- Email: [email protected]
Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013
American Journal of Emergency Medicine xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
American Journal of Emergency Medicine journal homepage: www.elsevier.com/locate/ajem
Brief Reports
Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013 Ari B. Friedman, BA, MS a,⁎, D. Daphne Owen, MD b, Victoria E. Perez, PhD c a b c
Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia, PA Department of Emergency Medicine, University of Pennsylvania, Philadelphia, PA School of Public and Environmental Affairs, Indiana University, Bloomington, IN
a r t i c l e
i n f o
Article history: Received 19 February 2016 Accepted 5 April 2016 Available online xxxx
a b s t r a c t Study Hypothesis: Low reimbursement from the uninsured has been claimed to threaten hospital finances and even hospital emergency department (ED) closure. We hypothesized in advance of beginning data collection that states that expanded Medicaid (“expansion states”) under the 2010 Patient Protection and Affordable Care Act would experience a reduced rate of ED closure compared with states that did not. Methods: We compiled a national census of EDs from 2006 through 2013 from federal databases and manually confirmed each closure. We used difference-in-differences regression on this longitudinal panel to compare the probability over time that a hospital was in operation in expansion states to nonexpansion states. Results: The number of hospitals grew every year nationally and in nonexpansion states. In expansion states, the number fell from 2027 in 2009 to 2019 in 2010, not surpassing the 2009 peak until 2012. In regression estimates, hospitals in expansion states were 2.2% (95% confidence interval, 0.3%-4.1%) less likely to be in operation after 2010 compared with the trend in nonexpansion states. Conclusions: States that expanded Medicaid experienced increased, rather than reduced, ED closure rates from 2010 through 2013. The financial benefits of the Affordable Care Act may be poorly targeted to the hospitals most vulnerable to closure. © 2016 Elsevier Inc. All rights reserved.
1. Introduction 1.1. Background Low levels of reimbursement from the uninsured have long been claimed to be a cause of poor hospital finances and even hospital emergency department (ED) closure. The Patient Protection and Affordable Care Act (ACA) aimed in part to reduce such uncompensated care.
Findings from this project, although not these specific results, were presented at the American Economics Association Meeting; San Francisco, CA; January 5, 2015. The authors received no grant funding and have no conflicts of interest to declare. A.B.F. and V.E.P. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. A.B.F. and V.E.P. conceived the study; designed the analysis; supervised the conduct of the trial and data collection; managed the data, including quality control; provided statistical advice on study design; and analyzed the data. D.D.O. and V.E.P. verified hospital closures and provided substantial input into the inclusion/exclusion criteria and search strategy. A.B.F. drafted the manuscript, and all authors contributed substantially to its revision. A.B.F. takes responsibility for the manuscript as a whole. ⁎ Corresponding author at: Leonard Davis Institute of Health Economics, University of Pennsylvania, 3641 Locust Walk, Philadelphia, PA 19104. Tel.: +1 215 284 5196. E-mail address: [email protected] (A.B. Friedman).
