HEALTH POLICY/ORIGINAL RESEARCH
Effect of the Affordable Care Act Medicaid Expansion on Emergency Department Visits: Evidence From State-Level Emergency Department Databases Sayeh Nikpay, PHD, MPH*; Seth Freedman, PHD; Helen Levy, PHD; Tom Buchmueller, PHD *Corresponding Author. E-mail:
[email protected], Twitter: @saynikpay.
Study objective: We assess whether the expansion of Medicaid under the Patient Protection and Affordable Care Act (ACA) results in changes in emergency department (ED) visits or ED payer mix. We also test whether the size of the change in ED visits depends on the change in the size of the Medicaid population. Methods: Using all-capture, longitudinal, state data from the Agency for Healthcare Research and Quality’s Fast Stats program, we implemented a difference-in-difference analysis, which compared changes in ED visits per capita and the share of ED visits by payer (Medicaid, uninsured, and private insurance) in 14 states that did and 11 states that did not expand Medicaid in 2014. Analyses controlled for state-level demographic and economic characteristics. Results: We found that total ED use per 1,000 population increased by 2.5 visits more in Medicaid expansion states than in nonexpansion states after 2014 (95% confidence interval [CI] 1.1 to 3.9). Among the visit types that could be measured, increases in ED visits were largest for injury-related visits and for states with the largest changes in Medicaid enrollment. Compared with nonexpansion states, in expansion states the share of ED visits covered by Medicaid increased 8.8 percentage points (95% CI 5.0 to 12.6), whereas the uninsured share decreased by 5.3 percentage points (95% CI –1.7 to –8.9). Conclusion: The ACA’s Medicaid expansion has resulted in changes in payer mix. Contrary to other studies of the ACA’s effect on ED visits, our study found that the expansion also increased use of the ED, consistent with polls of emergency physicians. [Ann Emerg Med. 2017;-:1-11.] Please see page XX for the Editor’s Capsule Summary of this article. 0196-0644/$-see front matter Copyright © 2017 by the American College of Emergency Physicians. http://dx.doi.org/10.1016/j.annemergmed.2017.03.023
SEE EDITORIAL, P. XXX. INTRODUCTION Background and Importance Since its passage in 2010, the Patient Protection and Affordable Care Act (ACA) has reduced the number of uninsured from 49 to 29 million1 through newly created private health insurance markets and an expansion of the Medicaid program. Changes in coverage have coincided with improvements in self-reported measures of access to care and health status, and reduced out-of-pocket medical expenditures.2-5 These changes in coverage have generally had positive effects on hospitals. In states that expanded Medicaid, hospitals have experienced shifts in payer mix from uninsured to Medicaid,6,7 leading to reductions in hospital uncompensated care8,9 and improved financial positions.10 In addition to changing payer mix, many experts predicted that the ACA would increase use of medical care, potentially straining the existing supply of health care Volume
-,
no.
-
:
-
2017
providers,11 including emergency departments (EDs).12 However, the conclusions of early evidence on the ACA’s Medicaid expansion on use of medical care have been mixed. Evidence from hospital discharge data suggests that inpatient discharges have not increased more in states that have expanded Medicaid over those that did not, but evidence from population-based surveys suggests that inpatient visits increased (S. Nikpay et al, unpublished data, 2016).7 However, early evidence from populationbased surveys5 and hospital discharge data from selected hospitals13,14 and states7,15 suggests that ED visits have not changed. This finding is surprising because some expected ED visits to increase substantially among patients newly eligible for Medicaid.16 In fact, a recent poll by the American College of Emergency Physicians found that 75% of emergency physicians reported that they experienced an increase in patient volume after 2014, and 56% reported they experienced an increase in Medicaid volume specifically.17 Annals of Emergency Medicine 1
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Editor’s Capsule Summary
What is already known on this topic Relatively little is known about the effect of the expansion of Medicaid under the Patient Protection and Affordable Care Act (ACA) on emergency department (ED) utilization. What question this study addressed Changes in ED visit rates and ED payer mix (Medicaid, private insurance, and uninsured) were associated with the introduction of Medicaid expansion under the ACA. What this study adds to our knowledge In 14 expansion states and 11 nonexpansion states, ED visit rates per capita increased more in Medicaid expansion states than in nonexpansion states after 2014. Compared with nonexpansion states, in expansion states the proportion of ED visits covered by Medicaid increased, whereas the uninsured share decreased. How this is relevant to clinical practice The results of this study are not expected to change clinical practice but inform the discussion of ED utilization under various insurance plans.
