Private Equity Acquisition Of Substance Use Treatment Centers Increases Probability Of Public Health Insurance Acceptance

Abstract

Private equity (PE) firms have actively acquired substance use disorder treatment facilities in the last decade. Evidence on whether these acquisitions affect establishment operations is limited. Using a novel catalog of treatment facilities acquired by PE firms and a difference-in-differences research design, we identified a relative increase in the probability that establishments accept public health insurance after PE acquisition. We did not observe a differential change in the probability that the facility offers common forms of medication treatment. Our results suggest that PE acquisitions may increase participation in public insurance programs to expand the revenue sources at acquired facilities without necessarily changing the scope of service lines available.

In 2021, 48.7 million adults and adolescents in the United States reported behavior consistent with a substance use disorder. Despite this, 85.1 percent of these people did not receive care in a specialized substance use disorder (SUD) treatment facility. Cost concerns were among the most commonly reported reasons among this population for not seeking care, with 47.9 percent of people aged 18 or older believing that insurance would not cover treatment.¹ Insofar as insurance is an arbiter of care affordability, identifying barriers to insurance acceptance may be an important lever for addressing SUD in the United States.

Private equity (PE) firm acquisitions of SUD treatment establishments have grown in recent years, with approximately 7 percent of all such facilities being owned by a PE firm by 2023.² Although PE’s penetration into other areas of care delivery has been well documented, its effects on SUD treatment establishments is less well understood.³ PE acquisition of physician practices has resulted in changes in organizational scale and care spending, likely in an effort to increase profitability.^4,5 These practices are often acquired as part of a “rollup” strategy, where multiple establishments are acquired in a single market to strengthen bargaining leverage with private insurers.⁶ A similar motivation may underlie the geographically concentrated nature of PE acquisitions of SUD treatment establishments.^7,8

PE firms may further expand profitability by making organizational changes that allow for reimbursement from public health insurers. Payments from traditional Medicare and Medicaid for SUD treatment are not negotiated, and instead follow a fixed reimbursement schedule.⁹ Although these reimbursements are often smaller than those from private insurers, these government programs have become central to SUD care settings. In 2009, 54 percent of SUD treatment facilities accepted Medicaid, and only one-third accepted Medicare.¹⁰ By 2019, Medicaid was the most common form of insurance among those with a diagnosed SUD.¹¹ In addition, 1.1 million Medicare beneficiaries have been diagnosed with opioid use disorder (OUD) as of 2023.¹²

PE firms often predicate acquisitions on the basis of claims of improving care quality, although evidence on this is mixed.¹³ In the context of hospitals, PE firm acquisitions have led to a prioritization of especially profitable service lines.^14,15 This could similarly manifest in SUD care settings if medication treatment is prioritized over services that might not be reimbursable, such as vocational support or twelve-step programs. This is significant because medication treatment is the primary evidence-based form of OUD treatment, yet its adoption has not kept up with the growth rate of OUD diagnosis in the US.¹⁶ Expanding the availability of buprenorphine and methadone, two drugs primarily associated with medication treatment, could represent an area in which financial incentives are aligned with increasing access to evidence-based treatment.

This study brings evidence to bear on these questions by linking a catalog of PE-acquired SUD treatment facilities to a detailed panel of establishment surveys. We document changes in insurance acceptance and changes in the availability of medication treatments spurred by PE acquisition. Although facilities are not randomly exposed to PE ownership, the exact timing of the acquisition is plausibly exogenous. Thus, we measure the differential trends in outcomes between facilities that are acquired in a particular year, relative to a matched set of comparison facilities.

Study Data And Methods

Data Sources And Sample

Operations of SUD treatment establishments were observed between 2009 and 2022, using the annual National Survey of Substance Abuse Treatment Services (N-SSATS) produced by the Substance Abuse and Mental Health Services Administration (SAMHSA). The N-SSATS data are the highest-quality publicly available data regarding facility treatment offerings, accepted forms of payment, and facility accreditations. The N-SSATS sampling frame includes “all facilities in the United States, both public and private, that provide substance abuse treatment.” recognized by SAHMSA in the Inventory of Substance Use and Mental Health Treatment Facilities. Surveys are sent to all establishments in this inventory, where establishments are distinguished by physical address. The N-SSATS had a 92 percent average survey response rate during our sample period.¹⁷