Recent studies show improving finances among hospitals in states that elected to expand Medicaid through the ACA [1]. Similarly, hospital associations and the popular press [2] have pointed to isolated hospital closures as evidence that failure to expand Medicaid has increased ED closures. 1.2. Importance Hospital ED closures may reduce access to care, increase transportation costs for patients, and disrupt local economies. Emergency department closures have also been associated with increased morbidity and mortality among the population living nearby in some studies [3], but not in others [4]. Because most (96.9% in 2012, authors' calculations) of acute care and critical access hospitals have an ED, we assume that ED closures are driven by hospital closures. 1.3. Goals of this investigation A longstanding trend in ED closures was documented before implementation of the ACA (1998-2008) and showed that safety-net status and low-profit margins were associated with increased risk of ED closure [5]. However, no studies have evaluated hospital closures since the ACA shifted hospital expectations about future uninsurance and Medicaid volumes. We hypothesized that states that expanded
http://dx.doi.org/10.1016/j.ajem.2016.04.006 0735-6757/© 2016 Elsevier Inc. All rights reserved.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006
2
A.B. Friedman et al. / American Journal of Emergency Medicine xxx (2016) xxx–xxx
Table 1 Baseline characteristics of hospitals in expansion vs nonexpansion states
DSH ($) Tenure (y) Hospitals in ZIP (#) CAH (%) HHI
Nonexpansion
Expansion
t Test (P value)
1 989 546 (4 318 104.0) 7.6 (1.3) 1.2 (0.5)
2 908 721 (5 771 765.0) 7.7 (1.1) 1.2 (0.4)
b.01 b.01 .85
25.4 (43.6) 2210 (1429)
24.8 (43.2) 2247 (1422)
2.2. Inclusion and exclusion criteria
.62 .53
The table presents mean (SD) characteristics of hospitals in 2006, the first year of the study. DSHs are the mean disproportionate share payments per hospital, in dollars. Tenure is the mean number of years hospitals were open. HHI is the Herfindahl-Hirschman Index, a measure of market concentration (10 000 is a monopoly, 0 is perfect competition), by bed share of Hospital Referral Region. CAH is the percent of hospitals with a critical access hospital designation.
Medicaid, either directly or through a state-specific waiver, experienced a trend toward a greater number of nonspecialty hospitals with EDs compared with states that had not yet expanded Medicaid by 2013. The hypothesis was formulated in advance of beginning data collection. Manually verified panel data from the Healthcare Cost Reporting Information System (HCRIS) provided a longitudinal census of open hospitals and were not subject to the nonresponse bias and imputation of the more commonly used American Hospital Association Annual Survey. 2. Methods 2.1. Data source for hospital status Data from the Centers for Medicare and Medicaid Services (CMS) HCRIS provides a national census of hospital financial reports from 2006 to 2013. Provider identifiers in this data set uniquely identify facilities, even through mergers and acquisitions. We supplemented HCIRS with 684 hospital-years meeting inclusion criteria from the Hospital Compare database and with 216 hospital-years of data from the CMS “late report,” consisting of hospitals that notified CMS that they remained open despite delays in filing a financial report. We identify potential openings/closures through the appearance/ disappearance of each provider identifier. We confirmed each closure by searching news databases (Google News, Lexis Nexis) for definitive local news reports. Closures that remained unconfirmed by this process were then confirmed through a direct phone call.
A
We restricted the sample to acute care, critical access, and federal hospitals with a 24-hour ED. We excluded all Indian Health Service hospitals (292 hospitalyears). We excluded specialty hospitals by searching their hospital names for common terms (eg, “ortho,” “surgical”) from HCRIS and the Indirect Medical Education to identify 242 specialty hospitals. We then confirmed these specialty hospitals by visiting Web sites, calling the facilities, and checking the National Plan and Provider Enumeration System. 2.3. Additional covariates Additional information on whether a hospital qualifies for the Critical Access program and on the number of staffed beds was obtained through the Hospital Compare and IME and Graduate Medical Education systems by merging on the same unique provider identifier. 2.4. Statistical models t Tests were used to compare individual characteristics across expansion vs nonexpansion states (Table 1). To assess the probability of hospitals remaining open, we used a difference-in-differences specification, implemented via linear probability model regression with year fixed effects. To account for the fact that the models regress hospital outcomes on state-level measures of Medicaid adoption, we cluster standard errors by state. Because hospital closure decisions likely take into account expectations regarding financial challenges, we used 2010, the year in which the ACA passed, as the beginning of the treatment period. Sensitivity analyses used a logit specification and treatment period beginning in 2012. 2.5. Ethical approval This study did not assess human subjects and was therefore exempt from institutional review board review.
