non-birth related admissions originated from the ER; column 3: number of non-birth-related admissions that did not originate from the ER; column 4: number of total mental illness admissions; column 5: number of mental illness admissions originated from the ER; and column 6: number of mental illness admissions that did not originate from the ER. (6) Other regressors are an indicator for each age, year fixed effects, quarterly fixed effects, state-specific linear time trend, hospital-specific fixed effects, quarterly national-level unemployment rate, interaction of unemployment and an indicator for each age. (7) Means of dependent variables are obtained for t ACA im t t leve

c m i

a

i

T F

N n o a i t

reatment and control groups before ACA enactment (2007 Q1–2010 Q1) and after he 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percen

odes (up to nine codes), an indicator for each of 27 comorbidity

easures calculated using the HCUP Comorbidity Software, and

ndicator variables for Major Diagnostic Categories (MDCs). As with our study of the number of hospitalizations, we consider

ll non-birth inpatient visits (Table 5) and visits related to mental

t l n e

able 4 raction of admissions insured: non-birth related admissions and mental illness admissio

All non-births admissions

Private insurance

Uninsured

ACA enactment effect (2010 Q2–Q3) 0.021***

(0.007) −0.010* (0.006)

ACA implementation effect (2010 Q4–) 0.060***

(0.005) −0.029*** (0.004)

Dependent variable means Treatment, before ACA enactment 0.392 0.228 Control, before 0.394 0.196 Treatment, after ACA implementation 0.398 0.211 Control, after 0.335 0.210

Mental illness admissions

Private insurance

Uninsured

ACA enactment effect (2010 Q2–Q3) 0.006 (0.011)

−0.001 (0.008)

ACA implementation effect (2010 Q4–) 0.058***

(0.009) −0.013* (0.007)

Dependent variable means Treatment, before ACA enactment 0.326 0.179 Control, before 0.273 0.165 Treatment, after ACA implementation 0.346 0.170 Control, after 0.224 0.179

otes: (1) Number of observations is 117,771 in the first set of rows and 40,469 in the sec on-birth related admissions (for the first panel) and mental illness admissions (for the s lds who are in neither control nor treatment. (3) Dependent variables—column 1: ratio dmissions; column 3: ratio of admissions covered by Medicaid; column 4: ratio of adm nsurance. (4) See Note (7) under Table 2 and Notes (3), (4), (6) and (7) under Table 3. *** S he 10 percent level.

plementation (2010 Q4 and onward). See Note (7) under Table 2. *** Significant at l.

llness (Table 6). Starting in columns 1 and 2 of Table 5, we find that

here is no statistically significant effect of the law on the levels or og transformation of length of stay. There is also no effect on the umber of procedures. We find marginally statistically significant ffect on total charges (but not log charges), with the enactment

ns.

Medicaid Medicare Other insurance

0.005 (0.006)

−0.012*** (0.004)

−0.003 (0.003)

−0.020*** (0.005)

−0.005 (0.003)

−0.006** (0.003)

0.265 0.040 0.070 0.256 0.086 0.062 0.276 0.042 0.066 0.291 0.093 0.064

Medicaid Medicare Other insurance

0.006 (0.011)

−0.006 (0.009)

−0.004 (0.006)

−0.031*** (0.009)

−0.003 (0.008)

−0.009* (0.005)

0.341 0.067 0.081 0.323 0.169 0.064 0.340 0.060 0.076 0.358 0.161 0.070

ond set of rows. (2) Quarterly hospital-age level variables are calculated using the econd panel) of young adults aged from 19 to 29, except for the removal of 26 year

of admissions covered by private health insurance; column 2: ratio of uninsured issions covered by Medicare; and column 5: ratio of admissions covered by other ignificant at the 1 percent level. ** Significant at the 5 percent level. * Significant at

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 181

Table 5 Effect of mandate on intensity of treatment for all non-births admissions.

LOS Log of LOS Number of procedures

Total charges Log of total charges

Risk-adjustment variables (including individuals’ demographic characteristics) are excluded ACA enactment effect (2010 Q2–Q3) 0.021

(0.046) 0.002 (0.005)

0.018 (0.012)

730.5* (410.2)

0.008 (0.007)

ACA implementation effect (2010 Q4–) −0.009 (0.037)

0.000 (0.005)

−0.003 (0.009)

−537.5* (304.2)

−0.009 (0.006)

Risk-adjustment variables are included ACA enactment effect (2010 Q2–Q3) 0.012

(0.043) 0.002 (0.005)

0.007 (0.011)

530.9 (369.7)

0.004 (0.006)

ACA implementation effect (2010 Q4–) −0.027 (0.031)

−0.003 (0.004)

0.008 (0.007)

−322.0 (259.6)

−0.002 (0.005)

Dependent variable means Treatment, before ACA enactment 4.38 1.39 1.08 24,298 9.54 Control, before 4.33 1.40 1.15 24,333 9.58 Treatment, after ACA implementation 4.36 1.39 1.08 28,972 9.71 Control, after 4.35 1.40 1.15 29,648 9.75