Establishments have three means of entering the N-SSATS sampling frame. Most enter through administrative data-sharing efforts between SAHMSA and state substance use agencies that are responsible for accreditation and monitoring of such establishments. Second, establishments can independently seek inclusion in the federal inventory, potentially to advertise on the SAHMSA-operated platform FindTreatment.gov, which launched in 2019.¹⁸ Finally, SAHMSA verifies the completeness of their inventory using other surveys and external data sets, such as those produced by the American Hospital Association. Starting in 2021, the N-SSATS was renamed the National Substance Use and Mental Health Services Survey, and the sampling frame was adjusted to additionally include establishments that provide mental health care without necessarily offering SUD treatment. The change in sampling frame is not pertinent to this research, as these new establishments, by virtue of not providing SUD treatment, are not eligible for inclusion in the analysis sample. For convenience, we refer to both data sets as the N-SSATS.

We compiled a catalog of PE acquisitions of SUD treatment establishments between 2014 and 2022 from two proprietary data sources: Pitchbook Inc. and the Irving Levin Associates Healthcare M&A database. Acquisitions of SUD treatment facilities were identified according to key word searches and validated using contemporaneous press releases and news media coverage. Further details regarding the catalog’s creation are provided in the online appendix.¹⁹ If an establishment was involved in more than one transaction during our study period, we focused on the first acquisition event. If a PE-controlled platform company opens a new establishment, that establishment has no preacquisition period, and so it was also excluded from the analysis. The final acquisition catalog consisted of 1,182 PE transactions of 924 establishments.

To create our analysis sample, we began by linking this acquisitions catalog to the N-SSATS according to a combination of physical address, geographic coordinates, and establishment name. Once all string matches had been identified, to account for potential measurement error in data reporting, we “fuzzy matched” addresses that have a Levenshtein distance of fewer than five characters. Further details regarding the linkage are in the appendix.¹⁹

In total, 657, or 71 percent, of all PE-acquired establishments can be found in the N-SSATS. Appendix exhibit A1 compares the socioeconomic characteristics of the areas surrounding PE-acquired facilities, distinguishing between those that can and those that cannot be linked to the N-SSATS.¹⁹ We defined the surrounding area using the Census Bureau’s ZIP Code Tabulation Area boundaries. Demographic characteristics were sourced from the Census Bureau’s five-year American Community Survey. Relative to establishments that could not be linked to the N-SSATS, linked establishments were in lower-income areas with lower rates of bachelor’s degree attainment and a higher prevalence of Medicaid enrollment. Beyond these exceptions, facilities that can and cannot be linked to the N-SSATS are in substantively similar areas.

Of the 657 acquired facilities that could be linked, we omitted those identified in the N-SSATS as being part of hospitals, as their financing structure differs markedly from establishments that primarily treat SUD. Because our research design relied on observing an establishment both before and after its acquisition, we further omitted those that are not observed in both periods. Omitted establishments were in areas that are slightly more affluent, with a lower proportion of non-White residents (appendix exhibit A2).¹⁹ These facilities were also more likely to accept some form of health insurance and less likely to offer medications for opioid use disorder (MOUD).

Appendix exhibit A3 presents the frequency of linked PE-acquired establishments by calendar year of acquisition.¹⁹ We matched each acquired establishment with up to six comparison establishments by state and ZIP Code Tabulation Area uninsurance rate among residents ages 18–64 years.. Focusing on those establishments operating before 2014, the baseline period before the first PE acquisitions in our data, matched comparison facilities are in demographically similar areas, are slightly more likely to accept a major form of health insurance, and are less likely to offer some form of MOUD (appendix exhibit A4).¹⁹

Failure to observe an establishment in a given year could occur as a result of frequent openings and closures, location changes, or reorganizations.^20,21 The final analysis sample includes 321 establishments acquired by PE firms and 1,926 comparison facilities. The median PE-acquired establishment was observed for four years before acquisition and three years after.