B 3200
5400
C
5100
4800
4500
Percent of Open Hospital EDs (Compared to 2006)
105
Number of Open Hospital EDs
Number of Open Hospital EDs
Of 444 suspected closures, 134 hospitals were not acute care or critical access hospitals. Of the remaining 310 potential closures, 194 hospitals were open and had never closed. Of the 116 hospitals that closed, 5 reopened more than 6 months later and 4 moved to a different locations. For 11 of the closed hospitals, we were unable to locate additional data on when their closure occurred.
2800
2400
2000
104
103 ACA Status Non−Expansion Expansion
102
101
100 1600
4200 2006
2008
2010 Year
2012
2006
2008
2010 Year
2012
2006
2008
2010
2012
Year
Figure. Number of open hospital EDs over time. Trends in number of hospitals open over time (A), broken down by state Medicaid expansion status (B), and percentage of number of hospitals open over time (C) relative to 2006 baseline, by state Medicaid expansion status.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006
A.B. Friedman et al. / American Journal of Emergency Medicine xxx (2016) xxx–xxx Table 2 Differences-in-differences regression results
Post-2010 Expansion Post × expansion
Coefficients (95% CI)
P
0.0419⁎⁎ (0.0232 to 0.0607) 0.0223 (−0.00083 to 0.045) −0.0219⁎ (−0.041 to −0.0030)
b.01 .061 .026
Linear probability model difference-in-difference regression results (n = 40 275 hospitalyears). The dependent variable is a binary indicator of whether a hospital was open in a given year. Standard errors are clustered at the state level. Model includes year fixed effects. ⁎ Significance level: P b .05. ⁎⁎ Significance level: P b .01.
3. Results 3.1. National trends There were, on average, 3521 acute care hospitals and 1289 critical care access hospitals in operation each year, for a total of 38 480 hospital-year observations between 2006 and 2013. More hospitals were open at the end of the study period than at the beginning. In the 4 years from 2006 through 2009, there were an average of 4765 hospital EDs open per year, compared with 4855 in the 4 years following (1.9% increase in number of open EDs). Figure, Panel A demonstrates that there were no years in which the number of hospital EDs declined nationally during the study period. 3.2. Comparison across Medicaid expansion status Table 1 provides characteristics of hospitals in expansion and nonexpansion states. Hospitals in expansion states were substantively similar to those in nonexpansion states in terms of number of competitor hospitals and probability of being a critical access hospital. Disproportionate Share Hospital (DSH) payments per hospital and average number of open years were lower in nonexpansion states. In 2006, there were 2006 hospitals in states that would eventually expand Medicaid, compared with 2718 in nonexpansion states. By 2013, there were 2040 in expansion states, compared with 2854 in nonexpansion states. In expansion states, the number of hospitals dipped briefly from 2027 in 2009 to 2019 in 2010, but recovered to 2031 by 2012. Panels B and C in Figure demonstrate a similar growth rate in expansion states and nonexpansion states before 2010, followed by a break in the trend that occurred only in nonexpansion states. Results of the differences-in-differences regression are shown in Table 2. Hospitals in expansion states were 2.2% (95% confidence interval [CI], 0.3%-4.1%) less likely to be in operation post-2010, controlling for both the trend in nonexpansion states to be 4.2% more likely to be in operation after 2010 (95% CI, 2.3%-6.1%) and for a 2.2% higher baseline rate of the remaining in operation in expansion states (95% CI, −0.083% to 4.5%). Results were robust to sensitivity analyses involving model specification (logit) and time period (2012 start date). 4. Discussion A number of news reports document hospital closures, particularly in states that have not expanded Medicaid eligibility. Whether these closures were occurring at an accelerated rate after passage of the ACA or in states that did not expand Medicaid remained unknown. In this article, we provide an overview of how hospital closures in recent years have differed from pre-ACA trends, both overall and by Medicaid expansion status. Hsia et al [5] documented a long-term trend in nonrural hospital ED closures from 1998 to 2008. Our analysis from 2006 to 2013 demonstrates that this trend seems to have reversed itself after passage of