Notes: (1) Number of observations is 1,246,517 in the first four columns and 1,242,770 in the last two columns. (2) Observations are non-birth-related admissions of young adults aged from 19 to 29, except for the removal of 26 year olds who are in neither control nor treatment. Observations in which length of stay exceeds 90 days are excluded. (3) Dependent variables—column 1: length of stay; column 2: log of the sum of length of stay and one; column 3: number of procedure codes (up to six codes); column 4: total charges; and column 5: log of the sum of total charges and one. (4) Other regressors—an indicator for the period after ACA enactment but before implementation, an indicator for the period after ACA implementation, an indicator for treatment group, year-specific fixed effects, quarter-specific fixed effects, quarterly state-specific linear time trends, hospital-specific fixed effects, quarterly national unemployment rate, and interaction of unemployment rate and an indicator for each age are included in the both sets of regressions. Risk-adjusted variables that are included in the second set of the regressions are an indicator for each year of age, gender, race/ethnicity, an indicator f des (u H . (5) S a cent l

a O c m

i d n o f w a t a

7

t i y p g i (

T E

N a ( d t

or admission occurred on the weekend, Charlson Index, the number of diagnosis co CUP Comorbidity Software, and indicator variables for major diagnostic categories t the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 per

nd implementation showing opposite signed and small changes. ur results for total charges are sensitive to the inclusion of patient haracteristics and functional form assumption, but overall suggest inimal change. In Table 6, we present the impact of the ACA mandate on

ntensity of mental-health-related inpatient visits. Across the ifferent specifications, only two coefficients are statistically sig- ificant, and one of them is significant at the 10 percent level nly. Moreover, both results are small and are sensitive to unctional form (in one case the log specification is significant

hile the level is not, and vice versa in the other case). Over-

ll we find a small and mainly statistically insignificant effect of he young adult mandate on the treatment intensity of young dults.

s v c

able 6 ffect of mandate on intensity of treatment for mental illness admissions.

LOS Log of LOS

Risk-adjustment variables (including individuals’ demographic characteristics) are excluded ACA enactment effect (2010 Q2–Q3) −0.147

(0.131) −0.011 (0.011)

ACA implementation effect (2010 Q4–) −0.129 (0.103)

−0.015 (0.009)

Risk-adjustment variables are included ACA enactment effect (2010 Q2–Q3) −0.154

(0.124) −0.011 (0.011)

ACA implementation effect (2010 Q4–) −0.160 (0.099)

−0.018** (0.009)

Dependent variable means Treatment, before ACA enactment 6.49 1.73 Control, before 6.63 1.76 Treatment, after ACA implementation 6.40 1.73 Control, after 6.64 1.76

otes: (1) Number of observations is 214,785 in the first four columns and 214,403 in the ged from 19 to 29, except for the removal of 26 year olds who are in neither control n 3) Other regressors are the same as those listed in Note (4) under Table 4 except that w iagnostic categories. (4) See Note (7) under Table 2, Notes (3), (4), (5) and (7) under Table he 5 percent level. * Significant at the 10 percent level.

p to nine codes), an indicator for each of 27 comorbidity measures obtained by the ee Note (7) under Table 2 and Notes (3), (4), (5) and (7) under Table 3. *** Significant evel.

. Heterogeneous effect of the mandate

Given prior research findings from Massachusetts that men- al health results can vary by exact type, and that substance use s a closely related disorder that is also highly relevant to the oung adult population, we explore this heterogeneity in analysis resented in Table 7. We separated the MDC 19 mental health cate- ory into subcategories of depression, psychoses and other mental llness, using the relevant International Classification of Diseases ICD) 9 codes.

In Table 7, we see that for depression, there are sizable and tatistically significant increases in all inpatient admissions and isits through the ER. Using unreported baseline means, the coeffi- ient magnitudes are on the order of 8.5 percent and 11.3 percent

Number of procedures

Total charges Log of total charges

−0.007 (0.011)

−130.6 (284.8)

−0.005 (0.013)

0.008 (0.007)

−297.7 (220.6)

−0.003 (0.010)

−0.008 (0.011)

−150.2 (273.3)

−0.006 (0.012)

0.008 (0.007)

−362.6* (212.5)

−0.008 (0.009)

0.24 13,135 9.06 0.27 13,399 9.10 0.26 15,093 9.23 0.27 15,568 9.26

last two columns. (2) Observations are mental illness admissions of young adults or treatment. Observations in which length of stay exceeds 90 days are excluded. e include an indicator variable for DRGs instead of an indicator variable for major

3, and Note (3) under Table 4. *** Significant at the 1 percent level. ** Significant at

182 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Table 7 Effect of mandate on mental illness admissions subcategories, and substance abuse admissions.

Number of admissions

Number of admissions through ER

Number of admissions not through ER

Private insurance

Uninsured Medicaid Medicare Other insurance

Mental illness admissions Depression ACA enactment effect (2010

Q2–Q3) 0.010 (0.041)

−0.020 (0.030)

0.029 (0.028)

−0.005 (0.017)

0.014 (0.013)

−0.007 (0.018)

0.0015 (0.0090)

−0.005 (0.009)

ACA implementation effect (2010 Q4–)

0.096** (0.037)

0.072*** (0.027)

0.024 (0.022)

0.092*** (0.014)

−0.035*** (0.011)

−0.035*** (0.013)

0.0002 (0.007)

−0.018** (0.007)

Psychoses ACA enactment effect (2010

Q2–Q3) −0.032 (0.058)

−0.017 (0.042)

−0.015 (0.032)

0.002 (0.014)

0.013 (0.012)

0.003 (0.018)

−0.0015 (0.0156)

−0.015 (0.009)

ACA implementation effect (2010 Q4–)

0.167*** (0.049)

0.145*** (0.037)

0.022 (0.029)