Ex ante, it is not clear how insurance expansions, public or private, should affect total ED visits. One view holds that giving previously uninsured individuals access to primary care will reduce their use of the ED by shifting care to other sites. On the other hand, gaining insurance may simply increase the use of all types of care.18 Research on previous health insurance expansions has produced various results. Studies on the Massachusetts health reform, which closely resembles the ACA in that it included both a Medicaid and private insurance expansion, suggest that the coverage expansion led to a reduction in ED visits by improving access to outpatient care.19-21 In contrast, the Oregon Health Insurance Experiment, which evaluated the effect of expanding Medicaid to low-income adults, found a positive and sustained effect of Medicaid coverage on ED visits.22,23 Results from individual states’ Medicaid expansions or contractions show a positive relationship between expansion and ED use as well.24,25 Provisions of the ACA that went into effect before 2014 also yield different conclusions about the effect of coverage expansion on ED visits. The ACA-mandated expansion of private employer-sponsored dependent coverage to children aged 26 years or younger was associated with a small decrease in 2 Annals of Emergency Medicine
ED visits.26-28 In contrast, a study of California’s early Medicaid expansion in 2011 found that ED visits increased, although the effect was temporary.29 Summarizing the previous literature, Medford-Davis et al16 suggested that the effect on ED visits depends on the type of coverage the patient has. They expected that patients who gain insurance through Medicaid, as opposed to marketplace coverage, should increase use of the ED because there is little cost sharing associated with ED use in Medicaid. However, patients who gain marketplace coverage should use the ED less than before because many marketplace plans have large deductibles and cost-sharing requirements. Although the ACA’s Medicaid expansion creates one income eligibility level in all states expanding Medicaid, the effect of this change on Medicaid enrollment depends on the state’s preexisting Medicaid income eligibility criteria. The policy led to a much larger increase in insurance coverage in states that had very low income eligibility limits before 2014 compared with those with already generous eligibility. Larger increases in coverage should translate to larger effects on ED visits. Therefore, although a simple comparison of states that did and did not implement the ACA Medicaid expansion is of interest, to properly understand the effect of the Medicaid expansion on ED visits, it is important to account for heterogeneity among expansion states in terms of coverage changes. Goals of This Investigation The goal of this study was to assess whether changes in Medicaid eligibility after 2014 were associated with changes in ED use and payer mix. To this end, we compared the change in total ED visits and the share of ED visits by payer types directly affected by the ACA (Medicaid, uninsured, and private) between states that did and did not expand Medicaid in 2014. Because we hypothesized that the size of the change in total visits and payer share should depend on the size of the population gaining coverage, we accounted for heterogeneity by separately analyzing the changes in visits for states with large and small increases in Medicaid enrollment between 2013 and 2014. MATERIALS AND METHODS Study Design The study used a difference-in-differences design to compare the change between the pre- and postexpansion periods between states that did and did not expand Medicaid in 2014. Our analysis sample included 14 expansion states (Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
New Jersey, Nevada, New York, Rhode Island, and Vermont) and 11 nonexpansion states (Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin). All expansion states in our sample expanded Medicaid on January 1, 2014, except for California, Minnesota, and New Jersey. These 3 states conducted partial expansions of the Medicaid program between 2010 and 2014. We included them in our analyses; however, when we excluded them the results did not change. This study design assumes that preexisting trends in ED use were similar in expansion and nonexpansion states, allowing nonexpansion states to serve as a control group for what would have occurred in expansion states. Difference-in-differences assumes no other concurrent policy changes that differentially affected the treatment versus the control states. We, and the numerous studies that use this design,2,4,13,30-33 have no reason to believe that such concurrent changes are occurring. However, to assess the validity of this assumption, we tested whether pretrends differed between expansion and nonexpansion states. A general concern with a difference-in-differences research design is that the estimates may be picking up the effect of other shocks that differentially affected the demand for or supply of emergency care in expansion and nonexpansion states. One way to test for this type of omitted variable bias is to estimate the same models on a population that was not directly affected by the Medicaid expansion but which should be subject to other factors affecting emergency care in expansion states. Elderly patients covered by Medicare represent an ideal group for conducting a “falsification test.” Visits from all 4 quarters of 2012 and the first 3 quarters of 2013 constituted our pre-expansion period, and visits from the first through fourth quarters of 2014 constituted our postexpansion period. We dropped the fourth quarter of 2013 from the analysis. Although Medicaid expansions went into effect January 1, 2014, Medicaid coverage began to increase in the final quarter of 2013 for most states. Because the open enrollment period for the ACA’s private health insurance options began in the third quarter of 2013, patients who discovered they were already eligible for Medicaid under pre-ACA rules signed up for Medicaid.34 Because we hypothesized that ED visits in states with larger proportional changes in Medicaid coverage should have been more affected than in states with smaller changes, we also stratified our analysis by the proportional change in the statewide Medicaid coverage between 2013 and 2014. Specifically, we divided expansion states into terciles based on the change in total Medicaid enrollment between 2013 and 2014 and estimated separate difference-in-differences regressions that compared changes Volume
-,
no.