3
the ACA. Direct comparison between studies is impossible, however, because our census of hospital EDs uses a different hospital database, includes rural and critical access hospitals, excludes specialty care and Indian Health Service hospitals, and manually verifies each closure. In addition to finding that nationally the number of hospital EDs is growing, we find that states that expanded Medicaid by 2013 experienced a lower rate of hospital growth (fewer hospitals in operation relative to the trend) than states that did not. This finding is contrary to our initial hypothesis and may be explained by complex interactions of policy, billing, and hospital finances. The first possibility is that nonexpansion states are systematically different from expansion states in ways that make them less likely to experience hospital closure, such as more “business-friendly” environments, more for-profit hospitals, or fewer certificate-of-need regulations. For-profit hospitals close more rapidly when patient mix becomes unprofitable, but they also open more rapidly when it becomes profitable [6]. The second possibility is that hospitals most at risk for closure may not be those experiencing improved financial outcomes. For instance, these hospitals might serve a population largely ineligible for Medicaid expansion, such as US citizens above 138% of the poverty line or undocumented immigrants. A related explanation is that hospitals at risk for closure might have more patients formerly on private insurance switch to Medicaid, with resultant lower reimbursement [7]. The third possibility is that Medicaid expansion is merely one piece of the post-ACA health insurance scheme, and the effect of expansion on reducing ED closure may be positive but dominated by other factors. Most notably, previously eligible individuals signed up for traditional Medicaid at greater rates due to publicity surrounding the ACA [7], and the ACA-established health insurance exchanges were created in all states. Because nonexpansion states generally had lower baseline insurance rates, these changes may have offset the Medicaid expansion. The apparent reversal of the overall trend from ever fewer EDs documented in earlier work to ever more EDs documented in ours may result from these effects. Finally, given that only 4 years of postpassage data and 2 years of postimplementation data were available, these effects may simply be short-term adjustments. However, considering that hospital closure is typically permanent, even transient effects matter for access to emergency care. These findings support claims that DSH payments should be more carefully targeted to vulnerable hospitals. Acknowledgments The authors wish to acknowledge helpful comments by Kosali Simon, PhD, and Seth Freedman, PhD, of IU SPEA, and Karin V. Rhodes, MD, of NorthShore-LIJ. The authors gratefully acknowledge Terry Huynh, MBA, of CMS for providing the late reports and information about the data collection process. The authors thank Felipe Lozano-Rojas, BA, of IU SPEA for research assistance. References [1] Garthwaite C, Graves J, Gross T, Notowidigdo M. The effect of the ACA on access to health care services; 2015[Working Paper]. [2] Olorunnipa T. Obamacare cutbacks shut hospitals where Medicaid went unexpanded. Bloomberg 2013 http://www.bloomberg.com/news/articles/2013-11-25/obamacarecutbacks-shut-hospitals-where-medicaid-went-unexpanded. Accessed December 14, 2015. [3] Liu C, Srebotnjak T, Hsia RY. California emergency department closures are associated with increased inpatient mortality at nearby hospitals. Health Aff 2014;33:1323–9. [4] Joynt KE, Chatterjee P, Orav EJ, Jha AK. Hospital closures had no measurable impact on local hospitalization rates or mortality rates, 2003–11. Health Aff 2015;34:765–72. [5] Hsia RY, Kellerman A, Yu-Chu S. Factors associated with closures of emergency departments in the United States. JAMA 2011;305:1978–85. [6] Norton EC, Staiger DO. How hospital ownership affects access to care for the uninsured. Rand J Econ 1994;25:171–85. [7] Sommers BD, Kenney GM, Epstein AM. New evidence on the Affordable Care Act: coverage impacts of early Medicaid expansions. Health Aff 2014;33:78–87.