0.035*** (0.011)

0.001 (0.009)

−0.025* (0.013)

−0.0109 (0.012)

−0.001 (0.006)

Other mental illness ACA enactment effect (2010

Q2–Q3) 0.055 (0.041)

0.030 (0.030)

0.025 (0.027)

0.033* (0.017)

−0.024* (0.014)

0.001 (0.019)

−0.0027 (0.0136)

−0.005 (0.011)

ACA implementation effect (2010 Q4–)

0.052 (0.033)

0.017 (0.024)

0.035* (0.020)

0.070*** (0.014)

−0.018 (0.011)

−0.038*** (0.014)

0.0008 (0.010)

−0.013* (0.008)

Substance abuse admissions ACA enactment effect (2010

Q2–Q3) 0.082 (0.060)

0.042 (0.031)

0.039 (0.046)

0.065*** (0.019)

−0.038** (0.018)

−0.020 (0.019)

−0.0201** (0.0097)

0.011 (0.011)

ACA implementation effect (2010 Q4–)

0.064 (0.047)

0.030 (0.028)

0.035 (0.033)

0.102*** (0.015)

−0.049*** (0.014)

−0.039*** (0.015)

−0.0092 (0.007)

−0.005 (0.008)

Note: (1) Number of observations is 57,930 in the first three columns and 27,924 in the last five columns in the first three panels. Number of observations is 62,970 in the first three columns and 22,295 in the last five columns in the last panel. (2) Quarterly hospital-age level variables are calculated using admissions related to each condition of young adults aged from 19 to 29, except for the removal of 26 year olds who are in neither control nor treatment. (3) Dependent variables—column 1: number of admissions; column 2: number of admissions originated from the ER; column 3: number of admissions that did not originate from the ER; column 4: ratio of admissions covered by p of adm a er Tab

i o s i c v i t h a t p a a M o o a i o c M

l g y a t a t o i t

r b E f i

8

v g c t c f p t e W s s t m q o t s a

rivate health insurance; column 5: ratio of uninsured admissions; column 6: ratio nd column 8: ratio of admissions covered by other insurance. (3) See Note (7) und

ncreases. There are also sizable insurance shifts in this category f hospitalizations, with private insurance increasing, and unin- urance and Medicaid decreasing by similar magnitudes; other nsurance also decreases. We also observe sizable increases in psy- hoses admissions after implementation, stemming mostly from isits through the ER. There is no statistically significant increase n other mental illness admissions, although insurance composi- ion shifts toward private coverage and away from Medicaid. The ealth insurance shifts are not surprising given the strong neg- tive link between mental health and employment that prior to he law, young adults who suffer from mental illness are likely to articipate in the Medicaid program. In Table 7, we also examine nother category of admission that is of high relevance for young dults: substance abuse admissions (Frank and McGuire, 2000; eara et al., 2014). Somewhat surprisingly, there is no indication

f increases in inpatient admissions, although consistent with all ur other findings, we document evidence of strong private insur- nce gains. The lack of increase in substance use admissions could ndicate that becoming insured allows better management of care utside the hospital for this disorder, as this is the category of are for which inpatient admissions decreased the most after the assachusetts health reform (Meara et al., 2014). Another dimension of heterogeneity we explore is at the age

evel. Although our main analysis uses treatment and control roups that are relatively close in age (19–25-year-olds vs. 27–29- ear-olds), a closer comparison would use those just below and bove the age cutoff of 26. Specifically, we evaluate the impact of he mandate by comparing 25-year-olds to 27-year-olds. These two ge groups are likely the most similar on observable characteris-

ics and the influence of macroeconomic factors. We re-estimate ur analysis reported in Table 3 (inpatient admissions volume) n Table 8, and find estimates that are consistent with but larger han our main results in Table 3. Compared to 27-year-olds and

p c o s

issions covered by Medicaid; column 7: ratio of admissions covered by Medicare; le 2 and Notes (3), (4), (6) and (7) under Table 3.

elative to the period before the law, among 25-year-olds, the num- er of admissions increased by 5.9 percent; admissions through the R increased by 5.3 percent, and admissions that did not originate rom the ER increased by 7.1 percent. The corresponding numbers n Table 3 are 3.5 percent, 6 percent and 9 percent.

. Further robustness checks

An implicit assumption we make in comparing the change in olume of inpatient admissions for the treatment versus control roup is that the relative cohort size of the two groups does not hange during our sample period. However, an alternative explana- ion for the 3.5-percent increase in the number of visits we estimate ould be that the population of 19-to-25-year-olds increased aster than the population of 25-to-29-year-olds. We verify that opulation trends do not drive our results by aggregating our data o the national level (by age and time) and dividing by population stimates available from U.S. Census Bureau from 2007 to 2011. e also use this specification to implement the standard error

olution suggested by Cameron et al. (2008) in the presence of a mall number of groups across which variation occurs. We present hese results in Appendix Table A1, along with a test of the com-

on trends assumption for this model. We see that the results are ualitatively very similar to those in Table 3. Statistical significance f results in Appendix Table A1 is slightly higher in one case, but he same in all other cases, relative to Table 3. Inpatient admis- ions now increase by 2.8 percent (relative to 3.5 percent earlier), nd mental health admissions increase by 7.3 percent (relative to 9

ercent earlier); results for other specifications are also similar. The ommon trends tests also remains fairly plausible, but there is now ne statistic (for admissions not through the ER) that is marginally tatistically significantly different.

