Hospital Mergers and Acquisitions From 2010 to 2019: Creating a Valid Public Use Database

ABSTRACT

Objective

To create, analyze, and distribute the Strategic Hospital Mergers & Acquisitions (M&A) Database, a detailed resource of hospital M&As from 2010 to 2019.

Study Setting and Design

We conducted more than 2000 Internet searches to supplement, verify, and correct M&A identifications of American Hospital Association (AHA) survey data. We assessed the accuracy of the AHA survey and performed staggered difference‐in‐differences analyses to estimate the impact of measurement error on treatment effects capturing shifts in our measure of hospital market power.

Data Sources and Analytic Sample

We analyzed 1537 M&A‐related ownership changes from 2010 to 2019 from our analytic sample of 4896 unique acute care general hospitals or critical access hospitals derived from the AHA Annual Survey dataset.

Principal Findings

The AHA survey dataset correctly identified the M&A deal completion year for only 40.1% of M&A‐related ownership changes. The improved accuracy and granular treatment indicators of our database corrected for underestimations of the impact of hospital consolidation on hospital market power, yielding an effect estimate over 200% higher than the uncorrected data.

Conclusions

By reducing errors in hospital M&A identification, our database can enhance the quality of studies investigating the effects of hospital consolidation on healthcare access and health outcomes.

Summary.

• What is known on this topic

  • ○
    Hospital mergers and acquisitions (M&As) have steadily increased over the past decade.
  • ○
    M&As affect hospital market competition within regions.
  • ○
    Data inaccuracies may obscure the true impact of hospital consolidations on outcomes.
• What this study adds

  • ○
    Our database offers more accurate and granular identification of hospital consolidation events compared to American Hospital Association data.
  • ○
    The study highlights biases in measuring M&A effects with uncorrected data.
  • ○
    Publicly available data can enable more robust research into hospital consolidations' effects on healthcare outcomes.

1. Introduction

Hospital mergers and acquisitions (M&As) have returned to pre‐pandemic levels; the aftermath of COVID‐19 may further intensify this trend [1, 2, 3]. While hospital M&As can lead to higher private health insurance reimbursements, they can also revive underperforming hospitals [3, 4] and increase efficiency in some aspects of hospital performance [5, 6, 7]. Recent empirical studies have examined the impact of hospital M&As, yielding two major strands of studies [5, 8, 9, 10, 11, 12, 13]. One strand discusses the effects of hospital M&As and competition on prices [7, 10, 14] and another focuses on care delivery and quality [11, 12, 13, 15, 16]. While there seems to be a clear impact on prices, the impact on care quality is inconclusive [13].

Because of the inconclusiveness of prior studies—potentially driven by measurement error in timing and identification of M&A transactions—and ongoing hospital consolidation in the US, research on hospital M&As must have the most accurate data possible. The Agency for Healthcare Research and Quality (AHRQ) and Centers for Medicare & Medicaid Services (CMS) have recognized this need, disseminating hospital change of ownership data for 2016 and later [17, 18]. However, a barrier to performing comprehensive research is that hospital M&A databases from earlier years are expensive or—like the American Hospital Association (AHA) annual surveys—error‐prone and incomplete. In their study of private insurance prices and spending, Cooper et al. [10] mention the potential inaccuracies of AHA hospital survey data, one of the most widely used datasets for identifying hospital mergers and system ownership changes [7, 10]. They do not quantify these inaccuracies in their study, however [10].

We thus constructed the “Strategic Hospital M&A Database.” It contains the month and year of M&A deal completion dates for hospital M&As between 2010 and 2019. In this paper, we present various analyses that serve as validity checks of the AHA survey system of accounting for M&As using AHA system ID numbers [19]. A change in a system ID number signals a potential change in ownership.

Our database provides details that cannot be gleaned from AHA data alone. For example, identifying the month and year of deal completions allows for more granular analyses with medical claims or similar data that offer more temporal granularity than annual data. We also identified the type of system change that occurred for each hospital. Using AHA data alone, one would not be able to appreciate the different reasons why a system ID might change. We emphasize the importance of this validation process by examining hospital system market shares before and after M&A transactions within a staggered difference‐in‐differences framework.