Analytical Approach

We first plotted the mean outcomes of facilities by year and acquisition status. We then estimated the differential change in service offerings or forms of insurance accepted at PE-acquired establishments relative to comparison establishments. Service offerings are distinguished between forms of medication treatment and forms of therapy often associated with SUD treatment, such as twelve-step programs and cognitive behavioral therapy.

We leveraged the panel data structure and adopted a difference-in-differences research design with two-way fixed effects. Facility fixed effects account for unobserved time-invariant establishment characteristics, and year fixed effects account for time shocks common to all facilities. During our sample period, the Patient Protection and Affordable Care Act authorized states to expand their Medicaid programs. To account for these expansions, which started in 2014, we included state-by-year fixed effects. Standard errors are clustered by establishment throughout. Our primary specification was a modified difference-in-differences estimator introduced in the referenced materials to account for potential bias resulting from staggered acquisition timing.²² In contrast to the conventional difference-in-differences approach, only establishments that are never acquired in our sample are used as comparison units.

We interpreted the coefficient on an indicator for the postacquisition period as the effect of acquisition on acquired facilities. This interpretation rests on the assumption that acquired and comparison facilities would have persisted on “parallel trends” in the absence of the acquisition. We lent support for this approach by estimating an analogous “event study” specification that includes coefficients for each calendar year relative to the year before acquisition. Formally and visually testing that acquired facilities did not diverge in the years preceding acquisition lent credence to the parallel trends assumption and the assumption that facilities did not anticipate acquisition.

In addition to subsample analysis, we demonstrated the stability of our primary results to alternative matching criteria. Because of the limited scope of the N-SSATS survey, we were not able to use establishment-level characteristics in our match without severely limiting our sample size. However, we could estimate the same models, leveraging a richer set of demographic matching criteria. We then presented results from an unmatched sample of facilities. Finally, we addressed the possibility that acquired facilities differ systematically from comparison facilities in unobserved ways by reestimating treatment effects only using the subsample of facilities that are acquired at some point in our sample.

This study was deemed exempt from Institutional Review Board review because it did not involve human participants. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Limitations

This study has two limitations of note. First, we could not rule out potentially biased survey nonresponse as a reason for missing data. However, we are assured that acquired and matched comparison facilities have qualitatively similar propensities to drop out of the N-SSATS from one year to the next (appendix exhibit A5).¹⁹ Second, we were only able to observe changes in the extent or diversity of services offered without the ability to measure changes in types of patients treated, care use, or care quality. We view this second point as an important consideration for future research.

Study Results

Sample Characteristics

Between 2009 and 2019, PE-acquired establishments were less likely to accept three major forms of health insurance (Medicare, Medicaid, and private insurance) than nonacquired establishments (exhibit 1). The opposite was true between 2020 and 2022, when insurance acceptance was similarly or more commonly accepted among these PE-acquired establishments. Approximately 17 percent of all acquired facilities did not accept any form of insurance in the year preceding acquisition.

Exhibit 1.

Association between private equity (PE) acquisition and forms of insurance accepted at substance use disorder treatment establishments, 2009–22

Source/Notes: SOURCE Authors’ analysis of data from the National Survey of Substance Abuse Treatment Services, paired with a data set of PE acquisitions. NOTES This figure plots changes in the probability that facilities accepted different forms of insurance.

Medicare became more widely accepted among PE-acquired facilities starting in 2020. This was driven by establishments that were acquired in or before 2020, rather than a change experienced by all PE-acquired establishments simultaneously (appendix exhibit A6).¹⁹ This trend break coincides with a regulatory change to the Medicare program in which enrollees began receiving coverage for care received at opioid treatment programs (OTPs). This reform could have encouraged PE acquisition of facilities that were well poised to take advantage of the policy.

Acquired facilities were more likely to offer buprenorphine and methadone treatment relative to comparison facilities (appendix exhibit A7, panel A).¹⁹ The availability of methadone was near perfectly correlated with recognition as an OTP with a correlation coefficient of 0.979. This is reassuring, as OTPs are treatment facilities that are federally accredited to dispense methadone. Comparison establishments were slightly more likely to offer group counseling, cognitive behavioral therapy, peer counseling, and twelve-step programs (appendix exhibit A7, panel B).¹⁹

Establishment Outcomes

Acquisition by a PE firm was associated with increased acceptance of both major forms of public health insurance (exhibit 2). PE acquisition increased the likelihood that establishments accepted Medicare by 25 percentage points off a base of 30.1 percent and increased the likelihood that establishments accepted Medicaid by 15.4 percentage points off a base of 62.6 percent. The event study estimates demonstrate that these effects are present from the year of acquisition and persist for at least five years after (exhibit 3). We fail to reject the null hypothesis that the preperiod coefficients for these outcomes are jointly distinct from zero, affirming our research design.

