Hospital behavior over the private equity life cycle

Abstract

Private equity is an increasing presence in US healthcare, with unclear consequences. Leveraging unique data sources and difference-in-differences designs, we examine the largest private equity hospital takeover in history. The affected hospital chain sharply shifts its advertising strategy and pursues joint ventures with ambulatory surgery centers. Inpatient throughput is increased by allowing more patient transfers, and crucially, capturing more patients through the emergency department. The hospitals also manage shorter, less treatment-intensive stays for admitted patients. Outpatient surgical care volume declines, but remaining cases focus on higher complexity procedures. Importantly, behavior changes persist even after private equity divests.

1. Introduction

A specific sector of the US economy that has been a prominent and growing target for private equity funds and associated controversy is healthcare. In 2018 alone, $100 billion of private equity funding flowed into the sector––a roughly 20-fold increase compared to two decades prior (Appelbaum and Bratt 2020). Over the past ten years, nearly $800 billion of private equity capital has been invested into US healthcare companies (Scheffler, Alexander, and Godwin 2021).¹ The well-documented aggressive push by private equity into healthcare is consistent with the industry perception that the opportunities and performance of these assets tend to eclipse the non-healthcare companies found within many private equity portfolios (Bain & Company 2022). However, a variety of industry participants, experts, advocates, and regulators have voiced concerns, if not objections, to this contemporary circumstance since private equity’s business objectives could conflict with patients’ best interests (e.g., see Gondi and Song 2019; Gustafsson, Seervai, and Blumenthal 2019; Sanger-Katz, Creswell, and Abelson 2019; Sharfstein and Slocum 2019).²

Private equity’s strong profit motives, coupled with set financial endpoints over relatively short time horizons (e.g., selling the company within 5–10 years), create high-powered incentives that may encourage rent-seeking behavior as well as efforts to distort provider agency away from what is optimal for patients. On the other hand, private equity could benefit healthcare firms by providing needed capital infusions as well as improving business management––interventions that have the potential to strengthen performance over the longer run. Given the theoretically ambiguous and opposing possibilities for private equity in US healthcare, a nascent literature has emerged to empirically examine their implications. Recent research has explored private equity effects focused on nursing homes (Huang and Bowblis 2019; Braun et al. 2020, 2021a; Gandhi, Song, and Upadrashta 2020a, 2020b; Gupta et al. 2024), physician practices (Tan et al. 2019; Konda et al. 2019; Braun et al. 2021b; Singh et al. 2022; Bruch et al. 2023), and ambulatory surgery centers (Bruch et al. 2022; Lin et al. 2023). There are indications that the involvement of private equity investors can lead to higher service prices and lower quality care within these industries, but the findings are also mixed and inconclusive across studies.

A complementary set of studies on private equity investments involves US hospitals. Not only is the hospital industry vital to the healthcare system, but in November 2006, financial history was made when the largest leveraged buyout deal across all sectors of the US economy took place as Bain Capital, Kohlberg Kravis Roberts (KKR), and Merrill Lynch Global Private Equity collectively acquired the Hospital Corporation of America (HCA) for $33 billion (HCA Healthcare 2006; Sorkin 2006; Dowd 2017).³ At the time of the transaction, HCA operated as the largest for-profit hospital chain in the US and was a publicly listed US company (NYSE: HCA). After the historic “mega deal”, HCA and its roughly 170 hospitals across 21 states remained under private equity control until returning to public markets in March 2011. The initial public offering (IPO) in 2011 is estimated to have netted Bain Capital over $1 billion (on an initial investment of just $64 million) and generated a return of over 200% for the remaining investors (Dowd 2017).⁴ Despite the clear significance of this monumental acquisition and the prominence of the targeted company within US healthcare, surprisingly little is known about how this massive private equity takeover affected HCA hospitals.

A few indirectly related studies have explored private equity involvement in hospitals, more generally. Bruch, Gondi, and Song (2020) find that in the first three years following a private equity acquisition hospitals’ charges and net income are higher, with virtually no changes in payer mix.⁵ Similarly, Cerullo et al. (2021) interpret their findings as hospitals substituting toward more profitable service line offerings soon after a private equity investment is made. However, both studies have important limitations, including the reliance on self-reported and annual financial data, short-run analyses, and inability to observe actual care delivery outcomes. Cerullo et al. (2022) extend these previous two studies by examining utilization and health outcomes; however, the authors are confined to hospital stays for just five conditions and a single payer (Medicare), where they find no clear effects. They also fail to differentiate between investment and divestment activities by private equity firms. Finally, Liu (2022) brings commercially insured claims data to bear on this question. The author benefits from observed transaction prices, which are positively associated with private equity ownership, but the data are restricted to outpatient claims (i.e., no inpatient data) over the 2013–2019 period. The data therefore miss the period when HCA was under private equity control and barely span the typical ownership duration for a given private equity fund and its portfolio company.

We take a different approach. First, since the majority (57%) of all US hospitals undergoing private equity ownership over the past two decades belonged to HCA (Offodile et al. 2021), we focus our empirics on this specific healthcare market shock. Second, we intentionally examine the effects of private equity investment and divestment to capture hospital behavior changes over the full life cycle of private equity ownership. Third, we compile and leverage several data sources unique to this literature that allow for a more detailed understanding of firm strategy and patient outcomes than exists in other studies. We specifically benefit from proprietary data on hospital advertising expenditures as well as data that detail ownership structures for the universe of ambulatory surgery centers (ASCs) operating across the US. ASCs historically rivaled hospitals for profitable business lines, but the two industries are becoming more financially integrated over time.⁶ These novel data assets consequently allow us to investigate if the infusion of private equity capital and private equity management causes HCA hospitals to revise/expand its marketing activities and/or to engage in joint ventures with non-hospital companies. The latter strategic maneuver could create important opportunities for hospitals to diversify their revenue streams and/or recapture valuable referrals that would otherwise be lost to competitors. Neither hospital marketing nor outside investment responses to private equity has been examined in the literature to date.

We then complement these national analyses with over a decade of all-payer encounter data from Florida––the state with the greatest density of HCA hospitals, accounting for nearly 30% of HCA aggregate hospital revenues, at the time of the private equity acquisition in 2006. Crucially, these all-payer data capture changes in payer mix, case mix, and treatment intensity across the two most important domains of hospital-based care (inpatient stays and outpatient surgical services), which directly speak to if, and how, hospital care delivery is impacted by the private equity funds’ acquisition as well as their eventual liquidation of their ownership positions.

We rely on standard difference-in-differences (DD) empirical strategies across all distinct data assets and components of our investigation, which benefit from being analytically straightforward and transparent. However, we are also cognizant of potential spillover effects to neighboring hospitals, if local competitors are aware of and respond to changes in HCA hospitals’ behavior while under private equity ownership. Since such spillovers could distort our DD estimates and interpretations, we formally test for changes in neighboring hospitals’ behavior in Section 7 and determine that our inferences from our main DD analyses are robust to accounting for any spillover effects.⁷

We ultimately find that the arrival of private capital and control leads HCA to adopt various new business strategies and deploy available financial resources in new ways among its corporate chain of hospitals. Specifically, the hospital chain transitions from committing virtually no advertising funds for outdoor mediums (e.g., billboards) to spending over $1 million per quarter during and after its private equity acquisition. This marketing shift does not come at the expense of the other most common hospital advertising medium (television) and is in sharp contrast to other major hospitals and health systems throughout the US. The chain also begins to actively invest in an otherwise rival industry (i.e., ASCs) whose firms regularly compete away profitable outpatient procedural cases from hospitals. Prior to private equity ownership, HCA infrequently entered into ASC-related joint ventures. After being taken private, the company consistently acquires new ASC ownership stakes year-over-year, with a peak of 41 investments in 2010 alone. Similar to the findings for advertising expenditures, HCA’s behavior while under private equity ownership (and beyond) is a clear departure from what is observed for other hospitals, health systems, and prominent ASC chains. Importantly, a new, more aggressive advertising campaign as well as acquiring more diversified revenue streams via strategic joint ventures are consistent with new management strategies and new sources of capital after coming under private equity ownership and direction––and aligns with known positive effects on non-healthcare firms’ information technology investments following a private equity buyout (Agrawal and Tambe 2016).

