Background and importance

U.S. acute and hospital-based care capacity have evolved over time, and their relevance is all the more salient given the challenges of the COVID-19 pandemic. This evolution occurs in the context of recent hospital and health system consolidation, with larger facilities capturing greater market share and consolidating care among fewer, more integrated systems nationwide [1]. Consolidation had been driven by numerous economic forces and uncertainty in the political climate following passage of the Affordable Care Act [2,3]. News reports have suggested that amidst this wave of consolidation hospitals and their associated emergency departments (EDs) may often be closed even in areas where readily-available alternative care sites are not present [4], though there is little empirical investigation to support this claim. Prior work has documented several impacts of consolidation including loss of market competition [5–7], incentivized coordination of care [8], and in select geographies, reduced access to hospital-based emergency care and poorer outcomes [9].

Access to care remains a challenge in the U.S., and this challenge is more than simply a question of having health insurance. Access is a complex multidimensional construct inclusive of coverage, a source for healthcare services, timely availability of services and a capable workforce to deliver services [10]. Although the Emergency Medical Treatment and Labor Act (EMTALA) protects patients’ access to emergency care-in that all who present must receive a medical screening exam, meaningful access to timely hospital and emergency care services is primarily driven by geography, or simply put the presence of a hospital-based emergency department close by [11].

Prior work based on national survey data found that between 1990 and 2009, the number of hospital-based EDs declined by 27% in non-rural areas, and for-profit owned EDs, those in competitive markets and those with safety-net status were more likely to close [12]. In the decade since, the financial underpinnings of emergency and hospital care in the U.S. has changed substantially as a result of growing profits from hospital-based ambulatory diagnostic and treatment centers as well as expansion of both Medicaid and Affordable Care Act-based insurance exchange populations [13]. Amidst a wave of consolidation, some hospital-based EDs may remain open as mergers favor health system network expansion rather than closure of a single facility [14].

Timely access to hospitals and EDs has been associated with outcomes after trauma and myocardial infarction, as well as inpatient mortality [9,15,16]. Both hospital and ED closures have been described as limiting access and associated with worse clinical outcomes at the state level in cardiovascular and trauma care [9,16]. Furthermore, for EDs that remain open, ED visit volumes have steadily increased year-over-year [17], and this may be one contributor to ED crowding and its downstream consequences [18,19]. Despite the importance of timely access to EDs with appropriate capacity, there has been no recent national assessment of ED closures and mergers as well as hospital capacity that accounts for both recent health system consolidation and the passage of the Affordable Care Act in 2010. The pressures placed on health systems, and the likelihood of ongoing surges in the Covid-19 outbreak, make the capacity of EDs and hospitals all the more salient.

Goals of this investigation

The goal of this investigation was to describe hospital-based ED closures and mergers in the period from 2005–2015. We characterized closures and mergers by urban-rural designation, ED visit volumes, as well as facility teaching and for-profit status. Our secondary objective was to characterize yearly ED visits per hospital-based ED and yearly hospital admissions per hospital bed over time and across urban-rural designation.

Study design and setting

This was a secondary analysis of the American Hospital Association (AHA) Annual Survey Database with data from years 2005 through 2015 [20]. The Annual Survey Database is a proprietary dataset maintained by the AHA compiling yearly results of an AHA-directed nationwide survey of all hospitals in the United States (including the 50 states, and excluding Puerto Rico, Mariana Islands, Marshall Islands, Virgin Islands, Guam, American Samoa) as well as data from the U.S. Census Bureau and hospital accrediting bodies. Response rates among hospitals ranged from 84% to 92% in the years analyzed. We specifically excluded Veterans Affairs and other hospitals related to the federal government/military (government, nonfederal facilities were included) as defined by control code in the AHA database (excepting the Indian Health Service, which was included). Our analysis included all short-term acute care hospitals providing general medical and surgical services, as defined by service code, that are listed in at least two contiguous years of data. Stand-alone pediatric facilities are identified by a different service code and therefore excluded from the present analysis. S1 Fig gives counts for included hospitals by exclusion criteria.

Measurements

Hospitals were linked year-over-year, and we defined a hospital-based ED for those hospitals that reported at least two consecutive years with >500 ED visits per year. This approach, requiring 2 years of consecutive data and >500 ED visits per year, was set to exclude potentially spurious sites in this dataset composed of aggregated hospital survey responses. As a point of reference, there was an estimated 28.4 million ED visits and 1,855 rural EDs in 2016, so that the average yearly ED visits per rural ED would be 15,309, while sites with <500 ED visits per year were felt to be less likely to represent a bona fide point of emergency care services. AHA data were combined with American Community Survey data [21] from the U.S. Census Bureau at the zip-code level to characterize population level characteristics as they are related to proximity to an ED, as well as Rural Urban Commuting Area (RUCA) code, a scheme for delineating urban and rural areas based on the American Community Survey. We defined four categories (designated according to recommendation of the Rural Health Research Center): urban (RUCA codes 1.0, 1.1, 2.0, 2.1, 3.0, 4.1, 5.1, 7.1, 8.1, 10.1), large rural (4.0, 4.2, 5.0, 5.2, 6.0, 6.1), small rural (7.0, 7.2, 7.3, 7.4, 8.0, 8.2, 8.3, 8.4, 9.0, 9.1, 9.2), and isolated (10.0, 10.2, 10.3, 10.4, 10.5, 10.6) [22]. These groupings are standardized from the Rural Health Research Center, and sort tracts according to patterns of commute (for example, a rural tract with 30–50% of commuting secondary flow to an urban area is categorized as urban) [23]. We defined ED visit volume categories for analysis consistent with the Emergency Department Benchmarking Alliance and the American College of Emergency Physicians Clinical Emergency Data Registry (500 to 20k, 20-40k, 40-60k, 60-80k, 80k+) [24]. Hospital-based EDs were characterized by each of ED visit volume categories, urban-rural designation, teaching status, and for-profit status.