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006
Contents lists available at ScienceDirect
American Journal of Emergency Medicine journal homepage: www.elsevier.com/locate/ajem
Brief Reports
Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013 Ari B. Friedman, BA, MS a,⁎, D. Daphne Owen, MD b, Victoria E. Perez, PhD c a b c
Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia, PA Department of Emergency Medicine, University of Pennsylvania, Philadelphia, PA School of Public and Environmental Affairs, Indiana University, Bloomington, IN
a r t i c l e
i n f o
Article history: Received 19 February 2016 Accepted 5 April 2016 Available online xxxx
a b s t r a c t Study Hypothesis: Low reimbursement from the uninsured has been claimed to threaten hospital finances and even hospital emergency department (ED) closure. We hypothesized in advance of beginning data collection that states that expanded Medicaid (“expansion states”) under the 2010 Patient Protection and Affordable Care Act would experience a reduced rate of ED closure compared with states that did not. Methods: We compiled a national census of EDs from 2006 through 2013 from federal databases and manually confirmed each closure. We used difference-in-differences regression on this longitudinal panel to compare the probability over time that a hospital was in operation in expansion states to nonexpansion states. Results: The number of hospitals grew every year nationally and in nonexpansion states. In expansion states, the number fell from 2027 in 2009 to 2019 in 2010, not surpassing the 2009 peak until 2012. In regression estimates, hospitals in expansion states were 2.2% (95% confidence interval, 0.3%-4.1%) less likely to be in operation after 2010 compared with the trend in nonexpansion states. Conclusions: States that expanded Medicaid experienced increased, rather than reduced, ED closure rates from 2010 through 2013. The financial benefits of the Affordable Care Act may be poorly targeted to the hospitals most vulnerable to closure. © 2016 Elsevier Inc. All rights reserved.
1. Introduction 1.1. Background Low levels of reimbursement from the uninsured have long been claimed to be a cause of poor hospital finances and even hospital emergency department (ED) closure. The Patient Protection and Affordable Care Act (ACA) aimed in part to reduce such uncompensated care.
Findings from this project, although not these specific results, were presented at the American Economics Association Meeting; San Francisco, CA; January 5, 2015. The authors received no grant funding and have no conflicts of interest to declare. A.B.F. and V.E.P. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. A.B.F. and V.E.P. conceived the study; designed the analysis; supervised the conduct of the trial and data collection; managed the data, including quality control; provided statistical advice on study design; and analyzed the data. D.D.O. and V.E.P. verified hospital closures and provided substantial input into the inclusion/exclusion criteria and search strategy. A.B.F. drafted the manuscript, and all authors contributed substantially to its revision. A.B.F. takes responsibility for the manuscript as a whole. ⁎ Corresponding author at: Leonard Davis Institute of Health Economics, University of Pennsylvania, 3641 Locust Walk, Philadelphia, PA 19104. Tel.: +1 215 284 5196. E-mail address: [email protected] (A.B. Friedman).