2. Study Data and Methods

2.1. Constructing the Database

2.1.1. The System ID

We first used the American Hospital Association (AHA) Survey dataset [19] from 2009 to 2019, which contains a hospital system ID variable coded as “missing” if the hospital is independent. Changes in this variable signal potential changes in system ownership. We are thus concerned with ownership consolidation [16]. To capture system ID changes that occurred between 2009 and 2010, we included 2009 data in our M&A identification process. Because the AHA dataset stores data at the hospital‐year level, hospitals that closed an M&A deal on January 1, 2015, will have the same year‐level deal completion indicator as those that closed on December 31, 2015 (almost 12 months difference). We also found that the year of the AHA system ID change often corresponded with the year of an M&A announcement (which could also mean an intention to merge, rather than an apparent M&A deal completion). A deal completion date is the official start date of a newly consolidated hospital's operations under new ownership. We constructed our database using three steps.

2.1.1.1. Step 1: Identify Year‐Level AHA System ID Changes

We included short‐term acute care general hospitals (from here on: general hospitals) and critical access hospitals (CAHs) as classified by the Medicare provider of service files. There were 4896 unique AHA respondent hospitals in our sample spanning 2010–2019. We identified when each hospital's AHA system identification numbers changed between 2009 and 2019. Changes could indicate an acquisition by a hospital or hospital system, a merger of equals, a divestiture/exit, or an arrangement unrelated to an M&A Appendix Table A1 illustrates this algorithm.

2.1.1.2. Step 2: Google Searches to Identify the Exact Date of Each System ID Change

Next, we performed over 2000 Google searches of the hospitals and their eventual systems. For example, according to the AHA data, “Marian Health Systems” became a part of “Ascension Health” in 2012. After studying various online sources describing the transaction, we found that Ascension Health completed their acquisition of Marian Health Systems in April 2013 [20], over a year later than what the AHA data provided.

When we could not find an exact deal completion date, we recorded the deal announcement date. If no online sources discussed an announcement, we assumed that the deal completion occurred in January of the AHA‐identified year. We include our online sources and URLs in the database.

2.1.1.3. Step 3: Categorize System ID Changes

We categorized system ID changes into one of the following (see Appendix Table A4 for our definitions): Acquisition by hospital/hospital system; Merger of equals; Divestitures or system exits; Bankruptcies or major financial disruptions; False signals; Consulting or management company contracts or other reasons.

We ran the system ID change algorithm on the AHA data. After excluding hospitals not in the 50 US states + DC and facilities not considered general hospitals or CAHs, we yielded a database of 2026 observations. We then identified hospitals involved in an acquisition (as an acquirer or a target) or merger of equals; we defined the union of these, “consolidation events.”

2.2. Analytical Methodology

We compared the M&As identified using the data from our database and the AHA dataset alone. We first examined the proportion of hospitals involved in a consolidation event for each year according to each dataset. Next, we compared the proportion of hospital consolidation events for each year.

Finally, we performed staggered difference‐in‐differences analyses suggested by Callaway and Sant'Anna [21] to assess the dynamic impacts of hospital consolidation events on hospital market power, measured by calculating hospital system market shares. Following prior literature, we used hospital bed ownership as the proxy for market share [22, 23, 24]. We first constructed a panel dataset at the hospital/quarter level. We then derived quarterly indicators for consolidation events corresponding to system ID changes of target or merged hospitals. We calculated hospital system market shares using the total number of hospital beds in an HRR as the denominator. For example, in an HRR with four hospitals of 100 beds each, where two belong to the same system and two are independent, the denominator for each hospital is 400 beds. The numerators are the total beds owned by the same system: 200 for system hospitals and 100 for independent hospitals. We use this measure as an indicator of hospital market power due to its implications for hospital‐insurer bargaining. A two‐hospital system leverages the collective market share of two hospitals in negotiations. If it acquires an independent hospital, forming a three‐hospital system, the newly acquired hospital can demand higher reimbursement rates, leveraging the expanded system's bargaining power.