Exhibit 2:

Average differential change associated with private equity acquisition of substance use disorder treatment establishments, 2009–22

	Coefficient	Standard error	Constant	Observations
Insurance accepted
Medicare	0.250****	0.023	0.301	15,654
Medicaid	0.154****	0.026	0.626	15,654
Private	0.200****	0.026	0.668	15,654
Medication treatments offered
Buprenorphine	−0.029	0.025	0.306	15,654
Methadone	−0.015*	0.009	0.176	15,654
Naltrexone	−0.009	0.025	0.263	12,777
Therapy services offered
Cognitive behavioral therapy	−0.005	0.015	0.921	10,470
Group counseling	−0.001	0.015	0.928	10,470
Twelve-step program	−0.014	0.025	0.653	9,201
Mentoring or peer support	−0.050*	0.028	0.607	9,201

SOURCE Authors’ analysis of data from the National Survey of Substance Abuse Treatment Services paired with a data set of private equity acquisitions. NOTES Coefficients document the association between an indicator for the post–private equity acquisition period and the outcome indicator noted in each column, using a difference-in-differences estimator proposed in the cited materials (see Sun L, Abraham S. Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. See note 22 in text). Establishment, year and state-by-year fixed effects are included in all estimates. Values in brackets are 95% confidence intervals computed from standard errors clustered by establishment. Data reporting of naltrexone availability starts in 2013. Data reporting on cognitive behavioral therapy and group counseling start in 2015. Data reporting of twelve-step programs and peer counseling start in 2016.

^*p < 0.1

^****p < 0.001

Exhibit 3.

Differential change in types of insurance accepted associated with private equity (PE) acquisition of substance use disorder treatment facilities, 2009–22

Source/Notes: SOURCE Authors’ analysis of data from the National Survey of Substance Abuse Treatment Services paired with a data set of PE acquisitions. NOTES This exhibit plots event-time coefficients documenting the association between PE acquisition in relative year zero and three outcome indicators. All values are derived using the discussed modified difference-in-differences estimator. Establishment, year, and state-by-year fixed effects are included in all estimates. Outcome indicators distinguish whether a given form of insurance is accepted at a substance use disorder treatment establishment. All coefficients are estimated in comparison with the year immediately before acquisition. Vertical bars indicate 95% confidence intervals calculated from standard errors clustered by establishment. Reported p values are from an F test of the significance of all preacquisition coefficients: private insurance pre-period, p < 0.001; Medicare pre-period, p = 0.131; Medicaid pre-period, p = 0.108.

The abrupt upward trend in Medicare acceptance plotted in exhibit 1 leaves open the possibility that these results reflect a secular trend spurred by the 2020 Medicare reform regarding OTPs. However, the differential increase in Medicare acceptance among PE-acquired establishments is apparent one year after acquisition for those acquired both before and after 2020 (appendix exhibit A8).¹⁹ The effect is also observable among the 137 establishments recognized as OTPs in the year before acquisition (appendix exhibit A9).¹⁹ Collectively, this suggests that changes in Medicare acceptance were happening before the 2020 Medicare reform, although the policy may have accelerated this trend.

Although we observe an overall increase in the likelihood that PE-acquired establishments accept private health insurance, the preperiod coefficients in exhibit 3 suggest that this divergence may be driven by factors other than PE acquisition. We fail to detect an effect of PE acquisition on the availability of buprenorphine, methadone, and naltrexone (exhibit 4).

Exhibit 4.