Recent studies show improving finances among hospitals in states that elected to expand Medicaid through the ACA [1]. Similarly, hospital associations and the popular press [2] have pointed to isolated hospital closures as evidence that failure to expand Medicaid has increased ED closures. 1.2. Importance Hospital ED closures may reduce access to care, increase transportation costs for patients, and disrupt local economies. Emergency department closures have also been associated with increased morbidity and mortality among the population living nearby in some studies [3], but not in others [4]. Because most (96.9% in 2012, authors' calculations) of acute care and critical access hospitals have an ED, we assume that ED closures are driven by hospital closures. 1.3. Goals of this investigation A longstanding trend in ED closures was documented before implementation of the ACA (1998-2008) and showed that safety-net status and low-profit margins were associated with increased risk of ED closure [5]. However, no studies have evaluated hospital closures since the ACA shifted hospital expectations about future uninsurance and Medicaid volumes. We hypothesized that states that expanded
http://dx.doi.org/10.1016/j.ajem.2016.04.006 0735-6757/© 2016 Elsevier Inc. All rights reserved.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006
2
A.B. Friedman et al. / American Journal of Emergency Medicine xxx (2016) xxx–xxx
Table 1 Baseline characteristics of hospitals in expansion vs nonexpansion states
DSH ($) Tenure (y) Hospitals in ZIP (#) CAH (%) HHI
Nonexpansion
Expansion
t Test (P value)
1 989 546 (4 318 104.0) 7.6 (1.3) 1.2 (0.5)
2 908 721 (5 771 765.0) 7.7 (1.1) 1.2 (0.4)
b.01 b.01 .85
25.4 (43.6) 2210 (1429)
24.8 (43.2) 2247 (1422)
2.2. Inclusion and exclusion criteria
.62 .53
The table presents mean (SD) characteristics of hospitals in 2006, the first year of the study. DSHs are the mean disproportionate share payments per hospital, in dollars. Tenure is the mean number of years hospitals were open. HHI is the Herfindahl-Hirschman Index, a measure of market concentration (10 000 is a monopoly, 0 is perfect competition), by bed share of Hospital Referral Region. CAH is the percent of hospitals with a critical access hospital designation.
Medicaid, either directly or through a state-specific waiver, experienced a trend toward a greater number of nonspecialty hospitals with EDs compared with states that had not yet expanded Medicaid by 2013. The hypothesis was formulated in advance of beginning data collection. Manually verified panel data from the Healthcare Cost Reporting Information System (HCRIS) provided a longitudinal census of open hospitals and were not subject to the nonresponse bias and imputation of the more commonly used American Hospital Association Annual Survey. 2. Methods 2.1. Data source for hospital status Data from the Centers for Medicare and Medicaid Services (CMS) HCRIS provides a national census of hospital financial reports from 2006 to 2013. Provider identifiers in this data set uniquely identify facilities, even through mergers and acquisitions. We supplemented HCIRS with 684 hospital-years meeting inclusion criteria from the Hospital Compare database and with 216 hospital-years of data from the CMS “late report,” consisting of hospitals that notified CMS that they remained open despite delays in filing a financial report. We identify potential openings/closures through the appearance/ disappearance of each provider identifier. We confirmed each closure by searching news databases (Google News, Lexis Nexis) for definitive local news reports. Closures that remained unconfirmed by this process were then confirmed through a direct phone call.
A
We restricted the sample to acute care, critical access, and federal hospitals with a 24-hour ED. We excluded all Indian Health Service hospitals (292 hospitalyears). We excluded specialty hospitals by searching their hospital names for common terms (eg, “ortho,” “surgical”) from HCRIS and the Indirect Medical Education to identify 242 specialty hospitals. We then confirmed these specialty hospitals by visiting Web sites, calling the facilities, and checking the National Plan and Provider Enumeration System. 2.3. Additional covariates Additional information on whether a hospital qualifies for the Critical Access program and on the number of staffed beds was obtained through the Hospital Compare and IME and Graduate Medical Education systems by merging on the same unique provider identifier. 2.4. Statistical models t Tests were used to compare individual characteristics across expansion vs nonexpansion states (Table 1). To assess the probability of hospitals remaining open, we used a difference-in-differences specification, implemented via linear probability model regression with year fixed effects. To account for the fact that the models regress hospital outcomes on state-level measures of Medicaid adoption, we cluster standard errors by state. Because hospital closure decisions likely take into account expectations regarding financial challenges, we used 2010, the year in which the ACA passed, as the beginning of the treatment period. Sensitivity analyses used a logit specification and treatment period beginning in 2012. 2.5. Ethical approval This study did not assess human subjects and was therefore exempt from institutional review board review.