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 183

Table 8 Effect of the mandate on the number and sources of all non-birth admissions among 25 and 27 years old.

Number of admissions

Number of admissions through ER

Number of admissions not through ER

ACA enactment effect (2010 Q2–Q3) 0.168 (0.152)

0.039 (0.108)

0.129 (0.094)

ACA implementation effect (2010 Q4–) 0.470*** (0.119)

0.273*** (0.092)

0.197*** (0.066)

Dependent variable means Treatment, before ACA enactment 7.90 5.14 2.76 Control, before 8.53 5.41 3.13 Treatment, after ACA implementation 8.25 5.40 2.85 Control, after 8.52 5.47 3.04

Notes: (1) Number of observations is 31,748. (2) Quarterly hospital-age level variables are calculated using the non-birth related admissions of young adults aged 25 and 27. ( lumn n Note ( p el.

s r 1 c i y i g y d f

a q J w a p r

m i a s o t O s o A s o s p e

i t w i w c I i 1 p t

c t n v a c

9

i a p f p t m c a n d o t t e m a y a a o o n c

i e t a i i

s (

2) Dependent variables—column 1: number of all non-birth related admissions; co umber of non-birth-related admissions that did not originate from the ER. (3) See ercent level. ** Significant at the 5 percent level. * Significant at the 10 percent lev

While in Table 8 we showed results for the 25 year old sub- et of the treatment group, in Appendix Tables A2 and A3, we show esults for further subsets. In Appendix Table A2 we find that among 9–22 year olds (relative to 27–29 year olds), there is a 3.9 per- ent increase in inpatient admissions and a 11.3 percent increase n mental health visits. Appendix Table A3 shows that among 23–25 ear olds, relative to 27–29 year olds, there is a 2.9 percent increase n inpatient visits and 5.8 percent increase in mental health, sug- esting that there are statistically significant effects among both ounger and older segments of the treatment group. There is evi- ence of common trends between treatment and control groups or all relevant statistics except one.

Last, we conduct a set of placebo tests where we estimate several dditional models falsely assuming that the reform took place in uarters prior to March 2010. For each of the 11 quarters between

anuary 2007 and March 2010 and using data from this time period, e re-estimate Eq. (1) assuming a placebo date for the ACA law

nd create a distribution of the results from the replications. We erform this test for both all and mental illness admissions, and eport results in Appendix Tables A4-A6.

We first examine the mean and standard deviation of the esti- ates reported in Appendix Table A4 relative to the values obtained

n Tables 2 and 3 for our sample of all and mental-health-related dmissions, for volume of visits as well as the insurance compo- ition of visits. When we consider our sample of all admissions, nly one specification out of eight has two or more placebo results hat are statistically significant at the 10-percent-or-smaller level. ne specification contains one placebo result that is statistically

ignificant at the 10% percent level. We note that the coefficients f placebo laws are smaller than those we obtain in Tables 3 and 4. mong the mental health illness categories, only four out of 88 pos- ible estimates are statistically significantly different from zero: ne at the 1-percent level, and three at the 10-percent level. Since ignificance at the 10 percent level contains zero, we interpret our lacebo analysis as suggesting that the impact of the law that we stimate is not due to chance.

As with our trends tests, even though our estimates of the mpact of the law on treatment intensity are sensitive to func- ional form assumptions and are mainly not statistically significant, e find instances of statistically significant tests for our treatment

ntensity variables in our placebo analysis. In Appendix Table A6, e show that in total there are 20 placebo results that are statisti-

ally significant at the 5% or lower level, out of a possible 110 cases. n Appendix Table 7, we present placebo tests for the treatment

ntensity of mental illness admissions. We see here that there are 5 placebo tests that show statistically significant effects, out of a ossible 110. In summary, the validity tests in this section confirm hat our study design is reasonable and that our estimates likely

i W m u

2: number of non-birth related admissions originated from the ER; and column 3: 7) under Table 2 and Notes (3), (4), (6) and (7) under Table 3. *** Significant at the 1

apture the effects of the mandate rather than the difference in rends between the treatment and control groups. Even though the umber of estimates that are significant for our treatment intensity ariables is not very far from what would be expected by chance lone, we note that these are also the outcomes that show the least onsistent effects due to reform.

. Discussion and conclusion

We present the first estimates of the impact of the ACA health nsurance expansion on inpatient medical care received by young dults using a nationally representative database of inpatient hos- ital visits. We estimate the impact of the law on 19–25-year-olds, or whom access to parental insurance coverage was especially oor prior to the ACA. We find evidence of greater demand response o health insurance coverage for mental health care than for general

edical care. Compared to another group of individuals who are lose in age but excluded from the law (27–29-year-olds), young dults who benefitted from the mandate increased their overall umber of non-birth inpatient visits by 3.5 percent. This result is riven mainly by admissions not through the ER. In our exploration f the impact of the mandate on mental health visits, an impor- ant component of inpatient care use by young adults, we find hat young adults increased their visits by 9 percent. Unlike gen- ral admissions, the increase in mental health admissions arises ainly through the ER. We also find strong evidence that the law

chieved its intended purpose of decreasing uninsurance among oung adults who use hospital-based care. The fraction of young dult hospitalizations that is uninsured decreased by 2.9 percent- ge points (12.5 percent) due to the law. Our analysis of the impact f the mandate on the intensity of medical care shows no consistent r significant change in how long young adults stay in a hospital, the umber of procedures that are performed on them, or on hospital harges.