-
:
-
2017
in the lowest tercile to nonexpansion states and changes in the highest tercile to nonexpansion states. Data Collection and Processing Our ED visit data came from the Fast Stats database, an early-release, aggregated version of the State Emergency Department Databases and State Inpatient Databases.* These databases are all-capture, longitudinal, state databases collected by the state and compiled by the Agency for Healthcare Research and Quality (AHRQ) and are typically available at the patient level, with a several-year lag. The Fast Stats data are therefore a valuable tool for timely policy evaluation. The aggregated Fast Stats data include ED visits that did and did not end in an admission to the hospital. All community, nonrehabilitation hospitals are included. The Fast Stats data were adjusted by AHRQ for missing data, using information from several other data sets, including American Hospital Association data, the Trauma Information Exchange Program database, and the American Trauma Society.35 This study involved publicly available, state-level, and thus deidentified data and was therefore granted exempt status by the Vanderbilt University Medical Center Institutional Review Board. The Fast Stats data also provide visit counts by expected payer and condition. Expected payer included Medicare (for patients 65 years), Medicaid (for patients 19 to 64 years), private insurance (for patients 19 to 64 years), and uninsured (for patients aged 19 to 64 years). Uninsured is defined as self-pay, charity, or other state and local indigent programs that do not constitute health insurance. Condition categories were defined by the primary listed clinical classification code and included abdominal pain, back or neck pain, headache, mental health or substance use, skin infections, dental, and injury-related visits defined by primary listed International Classification of Diseases, Ninth Revision (ICD-9) codes.35 See Table E1 (available online at http://www.annemergmed.com) for a list of all clinical classification codes and ICD-9 codes used to create ED visit categories. These categories accounted for 33% of all ED visits in 2013 and therefore are not exhaustive of all types of ED visits. Our measures of total visits include these categories and all other ED visits as well. Data on the Medicaid expansion status of states and the change in Medicaid enrollment between 2013 and 2014 were obtained from the Kaiser Family Foundation’s State Health Facts Web site.36,37 *The Fast Stats data are available on the AHRQ Web site: https://www. hcupus.ahrq.gov/faststats/statepayer/statesED.jsp. Data can be downloaded by selecting any state, selecting “Show Data Export Options,” and selecting “Excel Export.”
Annals of Emergency Medicine 3
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Selection of Participants Not every state reports data to the Fast Stats program, and among those that report, data are not available from all years. Our analysis used quarterly data from all 25 states for which pre- and post-2014 data were available.† These states represent 68% of the US adult, non-elderly population.38 The final analysis sample consisted of 275 state-quarter-year observations, with an average of roughly 50.9 ED visits per 1,000 population per quarter-year in the pre-period. Methods of Measurement Our analysis focused on 2 ED visit measures: total visits per capita (Figure E1 and E2 [available online at http:// www.annemergmed.com]) and payer shares. For the first measure, we divided total ED visits by estimates of annual state population (in thousands) published by the Bureau of Economic Analysis.39 For the second measure, we divided payer-specific ED visits (Medicaid, private insurance, and uninsured) by total non-Medicare ED visits. We chose to exclude Medicare ED visits from the total and payer-specific share of total variables because the ACA is not expected to and did not empirically affect the number of Medicare visits. As a falsification test of our analysis, we estimated the association of the ACA Medicaid expansion with Medicare visits. We constructed total visits per capita and payer shares for all visits and condition-specific visits for each of the conditions available. Primary Data Analysis We estimated unadjusted and adjusted difference-in-differences ordinary least squares regressions. Specifically, our unadjusted estimates are from linear multivariable regressions, with an indicator for observations from the postexpansion period (2014), an indicator for Medicaid expansion status, and their interaction. In adjusted regressions, we included controls for national trends in ED visits by including year-by-quarter fixed effects, allowing us to remove any linear or nonlinear patterns in ED usage that are common to both expansion and nonexpansion states. We controlled for characteristics of states that were constant over time, using state fixed effects. We also included state-by-calendar-quarter controls for time-varying state characteristics, such as the unemployment rate, and the income, sex, race, and educational distributions of the population from the monthly Current Population Survey, a survey of †
The State Emergency Department Databases and State Inpatient Databases contain information on 35 states, but only 27 allow AHRQ to include their data in Fast Stats. The sample included 25 states because 2 states, Maine and Utah, did not report any data for 2014.