B 3200
5400
C
5100
4800
4500
Percent of Open Hospital EDs (Compared to 2006)
105
Number of Open Hospital EDs
Number of Open Hospital EDs
Of 444 suspected closures, 134 hospitals were not acute care or critical access hospitals. Of the remaining 310 potential closures, 194 hospitals were open and had never closed. Of the 116 hospitals that closed, 5 reopened more than 6 months later and 4 moved to a different locations. For 11 of the closed hospitals, we were unable to locate additional data on when their closure occurred.
2800
2400
2000
104
103 ACA Status Non−Expansion Expansion
102
101
100 1600
4200 2006
2008
2010 Year
2012
2006
2008
2010 Year
2012
2006
2008
2010
2012
Year
Figure. Number of open hospital EDs over time. Trends in number of hospitals open over time (A), broken down by state Medicaid expansion status (B), and percentage of number of hospitals open over time (C) relative to 2006 baseline, by state Medicaid expansion status.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006
A.B. Friedman et al. / American Journal of Emergency Medicine xxx (2016) xxx–xxx Table 2 Differences-in-differences regression results
Post-2010 Expansion Post × expansion
Coefficients (95% CI)
P
0.0419⁎⁎ (0.0232 to 0.0607) 0.0223 (−0.00083 to 0.045) −0.0219⁎ (−0.041 to −0.0030)
b.01 .061 .026
Linear probability model difference-in-difference regression results (n = 40 275 hospitalyears). The dependent variable is a binary indicator of whether a hospital was open in a given year. Standard errors are clustered at the state level. Model includes year fixed effects. ⁎ Significance level: P b .05. ⁎⁎ Significance level: P b .01.
3. Results 3.1. National trends There were, on average, 3521 acute care hospitals and 1289 critical care access hospitals in operation each year, for a total of 38 480 hospital-year observations between 2006 and 2013. More hospitals were open at the end of the study period than at the beginning. In the 4 years from 2006 through 2009, there were an average of 4765 hospital EDs open per year, compared with 4855 in the 4 years following (1.9% increase in number of open EDs). Figure, Panel A demonstrates that there were no years in which the number of hospital EDs declined nationally during the study period. 3.2. Comparison across Medicaid expansion status Table 1 provides characteristics of hospitals in expansion and nonexpansion states. Hospitals in expansion states were substantively similar to those in nonexpansion states in terms of number of competitor hospitals and probability of being a critical access hospital. Disproportionate Share Hospital (DSH) payments per hospital and average number of open years were lower in nonexpansion states. In 2006, there were 2006 hospitals in states that would eventually expand Medicaid, compared with 2718 in nonexpansion states. By 2013, there were 2040 in expansion states, compared with 2854 in nonexpansion states. In expansion states, the number of hospitals dipped briefly from 2027 in 2009 to 2019 in 2010, but recovered to 2031 by 2012. Panels B and C in Figure demonstrate a similar growth rate in expansion states and nonexpansion states before 2010, followed by a break in the trend that occurred only in nonexpansion states. Results of the differences-in-differences regression are shown in Table 2. Hospitals in expansion states were 2.2% (95% confidence interval [CI], 0.3%-4.1%) less likely to be in operation post-2010, controlling for both the trend in nonexpansion states to be 4.2% more likely to be in operation after 2010 (95% CI, 2.3%-6.1%) and for a 2.2% higher baseline rate of the remaining in operation in expansion states (95% CI, −0.083% to 4.5%). Results were robust to sensitivity analyses involving model specification (logit) and time period (2012 start date). 4. Discussion A number of news reports document hospital closures, particularly in states that have not expanded Medicaid eligibility. Whether these closures were occurring at an accelerated rate after passage of the ACA or in states that did not expand Medicaid remained unknown. In this article, we provide an overview of how hospital closures in recent years have differed from pre-ACA trends, both overall and by Medicaid expansion status. Hsia et al [5] documented a long-term trend in nonrural hospital ED closures from 1998 to 2008. Our analysis from 2006 to 2013 demonstrates that this trend seems to have reversed itself after passage of