Our clinical care findings from affected Florida hospitals indicate that, during private ownership and the return to public markets, there is a focus on increasing inpatient throughput––but not evenly across patient types. As the number of non-pregnancy-related admissions steadily rises, peaking at an approximately 25% increase over baseline levels, the share of pregnancy-related admissions falls by 13–19%. Higher inpatient volumes are in part due to a sharp doubling of the rate of transfer patients in the lead up to the 2011 IPO. However, the primary driver of greater hospitalizations throughout the ownership transitions is drawing more patients through the hospitals’ emergency departments. The share of hospital stays originating in the emergency department is more than doubled for pregnancy-related admissions and as much as 10% higher for non-pregnancy-related admissions compared to the baseline (pre-private equity) rate. Moreover, the increase in hospitalizations via the emergency department channel is strongly due to a nearly immediate and then persistent change in the propensity to admit a patient arriving at the emergency department after the chain has been taken private.

While HCA inpatient volumes are growing, hospital stays are simultaneously becoming shorter and less treatment intensive. Non-pregnancy-related patients spend 3–12% less time in the hospital overall, with the mean level effects driven by changes in the left-tail of the length of stay distribution. For example, the rate of same day discharges increases by approximately 30–50% when compared to the pre-private equity period. These patients are also as much as 10% less likely to have any procedure performed during the stay after the private equity ownership transition. Despite the reduction in treatment intensity per stay, the inpatient case mix is unchanged, and we find no evidence of care quality erosion (i.e., worsening mortality rates).⁸ Importantly, these hospital behavior changes are generally evident across payer markets and extend beyond the period of private equity control. They are also not explained by changes in hospitalized patients’ health risk status during the private equity investment or divestment periods. Finally, following the introduction of private equity ownership and direction, affected hospitals dial back their volume of outpatient surgery cases by 18 to 21%––with remaining cases focused on higher complexity procedures.

Overall, and in contrast with the most closely related literature to date, our more comprehensive empirical estimates reveal a variety of substantive hospital behavior changes during and after private equity ownership. Many of these strategic shifts continued even after HCA reemerged as a public company, suggesting that private equity ownership drives corporate-wide changes that managers and shareholders expect to be profitable over the long-run. Although these findings are inherently tied to a single transaction involving one large hospital chain, the observed effects seem to align with what could be expected in other settings involving private equity management.

2. Advertising expenditures

2.1. Data

To measure changes in advertising intensity, we use proprietary information collected by Kantar Media.⁹ The data exist for a host of economic sectors, firm types, product types, and mediums (e.g., outdoor, radio, television, etc.). We focus on hospitals and medical centers’ advertising behavior in the outdoor (e.g., billboards) and spot television (TV) domains from 2003 through 2017 at the quarterly level. For each distinct advertising entity used in our analyses (described next), we calculate the aggregate amount spent by advertising domain in nominal (‘000) dollars per quarter-year across all media markets (“DMAs”). DMAs are a longstanding industry construct that reflect collections of counties where common programming and accompanying advertisements take place.

2.2. Empirical strategy and estimation

The treatment group is straightforward in our analytic context––i.e., it is HCA and all associated outdoor and TV advertising conducted in a given quarter-year. The inclusion criteria for control group hospitals/medical centers takes into account the fact that HCA is a national hospital chain with significant advertising activity throughout this 15-year study period. For example, in the early years of our analytic window, HCA advertises in 17 unique DMAs but in as many as 40 DMAs in later years. Thus, we require members of the control comparison group to advertise in more than 5 but less than 80 DMAs. The upper limit is to ensure that we include other large hospital chains (e.g., Ascension advertises in 64 DMAs and Tenet advertises in 75 DMAs) but also to exclude national campaigns tied to hospital philanthropy seeking, advocacy efforts, and the like (e.g., American Hospital Association, St. Jude’s network, etc.). We additionally limit the control group to advertisers with nonzero expenditures in all 60 quarters spanning 2003 to 2017. Doing so leaves us with 181 unique control group hospital systems/medical centers for the outdoor ad spending analyses and 182 unique controls for the TV ad spending analyses.

In the interest of transparency, we begin our empirics by plotting the raw data trends for HCA across the two advertising mediums of interest. We then do likewise for five prominent hospital chains (i.e., Ascension, LifePoint, Tenet, Trinity Health, and Universal Health Services) to implement a crude comparison of the raw data trends from 2003–2017. We then move to a standard difference-in-differences (DD) event study estimation framework:

$$Y_{at}=\sum _{\begin{array}{c}j=-16\\ j\ne -8\end{array}}^{42}{\delta }_j\mathbf{1}\left[Treate{d}_a\times \left(Time=j\right)\right]+{\lambda }_a+{\gamma }_t+{\epsilon }_{at}$$ (1)

The parsimonious DD specification includes hospital advertiser ($\lambda$) and quarter-year ($\gamma$) fixed effects. Our hospital advertiser-by-time outcomes ($Y$) for each hospital advertiser ($a$) at quarter-year ($t$) are as described above, and the $Treated$ variable is equal to one for HCA and zero otherwise. The resulting series of delta coefficients $\left({\delta }_j\right)$ can inform the presence or absence of differential trending across the treatment and control groups prior to the private equity event ($t=0$) as well as any differential behavior (and any dynamics in the effects) after HCA newly enters private equity ownership during the fourth quarter of 2006 and when private equity ownership is terminated by the second quarter of 2011 ($t=16$). The standard errors are clustered at the advertiser level (182 and 183 distinct entities for the outdoor advertising expenditures and the TV advertising expenditures estimations, respectively).

2.3. Results

Figure 1 describes HCA national advertising expenditures per quarter for each of the two mediums of interest. Interestingly, from 2003 through 2006, HCA outdoor advertising was limited (typically less than $100,000 per quarter) while TV advertising was often several multiples of the outdoor advertising levels––suggesting that the company marketing strategy was primarily focused on reaching consumers through TV, rather than outdoor signage (e.g., billboards). During the second year of private equity ownership, the HCA marketing strategy appears to take a dramatic and permanent shift. The TV trend is fairly consistent over the remaining years of data (though with seasonality), but outdoor quarterly advertising increases to over $1 million by the conclusion of private equity ownership in early 2011 and remains on an upward trajectory that eventually peaks at approximately $1.7 million in quarterly expenditures by 2017. Importantly, the more than 10-fold increase in outdoor advertising spending (in nominal terms) is not simultaneously occurring with HCA hospital chain expansion. In fact, when examining the aggregate supply of HCA hospitals across the US and over time, there is little indication that the company begins to agressively buy-up or develop new hospitals when coming under private equity ownership. The total number of HCA hospitals actually declines from its highest point in 2004 to a nadir in 2012 (Appendix Figure A1). Thus, higher HCA advertising expenditures are not driven by an increase in the number of hospitals belonging to the chain.

Fig 1. National HCA Advertising Spending over Time by Medium.

Notes: Advertising data are from Kantar Media and span the first quarter of 2003 through the final quarter of 2017. Advertising expenditures are in thousands of dollars and nominal terms. The vertical dashed lines indicate the beginning and end of private equity ownership of the HCA hospital chain.

Appendix Figure A2 also demonstrates that the abrupt and large advertising strategy shift by HCA is not observed by other prominent hospital chains. HCA’s outdoor advertising activity is below average among this subset of hospital chains during the 2003–2006 period, and its upward climb starting in 2008 is also not mirrored by the other chains. The non-HCA trends in Appendix Figure A2 show seasonal fluctuations but are otherwise largely flat for this 15-year period. Appendix Figure A3 further illustrates the uniqueness of the HCA experience by plotting its corporate trend against our control group average outdoor expenditures over time. Figure 2 offers formal tests for HCA differential advertising behavior using our full set of control comparison units described in Section 3.2 and the event study specification given by Equation (1). It is clear from panel (a) and panel (b) in Figure 2 that even with the inclusion of many more control comparison units, the DD event study estimates largely recover the same pattern for HCA advertising behavior change as revealed in Figure 1 and Appendix Figures A2–A3. Specifically, the abrupt trend change and magnitude of outdoor advertising expenditures appears unique to HCA over this period and suggests a tactical management change (and possibly use of fresh capital) after the company becomes privately held. We do note that the standard errors on the estimates are quite small for the differential change in outdoor advertising (panel (a), Figure 2); however, as we demonstrate in the following section (Section 2.4), this is a consequence of the extraordinarily large increase in advertising spending by HCA compared to all other hospital systems and medical centers in the analytic sample.

Fig 2. Diff-in-Diff Event Study Estimates for Private Equity Ownership Effects on HCA Advertising Expenditures.

Notes: Advertising data are from Kantar Media. There are 181 control group units in panel (a), and 182 control group units in panel (b). Controls are comprised of large hospital chains and health systems.