Our findings contribute to the literature on the effect of health nsurance coverage on access and medical care utilization. The gen- ral consensus in the literature is that health insurance increases he use of medical services, including inpatient care. Our findings re consistent with this hypothesis as we find that young adults ncreased their overall use of inpatient services in response to gain- ng access to health insurance.

However, our results differ from the experience of Mas- achusetts which saw no change in general inpatient admissions Kolstad and Kowalski, 2012) and no change or small decreases

n use of mental health care by young adults (Meara et al., 2014).

e note that the health insurance expansion we study here is uch smaller than other large-scale efforts such as the near-

niversal health insurance expansion in Massachusetts. In addition,

184 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Table A1 DD results using aggregated quarterly data and wild cluster bootstrap-t procedure.

Non-birth related Mental illness

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of admissions

Number of admissions through ER

Number of admissions not through ER

Effect of mandate on the number and sources of admissions ACA enactment effect (2010 Q2–Q3) 0.046

(0.602) −0.003 (0.876)

0.049 (0.532)

0.004 (0.882)

−0.009 (0.508)

0.013 (0.256)

ACA implementation effect (2010 Q4–) 0.186*** (0.002)

0.050 (0.204)

0.136*** (0.002)

0.087*** (0.000)

0.067*** (0.000)

0.021* (0.078)

Dependent variable means Treatment, before ACA enactment 6.74 4.42 2.33 1.20 0.71 0.48 Control, before 8.42 5.30 3.12 1.28 0.76 0.52 Treatment, after ACA implementation 6.72 4.38 2.35 1.29 0.77 0.53 Control, after 8.15 5.19 2.96 1.29 0.74 0.54

Notes: (1) The number of observations is 200. (2) Outcome variables are the number of admissions per 1000 people for each age per quarter. (3) P-values are in parentheses. We cluster on year-quarter and perform wild cluster bootstrap-t test with 999 replications, following Cameron et al. (2008).

Non-birth related admissions Mental illness admissions

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Test for equality of pre-reform trends for number of admissions Interaction of time trend and a dummy

variable for treatment group 0.015 (0.022)

−0.009 (0.016)

0.024* (0.013)

0.003 (0.008)

0.000 (0.006)

0.003 (0.005)

Notes: (1) The number of observations is 200. (2) See Notes (2) and (3) under the panel above.

Table A2 19–22 years old vs. 27–29 years old: extensive margin results and trend test.

Non-birth related Mental illness

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of admissions

Number of admissions through ER

Number of admissions not through ER

Effect of mandate on the number and sources of admissions ACA enactment effect (2010 Q2–Q3) 0.170

(0.144) 0.073 (0.096)

0.097 (0.078)

0.070 (0.108)

0.047 (0.080)

0.023 (0.061)

ACA implementation effect (2010 Q4–) 0.276* (0.142)

0.079 (0.095)

0.197*** (0.076)

0.398*** (0.099)

0.280*** (0.072)

0.118** (0.053)

Dependent variable means Treatment, before ACA enactment 6.98 4.56 2.43 3.51 2.07 1.44 Control, before 8.76 5.50 3.26 3.64 2.15 1.49 Treatment, after ACA implementation 7.48 4.83 2.65 4.20 2.47 1.74 Control, after 8.99 5.73 3.26 3.89 2.25 1.65

Notes: (1) Number of observations is 111,118 in the first three columns and 40,551 in the last three columns. (2) Quarterly hospital-age level variables are calculated using the non-birth related admissions (for columns 1–3) and mental illness admissions (for columns 4–6) of young adults aged from 19 to 22, and 27–29 year olds. (3) Cells of the table contain: coefficients and standard errors in parentheses. Coefficients in the first row are from the interaction of a dummy variable for treatment group (19–22 years old) and a dummy variable for the period after ACA enactment but before implementation (the second and third quarters of 2010); coefficients in the second row are from the interaction of a dummy variable for treatment group and a dummy variable for the period after ACA implementation (the fourth quarter of 2010 and onwards). (4) See Note (7) under Table 2 and Notes (4)–(7) under Table 3.

Non-birth related admissions Mental illness admissions

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Test for equality of pre-reform trends for number of admissions Interaction of time trend and a dummy

variable for treatment group −0.038 (0.056)

−0.047 (0.038)

0.008 (0.029)

0.004 (0.037)

−0.005 (0.025)

0.009 (0.023)

Notes: (1) Number of observations is 66,528 in the first three columns and 24,311 in the last three columns. (2) See Notes (3), (5), (6) and (7) under Table 2, Note (5) under Table 3, Note (2) under the panel above.

Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187 185

Table A3 23–25 years old vs. 27–29 years old: extensive margin results and trend test.