4 Annals of Emergency Medicine
noninstitutionalized US residents. (See Appendix E1, available online at http://www.annemergmed.com, for more details on the regression specification.) The analysis was also performed separately for expansion states with the highest and lowest proportional change in Medicaid coverage between 2013 and 2014. A key assumption in difference-in-differences analysis is that trends in ED visits in states that did and did not expand Medicaid were evolving similarly before 2014. To test this assumption, we investigated whether the trends in ED visits for expansion and nonexpansion states were parallel before the fourth quarter of 2013. We found no evidence that the trends were not parallel. A statistical test of common trends for each dependent variable is presented in Table E4, available online at http://www.annemergmed. com. To account for correlations within states over time, we clustered our standard errors at the state level. Because our analysis included only 25 states, which is less than recommended for clustering, we used the wild-cluster bootstrap procedure to estimate state-clustered standard errors.40 To determine whether our results were attributable to temporary changes in a particular quarter, to states that expanded Medicaid under the ACA before 2014, or to differences in trends by region, we conducted several sensitivity tests, including omitting the first and second quarter of 2014, excluding states that expanded Medicaid under the ACA before 2014, and controlling for regional time trends. To determine the robustness of our results to the choice of preperiod, we re-estimated the results, including data from 2011. Finally, to determine whether our results were due to the particular set of states in our data set, we also re-estimated the results, limiting to states that overlapped with the sample from 2 recent articles. (Appendix E3 [available online at http://www. annemergmed.com]; results from these sensitivity analyses are presented in Table E5, available online at http://www. annemergmed.com.) All analyses were performed with Stata (version 14; StataCorp, College Station, TX). RESULTS Figure 1 presents total annual visits per 1,000 population in Medicaid expansion and nonexpansion states, adjusted for seasonality and weighted by state population in 2014. In the pre-expansion period, expansion states had approximately 10 fewer ED visits per 1,000 population each quarter than nonexpansion states. Total visits declined by approximately 2 visits per 1,000 population per quarter between the beginning of 2012 and Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
the end of 2013 in both groups and then diverged from what pre-expansion trends would have predicted in the first quarter of 2014. After expansion, visits per capita increased from 46 to 50 per 1,000, or 4 per 1,000 population, in expansion states. During the same period, visits in nonexpansion states initially decreased by 2 and then increased by 3 per 1,000 population, returning to slightly above the pre-expansion level by the end of 2014. Figure 2 and Figure E3 (available online at http://www. annemergmed.com) presents trends in payer mix in expansion and nonexpansion states. Consistent with other recent studies,7,15 we found that in expansion states the share of Medicaid-covered visits increased sharply, from 35% in 2013 to approximately 48% by the end of 2014. In contrast, the Medicaid share of visits in nonexpansion states increased only slightly in 2014. The pattern for uninsured discharges was approximately the mirror image: uninsured discharges decreased in expansion states by approximately 5 percentage points after 2013 and declined from 23% to 11% by the end of 2014. The uninsured share of visits also decreased in nonexpansion states, although less so, from approximately 35% to 31%. This pattern is consistent with the hypothesis that the change in total visits observed in Figure 1 was caused by an increase in Medicaid coverage.
Figure 1. ED visits per capita by 2014 Medicaid expansion status. Source: AHRQ Fast Stats ED data. The figure plots mean seasonality-adjusted total ED visits per 1,000 state population in each calendar quarter, with 95% CIs. Data are weighted by 2014 state population. The solid line represents the best-fit line in the pre-expansion period (2012 Q1 to 2013 Q3), and the dashed line represents the projection of the best-fit line into the postperiod (2013 Q4 to 2014 Q4). Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Volume
-,
no.
-
:
-
2017
The share of visits covered by private insurance remained constant for expansion states and increased by several percentage points for nonexpansion states. This result is consistent with the fact that the gains in insurance coverage in nonexpansion states were almost entirely in the form of private coverage. The visual findings of Figures 1 and 2 were confirmed by the results of the regression analysis in Table 1 and E2 (available online at http://www.annemergmed.com). For total visits, the unadjusted model yielded a relative increase of 2.35 visits per 1,000 population per quarter (95% confidence interval [CI] –0.58 to 4.12). Adding controls for state characteristics and national trends in ED visits over time had little effect on the estimated association (2.47; 95% CI 1.06 to 3.88). In terms of shares, the adjusted and unadjusted models led to similar results. The adjusted relative changes in the shares of discharges were 8.8 percentage points for Medicaid (unadjusted 0.088; 95% CI 0.050 to 0.126) and approximately 5.3 percentage points for the uninsured (adjusted –0.053; 95% CI –0.089 to –0.017), suggesting that most but not all of the new Medicaid patients would have otherwise been uninsured. The share of ED patients covered by private insurance decreased by approximately 4 percentage points in expansion states relative to nonexpansion states (adjusted –0.035; 95% CI –0.050 to –0.020). In the bottom row of Table 1, we report results from our falsification tests, in which the dependent variable is the number of ED visits per 1,000 population among adults aged 65 years and older. A finding that the implementation of the ACA Medicaid expansion led to an increase in ED visits among this population would argue against interpreting our main results as representing the effect of the policy. In fact, the regression results for the older population indicated no statistically significant change in ED visits in expansion states relative to nonexpansion states. We also tested for differential pretrends for all outcomes and failed to reject the null hypothesis that the outcomes we examine were trending in a similar fashion in expansion and nonexpansion states (see Table E4, available online at http://www.annemergmed.com, for results). The analysis presented thus far has been based on a coarse categorization of states, which obscures important heterogeneity among expansion states. We sharpened the analysis by taking into account heterogeneity across Medicaid expansion states. If the results in Figure 1 and Table 2 reflect a causal effect of the Medicaid expansion on ED utilization, we would expect to observe larger changes in states in which the ACA led to a greater increase in Medicaid enrollment. To test this hypothesis, we estimated Annals of Emergency Medicine 5
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Figure 2. ED payer mix by 2014 Medicaid expansion status. Source: AHRQ Fast Stats ED data. The figure plots the mean seasonality-adjusted share of all non-Medicare ED visits covered by Medicaid (A), with no source of coverage (B), and covered by private insurance (C), along with 95% CIs. Data are weighted by state population in 2014. The dashed vertical line represents the first quarter of the ACA’s Medicaid expansion, Q1. 2014 Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin.