2.4. Robustness

To examine the robustness of our inferences from Figure 2, we implement an empirical exercise that leverages a standard (i.e., “2×2”) DD estimation and cycles through 182 sequential estimations of the DD model for outdoor advertising spending, which ultimately allows each unique advertiser from the analytic sample to serve as the treated unit during a single run. One run will include the true treatment unit (i.e., HCA), and the other 181 runs will each include a different placebo treatment unit. The estimating equation is as follows:

$$Y_{at}={\delta }_1{PEOwnership}_{at}+{\delta }_2{PEDivestment}_{at}+{\lambda }_a+{\gamma }_t+{\epsilon }_{at}$$ #(2)

In Equation (2), $PEOwnership$ turns on for the treated unit belonging to a given estimation during HCA’s private equity ownership period (Q4 2006 through Q1 2011), with $PEDivestment$ turning on immediately after (Q2 2011) when HCA has returned to public markets. We then plot in Appendix Figure A4 the resulting investment period and divestment period DD coefficients $\left({\delta }_1,{\delta }_2\right)$ from all 182 estimations to show where the “true” DD estimates (i.e., those from using HCA as the treated unit) fall among the full distribution. Panel (a) and panel (b) of Appendix Figure A4 show that the DD estimates with HCA as the treated unit are the largest in the resulting distributions. During the ownership period, most of the estimates fall between −$100,000 and +$100,000 in outdoor advertising while the HCA estimate is nearly +$300,000. Moving to the second post-period, the gap between the HCA-specific estimate (nearly +$1,000,000) and the rest of the distribution is more pronounced. Even the farthest non-HCA outliers have DD estimates that are only around half of the HCA-specific estimate. Taken together, the DD distribution patterns in Appendix Figure 4 support our prior interpretations.

3. ASC ownership stakes

3.1. Background and data

Medical services have been rapidly migrating to outpatient delivery settings for many years (Munnich and Parente 2018; Baker, Bundorf, and Kessler 2019)––with even the hospital industry demonstrating inpatient and outpatient revenue streams that are now roughly equal in size (AHA 2020). Hospitals are not the sole providers of outpatient procedures and surgeries, however. Ambulatory surgery centers (ASCs) rival hospitals and often steal profitable business belonging to traditional Medicare and privately insured patients from hospitals’ outpatient surgical departments (Munnich and Parente 2014; MedPAC 2021). The ASC industry currently captures 60% or more of all outpatient procedural care (Frack, Grabenstatter, and Williamson 2017) and is composed of over 5,000 individual firms spread out across the US (Munnich and Richards 2022), with a total market value approaching $30 billion.¹⁰ ASCs are also overwhelmingly privately held, for-profit firms where physicians’ financial interests are known to directly influence the choice over treatment setting––i.e., opting for an ASC versus a hospital outpatient department for a given case (e.g., Munnich et al. 2021; Geruso and Richards 2022; Richards, Seward, and Whaley 2022). Hospitals are also known to pursue joint ventures with the ASC industry––consistent with a “can’t beat them, join them” business strategy in contested markets. Doing so can diversify hospitals’ revenue streams and even allow hospitals to share in the financial gains that surrounding ASCs enjoy at the expense of rival hospitals.

To test if private equity ownership influences HCA’s strategic decisions concerning ASC joint ventures, we leverage data on ASC ownership details that was obtained by a FOIA request to the Centers for Medicare and Medicaid Services (CMS) in April 2019. The data provide an exhaustive list of individual owners (most commonly physicians) as well as organizational owners (e.g., hospital, health systems, ASC corporate chains, limited liability corporations, as well as institutional investors) belonging to a uniquely identified ASC. A complete ownership record is observed so long as the relevant ASC was certified by Medicare and operational by January 1, 2005 or later. This latter data limitation means that we cannot observe ASCs that shutdown before January 2005; however, for ASCs still open as of January 2005, even if their market debut was many years prior to 2005, we are able to reconstruct their complete historical ownership record (both the name of the owner and the timing of ownership).¹¹

3.2. Empirical strategy and estimation

Paralleling our empirics for advertising expenditures, we begin by presenting the raw counts of new ASC ownership events by year for HCA as well as seven other nationally recognized hospital chains present in the ASC ownership data. The seven comparison chains are Ascension, Catholic Health Initiatives, Community Health Systems, LifePoint, Tenet, Trinity, and Universal Health Services. We then extend the analyses to a DD event study setup. A challenge in doing so is that the FOIA data does not classify owners with any granularity beyond “individual” and “organization”. Thus, we cannot simply subset to hospitals and health systems. Instead, we manually examine over 6,000 distinct organization names belonging to the “organization” subgroup and retain 125 unique entities (with HCA as one of those entities) that are recognizable as either large hospital systems, hospital chains, ASC chains, or national institutional investors––excluding the three private equity firms involved in the HCA takeover in 2006. Of note, the overwhelming majority of entities were small private firms (i.e., “LLCs” and “LLPs”) and financiers.¹²

For each ASC investor entity, we calculated the total number of new ASC ownership investments made per year from 2000 through 2014, which includes true zeros for a given investor-year pairing. The accompanying DD estimating equation is as follows:

$$Y_{it}=\sum _{\begin{array}{c}j=-7\\ j\ne -1\end{array}}^7{\delta }_j\mathbf{1}\left[Treate{d}_i\times \left(Time=j\right)\right]+{\theta }_i+{\gamma }_t+{\epsilon }_{at}$$ (3)

The specification in Equation (3) includes ASC investor (θ) and year ($\gamma$) fixed effects. The outcome ($Y$) for ASC investor ($i$) in year ($t$) is the aggregate number of new ASC investments, including zero when appropriate, and the $Treated$ variable is equal to one for HCA and zero otherwise. Just as before, we use the resulting series of delta coefficients (${\delta }_j$) to examine differential trends in the outcome for HCA before, during, after its private equity takeover. The standard errors are clustered at the ASC owner entity level––125 in total.

3.3. Results

Figure 3 shows that HCA engaged in relatively few new joint ventures with its ASC rivals prior to 2007––a trend common amongst the other seven hospital chains shown in Figure 3. However, once HCA becomes a privately held company, there is a striking uptick in ASC joint ventures, with 17 formed in 2008 alone. New ASC investments more than double in 2010 for HCA, with 41 new joint ventures in total, and then decrease in 2011 when 13 investments occur for HCA.¹³ The strategic shift also appears to persist after HCA re-emerges as a public company after early 2011. Importantly, none of the other seven hospital chains demonstrate comparable behavior over this period (Figure 3), and the aggregate control group average likewise departs from the HCA trend (Appendix Figure A5). HCA investment activity in the rival ASC industry is a clear outlier, which we formally test with Equation (3) and our wider set of control comparison ASC investors described above. The event study DD estimates are presented in Figure 4 and reaffirm what is seen in the raw data trends belonging to Figure 3 and Appendix Figure A5. HCA demonstrates no differential behavior in the lead up to private equity ownership, but then becomes differentially involved in ASC joint ventures during the private equity period and continues to do so even after private equity returns the hospital chain to public markets.

Fig 3. Comparing HCA Ownership Stakes in Ambulatory Surgery Centers to Other Prominent Health Systems 2000 – 2014.

Notes: Count of new ambulatory surgery center (ASC) ownership stakes made per year. ASC ownership information is from a FOIA request to the Centers for Medicare and Medicaid Services (CMS). Vertical dashed lines demarcate the years of private equity ownership for HCA.

Fig 4. Diff-in-Diff Event Study Estimates for Private Equity Ownership Effects on HCA ASC Ownership Stakes.

Notes: Ownership data are from a FOIA request to CMS. There are 124 distinct control group units in the underlying estimation, which is a mix of large hospitals, health systems, ASC chains, and institutional investors.

3.4. Robustness

To examine the robustness of our inferences from Figure 4, we closely follow the exercise described in Section 2.4, relying on a standard (i.e., “2×2”) DD estimation that cycles through 126 sequential estimations to allow each unique ASC investor from the analytic sample to serve as the treated unit during a single run. The estimating equation is identical to Equation (2), with the exception of the unit fixed effects $\left({\theta }_i\right)$:

$$Y_{it}={\delta }_1{PEOwnership}_{it}+{\delta }_2{PEDivestment}_{it}+{\theta }_i+{\gamma }_t+{\epsilon }_{it}$$ #(4)

The definitions of $PEOwnerhip$ and $PEDivestment$ in Equation (4) also reflect the annualized data for ASC ownership so that they capture the 2007–2011 and 2012–2014 periods, respectively.