Non-birth related Mental illness

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of admissions

Number of admissions through ER

Number of admissions not through ER

Effect of mandate on the number and sources of admissions ACA enactment effect (2010 Q2–Q3) 0.063

(0.104) 0.012 (0.074)

0.051 (0.059)

−0.018 (0.096)

−0.079 (0.070)

0.061 (0.057)

ACA implementation effect (2010 Q4–) 0.223** (0.093)

0.108* (0.065)

0.115** (0.053)

0.205** (0.089)

0.173*** (0.062)

0.033 (0.056)

Dependent variable means Treatment, before ACA enactment 7.64 5.00 2.64 3.52 2.10 1.42 Control, before 8.76 5.50 3.26 3.64 2.15 1.49 Treatment, after ACA implementation 8.00 5.26 2.74 3.92 2.33 1.58 Control, after 8.99 5.73 3.26 3.89 2.25 1.65

Notes: (1) Number of observations is 95,244 in the first three columns and 34,758 in the last three columns. (2) Quarterly hospital-age level variables are calculated using the non-birth related admissions (for columns 1–3) and mental illness admissions (for columns 4–6) of young adults aged from 23 to 29, except for the removal of 26 year olds who are in neither control nor treatment. (3) Cells of the table contain: coefficients and standard errors in parentheses. Coefficients in the first row are from the interaction of a dummy variable for treatment group (23–25 years old) and a dummy variable for the period after ACA enactment but before implementation (the second and third quarters of 2010); coefficients in the second row are from the interaction of a dummy variable for treatment group and a dummy variable for the period after ACA implementation (the fourth quarter of 2010 and onwards). (4) See Note (7) under Table 2 and Notes (4)–(7) under Table 3.

Non-birth related admissions Mental illness admissions

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Number of all admissions

Number of admissions through ER

Number of admissions not through ER

Test for equality of pre-reform trends for number of admissions Interaction of time trend and a dummy

variable for treatment group −0.060 (0.039)

−0.055** (0.027)

−0.005 (0.022)

−0.020 (0.032)

−0.011 (0.023)

−0.008 (0.020)

Notes: (1) Number of observations is 57,024 in the first three columns and 20,838 in the last three columns. (2) See Notes (3), (5), (6) and (7) under Table 2, Note (5) under Table 3, Note (2) under the panel above.

Table A4 Effects of placebo laws on the number of admissions.

Distribution of the coefficients of the placebo laws

Number of coefficient estimates that are significant in the placebo law regressions (out of 11 estimates for each row)

Estimated effects in the main specification

Mean Standard deviation

1% level 5% level 10% level Enactment effect (2010 Q2-Q3)

Implementation effect (2010 Q4-)

Non-birth related admissions Number of total admissions −0.026 0.114 0 0 0 0.124 0.253** Number of admissions through ER −0.028 0.119 1 0 2 0.047 0.092 Number of admissions not through ER 0.002 0.056 0 0 0 0.077 0.162*** Rate of private health insurance 0.000 0.004 0 0 0 0.021*** 0.060*** Rate of uninsured 0.000 0.004 0 0 0 −0.010* −0.029*** Rate of medicaid −0.002 0.005 0 0 0 0.005 −0.020*** Rate of medicare 0.001 0.004 0 0 0 −0.012*** −0.005 Rate of other insurance 0.001 0.004 0 0 1 −0.003 −0.006** Mental illness admissions Number of total admissions 0.000 0.072 0 0 0 0.032 0.315*** Number of admissions through ER −0.008 0.056 0 0 0 −0.007 0.234*** Number of admissions not through ER 0.009 0.044 0 0 0 0.039 0.081* Rate of private health insurance 0.005 0.013 0 0 1 0.006 0.058*** Rate of uninsured −0.001 0.014 1 0 0 −0.001 −0.013* Rate of medicaid −0.002 0.014 0 0 0 0.006 −0.031*** Rate of medicare −0.001 0.012 0 0 1 −0.006 −0.003 Rate of other insurance −0.001 0.009 0 0 1 −0.004 −0.009*

Note: (1) Data: The NIS data for the period from the first quarter of 2007 to the fourth quarter of 2009, which is prior to the passage of the ACA in March 2010. The data from California, Maine and Texas are excluded because precise information on age was not available. (2) We select each possible quarter between the second quarter of 2007 and the fourth quarter of 2009 one at a time. We then estimate the main model using each separate placebo date for defining the “Implement” variable. We show here the means and standard deviations of the coefficients we obtain. (3) The last two columns repeat estimates from Tables 3 and 4 for comparison.

186 Y. Akosa Antwi et al. / Journal of Health Economics 39 (2015) 171–187

Table A5 Effects of placebo laws on treatment intensity, non-birth related admissions.

Distribution of the coefficients of the placebo laws

Number of coefficient estimates that are significant in the placebo law regressions (out of 11 estimates for each row)

Estimated effects in the main specification

Mean Standard deviation

1% level 5% level 10% level Enactment effect (2010 Q2–Q3)

Implementation effect (2010 Q4–)

Risk-adjustment variables (including individuals’ demographic characteristics) are excluded Length of stay (LOS) −0.017 0.047 0 0 0 0.021 −0.009 Log of LOS −0.001 0.008 0 1 0 0.002 0.000 Number of procedures −0.003 0.017 0 1 0 0.018 −0.003 Total charges −171.2 567.3 1 2 1 730.5* −537.5* Log of total charges −0.002 0.015 2 3 0 0.008 −0.009 Risk-adjustment variables are included Length of stay (LOS) −0.017 0.046 0 0 0 0.012 −0.027** Log of LOS −0.002 0.008 0 2 1 0.002 −0.003** Number of procedures −0.002 0.013 0 1 0 0.007 0.008 Total charges −152.9 436.2 0 2 1 530.9 −322.0 Log of total charges −0.002 0.014 1 4 0 0.004 −0.002

Note: (1) See Notes (1) and (2) under Appendix Table A4. (2) The last two columns repeat estimates from Table 5 for comparison.

Table A6 Effects of placebo laws on treatment intensity, mental illness admissions.