Because hospitals experience consolidation events at different points, we used Callaway and Sant'Anna [21] to account for staggered adoption of consolidation. We assumed treatment irreversibility following the approach outlined by Callaway and Sant'Anna [21] and clustered standard errors at the hospital level (see the Supporting Information: Appendix for further explanation of the empirical approach). Furthermore, we estimated how measurement error could impact these estimates without correcting for proper deal completion timing and consolidation classifications. For hospitals that underwent multiple consolidation events, we allowed system market shares to grow with consolidation activity, but identified the last instance of consolidation as treatment, providing the most conservative estimators possible.

We ran three models, each with the same corrected system market shares outcome but varied inaccuracies in treatment specification and timing. (Model 1) with our database; (Model 2) with consolidation events properly identified but without corrected deal completion dates; (Model 3) with the AHA‐only system ID change algorithm (no corrections to deal timing and categorization). Please see the Supporting Information: Appendix for further details of our market power analysis.

3. Study Results

During our analysis of system ID changes between 2010 and 2019, we identified 1115 system ID changes corresponding to hospital acquisitions and 422 corresponding to mergers of equals (1537 total system ID changes corresponding to consolidation events). These reflect heavy M&A activity by systems such as HCA Healthcare (in Florida and the Southeast), CommonSpirit Health (in the Southwest), and Beth Israel Deaconess Medical Center (in the Northeast). Appendix Figure A2 illustrates hospital consolidation activity in the US between 2010 and 2019.

3.1. Accuracy Analysis and Descriptive Statistics

Of the 2026 system ID changes by a sample hospital, 1573 (77.6%) had associated articles online, and 1537 (75.9%) of them described consolidation events. Of the 1595 general hospital records, 1301 (81.6%) had a completed deal date documented online. Of the 1301 records of general hospitals with articles online, only 522 (40.1%) of the deal completion years in the AHA database were accurate. Of the 431 CAH records, only 272 (63.1%) had a completed deal date documented online. Of the 272 records of CAHs with articles online, only 112 (41.2%) of the deal completion years in the AHA database were accurate.

The most common reasons for system ID changes were due to acquisitions by other hospitals (1115) and mergers of equals (422). Most false signals (181 total) were likely due to hospital errors in completing the AHA survey. We also observed many system splits of exit events (40) in 2016 when Community Health Systems spun off Quorum Health Corporation [25]. Appendix Table A5 summarizes the distribution of reasons for system ID changes.

Next, we present the frequencies of each hospital consolidation event across the years 2010–2019 in Table 1. Out of the 4896 unique hospitals in the AHA surveys identified over this period, almost half were on the acquiring side of an M&A at least once between 2010 and 2019. Approximately 8% were on the acquiring side of an M&A transaction at least 10 times.

TABLE 1.

Summary of consolidation events and their characteristics over 10 years of the study.

Frequency	Acquirer	Acquirer as system	Targeted	Targeted as system	Mergers	Merger of systems
1	630 (13%)	606 (12%)	845 (17%)	446 (9%)	412 (8%)	361 (7%)
2–4	774 (16%)	759 (16%)	129 (3%)	82 (2%)	5 (0%)	4 (0%)
5–9	532 (11%)	531 (11%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
10 or over	395 (8%)	393 (8%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
No event	2565 (52%)	2607 (53%)	3922 (80%)	4368 (89%)	4479 (91%)	4531 (93%)
Total unique hospitals	4896

Note: Authors' analysis of deal completion data from their final analytical dataset featuring the Strategic Hospital M&A Database, 2010–2019 hospital data from the American Hospital Association Annual Survey, and 2010–2019 data from the Centers for Medicare & Medicaid Services Provider of Services files. The authors tabulated the frequencies for each consolidation event type. Being an acquirer as a system means that a hospital was part of a multi‐hospital system that acquired a hospital or hospital system. Being targeted as a system means being a hospital part of a targeted multi‐hospital system. A merger of systems means two distinct hospital systems merging into one system.