The resulting distributions of DD estimates are displayed in Appendix Figure A6. The “true” DD estimates from considering HCA as the treated unit are not the stark outliers they were in Appendix Figure A4 (i.e., for outdoor advertising expenditures), which somewhat tempers the conclusions that can be drawn from Figures 3 and 4. However, the estimate from the private equity ownership period is at the 95^th percentile of the distribution in panel (a), and the estimate for the divestment (i.e., public ownership) period is at the 99^th percentile in panel (b). We also note the largest differential changes belonging to the two distributions in Appendix Figure A6 (i.e., the coefficients that are larger than those belonging to HCA) are not from a common ASC investor, so in this regard, HCA is seemingly unique in demonstrating a large differential change during 2007–2011 as well as 2012–2014.

4. Hospital inpatient care

4.1. Data

We benefit from the universe of inpatient and outpatient (ambulatory) surgery discharge records that encompass all hospitals and all payers in Florida (including the self-insured and bed/debt charity care groups). The data are maintained and distributed by the Florida Agency for Health Care Administration (AHCA) and span 2003 through 2013. The encounter data are also at the quarter-year level and contain rich information on patient characteristics, services received, and payer type. Such historical and comprehensive data are crucial to studying HCA hospitals’ behavior over the full private equity lifecycle (i.e., private equity investment and divestment financial endpoints), especially to observe sufficient pre-investment and post-divestment time windows.

Additionally, Florida is a key state for the national HCA hospital chain at the time of its private equity takeover. According to the American Hospital Association (AHA) annual survey, Florida had the greatest density of HCA hospitals of any state in 2006, with only Texas having a comparable number—even among the top-5 other states for HCA hospital density in 2006 (Appendix Figure A7).¹⁴ Florida HCA hospitals also accounted for $24.7 billion in aggregate hospital revenues (29% of national HCA hospital revenues) in that year. On average, Florida HCA hospitals are similar in size (in terms of bed counts) as well as the quantity of full-time workers when compared to HCA hospitals found elsewhere around the US (Appendix Table A1). Both sets of HCA hospitals have just under a 6% average operating margin in 2006 as well (Appendix Table A1). Taken together, the quality of the historical inpatient and outpatient data, coupled with the relevance of Florida to the HCA company at the time of private equity acquisition, suggests that these analyses can complement the empirics from Sections 2 and 3 and generate valuable insights regarding actual clinical care delivery adjustments made by HCA hospitals during and after private equity ownership––something that is conspicuously absent from the existing empirical literature.

4.2. Empirical strategy and estimation

We first match hospitals using hospital name and exact address from the discharge data to HCA ownership information (by year) according to the corresponding AHA survey data. Then, for our treatment (HCA) and control (non-HCA) Florida hospitals, we restrict to general acute care hospitals consistently present in the Florida discharge data from Q1 2003 through Q4 2013. Doing so leaves us with 35 HCA hospitals and 103 control hospitals observed in the analytic data.¹⁵ Across all discharge data sets, we collapse the data to the hospital-quarter-year level for our DD analyses.

We implement two versions of a DD estimation. The first is the standard “2×2” setup, with a slight modification to give us two distinct post-periods in the specification (i.e., one for during and one for after private equity ownership). We then translate the DD analyses into a full, flexible event study spanning the 11-year analytic period (44 quarter-years in total). Equation (5) and Equation (6) include hospital ($\eta$) and quarter-year ($\gamma$) fixed effects and cluster the standard errors at the hospital level. $Y$ are the outcomes of interest for each hospital ($h$) in quarter-year ($t$).

$$Y_{ht}={\delta }_1{PEOwnership}_{ht}+{\delta }_2{PEDivestment}_{ht}+{\eta }_h+{\gamma }_t+{\epsilon }_{ht}$$ (5)

$$Y_{ht}=\sum _{\begin{array}{c}j=-15\\ j\ne -3\end{array}}^{28}{\delta }_j\mathbf{1}\left[{Treated}_h\times (Time\left.=j)\right]+{\eta }_h+{\gamma }_t+{\epsilon }_{ht}\right.$$ (6)

In Equation (5), the binary $PEOwnership$ variable is equal to one for the 35 treatment group HCA hospitals during HCA’s private equity ownership (Q4 2006 through Q1 2011), and the binary $PEDivestment$ variable is equal to one for these same 35 hospitals during the remainder of the analytic window (Q2 2011 through Q4 2013) when HCA has returned to public markets. Equation (6) has an identical interpretation as our previous event study models from Sections 2 and 3. We also note that, in the interest of conserving space as well as transparency of findings, we generally present the event study estimates in our main results and display the 2×2 estimates from Equation (5) in the appendix materials, with the exception of our most important findings for hospital care delivery changes.

4.3. Results for all admissions

When looking across all inpatient stays, we measure the total volume of admissions as well as the share of admissions devoted to pregnancy-related patients (i.e., newborn births and expecting mothers) for each hospital in each quarter-year.¹⁶ The “2×2” DD estimates (Equation (5)) are shown in Appendix Table A2, with the corresponding event study estimates (Equation (6)) in Figure 5.

Fig 5. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Appendix Table A2. Vertical bars bookend private equity ownership of HCA.

During the 2003–2005 period, HCA hospitals only averaged 80% of the aggregate inpatient volume per quarter-year as the average hospital in our control group. However, the gap narrows over time, and by the time HCA returns to public markets, their average inpatient admission volumes are up by 12% over baseline levels. The estimates in column 2 (Appendix Table A2) indicate that these additional admissions are less likely to be coming from births. Once HCA is privately held, the share of inpatient stays for a newborn birth falls by approximately 13%, with the effect larger (19% decline) after private equity liquidates its ownership position. To explore what could be thought of us a pure intensive margin effect, we further restrict the treatment and control hospitals to those having at least one birth in every quarter-year during our 11-year period (column 3, Appendix Table A2). Among HCA hospitals consistently providing newborn delivery care for all quarter-years, their share of admissions tied to pregnancy also fall by 11% and 20% during the private equity investment and divestment phases, respectively. The event study results in Figure 5 align with the inferences drawn from Appendix Table A2. Additionally, across all three panels of Figure 5, the effects on hospital admissions following the private equity take-over demonstrate increasing magnitudes over time. For example, the gain in inpatient volumes is approximately 16–19% over HCA baseline levels (Appendix Table A2) by the final year of our study period (panel (a) Figure 5).

In light of the findings from Appendix Table A2 and Figure 5, we separately subset to pregnancy-related and non-pregnancy-related admissions in Sections 4.4 and 4.5, respectively, to then examine any changes in treatment intensity as well as payer mix tied to these specific subgroups of hospitalizations.

4.4. Results for pregnancy-related admissions

After restricting to pregnancy-related admissions and hospitals caring for these patients throughout our study period, we measure the length of stay (LOS), share of stays involving a c-section surgery, and share of stays involving at least one procedure (any type) to capture treatment intensity during these specific hospital stays. We likewise calculate the share of these admissions originating in the hospital’s emergency department, the share that are transfer patients, the share admitted over the weekend, the average distance traveled by the patient, and the in-hospital mortality rate.¹⁷ We also examine any shifts in payer mix using a mutually exclusive and exhaustive set of payer-specific variables (i.e., bad debt/charity care, commercial (non-Medicare), Medicaid, Medicare Advantage, traditional Medicare fee-for-service (FFS), self-insured, and an ‘all others’ composite group).¹⁸

To conserve space, we reserve the empirical results and detailed write up for Appendix A and only briefly summarize the findings here. Specifically, the collection of findings for pregnancy-related admissions shows no evidence of a selective retention of patients (e.g., higher paying) or changes in treatment intensity (e.g., LOS or c-section rates). The clearest change is the proportion of pregnant women originating in the hospital’s own emergency department.

4.5. Results for non-pregnancy-related admissions

We next exclude pregnancy-related hospitalizations to assess HCA hospital behavior changes for all other admission types among our balanced panel of Florida hospitals. We also extend our previous set of treatment intensity measures to include whether the hospitalization includes laboratory testing, whether the hospital stay includes surgical suite (i.e., operating room) or intensive care unit (ICU) utilization, whether the patient is discharged with post-acute care home health services or simply to home, and the inpatient case mix. To capture changes in case mix (across all payers), we apply the publicly available 2006 Medicare FFS diagnosis related group (DRG) weights to the principal medical reason for the hospitalization as reported on the discharge record. Medicare bases its hospital reimbursements through its inpatient prospective payment system (IPPS) on DRGs and their associated weights, which reflect severity of illness and expected costs of care. After applying the DRG weights to each discharge record, we average the weights across all relevant admissions for each quarter-year belonging to a given hospital. Of note, we also rely on the DRG weights from a single year so that changes in the outcome can be solely attributed to changes in the actual hospitalization case mix (i.e., not confounded with changes in the administratively set DRG weights over time).