Distribution of the coefficients of the placebo laws

Number of coefficient estimates that are significant in the placebo law regressions (out of 11 estimates for each row)

Estimated effects in the main specification

Mean Standard deviation

1% level 5% level 10% level Enactment effect (2010 Q2–Q3)

Implementation effect (2010 Q4–)

Risk-adjustment variables (including individuals’ demographic characteristics) are excluded Length of stay (LOS) 0.019 0.196 0 0 0 −0.147 −0.129 Log of LOS 0.001 0.017 0 0 1 −0.011 −0.015 Number of procedures 0.001 0.016 1 1 0 −0.007 0.008 Total charges 63.9 357.6 0 1 0 −130.6 −297.7 Log of total charges 0.003 0.024 0 1 0 −0.005 −0.003 Risk-adjustment variables are included Length of stay (LOS) 0.026 0.238 0 0 1 −0.154 −0.160 Log of LOS 0.002 0.016 0 1 0 −0.011 −0.018** Number of procedures 0.001 0.016 1 1 0 −0.008 0.008 Total charges 79.7 399.6 0 1 1 −150.2 −362.6* Log of total charges 0.004 0.023 1 1 2 −0.006 −0.008

N eat e

w s t c c r c s c i m c

h w w s n u p m

s c

A

a B L K o R A H t C

ote: (1) See Notes (1) and (2) under Appendix Table 4. (2) The last two columns rep

e study the short-run effect of the ACA health insurance expan- ion on young adults, leaving for future work the examination of he long-run effects. The findings by Meara et al. (2014) of no hange or decreases in the use of hospital-based mental health are by young adults after the Massachusetts health insurance eform could reflect the availability of outpatient mental health are providers in Massachusetts that might be absent in other tates. This means that the interaction of demand and supply-side onstraints could determine the long and short run impact of health nsurance expansions. Cunningham (2009) finds that shortage of

ental health care providers is as much a barrier to mental health are access as lack of health insurance coverage.

There are several other limitations to our work. Since we use ospital discharge data that do not contain individual identifiers, e are unable to distinguish whether the increase in inpatient visits e document is the result of increased frequency of visits by the

ame patients or an increase in visits by new patients. We also do

ot observe visits to private mental health hospitals. Future work sing household- or individual-level surveys as well as more com- rehensive hospital data sets will help us better understand the echanism behind the increase in inpatient care use found in this

f I t (

stimates from Table 6 for comparison.

tudy, as well as shed light on the effects of the law on the use of are in other settings.

cknowledgements

We are grateful for helpful comments from the editor, an nonymous referee, Jim Marton, Ellen Meara, Lauren Nicholas, en Sommers, and seminar audiences at the Drexel University, ehigh University, University of California Irvine, University of ansas, Ball State University, University of Cincinnati, University f Mississippi, University of New Mexico, Agency for Healthcare esearch and Quality, RAND, 2013 Association of Public Policy nalysis and Management conference (APPAM), 2014 Midwest ealth Economics Conference (MHEC), the 23rd Annual Meeting of

he Midwest Econometrics Group, and IUPUI-IUB Mini-Conference. ontact information for corresponding author: Kosali Simon, Pro-

essor, The School of Public and Environmental Affairs (SPEA), ndiana University, Rm 359, 1315 East Tenth Street Blooming- on, IN 47405-1701, [email protected], +1 812-856-3850 w).

Health

A

R

A

A

A

B

B

B

C

C

C

C

C

D

D

F

F

G

G

G

H

H

H

K

K

L

M

M

M

M

M

M

P

R

S

S

S

T

T

W young. Wall Street Journal, Available at: http://online.wsj.com/news/articles/

Y. Akosa Antwi et al. / Journal of

ppendix.

eferences

kosa Antwi, Y., Moriya, A.S., Simon, K., 2013. Effects of federal policy to insure young adults: evidence from the 2010 affordable care act dependent coverage mandate. American Economic Journal: Economic Policy 5 (4), 1–28.

nderson, M., Dobkin, C., Gross, T., 2012. The effect of health insurance coverage on the use of medical services. American Economic Journal: Economic Policy 4 (1), 1–27.

nderson, M.L., Dobkin, C., Gross, T., 2014. The Effect of health insurance on emer- gency department visits: evidence from an age-based eligibility threshold. Review of Economics and Statistics 96 (1), 189–195.

arry, C.L., Ridgely, M.S., 2008. Mental health and substance abuse insurance parity for federal employees: how did health plans respond? Journal of Policy Analysis and Management 27 (1), 155–170.

ertrand, M., Duflo, E., Mullainathan, S., 2004. How much should we trust differences-in-differences estimates? Quarterly Journal of Economics 119 (1), 249–275.

uchmueller, T.C., Cooper, P.F., Jacobson, M., Zuvekas, S.H., 2007. Parity for whom? Exemptions and the extent of state mental health parity legislation. Health Affairs 26 (4), w483–w487.

ameron, A.C., Gelbach, J.B., Miller, D.L., 2008. Bootstrap-based improvements for inference with clustered errors. Review of Economics and Statistics 90 (3), 414–427.

antor, J.C., Monheit, A.C., DeLia, D., Lloyd, K., 2012. Early impact of the affordable care act on health insurance coverage of young adults. Health Services Research 47 (5), 1773–1798.

ard, D., Dobkin, C., Maestas, N., 2008. The impact of nearly universal insurance cov- erage on health care utilization and health: evidence from Medicare. American Economic Review 98 (5), 2242–2258.