3.2. Comparison With AHA Results

Figure 1a presents trends in the rate of M&As, contrasting the estimates obtained using ONLY the AHA data versus the estimates we obtained using our Strategic Hospital M&A Database. In most years, the AHA data overestimated the proportion of hospitals involved in a consolidation event. Of note, we see a major difference in percentages in 2010, directly attributable to several instances of erroneous omissions of the AHA system ID value in 2009 (thus, picking up erroneous consolidation events in 2010). Figure 1b presents corresponding trends in average Herfindahl–Hirschman Indices (HHIs) over the same period. Using uncorrected AHA data increased the identification of erroneous consolidation events, inflating the average HHI across the study period. We also present results for urban and rural markets separately in Appendix Figure A3 to illustrate how different market types are affected by M&A indicator accuracy.

FIGURE 1.

Comparison of hospital consolidation events and HHI trends. The Strategic Hospital M&A Database refers to the deal completion data from the final analytical dataset, which includes data from the Strategic Hospital M&A Database (2010–2019), the American Hospital Association (AHA) Annual Survey (2010–2019), and the Centers for Medicare and Medicaid Services Provider of Services files (2010–2019). The raw AHA‐only data refers to deal completion indicators generated solely using uncorrected AHA system ID data. For panel (a), the authors used the Strategic Hospital M&A Database and the AHA‐only consolidation event indicators to calculate the proportion of hospitals experiencing a consolidation event (as a target, part of an acquiring system, or part of a merger of equals). For the Herfindahl–Hirschman Indices trends in panel (b), the authors used the Strategic Hospital M&A Database to track changes in hospital system structure at the Hospital Referral Region level, compared to calculations relying solely on the uncorrected AHA data. The authors include all short‐term acute general hospitals and critical access hospitals in the calculations.

3.3. How Hospital Consolidation Events Impact Hospital Market Power

Finally, we present the results from our staggered difference‐in‐differences analyses [21] estimating the impact of hospital consolidation events on hospital market power. Figure 2 reveals the estimated effects over time. Estimators for Model 1 (M1) exhibit an immediate discrete jump following consolidation, while the less accurate data do not show this. There is a clear bias to the null with less accurate data, as M1, M2, and M3 yield the aggregated coefficients 0.0322, 0.0235, and 0.0094, respectively. This bias could impact the results of studies with shorter study periods, introducing substantial measurement error. In Appendix Table A6 we present short‐run estimates spanning eight quarters (2 years) after consolidation. We observe the substantive differences in the magnitudes of the dynamic estimates resulting from the data inaccuracies in Models 2 and 3. These inaccuracies biased the estimated relationship between consolidation events and hospital market power.

FIGURE 2.

Impact of hospital consolidation on system market share, by different consolidation identifiers. Model 1: Aggregated Coefficient = 0.0322*** (Std. Err. = 0.0033), Hospital‐quarters = 184,836. Model 2: Aggregated Coefficient = 0.0235*** (Std. Err. = 0.0026), Hospital‐quarters = 181,552. Model 3: Aggregated Coefficient = 0.0094*** (Std. Err. = 0.0025), Hospital‐quarters = 181,736. ***p < 0.01, **p < 0.05, *p < 0.1. The authors' analysis of deal completion data from their final analytical dataset featuring the Strategic Hospital M&A Database, 2010–2019 hospital data from the American Hospital Association (AHA) Annual Survey, and 2010–2019 data from the Centers for Medicare & Medicaid Services Provider of Services files. The number of hospital‐quarters varies by model because of differences in the number of “always‐treated” units. Always‐treated hospitals experienced their last (thus, their only) consolidation event during the first quarter of 2010. The Callaway and Sant'Anna [21] method ignores always‐treated units because they do not have pre‐treatment outcome data that can be used for estimation. The N varies by model because of erroneously labeled consolidation events and erroneous timing of the last instance of consolidation events (e.g., there could be properly labeled consolidation events with erroneous timing, or erroneously labeled consolidation events that affect when the final instance of consolidation occurs). The authors ran three staggered difference‐in‐differences models [21] (using the “csdid” command in Stata), each using different consolidation identifiers: (Model 1) with our database (properly identified consolidation events and corrected deal completion timing); (Model 2) with consolidation events properly identified but without corrected deal completion dates; (Model 3) with the AHA‐only system ID change algorithm. The authors ran all models at the hospital/quarter‐level and clustered standard errors at the hospital level.