We begin by examining changes in the volume of non-pregnancy-related admissions and sources of those admissions in Table 1. Across these margins, the average HCA hospital looks similar to the average non-HCA hospital in Florida during the baseline (2003–2005) period. However, during and after HCA’s private equity ownership, HCA demonstrates a 5% and 16% increase in hospitalizations, respectively, with more of those admissions originating from the hospital’s emergency department.¹⁹ There is also an uptick in the receipt of transfer patients and weekend admissions (columns 3 and 4, Table 1). Yet, HCA’s geographic catchment area appears unchanged (column 5, Table 1), with HCA patients still traveling just over 10 miles from their zip code of residence, on average.

Table 1.

Diff-in-Diff Estimates for Private Equity Investment and Divestment Effects on Hospital Non-Birth/Non-Pregnancy Related Admissions

	Number of Admissions	Admitted Through ED	Transfer Patient	Weekend Admission	Distance Traveled (miles)
	(1)	(2)	(3)	(4)	(5)
1[PE Ownership]	130.6** (66.8)	0.017** (0.008)	−0.001 (0.005)	0.005** (0.002)	0.025 (0.163)
1[PE Divestment]	431.2*** (126.9)	0.052*** (0.016)	0.031*** (0.009)	0.014*** (0.004)	0.115 (0.256)
Hospital Fixed Effects	Yes	Yes	Yes	Yes	Yes
Qtr-Year Fixed Effects	Yes	Yes	Yes	Yes	Yes
Unique Hospitals	138	138	138	138	138
Observations (N)	6,072	5,520	6,072	6,072	6,068
HCA 2003–2005 Outcome Mean	2,626	0.66	0.03	0.19	10.5
Controls 2003–2005 Outcome Mean	3,068	0.66	0.02	0.19	10.6

Notes: Analytic data are from the universe of Florida inpatient discharge records collapsed to the hospital-quarter-year-level. Analyses are restricted to general, short-term acute care hospitals consistently observed from Q1 2003 through Q4 2013. There are 35 unique hospitals in the HCA treated group. “PE” stands for private equity. All admissions for birth or pregnancy-related issues are excluded from these analyses. “ED” stands for emergency department. Four quarters are unusable for the outcome in column 3 due to a variable definition and reporting requirement transition; thus, those quarters are dropped from the analyses. Column 5 measures the zip centroid-to-zip centroid distance from the patient’s zip code to the hospital’s zip code. Standard errors clustered at the hospital level

^***P value at 0.01

^**P value at 0.05

The corresponding event study estimates in Figure 6 reveal important dynamics across the changes noted in Table 1. Specifically, the increase in inpatient admissions for the hospital chain begins and grows while it is privately held and continues after its return to public ownership. By the end of our study period, average hospitalization volumes are up by approximately 23% over HCA’s pre-period level, and the share originating from the emergency department is more than 10% higher (panels (a) and (b) in Figure 6). Interestingly, HCA hospitals’ willingness to accept more transfer patients does not occur until one year prior to the IPO. The sharp 100% increase over baseline (2003–2005) in the share of admissions transferred from other hospitals then stably persists for the reemerged public company.

Fig 6. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Table 1. Vertical bars bookend private equity ownership of HCA.

Table 2 focuses on these patients’ corresponding length of stay (LOS). Column 1 captures any differential changes in the average LOS while columns 2 and 3 home in on extremely short duration hospitalizations—a hospital behavior of known and longstanding concern among payers and regulators (e.g., see Shi (2023)). The average length of stay falls by 3% and 10% for the private equity investment and divestment periods, respectively, in column 1 of Table 2. However, it is clear from columns 2 and 3 that the average effect is driven by more admissions falling in the left-tail of the LOS distribution. Same day discharges (i.e., when the patient is admitted and discharged on the same calendar day) increase by 30–40% when averaged over the two relevant post-periods. Similarly, HCA hospitalizations spanning two days or less are elevated by as much as 13% according to the DD estimates reported in column 3 of Table 2. The event studies (Figure 7) reinforce these inferences. HCA inpatient durations are closely tracking non-HCA hospitals prior to private equity’s arrival, but soon after the chain has been taken private, its owned hospitals are generating many more short stays for patients—again consistent with a stronger managerial emphasis on hospital throughput. Importantly, these operational changes remain even after HCA returns as a public company.

Table 2.

Diff-in-Diff Estimates for Private Equity Investment And Divestment Effects on Hospital Course for Non-Birth/Non-Pregnancy Related Admissions

	Avg. Length of Stay (LOS)	Same Day Discharge	Short Stay (<=2 days)
	(1)	(2)	(3)
1[PE Ownership]	−0.143*** (0.056)	0.006*** (0.001)	0.018*** (0.006)
1[PE Divestment]	−0.516*** (0.078)	0.008*** (0.002)	0.046*** (0.008)
Hospital Fixed Effects	Yes	Yes	Yes
Qtr-Year Fixed Effects	Yes	Yes	Yes
Unique Hospitals	138	138	138
Observations (N)	6,072	6,072	6,072
HCA 2003–2005 Outcome Mean	4.96	0.02	0.35
Controls 2003–2005 Outcome Mean	4.90	0.02	0.36

Notes: Analytic data are from the universe of Florida inpatient discharge records collapsed to the hospital-quarter-year-level. Analyses are restricted to general, short-term acute care hospitals consistently observed from Q1 2003 through Q4 2013. There are 35 unique hospitals in the HCA treated group. “PE” stands for private equity. All admissions for birth or pregnancy-related issues are excluded from these analyses. “LOS” stands for length of stay.

Fig 7. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Table 2. Vertical bars bookend private equity ownership of HCA.

Appendix Table A5 displays our DD estimates for the intensity of care belonging to a given hospital stay, the inpatient case mix, and the in-hospital mortality rate. The first thing to note from Appendix Table A5 is that over the 2003–2005 period hospitals belonging to the HCA chain tend to manage their hospitalized patients in a manner that closely aligns with non-HCA Florida hospitals, on average. The second salient feature of Appendix Table A5 is that none of the DD estimates imply a greater intensity of inpatient care for HCA patients. By the time HCA is a public company again, its hospital stays are 10% less likely to involve the use of an operating room and 8% less likely to involve any medical procedure at all. Likewise, ICU utilization and the rate of patients being discharged with home health services are each down almost 20% over their pre-period rates following the 2011 IPO event (with routine discharges to home showing a commensurate rise). At the same time, the mix of health problems belonging to HCA patients is unchanged over this entire period (column 7), and the in-hospital mortality rate declines by 5–10% in comparison to the baseline rate (column 8).

The accompanying event study estimates in Figure 8 are uniformly well-behaved during the pre-period years and then show differential changes after HCA is taken over by private equity. The likelihood of utilizing an operating room decreases more gradually; however, the decline in the probability of receiving any inpatient medical procedure (i.e., an extensive margin effect) closely tracks with the sharp fall in the LOS outcome (Figure 7). Drops in ICU utilization as well as home health coordination do not clearly materialize until the immediate lead up to the IPO but then persist after HCA’s return to public ownership. The risk of in-hospital mortality declines soon after the private equity arrival and remains stably lower throughout our study period (panel (f) in Figure 8). However, as remarked in the notes for Appendix Table A5, the mortality findings are not robust to the inclusion of simple patient mix variables (e.g., age, gender, race, and number of comorbidities) that are commonly used for risk adjustment. Additionally, when restricting to acute myocardial infarction (AMI) hospitalizations, we do not observe any change in the hospital-level mortality rate during private equity ownership (results available by request). The conservative interpretation is that there is no evidence of worsening inpatient mortality when under private equity control.

Fig 8. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Appendix Table A5. Vertical bars bookend private equity ownership of HCA.

The DD results for payer mix changes in Appendix Table A6 are generally unremarkable. There are no detectable changes for commercially insured, self-insured, or ‘all other’ payer shares, and the patterns of findings for Medicare as well as Medicare Advantage groups are not clearly consistent over time––Medicaid event study estimates in Appendix Figure A11 (panel (b)) additionally provide equivocal findings. The only two compelling results are for bad debt/charity care and Medicare FFS payer groups. The former increases for HCA hospitals during and after private equity’s involvement while the latter claims a shrinking relative share of the average HCA hospital payer mix over time. The corresponding event study patterns in panels (a) and (d) of Appendix Figure A11 reinforce these interpretations. Affected hospitals witness the share of their payer mix devoted to uncompensated care multiply by 2–4 times their pre-period levels following the private equity ownership transition. And approximately a year after being privately held, the share of patients belonging to traditional Medicare begins to steadily fall––culminating in a roughly 10% decline relative to baseline levels (column (5), Appendix Table A6).²⁰

4.6. Heterogeneity for non-pregnancy-related admissions

While the increase in hospital inpatient volumes, coupled with shorter and less intensive hospital stays, is consistent with a strategy for increasing hospital throughput, the observed changes in inpatient treatment behavior (Figures 7–8) could also be explained, at least in part, by other dynamic factors. Specifically, as the hospitals expand their inpatient volumes, the payer(s) associated with the marginal patients and/or the relative health status of the marginal patients could shape subsequent treatment decisions. In other words, if affected hospitals are disproportionately attracting more business from lower reimbursing payer groups and/or patients with less severe illness, then there would be less incentive and/or need to engage in more aggressive care delivery.