the correlation between the percentage change from 2013 to 2014 in the population with Medicaid and the change in ED visits per capita. The results, reported in Figure 3, show that states with larger increases in Medicaid enrollment experienced the largest increases in ED visits. The states with the largest changes in Medicaid were expansion states, and those with smaller changes were generally nonexpansion states. Within the expansion states, those that had the least generous Medicaid income eligibility before the ACA had the largest increases in coverage. For example, Kentucky and Nevada, which had historically low Medicaid eligibility levels for adults, had the largest increases in ED visits. In contrast, expansion states such as New York and Hawaii, which already covered adults above the poverty level, experienced little change in ED visits. The change in these states was similar to the change in Medicaid enrollment in the nonexpansion states, which also experienced little change in enrollment and little change in ED visits. The estimated regression line was DED visits¼5.75DMedicaid Coverage–0.14, indicating that for each additional percentage point increase in Medicaid enrollment, total quarterly visits per 1,000 6 Annals of Emergency Medicine
population increased by approximately .06 visits. For example, this corresponded to an increase in ED visits by approximately 1 per 1,000 when considering a state such as Nebraska, which experienced no change in Medicaid coverage, to one such as Illinois, which experienced an increase of approximately 20 percentage points. Table 3 presents results of our regression analysis comparing states in the top (HI, IA, IL, NY, VT in Figure 4) or bottom (light blue states in Figure 4) tercile of Medicaid enrollment changes to nonexpansion states to formally test whether states with larger changes in Medicaid enrollment experienced larger changes in ED visits. Consistent with the hypothesis that the changes we observed in the full sample were driven by the ACA Medicaid expansion, all coefficient estimates had greater magnitude for states with larger Medicaid enrollment changes. Medicaid expansion increased visits by 3.4 (coefficient estimate 3.36; 95% CI 2.09 to 4.63) per 1,000 population per quarter in states with the largest Medicaid enrollment changes, but the effect was half as large (coefficient estimate 1.86; 95% CI 0.82 to 2.89) in states with the smallest Medicaid enrollment changes. Changes in shares followed the same Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Table 1. Difference-in-differences (DiD) regression analysis: total visits per capita and payer mix. Pre-Expansion Total visits per 1,000 population Expansion Nonexpansion Share Medicaid Expansion Nonexpansion Share uninsured Expansion Nonexpansion Share private Expansion Nonexpansion Medicare visits per 1,000 population Expansion Nonexpansion
46.74 57.09
Postexpansion
Difference
48.23 56.24
1.49 –0.85
Simple DiD and 95% CI 2.35
0.58 to 4.12
Adjusted DiD and 95% CI 2.47
1.06 to 3.88
0.316 0.266
0.421 0.276
0.105 0.010
0.095
0.057 to 0.134
0.088
0.050 to 0.126
0.257 0.366
0.167 0.334
–0.091 –0.032
–0.059
–0.092 to –0.026
–0.053
–0.089 to –0.017
0.427 0.368
0.413 0.390
–0.015 0.022
–0.037
–0.054 to –0.020
–0.035
–0.050 to –0.020
0.20
–0.33 to 0.74
0.19
–0.34 to 0.73
13.91 17.17
14.27 17.32
Source: AHRQ Fast Stats ED data. The table presents the results of the unadjusted and adjusted difference-in-differences analyses. Information on the adjusted specification may be found in Appendix E1 (available online at http://www.annemergmed.com). Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Total ED visits per capita are quarterly visits per 1,000 state population, and payer mix is the share of non-Medicare ED visits covered by Medicaid, with no source of coverage, and covered by private insurance. Results are weighted by using 2014 state population. Standard errors are heteroscedasticity robust and clustered at the state level. The results are also robust to using a wild-cluster bootstrap to estimate P values corrected for a small number of clusters (see Appendix E2, available online at http://www.annemergmed.com).
pattern in both sets of states, but were much more pronounced in states with the largest Medicaid enrollment changes. Figure 4 plots coefficients from the adjusted regression models for the association of Medicaid expansion on number of visits per capita and payer mix by type of ED visit. The change in number of visits is statistically different from zero for most visit types, except for dental visits. The relative increase in visits is largest for injuries; visits per 1,000 population increased by nearly half a visit per quarter in expansion states relative to nonexpansion states. All visit types experienced similar changes in payer mix: an increase in the Medicaid share, decreases in the uninsured share, and a small decrease in the private share. Dental, mental health and substance abuse, and skin-related visits experienced the largest increases in the Medicaid share. LIMITATIONS Our analysis has several important limitations. First, Fast Stats ED data do not contain information from all 50 states and Washington, DC. Therefore, our sample may not be representative of the effect of the ACA on ED visits in all states. However, the states in our sample constitute nearly 70% of the US population and cover all 4 regions of the United States. Second, because the Fast Stats data are aggregated to the state level, we could not investigate the effect of the Medicaid expansion on subgroups that should have been most affected by Medicaid expansion: low-income childless Volume
-,
no.