Differential change in types of medication therapies offered associated with private equity (PE) acquisition of substance use disorder treatment facilities, 2009–22

Source/Notes: SOURCE Authors’ analysis of data from the National Survey of Substance Abuse Treatment Services paired with a data set of PE acquisitions. NOTES This exhibit plots event-time coefficients documenting the association between PE acquisition in relative year zero and three outcome indicators. All values are derived using the discussed modified difference-in-differences estimator. Establishment, year, and state-by-year fixed effects are included in all estimates. Outcome indicators distinguish whether a given form of medication treatment is offered at a substance use disorder treatment establishment. All coefficients are estimated in comparison with the year immediately before acquisition. Vertical bars indicate 95% confidence intervals calculated from standard errors clustered by establishment. Reported p values are from an F-test of the significance of all preacquisition coefficients: Naltrexone pre-period, p = 0.066; Methadone pre-period, p = 0.708; Buprenorphine pre-period, p < 0.001.

We next probe the robustness of our results to alternative matching criteria. Our primary specification is estimated on a sample that is matched by state and ZIP Code Tabulation Area uninsurance rate (appendix exhibit A10).¹⁹ We find almost identical regression coefficients after additionally matching on ZIP Code Tabulation Area population and poverty rate, and our results are also largely unchanged if we remove all matching criteria and instead use the full N-SSATS sample.

Although measurement error in the exact timing of deals is possible, these primary results are unaffected when calendar years are selectively pruned to isolate the “short” and “medium-run” effects of PE acquisition on acquired establishments (appendix exhibit A10).¹⁹ Our primary results are also not driven by our choice of difference-in-differences estimator. We arrive at similar conclusions when we use an alternative estimator that does not allow for the inclusion of state by year fixed effects, but does additionally use acquired establishments as comparison units up until the year of acquisition.²³ Our results are also broadly unchanged using canonical difference-in-differences regression models with two-way fixed effects.

PE-acquired establishments may be systematically different from comparison establishments in unobserved ways. We can account for these unobserved differences by reestimating this alternative difference-in-differences model only on the establishments acquired at some point in our sample period—in other words, using establishments that have yet to be acquired as comparison units. These estimates are directionally consistent with our primary results; however, we are not able to distinguish whether the lack of precision can be attributed to selection concerns or statistical power constraints.¹⁹

Discussion

This study finds that acquisition by a PE firm, on average, increases the probability that SUD treatment establishments accept public health insurance. This expansion of coverage after acquisition helped bring PE-owned establishments to a rate of insurance acceptance comparable to that observed at nonacquired establishments. Rates of Medicare acceptance are markedly higher among facilities owned by PE firms. PE’s expansion of insurance coverage is consistent with the scale of operations being an important predictor of perceived value for later acquisitions.²⁴ PE-acquired facilities appear to be differentially sensitive to expansions in public insurance coverage for SUD treatment in particular. An implication of this research is that future policy efforts to increase care access at traditionally independent establishments may be moderated by changes in ownership structure.

Medicare and Medicaid accreditation is a costly process whereby an establishment’s care practices and procedures are comprehensively evaluated before it can participate in either program. PE firms might pursue such credentialing when otherwise independently owned establishments would not because of their financial resources or by leveraging experience from previous acquisitions. Thus, the barriers to accreditation may inadvertently favor certain ownership structures, including PE ownership. However, in many contexts, accreditation might not imply health care quality.²⁵ Also, participation in these programs might not alleviate access constraints if, in practice, only a small number of Medicaid managed care organizations are ultimately contracted with.²⁶ Further work is needed examining whether PE acquisition affects access to care and quality of care differentially across different payment sources.

At baseline, facilities acquired by PE firms were more likely to offer each medication treatment of interest relative to comparison facilities. This suggests that offering medication treatment could be attractive to prospective acquirers. Opioid treatment programs are common targets of PE,²⁷ with more than half of acquired facilities in our study offering methadone at baseline. Further research is needed to examine whether PE acquisition changes the quantity and mix of services rendered, as well as the mix of patients ultimately treated at affected establishments.

Conclusion

On average, we do not observe a change in service offerings at SUD treatment facilities acquired by a PE firm, including common forms of medication treatment. However, PE acquisition increased the likelihood that establishments accept two forms of public health insurance—Medicare and Medicaid—although we cannot determine whether this was associated with changes in patient insurance type in practice. These findings suggest that PE firms focus on changes in public program participation that are a first and necessary step to increasing the number of patients accessing covered care at SUD treatment facilities.