Recall, we only observe limited changes in payer mix composition (Appendix Table A6), and the case mix (i.e., medical reasons) for non-pregnancy-related hospitalizations is unchanged during the time of private equity investment as well as divestment (Appendix Table A5). Nevertheless, for completeness, we further explore these potential mediating factors by re-examining the treatment intensity outcomes as well as investigating each hospital’s aggregate patient health risk profile within payer and over time.

We specifically focus on commercially insured, Medicaid, Medicare Advantage, and Medicare FFS payer groups to best preserve adequate cell sizes and hospital panel lengths in the subsequent estimations since payer-specific outcome measures are incalculable for quarter-years when a given hospital lacks an inpatient stay for a particular (and usually small) payer. These four payer groups also account for 91% of all non-pregnancy-related admissions to HCA hospitals and 89% of all such admissions among our control group hospitals on average during the pre-period years (see Appendix Table A6). Our health risk outcomes of interest include average patient age, share female, share identifying as white race, average Charlson Comorbidity Index, and average Elixhauser Score for each payer-specific inpatient population.²¹

In Appendix Table A7, we can see that admitting more patients through the hospital’s emergency department is a common behavior across payer groups. Only Medicare advantage (Panel C) fails to have either DD estimate reach statistical significance at conventional levels. The effect magnitude is particularly large for the commercially insured group where hospitalized patients are as much as 15% more likely to have arrived through the hospital’s emergency department. The same cannot be said for the patient transfer channel. The increase observed in panel (c) of Figure 6 appears almost exclusively driven by Medicare FFS patients—with only a modest contribution from Medicare Advantage enrollees (Panels C and D). Additionally, the relative change in the likelihood of a Medicare beneficiary being a transfer patient around the timing of the IPO is 147% over the baseline rate for HCA hospitals. This is unsurprising since traditional Medicare patients are unencumbered by provider networks and allowing more of them to be transferred into HCA hospitals immediately preceding the IPO is a plausible means to boost admission volumes/revenues to bolster HCA’s public market valuation––and hence stock price. The corresponding event study findings in Figure 9 also align with the interpretations from Appendix Table A7 and demonstrate that the effect on the probability of a hospitalization originating from the emergency department actually grows in magnitude throughout our study period (e.g., panels (a) and (c) in Figure 9). Commercially insured hospital stays, for instance, are a full 20% more likely to have been admitted to the inpatient unit via the hospital’s emergency department by our final year of data (2013).

Fig 9. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Appendix Table A7. Vertical bars bookend private equity ownership of HCA.

The inpatient treatment intensity outcomes in Appendix Table A8 provide a similar pattern of findings. The DD estimates are uniformly negative across each payer subgroup and typically substantive in magnitude. The results also indicate that the estimates from Figures 6–8 (i.e., the effects across all non-pregnancy-related hospitalizations) are not a consequence of shifting payer mix. Shorter, less intensive hospital stays are evident within payer as HCA comes under private equity control and then reemerges as a public company. The lone exception in Appendix Table A8 is in-hospital mortality where the DD estimates only reach statistical significance at conventional levels for the commercially insured patient population (column 8, Panel A).

The DD estimates for patients’ health risk profile are generally lacking a compelling pattern in Appendix Table A9. The coefficients are typically small and lack statistical significance approximately half of the time. They are also inconsistently signed, with some results (e.g., Charlson Comorbidity Index and Elixhauser Score for Medicare Advantage and Medicare FFS patients) implying a less favorable health risk profile, on average, as opposed to a healthier patient population. The most compelling changes are for the average age outcome for each payer subgroup, but even then, the findings are not sharp (see event study results in Appendix Figure A12) or large in magnitude (often 1–5% relative changes in comparison to the pre-period means).

Taken together, the results from Sections 4.5 and 4.6 imply a new company-wide strategy that prioritizes hospital throughout (i.e., more (non-pregnancy) admissions but shorter stays) instead of maximizing service intensity/complexity for a given admission once private equity becomes involved. The declines in treatment intensity are not easily explained by shifts in payer mix or newly attracting a healthier patient population.

5. Hospital outpatient care

5.1. Data and estimation

When moving to the quarterly outpatient procedural discharge records (i.e., the hospital outpatient departments, or HOPDs) for 2003 through 2013, we measure the aggregate case volume, number of unique physicians performing cases at a given HOPD, a constructed case complexity index, the volume of procedures using laparoscopic technology, and the use of any robotic technology (i.e., extensive margin).²² Our analytic data and empirical setup parallel Section 4; however, two control group hospitals do not have a HOPD for surgical and procedural services and are consequently absent from the ambulatory surgery discharge database. Otherwise, the composition of treatment group and control group Florida hospitals is identical to Section 4. Estimating Equations (5) and (6) are again used throughout, and all standard errors remain clustered at the hospital level.

5.2. Results

When examining the outpatient surgery side of hospital care delivery (i.e., the HOPDs) in Appendix Table A10, we see substantive drops in total case volumes per quarter for HCA hospitals. The magnitude of the effect is roughly 18–21% compared to their pre-period average output and fairly stable during both the private equity ownership phase and the chain’s re-emergence as a public company (panel (a) Figure 10). Notwithstanding the lower levels of HOPD activity among affected hospitals, they strongly focus on higher complexity cases once under private equity ownership (column (3) in Appendix Table A10). The investment and divestment periods demonstrate relative effect sizes of 22–37%, respectively, and the behavior change is immediate and growing as evidenced by the event study results in panel (b) of Figure 10. Importantly, the differential shift toward higher complexity cases is present even when restricting to the two dominant payers within this space (Appendix Figure A13)—i.e., the commercially insured and traditional (FFS) Medicare markets (Hall et al. 2017). This strategic shift is perhaps unsurprising as well. Devoting more of their outpatient service delivery to higher complexity cases is a plausible means to better leverage the hospital’s comparative advantages in care production and to insulate themselves against business stealing by rivals. It also could reflect, at least in part, growing expectations of lower complexity procedures being referred to non-hospital HCA joint ventures, such as ASC settings (Section 3).

Fig 10. Diff-in-Diff Event Study Estimates.

Notes: Outcome definitions and analytic samples are identical to those reported in Appendix Table A10. Vertical bars bookend private equity ownership of HCA.

Despite the greater emphasis on higher complexity outpatient procedures, the volume of surgeries utilizing laparoscopic technology decreases by 17–32% over these same periods––indicating that the more complex cases are not tied to the use of the otherwise costly technology (which is often associated with higher reimbursements). While use of robotic technology for outpatient surgery is quite rare across all hospitals during much of this time span, HCA hospitals do seem to differentially adopt the technology––and hence make costly capital investments––as robotic surgery gains popularity toward the end of our study period (column 5 in Appendix Table A10 and panel (d) in Figure 10). Doing so at least aligns with a long-run strategy of targeting more complex and technologically advanced cases within the outpatient surgery market.

We next assess any payer mix changes for HCA outpatient surgeries in Appendix Table A11 and Appendix Figure A14. As these are uniformly elective services, the exposure to uncompensated care (i.e., bad debt/charity cases) is expectedly small and unchanged with private equity ownership (column 1 of Appendix Table A11). There also seems to be greater avoidance of Medicaid outpatient procedures, at least over the short-run. The share of outpatient procedures devoted to Medicaid patients declines by roughly 30% after the private equity acquisition (column 3 of Appendix Table A11); the share also does not rebound until the private equity divesture occurs in early 2011 (panel (a) in Appendix Figure A14). Not unlike what was evident for inpatient care in Section 4, there is a modest and gradual decline in traditional (FFS) Medicare’s relative share of the hospital chain’s HOPD payer mix, with a largely offsetting increase in exposure to the Medicare Advantage market. The greater relative tilt toward Medicare Advantage outpatient surgery cases, specifically, is sharp and increasing to some degree over time (panel (b) in Appendix Figure A14). The DD estimates in column 4 of Appendix Table A11 indicate that the relative changes are as much as an 80% increase over the affected hospitals’ (admittedly low) baseline rates. Interestingly, there is some suggestive evidence that, following the private equity acquisition, HCA hospitals provide more outpatient procedural care for cash-paying patients––though this is relatively rare at the outset, and the event study findings in panel (d) in Appendix Figure A14 lack sufficient precision to draw strong conclusions.