-
:
-
2017
adults. Similarly, we could not investigate the effect of the expansion on visits for specific types of conditions or severity levels within the broad categories contained in Fast Stats data. The categories we studied did not represent a complete set of conditions, and future analyses should estimate the effect of the ACA on the targeted group and specific groups of conditions once data become available. Third, our analysis could not fully separate the effect of the establishment of new health insurance marketplaces under the ACA, which happened in all states at the same time as Medicaid expansion. Some of the effect we identified in this article may have been due to exchanges if states that expanded Medicaid also experienced larger changes in coverage on marketplaces than nonexpansion states. Fourth, part of the difference in total visits between expansion and nonexpansion states was due to a decrease in nonexpansion states in the first quarter of 2014. However, we investigated whether the results were a result of the decrease and found that they were not; ED visits increased in expansion states, even with omission of the first and second quarters of 2014. Fifth, data were available only through the end of 2014. Therefore, the results of our analysis represent only the very earliest assessment of the Medicaid expansion on ED visits. DISCUSSION Some researchers predicted that the ACA’s coverage expansions would lead to a large short-run increase in ED Annals of Emergency Medicine 7
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits Table 2. Summary statistics for the analysis sample. Nonexpansion States Expansion States
Total
Characteristics
Mean
SD
Mean
SD
Mean SD
ED visits per 1,000 population Share uninsured Share Medicaid Share private Female patient High school or less Married Age, y 6–17 18–24 25–29 30–39 40–49 50–59 60–64 65–79 80 Household income, $ <5,000 5–24,999 25–74,999 75–100,000 >100,000 Unemployment rate No. state quarters
56.7
7.4
47.0
8.8
50.9 9.5
35.6 26.9 37.5 52.0 40.7
7.6 5.2 7.4 1.0 3.1
22.7 35.1 42.2 51.6 38.0
7.5 8.6 5.6 0.9 3.6
27.9 31.8 40.3 51.8 39.1
53.4
2.2
50.8
2.3
51.9 2.6
15.8 15.4 12.6 13.5 14.1 5.9 8.7 4.5 1.7
1.2 0.9 0.7 1.0 0.8 0.6 1.0 0.8 0.5
15.9 17.2 13.0 13.3 13.8 5.7 7.6 4.1 1.7
0.7 0.9 0.7 0.6 0.7 0.5 0.7 0.5 0.3
15.9 16.5 12.8 13.4 13.9 5.8 8.0 4.3 1.7
1.0 1.3 0.7 0.8 0.8 0.5 1.0 0.7 0.4
3.2 19.2 26.4 20.3 30.9 6.9
0.8 2.6 2.3 1.9 3.8 1.5
2.7 16.2 22.9 18.4 39.8 7.6
0.8 3.2 2.9 1.9 6.3 1.8
2.9 17.4 24.3 19.2 36.2 7.3
0.8 3.3 3.2 2.1 6.9 1.7
132
168
9.8 8.4 6.8 1.0 3.7
300
Source: Authors’ analysis of Fast Stats ED data and monthly current population survey data, weighted with 2014 state population. Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont. Nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin.
visits for Medicaid patients and a small reduction in ED visits for privately insured patients purchasing through state nongroup marketplaces.16 Consistent with these predictions, our results showed that Medicaid visits increased in 2014, particularly in states with the largest increases in Medicaid enrollment. We show that the large degree of variation in Medicaid eligibility criteria before 2014 matters for estimating changes in utilization of the ED after expansion. The states with the lowest pre-ACA income eligibility thresholds for childless adults had the largest changes in enrollment and also experienced the largest changes in total ED visits. Because low-income childless adults were the target population of the Medicaid expansion, changes in ED visits were presumably driven by changes in utilization among childless adults, although we were unable to explicitly test this hypothesis with the Fast Stats data. We found that the change in total visits was 8 Annals of Emergency Medicine
Figure 3. Relationship between changes in ED visits per capita and changes in Medicaid enrollment between 2013 and 2014. Source: AHRQ Fast Stats ED data. The y axis represents the change in total quarterly ED visits per 1,000 state population between 2013 and 2014, and the x axis represents the change in monthly Medicaid and Children’s Health Insurance Program enrollment between 2013 and 2014. Each state is labeled with its postal abbreviation, and the best-fit line (5.75x–0.14) is plotted across expansion and nonexpansion states.