Outcomes

Our primary outcomes were the number of hospital-based ED closures and hospital-based ED mergers (sites that remained open but were subject of a merger with another health system) by year. Available with each year of the AHA dataset, there is a supplemental document with a described reason for each hospital site removed or added to the data. A hospital-based ED closed if the site no longer reported >500 ED visits in a given year, or if the hospital closed. By defining closures with respect to ED visit counts in two contiguous years from available data in 2005 through 2015, closures were defined in the analysis only for the period from 2006–2014. If a site was removed from the dataset and their listed description was not for a merger, but some other change, these were independently reviewed by research staff to clarify if the site had indeed closed. A hospital-based ED was considered to have closed if the site was not reported in a subsequent year, even with a listed reason for removal as merger, if that geographic site was not found to be operating a hospital-based ED site in 2015 after independent review by research staff. Some hospitals may remain open while their associated ED is closed, and these fit our outcome definition for a hospital-based ED closure. Operation of ED sites at the specified locations were primarily assessed via published notes or histories from the associated healthcare system. When such information was not available, ongoing operation was assessed via a combination of calls to the appropriate healthcare system and internet searches for the appropriate locations. Any site deemed permanently closed was confirmed by at least two separate sources. If the hospital-based ED was in fact still in operation in 2015 as clarified by external sources, but removed from the dataset for a listed merger, then the most recently available year of data in AHA was copied forward to 2015. These sites are defined as merged for the purposes of our analysis (that is, the reported outcome for merger is a site that remained in operation). This approach ensured that an ED was not defined as closed when in fact the site remained in operation and simply consolidated with a larger system. Total ED visits rose year-over-year for nearly all sites (>90%), which suggests that this approach of copying data forward where sites were removed from the dataset but in fact remained open as part of a hospital or health system merger would provide a conservative estimate of total ED visit volume. Those sites that were listed as new in the dataset but ‘not previously registered’ were not listed as new hospitals. We include those listed as ‘new’ or ‘newly re-opened’ as opening facilities, with the caveat that these facilities must meet the same criteria for hospital-based ED as above.

We tabulated total ED visit volumes, total hospital-based EDs, total acute care admissions, and total inpatient beds by year. To summarize these changes over time, we then calculated the nationwide average yearly ED visits per hospital-based ED and the nationwide average yearly hospital admissions per hospital bed.

Analysis

For the primary analysis, year-over-year trends in hospital-based ED closures and mergers were characterized by urban-rural location, ED visit volumes, as well as facility teaching and for-profit status. Univariable and multivariable logistic regression models were constructed to estimate the likelihood of hospital-based ED closure and merger among the following covariates: urban-rural location, ED visit volume, facility teaching and for-profit status. Each observation for model estimation is a unique hospital-based ED, and we assessed for multicollinearity amongst independent variables by obtaining associated variance inflation factors. Both unadjusted and adjusted odds ratios and 95% confidence intervals were calculated for the relationship between urban-rural designation, ED visit volume, teaching status, as well as for-profit status and hospital-based ED closure and merger during the study period. The geographic location of closures and mergers was represented according to the hospital referral region (HRR) where the closure or merger took place, subject to the limitation that only those HRRs with 5 or greater hospital-based EDs were depicted to protect hospital identities according to AHA data use agreement. This was to appropriately limit the granularity of the results according the AHA’s terms of data use. We furthermore describe the year-over-year trends in ED visit and acute care hospitalization volumes according to urban-rural designation. Our analysis does not include time trends, as event rates for closure and merger were fairly uncommon to derive stable estimates, and variation in other ED characteristics generally static over time, to provide a definitive analysis of how factors associated with closure and merger have changed over time. The secondary analysis consisted of tabulating number of ED visits, number of hospital-based EDs, number of inpatient stays, and number of available hospital beds by year. The outcomes were then defined as the ratio of ED visits to EDs and hospital admissions to hospital beds. The percent change from 2005 baseline through 2015 in these two outcomes was calculated, stratified by urban-rural designation (urban, large rural, small rural, isolated). S1 Table provides the STROBE checklist for the present investigation. All analyses were performed in R (R Foundation, version 4.0.1).

Limitations

First, hospital reporting in the AHA dataset is not entirely standardized, and the dataset relies on self-report from a representative of each individual hospital responding to the survey. This introduces the risk of inaccurate reporting or missing data. Second, though we regard this also as a potential strength of our analysis, hospital mergers may also represent a limitation. It is possible that what appear as ED closures in our dataset are in fact simple hospital consolidation into larger systems with ED reporting absorbed in to the larger hospital. Our analysis attempted to specifically address this issue, as each site dropped from the dataset was individually reviewed. Third, an important aspect of the Affordable Care Act was the initiation of statewide Medicaid expansions in 2014, but because closure and merger are defined by two consecutive years of data for a given hospital-based ED, we do not have results after the expansions. Finally, though our analysis can address the distribution of EDs across urban and rural areas, it does not address more specific care processes such as time-to-arrival for ED evaluation, or the change in how far a person may need to travel to the next closest ED when one closes, that may have changed in the period we analyzed. There is a great deal of literature on the topic of care for time-sensitive conditions, and future analyses should explore the relationship between the trends observed in this work and specific patient outcomes for time-sensitive illness [9,15,16].