6. Emergency department channel

6.1. Data and estimation

We conclude our core empirics by turning our attention to emergency department medical decision-making for HCA hospitals during and after its transition to private equity ownership. Recall from Sections 4.3, 4.5, and 4.6, we observe stark increases in the share of inpatient stays originating from the hospital’s own emergency department. This result is evident across hospitalization types as well as payers and informs us about the relative contribution of the hospital’s emergency department compared to all other channels that can lead to an inpatient admission.

But these findings do not reveal the underlying cause of the increasing flow of patients from the emergency department. For instance, this pattern could materialize by attracting more patients to HCA emergency departments (e.g., via advertising campaigns) and admitting the same fraction of patients as before––i.e., no change in clinical behavior among emergency department physicians. Alternatively, the increase could be driven by increasing the share admitted among a stable quantity of emergency department visits. The former would be a more innocuous change of circumstances, while the latter (i.e., inappropriately admitting emergency department patients for inpatient care) could reflect a perverse behavior change that leads to inefficient medical spending. Wide dispersion in admission rates has long been documented (Sabbatini, Nallamothu, and Kocher 2014; Venkatesh et al. 2015), with many questioning whether meaningful health benefits accrue from these marginal admissions (Sabbatini, Nallamothu, and Kocher 2014; Currie and Slusky 2020). Empirical findings further suggest that excessive admitting behavior is perhaps more common among for-profit hospital chains (e.g., see Pines, Mutter, and Zocchi (2013); Howard and David (2021)), and HCA has been specifically accused of such perverse activity in recent years by consumer-centric groups as well as federal legislators.²³

To disentangle these two (though not mutually exclusive) possibilities and shed further light on HCA behavior change when under private equity ownership, we supplement our previous data with a third Florida AHCA discharge database that captures the universe of emergency department visits that do not result in the patient being admitted to the presenting hospital. We can then match these records to our universe of inpatient records reporting the hospital’s emergency department as the admission source to construct a hospital by quarter-year measure of total emergency department visit volume (by payer) as well as the fraction of those visits that ultimately resulted in an inpatient admission. One drawback for this empirical exercise is that the emergency department discharge database does not begin until the first quarter of 2005, so we necessarily sacrifice two years of pre-period data.²⁴ However, we still benefit from seven quarters of pre-period observations per hospital and can credibly estimate the following event study specification that spans 2005–2013 for our two emergency department outcomes of interest:

$$Y_{ht}=\sum _{\begin{array}{c}j=-7\\ j\ne -3\end{array}}^{28}{\delta }_j\mathbf{1}\left[Treate{d}_h\times (Time\left.=j)\right]+{\eta }_h+{\gamma }_t+{\epsilon }_{ht}\right.$$ (7)

Equation (7) has the same analytic setup and interpretations as Equation (6) from Section 4. All that is different is the modest truncation of the pre-period quarter-years available.

6.2. Results

In Table 3, we first summarize the 2005 hospital-level emergency department visit volumes and the propensity to admit patients within payer market for our treatment and control group hospitals. HCA hospitals attract fewer commercially insured and Medicaid patients to their emergency departments than non-HCA hospitals, on average, but also tend to have a greater volume of Medicare Advantage patients. Interestingly, within each of the four key payer groups in Table 3, HCA admits a lower share of patients to its inpatient units prior to becoming a privately held company. Put differently, the emergency medicine physicians staffing HCA emergency departments have a weaker propensity to admit patients from a given payer, at least according to the unadjusted rates (i.e., ignoring patient population risk profiles within an insurance subgroup).

Table 3.

Average Total Emergency Department Patients And Share Of Emergency Department Patients Admitted To The Same Hospital’s Inpatient Unit By Payer In 2005

	HCA Hospitals	Control Group Hospitals
Total ED Encounters
Commercial	2,768	3,273
Medicaid	1,815	2,133
Medicare Advantage	434	329
Medicare FFS	1,911	1,924
Share Admitted
Commercial	0.14	0.17
Medicaid	0.10	0.15
Medicare Advantage	0.40	0.46
Medicare FFS	0.45	0.48
Unique Hospitals	35	99

Notes: Restricts to treatment and control group hospitals from main analyses that also have emergency department (ED) encounter data in 2005. Four hospitals do not have relevant inpatient and ED outpatient admissions and are excluded from these empirical exercises. “Total ED Encounters” is the total number of inpatient admissions for the relevant payer that came through the hospital’s ED summed with the total number of outpatient ED encounters at the same hospital (i.e., ED visits that did not result in an inpatient admission to the hospital). It then serves as the denominator for calculating the share of relevant ED patients ultimately admitted to the hospital’s inpatient unit.

Figure 11 presents the event study results for total emergency department visits for each payer market. Only one of the four main payer groups demonstrates an increase in emergency department visit volumes. During the latter half of HCA’s private equity ownership, the commercially insured visit volumes increase by as much as 18% over the 2005 levels reported in Table 3 and maintain the elevated levels following HCA’s early 2011 IPO. Visit volumes are relatively stable for HCA Florida hospitals among the Medicaid, Medicare Advantage, and Medicare FFS patient populations. Of note, the increase in commercially insured patients coincides with HCA launching a more robust advertising campaign in Florida media markets, specifically (Appendix Figure A15). A supplementary analysis (Appendix Figure A16) also shows that the increase in commercially insured visits by HCA is via diversion, rather than market expansion (i.e., an intensive, instead of extensive, margin effect). There are no differential changes in the aggregate quantity of commercially insured patients seeking emergency department care in areas where HCA has a market presence in comparison to other Florida markets where HCA is absent. Thus, HCA appears to steal business from surrounding competitors.

Fig 11. Diff-in-Diff Event Study Estimates for Total Emergency Department Encounters by Payer.

Notes: Outcome definitions and analytic samples are identical to those reported in Table 3. Vertical bars bookend private equity ownership of HCA

Where we see clear and immediate behavior change across each major patient-payer population is the share of emergency department patients being admitted to the hospital’s inpatient unit (Figure 12). Trends in admitting behavior by HCA emergency department physicians are stable during the nearly two years prior to HCA’s ownership transition, but once under private equity control, the admission rate sharply increases and remains at an elevated level indefinitely. The effects are also large. Commercially insured, Medicaid, and Medicare FFS patients are each as much as 5-percentage points more likely to be admitted (36%, 50%, and 11% in relative terms, respectively). Medicare Advantage patients have roughly a 10-percentage point higher likelihood of being admitted when arriving to an HCA emergency department (25% relative increase). The magnitude of the change in physician behavior among HCA emergency departments eclipses what has been found when hospitals strategically contract with physician staffing companies (Cooper, Scott Morton, Shekita 2020) and eliminates the previous gaps in admitting behavior between HCA and non-HCA hospitals across payers (Table 3). The immediate increase in admitting to the inpatient unit is even evident among the much smaller bad debt/charity care patient population (Appendix Figure A17), which will ultimately generate uncompensated care costs for the hospital.²⁵ The behavior change displayed in Figure 12 is not easily explained by a sudden change in the types of patients arriving to the emergency department or an under-provision of inpatient care prior to private equity ownership. Recall from Section 4, the mix of underlying medical problems associated with the inpatient hospitalizations is unchanged and the patient risk profiles show, at most, small and gradual changes toward a healthier patient population over this time period (primarily in terms of becoming slightly younger). Additionally, affected hospitals are sharply providing less intensive care for their hospitalized patients over this same period.

Fig 12. Diff-in-Diff Event Study Estimates for Share of Emergency Department Encounters Resulting in Admission to Inpatient Unit by Payer.

Notes: Outcome definitions and analytic samples are identical to those reported in Table 3. Recall, four quarters are unusable for the inpatient variable capturing admission through the ED due to a variable definition and reporting requirement transition; thus, those quarters are dropped from the analyses. Vertical bars bookend private equity ownership of HCA.

For completeness, in Figure 13 and Appendix B, we focus on care delivered within the hospital emergency departments and examine staffing behavior as well as treatment intensity. The analytic setup and estimation closely follow the above, with the sole exception that only the ED discharge database is utilized for these supplementary results. The findings in Figure 13 (Panel A) demonstrate that HCA hospitals reduce their ED physician staffing by approximately 25% by the second year of private equity ownership. The decline in ED physician staffing is also maintained after private equity divestment in 2011. There is no strong evidence that non-physician labor (e.g., nurse practitioners or physician assistants) was used to substitute for physician labor in the affected EDs while under private equity control. The results in Appendix B also show no indications of more aggressive care management for these patients during this time. If anything, there are fewer services provided to these patients.