twice as large in a state such as Kentucky, in which most childless adults were ineligible for Medicaid at any income level before 2014, than in states such as Hawaii, in which childless adults were already eligible for Medicaid above the poverty line.37 This finding suggests that Medicaid expansions had a larger effect on the health care system in places in which more people were expected to gain coverage. Although our conclusions are consistent with those of the Oregon Health Insurance Experiment, in regard to the effect of ED visits on total volume, they differ from those of other studies of the early effects of Medicaid under the ACA. We believe the difference arose because of differences in the composition of states in the sample. For example, the states used to estimate the effect of Medicaid expansion in the study by Hempstead and Cantor,7 who found no differential change in ED visits between Medicaid expansion and nonexpansion states, include only 1 state that should have experienced a large change in Medicaid enrollment as a result of not having childless adult coverage before the ACA (Kentucky). Our results may also differ from those of Pines et al15 because that study used data from 478 hospitals in select states, whereas our sample included all hospitals from 25 states. If the ACA’s Medicaid expansion increased total use of the ED, an important question is whether the existing capacity of providers will be sufficient to meet increased demand. Although the existing supply of primary care Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Table 3. Difference-in-differences regression analysis: total visits per capita and payer mix by terciles of Medicaid enrollment change. Largest Medicaid Enrollment Changes (95% CI) Total visits per 1,000 population Share Medicaid Share uninsured Share private
3.36 0.115 –0.096 –0.020
(2.09 to 4.63) (0.013 to 0.218) (–0.201 to 0.010) (–0.029 to –0.011)
Smallest Medicaid Enrollment Changes (95% CI) 1.86 0.050 –0.021 –0.029
(0.82 to 2.89) (0.020 to 0.079) (–0.068 to 0.026) (–0.050 to –0.008)
Source: AHRQ Fast Stats ED data. The table presents regression-adjusted difference-in-difference estimates and their 95% CIs. The first 2 columns compare states with the largest change in Medicaid enrollment (top tercile: Kentucky, Minnesota, Nevada, and Rhode Island) and nonexpansion states before and after the ACA, and the second 2 columns compare states with the smallest change in Medicaid enrollment (bottom tercile: Hawaii, Iowa, Illinois, New York, and Vermont) and nonexpansion states before and after the ACA. Nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Information on the adjusted regression specification may be found in Appendix E1, available online at http://www.annemergmed.com. Total ED visits per capita are quarterly visits per 1,000 state population, and payer mix is the share of non-Medicare ED visits covered by Medicaid, with no source of coverage, and covered by private insurance. Standard errors are heteroscedasticity robust and clustered at the state level. Regression results are weighted with 2014 state population. The results are also robust to using a wild-cluster bootstrap to estimate P values corrected for a small number of clusters (see Appendix E2, available online at http://www.annemergmed.com).
providers is expected to be sufficient to meet increased demand for primary care,41 less is known about ED capacity. Our estimate of 2.5 additional visits per 1,000 people per quarter, or 10 per year, in expansion states implies that 1.13 million ED visits in 2014 could be attributed to Medicaid expansion in these states, given their total population of 112.7 million. This number is equivalent to 4.7% of total 2014 visits in expansion states. Put another way, the Medicaid expansion increased visits by 0.59 visits per new enrollee, given our estimate of 1.13 million new ED visits and the fact that expansion states had 1.9 million new enrollees in 2014.37 Our results’ numbers are slightly larger than those from the Oregon Health Insurance experiment, which found that Medicaid coverage increased visits among patients newly enrolled in Medicaid by 0.41 visits.23 Although Oregon’s increase in ED visits appears to be permanent,22 previous research suggests that the increase in ED visits may be temporary because of pent-up demand.16,29 Future research should revisit how ED visits continued to change beyond the first year of implementation in 2014. If the change is permanent, future research must also consider the effect of greater ED volume associated with the ACA on quality of care and other potential outcomes. One possibility is that the Medicaid expansion could result in ED crowding. Existing research suggests that ED crowding increases hospital mortality, length of stay, and costs.42 A recent poll by the American College of Emergency Physicians found that, after implementation of the ACA’s major coverage provisions in 2014, two thirds of emergency physicians said that demands on their time were increasing, and 70% believed that their EDs were not adequately prepared for large changes in volume.17 However, if additional patients treated in the ED are not admitted, then the Medicaid expansion could not result in crowding. Our results contain several interesting findings by visit type. First, some of the largest changes in payer mix were Volume
-,
no.