4. Discussion

Over half of all US general acute care hospitals are currently part of a hospital system [26], with some involved in 10 or more consolidation events between 2010 and 2019. This is primarily due to major hospital systems regularly acquiring other hospitals [10, 27, 28].

Hospital consolidation events have triggered significant changes in hospital system market shares within HRRs. In our difference‐in‐differences analysis of this phenomenon, we anticipated a noticeable increase in system market shares following hospital consolidation events. For example, consider three independent hospitals in a local market, each with an equal number of beds. Each hospital would initially control one‐third of the market. If two of these hospitals merged, the resulting system would control two‐thirds of the market, a discrete increase of one‐third.

While system market share responds immediately to consolidation events, health‐related outcomes—such as 30‐day unplanned readmissions—may be influenced by multiple care delivery factors, following a more gradual trajectory. As such, measurement errors in identifying M&As could obscure the true effects of these events. While our database's within‐year timing is best used with temporally granular data, such as medical claims data, it also provides substantive utility with data that only vary annually. A well‐regarded strategy to minimize potential bias around the treatment period is establishing a “washout period” where the year of treatment is excluded from the estimation sample [29]. Our database corrects most of the 60% of deals with a misidentified deal completion year derived from uncorrected AHA data, allowing researchers to establish more accurate and informed washout strategies. Let's say we find a deal completion year of 2015 with a deal completion month in March. The researcher can drop data from 2014 if they are concerned with potential pre‐treatment anticipatory behavior, or from 2015 if they wish to wash out most of the treatment year, following Owsley and Lindrooth [29]. The researcher is empowered with more informed analytical options.

Our database and study have limitations. First, the AHA survey has a response rate of approximately 80% [30]. Second, a few deals in our database did not have publicly available information online. Third, our database contains measurement error, including those stemming from ambiguous or restricted sources. Despite these limitations, our database offers substantial improvements over the AHA data alone.

Our database provides three key benefits for researchers studying hospital M&As. First, it offers accurate deal completion dates with more precise and granular M&A timing, reducing measurement error. This is crucial for research, policy, and antitrust agencies responding to the short‐term impacts of hospital M&As. For example, our database enables a more accurate measurement of the time between M&A deal completion and hospital–insurer bargaining over reimbursement rates. Second, it distinguishes transaction types, allowing researchers to differentiate between consolidation events and system exits, reducing biases from misidentified events. Finally, the 2010–2019 database will be publicly available, expanding opportunities for analyses and collaboration between researchers and government agencies. This resource complements the existing AHRQ Compendium [17] and CMS [18] databases, offering the most detailed publicly available M&A data from 2010 to 2019.

We plan to update the database annually to reflect ongoing consolidation trends and ensure relevance to current research. We aim to leverage our robust database and the latest artificial intelligence‐assisted methods to automate hospital consolidation event identification, enhancing data collection efficiency and accuracy. The Strategic Hospital M&A Database will empower public health researchers to conduct more accurate studies on the effects of hospital consolidation on health outcomes.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Oh H., Mor V., Kim D., Foster A., and Rahman M., “Hospital Mergers and Acquisitions From 2010 to 2019: Creating a Valid Public Use Database,” Health Services Research 60, no. 5 (2025): e14642, 10.1111/1475-6773.14642.

Data Availability Statement

To access the latest post‐publication version of the Strategic Hospital M&A Database, please visit: hyesunghaceoh.github.io/data.