Fig 13. Diff-in-Diff Event Study Estimates for Number of Unique Physicians Staffing the Emergency Department and the Likelihood of Using Any Nurse Practitioners or Physician Assistants in the Emergency Department.

Notes: Analytic samples and estimation parallel those from Figures 11 and 12. In 2005, the Florida HCA hospitals had 73.7 unique physicians (MDs or DOs) staffing their emergency department (ED) in a given quarter-year. An NP or PA was listed as the main ED provider in only 0.3% of ED visits in a given quarter-year for these same hospitals during the 2005 period. Vertical bars bookend private equity ownership of HCA.

7. Sensitivity checks and spillover analyses

The findings from Sections 4–6 demonstrate important and coherent strategic changes implemented within HCA hospitals during and after private equity ownership. As a final empirical exercise to support these inferences, we modify our analytic approaches to formally assess the presence of any spillovers that would indicate market-wide effects of private equity control among a prominent hospital chain. Such a finding would be interesting in its own right, but it could also reveal potential contamination for our previous DD estimates. Thus, we want to ascertain that our prior results from Sections 4–6 are not sensitive to revising the control comparison group of non-HCA Florida hospitals.

To implement the sensitivity check, we first identify all non-HCA hospitals that are geographically proximate to one or more HCA hospitals (i.e., found within the same county). Of the 67 Florida counties, 23 counties contain at least one HCA hospital (and up to four), and 63 of our control group hospitals are located in these same counties. We assume that these competing hospitals would have the highest propensity to follow any HCA strategic changes that are observable to competitors and perceived to be financially or operationally beneficial. We then remove these hospitals from the control group and re-estimate our key findings from Sections 4–6. The corresponding event studies are displayed in Appendix C.

Notwithstanding larger 95% confidence intervals for some outcomes, the qualitative patterns, point estimates, and inferences are virtually unchanged from our main results in Sections 4–6 when shrinking our control comparison set of hospitals by more than half. The lone exception across the key outcomes reported in Appendix C is for ED physician admitting behavior across patient-payer groups (Appendix Figure C5). The general pattern and interpretations match those from Figure 12 in Section 6. However, the magnitudes of the changes are reduced by as much as half. In other words, the private equity ownership effects are still present and substantive when excluding competing hospitals near HCA hospitals, but our previous estimations in Section 6 may be overstating the size of the effects.

Our final empirical exercise borrows from our control group sensitivity check but then modifies the estimation to test for private equity spillover effects directly. These analyses again focus on our key findings of interest and leverage the DD event study estimations that we have used throughout. The only modification is re-introducing the previously excluded non-HCA hospitals (due to their geographic proximity to HCA hospitals) but not as control hospitals; instead, they are now part of a ‘spillover treatment’ group. All HCA hospitals (i.e., the true treatment group) are dropped from the estimations. Any apparent differential changes during the relevant two post-periods belonging to our main analyses can then be interpreted as HCA-private equity spillovers onto neighboring hospitals.

Given what is evident in Appendix C, it is unsurprising that in Appendix D we almost uniformly fail to detect any changes in non-HCA hospital behavior among competing hospitals in close vicinity to one or more HCA hospitals. These non-HCA hospitals demonstrate no change in their hospitalization quantities or lengths of stay (Appendix Figures D1–D2). Their inpatient treatment intensity is also constant over time, apart from home health referrals which decline in a manner that parallels what is seen for HCA hospitals during this time frame (Appendix Figure D3). Their composition of outpatient procedural services is unchanged in terms of case mix complexity (Appendix Figure D4), and they staff their EDs with the same quantity of physicians over this period (Appendix Figure D6). The noteworthy differential changes for this potential spillover set of non-HCA hospitals localize to inpatient admitting rates for patients presenting to their EDs (Appendix Figure D5). Interestingly, for three of the four main payer groups, HCA competitors are less likely to admit patients when HCA hospitals are sharply more likely to admit their ED patients. The divergence in behavior across these two groups of hospitals in contested markets explains the magnitude discrepancies in the private equity effects shown in Figure 12 (Section 6) and Appendix Figure C5. It also could be indicative of a deliberate strategic choice to distance themselves from the abrupt behavior changes demonstrated by local HCA hospitals in order to be more favorably viewed by consumers, insurers, and/or regulators. Unfortunately, our data do not allow us to definitively establish why these hospitals chose to run counter to the HCA hospital behavior change. We can only speculate on their motivations to adjust almost exclusively along this margin.

8. Conclusions

Risks of anticompetitive effects in hospital markets tied to horizontal integration (e.g., see Gowrisankaran, Nevo, and Town 2015; Schmitt 2017; Cooper et al. 2019; Beulieu et al. 2020; Gaynor et al. 2021; Prager and Schmitt 2021) and/or vertical integration with physician practices (e.g., see Baker, Bundorf, and Kessler 2016; Carlin, Feldman, and Dowd 2016; Koch et al. 2017; Dranove and Ody 2019; Lin, McCarthy, and Richards 2021; Whaley et al. 2021; Richards, Seward, and Whaley 2022) are currently known. Yet, the organizational structures and implicit incentives within the hospital industry are continuously evolving. Private equity, specifically, is playing a growing role within US healthcare, with no sign of abating. Many stakeholders, policymakers, and regulators have raised questions about the potential effects on patients and the healthcare system. Some have already gone so far as to advocate for prohibiting private equity investments in key healthcare industries; though, the rhetoric and public discourse has so far outpaced the evidence.

To help close this knowledge gap, we examine hospital behavior over the full life cycle of private equity ownership by leveraging several advantageous data sources that are unique to the existing literature and allow us to go well beyond the few and limited empirical studies on private equity involvement in US hospitals known to date. While our DD findings show a variety of important––and sometimes large––behavior changes, they are consistent with the private equity owners deploying new strategic management decisions as well as capital to permanently shift the hospital chain into a better financially performing state. Inpatient volumes are elevated by increasing the flows of transfer patients as well as patients from the hospitals’ emergency departments. At the same time, there is no evidence of quality care erosion with respect to in-hospital mortality rates. For the outpatient procedural care side, private equity owners seem to emphasize playing to the hospitals’ comparative advantage––i.e., do fewer total cases but focus on higher complexity surgeries. Higher complexity cases are typically tied to higher reimbursements and may be more difficult for competitors to steal due to the more extensive physical and human capital inputs belonging to their production functions. The chain also adopts a new direct-to-consumer advertising campaign, which leads to much greater advertising expenditures in aggregate and plausibly more patients presenting to its emergency departments, at least among certain payer groups. The company simultaneously becomes an early adopter of the now common hospital-ASC joint venture strategy. Each of these substantive business decisions may have never occurred in the absence of outside management (i.e., private equity ownership) and possibly the absence of outside capital––e.g., if liquidity constraints were tighter prior to the influx of private funding. Interestingly, Bernstein, Lerner, and Mezzanotti (2019) find that private equity-backed companies across economic sectors experienced better investments and financial performance during the 2008–2009 global financial crisis, with private equity employees reporting more time devoted to their portfolio companies during the downturn. This broader private equity behavior aligns with the timing and substance of the capital expenditures we observe for our specific analytic context.

That said, the increase in the propensity to admit patients that present to an affected emergency department following the transition to private equity ownership––coupled with the shift toward shorter, less intensive hospital stays––raises the possibility of a perverse behavior change that could financially benefit the hospital at the expense of consumer welfare.²⁶ Capital investments and/or care delivery efficiency enhancements could also have alleviated pre-existing capacity constraints––and hence facilitated inpatient care that was previously infeasible for the hospital chain. However, the ED admitting behavior change appears soon after the ownership transaction, so any investment or management interventions would need fairly immediate impacts to fully explain the observed change in hospital throughput. It is also noteworthy that insurers (public or private) did not seem to pushback against the increase in hospitalizations, even years later.

In sum, the record-making leveraged buyout of HCA mattered for financial history as well as the short- and long-run behavior and performance of the hospital chain. The observed effects on hospital strategy and operations are admittedly specific to a single private equity transaction but also align with broader management approaches often linked to private equity ownership. The effects also outlive the hospital chain’s spell under private equity control.

Highlights.

• Study examines a record leveraged buyout of a large hospital chain

• Captures effects for private equity (PE) investment and divestment periods

• Finds PE effects on multiple dimensions of hospital strategy and operations