awley, J., Moriya, A.S., Simon, K., 2013. The impact of the macroeconomy on health insurance coverage: evidence from the great recession. Health Economics, http://dx.doi.org/10.1002/hec.3011.

unningham, P.J., 2009. Beyond parity: primary care physicians’ perspectives on access to mental health care. Health Affairs 28 (3), w490–w501.

afny, L., Gruber, J., 2005. Public insurance and child hospitalizations: access and efficiency effects. Journal of Public Economics 89 (1), 109–129.

oyle Jr., J.J., 2005. Health insurance, treatment and outcomes: using auto accidents as health shocks. Review of Economics and Statistics 87 (2), 256–270.

inkelstein, A., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J.P., Allen, H., Baicker, K., 2012. The Oregon health insurance experiment: evidence from the first year? Quarterly Journal of Economics 127 (3), 1057–1106.

rank, R.G., McGuire, T.G., 2000. Economics and mental health. Handbook of Health Economics 1, 893–954.

arfield, R.L., Zuvekas, S.H., Lave, J.R., Donohue, J.M., 2011. The impact of national health care reform on adults with severe mental disorders. American Journal of Psychiatry 168 (5), 486–494.

oldman, W., McCulloch, J., Sturm, R., 1998. Costs and use of mental health services

before and after managed care. Health Affairs 17 (2), 40–52.

rant, J.E., Potenza, M.N. (Eds.), 2010. Young Adult Mental Health. Oxford University Press.

ealth Care Cost Institute (HCCI), 2014. Impact of the Mental Health Par- ity and Addiction Equity Act on Inpatient Admissions, Available at:

Y

Economics 39 (2015) 171–187 187

http://www.healthcostinstitute.org/news-and-events/issue-brief-impact- mental-health-parity-and-addiction-equity-act-inpatient-admission

CUP Nationwide Inpatient Sample (NIS), 2011. Healthcare Cost and Utilization Project (HCUP). Agency for Healthcare Research and Quality, Rockville, MD www.hcup-us.ahrq.gov/nisoverview.jsp

organ, C.M., 1986. The demand for ambulatory mental health services from spe- cialty providers. Health Services Research 21 (2), 291–320.

olstad, J.T., Kowalski, A.E., 2012. The impact of health care reform on hospital and preventive care: evidence from Massachusetts. Journal of Public Economics 96 (11), 909–929.

uhn, A., Lalive, R., Zweimüller, J., 2009. The public health costs of job loss. Journal of Health Economics 28 (6), 1099–1115.

uo, F., Florence, C.S., Quispe-Agnoli, M., Ouyang, L., Crosby, A.E., 2011. Impact of business cycles on US suicide rates, 1928–2007. American Journal of Public Health 101 (6).

aclean, J.C., 2013. The health effects of leaving school in a bad economy. Journal of Health Economics 32 (5), 951–964.

anning, W.G., Wells, K., Buchanan, J., 1989. Effects of Mental Health Insurance: Evidence from the Health Insurance Experiment. RAND, Santa Monica, CA.

eara, E., Golberstein, E., Zaha, R., Greenfield, S.F., Beardslee, W.R., Busch, S.H., 2014. Use of hospital-based services among young adults with behavioral health diag- noses before and after health insurance expansions. JAMA Psychiatry 71 (4), 404–411.

iller, S., 2012. The effect of insurance on emergency room visits: an analysis of the 2006 Massachusetts health reform. Journal of Public Economics 96 (11), 893–908.

ojtabai, R., 2007. Americans’ attitudes toward mental health treatment seeking: 1990–2003. Psychiatric Services 58 (5), 642–651.

ulcahy, A., Harris, K., Finegold, K., Kellermann, A., Edelman, L., Sommers, B.D., 2013. Insurance coverage of emergency care for young adults under health reform. New England Journal of Medicine 368 (22), 2105–2112.

atel, V., Flisher, A.J., Hetrick, S., McGorry, P., 2007. Mental health of young people: a global public-health challenge. The Lancet 369 (9569), 1302–1313.

oberts, B., Terrell, D., 2014. A cohort component analysis of the 2007–2009 reces- sion. Monthly Labor Review.

ommers, B.D., Kronick, R., 2012. The affordable care act and insurance coverage for young adults. Journal of the American Medical Association 307 (9), 913–914.

ommers, B.D., Buchmueller, T., Decker, S.L., Carey, C., Kronick, R., 2013. The afford- able care act has led to significant gains in health insurance and access to care for young adults. Health Affairs 32 (1), 165–174.

turm, R., 1997. How expensive is unlimited mental health care coverage under managed care? Journal of the American Medical Association 278 (18), 1533–1537.

aube, C.A., Kessler, L.G., Burns, B.J., 1986. Estimating the probability and level of ambulatory mental health services use. Health Services Research 21 (2), 321–340.

efft, Nathan, 2011. Insights on unemployment, unemployment insurance, and mental health. Journal of Health Economics 30.2, 258–264.

eaver, C., Radnofsky, L., 2013. New health-care law’s success rests on the

SB10001424127887324263404578613700273320428 (accessed 26.11.13). u, J.W., Adams, S.H., Burns, J., Brindis, C.D., Irwin Jr., C.E., 2008. Use of mental health

counseling as adolescents become young adults. Journal of Adolescent Health 43 (3), 268–276.