3
the ACA. Direct comparison between studies is impossible, however, because our census of hospital EDs uses a different hospital database, includes rural and critical access hospitals, excludes specialty care and Indian Health Service hospitals, and manually verifies each closure. In addition to finding that nationally the number of hospital EDs is growing, we find that states that expanded Medicaid by 2013 experienced a lower rate of hospital growth (fewer hospitals in operation relative to the trend) than states that did not. This finding is contrary to our initial hypothesis and may be explained by complex interactions of policy, billing, and hospital finances. The first possibility is that nonexpansion states are systematically different from expansion states in ways that make them less likely to experience hospital closure, such as more “business-friendly” environments, more for-profit hospitals, or fewer certificate-of-need regulations. For-profit hospitals close more rapidly when patient mix becomes unprofitable, but they also open more rapidly when it becomes profitable [6]. The second possibility is that hospitals most at risk for closure may not be those experiencing improved financial outcomes. For instance, these hospitals might serve a population largely ineligible for Medicaid expansion, such as US citizens above 138% of the poverty line or undocumented immigrants. A related explanation is that hospitals at risk for closure might have more patients formerly on private insurance switch to Medicaid, with resultant lower reimbursement [7]. The third possibility is that Medicaid expansion is merely one piece of the post-ACA health insurance scheme, and the effect of expansion on reducing ED closure may be positive but dominated by other factors. Most notably, previously eligible individuals signed up for traditional Medicaid at greater rates due to publicity surrounding the ACA [7], and the ACA-established health insurance exchanges were created in all states. Because nonexpansion states generally had lower baseline insurance rates, these changes may have offset the Medicaid expansion. The apparent reversal of the overall trend from ever fewer EDs documented in earlier work to ever more EDs documented in ours may result from these effects. Finally, given that only 4 years of postpassage data and 2 years of postimplementation data were available, these effects may simply be short-term adjustments. However, considering that hospital closure is typically permanent, even transient effects matter for access to emergency care. These findings support claims that DSH payments should be more carefully targeted to vulnerable hospitals. Acknowledgments The authors wish to acknowledge helpful comments by Kosali Simon, PhD, and Seth Freedman, PhD, of IU SPEA, and Karin V. Rhodes, MD, of NorthShore-LIJ. The authors gratefully acknowledge Terry Huynh, MBA, of CMS for providing the late reports and information about the data collection process. The authors thank Felipe Lozano-Rojas, BA, of IU SPEA for research assistance. References [1] Garthwaite C, Graves J, Gross T, Notowidigdo M. The effect of the ACA on access to health care services; 2015[Working Paper]. [2] Olorunnipa T. Obamacare cutbacks shut hospitals where Medicaid went unexpanded. Bloomberg 2013 http://www.bloomberg.com/news/articles/2013-11-25/obamacarecutbacks-shut-hospitals-where-medicaid-went-unexpanded. Accessed December 14, 2015. [3] Liu C, Srebotnjak T, Hsia RY. California emergency department closures are associated with increased inpatient mortality at nearby hospitals. Health Aff 2014;33:1323–9. [4] Joynt KE, Chatterjee P, Orav EJ, Jha AK. Hospital closures had no measurable impact on local hospitalization rates or mortality rates, 2003–11. Health Aff 2015;34:765–72. [5] Hsia RY, Kellerman A, Yu-Chu S. Factors associated with closures of emergency departments in the United States. JAMA 2011;305:1978–85. [6] Norton EC, Staiger DO. How hospital ownership affects access to care for the uninsured. Rand J Econ 1994;25:171–85. [7] Sommers BD, Kenney GM, Epstein AM. New evidence on the Affordable Care Act: coverage impacts of early Medicaid expansions. Health Aff 2014;33:78–87.
Please cite this article as: Friedman AB, et al, Trends in hospital ED closures nationwide and across Medicaid expansion, 2006-2013, Am J Emerg Med (2016), http://dx.doi.org/10.1016/j.ajem.2016.04.006