-
:
-
2017
for dental visits. This change likely reflects that dental ED visits are most prevalent among low-income, nonelderly adults, the target population for the ACA’s Medicaid expansion.43 Private coverage for dental ED visits also changed the least. This result is consistent with the relatively low level of private dental coverage in the adult
Figure 4. Difference-in-differences estimates by ED visit type. Source: AHRQ Fast Stats ED data. The figure presents regression-adjusted difference-in-difference estimate and their 95% CI by ED visit type. Information on adjusted regression specification may be found in Appendix E2, available online at http://www.annemergmed.com. Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Total ED visits per capita are quarterly visits per 1,000 state population, and payer mix is the share of nonMedicare ED visits covered by Medicaid, with no source of coverage, and covered by private insurance. Standard errors are heteroscedasticity robust and clustered at the state level. Results are weighted by 2014 state population. The results are also robust to using a wild-cluster bootstrap to estimate P values corrected for a small number of clusters (see Appendix E2 and Table E3, available online at http://www. annemergmed.com). Annals of Emergency Medicine 9
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
population in general.43 Although many uninsured patients gained Medicaid coverage, there was no change in dental ED visits overall, likely because adult coverage under Medicaid is often minimal and caps the total amount of reimbursement for care.44 Still, gaining insurance for these patients could be beneficial because out-of-pocket dental costs were reported as one of the more unaffordable types of care among the target population for Medicaid expansion under the ACA.45 Mental health visits also experienced a large change in both the total number and the share covered by Medicaid. These disproportionately large changes may reflect greater prevalence of mental illness and substance abuse among lower-income populations46,47 and increasing prevalence of overdoses caused by legal and illegal opioid drugs.48 Work before the ACA enactment has come to differing conclusions about acuity of ED visits after expansion, depending on the setting. In Massachusetts, low-acuity visits decreased, but in Wisconsin they increased.21,24 Studying the first 2 quarters of data from 2014, others found that the complexity of patients increased.49 Unfortunately, the Fast Stats early-release data do not contain sufficient detail to estimate the effect of ACA Medicaid expansion on acuity of ED visits. The effect on acuity remains an open yet important question, and researchers must investigate this issue when discharge-level data become available. The shift from uninsured to Medicaid coverage after 2014 in states that expanded Medicaid leads to the question, what will happen to the financial circumstances of EDs? Medicaid typically pays a significantly lower fraction of charges than other payers,50 and some analyses suggest that the uninsured may pay more in self-pay amounts than Medicaid reimburses for select conditions.51 Furthermore, large cuts to Disproportionate Share Hospital payments are scheduled to occur in 2018, reducing Medicaid revenues to EDs.52 Simultaneously, other aspects of the ACA may positively affect EDs. For example, under the ACA, many uninsured patients will have gained private insurance, which reimburses more generously, and insurers are prohibited from denying charges if patients seek emergency care out of network. This should reduce ED bad debts and potentially help financial circumstances. In conclusion, we found that the ACA’s Medicaid expansion increased ED visits and shifted hospital payer mix away from uninsured patients toward Medicaid in 2014. We found that states in which the target population for Medicaid expansion was not already eligible experienced the largest increases in ED use, suggesting that the effect of the Medicaid expansion depends on Medicaid eligibility criteria in effect before the ACA. Future work must 10 Annals of Emergency Medicine
investigate the longer-term effects of Medicaid expansion and also whether expansion led to changes in acuity of visits. Supervising editor: David J. Magid, MD, MPH Author affiliations: From the Department of Health Policy, Vanderbilt University School of Medicine, Nashville, TN (Nikpay); School of Public and Environmental Affairs, Indiana University, Bloomington, IN (Freedman); Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI (Levy); and Ross School of Business, University of Michigan, Ann Arbor, MI, (Buchmueller). Author contributions: SN and SF prepared the analysis data, conceived the study design, analyzed the data, and drafted the article. HL and TB provided advice on the study design and reviewed the article. All authors contributed to article revisions. SN takes responsibility for the paper as a whole. All authors attest to meeting the four ICMJE.org authorship criteria: (1) Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; AND (2) Drafting the work or revising it critically for important intellectual content; AND (3) Final approval of the version to be published; AND (4) Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article as per ICMJE conflict of interest guidelines (see www.icmje.org). The authors have stated that no such relationships exist. Publication dates: Received for publication August 30, 2016. Revisions received November 30, 2016, and February 16, 2017. Accepted for publication March 10, 2017.
REFERENCES 1. Obama B. United States health care reform: progress to date and next steps. JAMA. 2016;316:525-532. 2. Sommers BD, Musco T, Finegold K, et al. Health reform and changes in health insurance coverage in 2014. N Engl J Med. 2014;371:867-874. 3. Sommers BD, Buchmueller T, Decker SL, et al. The Affordable Care Act has led to significant gains in health insurance and access to care for young adults. Health Aff (Millwood). 2013;32:165-174. 4. Sommers BD, Gunja MZ, Finegold K, et al. Changes in self-reported insurance coverage, access to care, and health under the Affordable Care Act. JAMA. 2015;314:366-374. 5. Wherry LR, Miller S. Early coverage, access, utilization, and health effects associated with the Affordable Care Act Medicaid expansions: a quasi-experimental study. Ann Intern Med. 2016;164:795-803. 6. Nikpay S, Buchmueller T, Levy HG. Affordable Care Act Medicaid expansion reduced uninsured hospital stays in 2014. Health Aff (Millwood). 2016;35:106-110. 7. Hempstead K, Cantor JC. State Medicaid expansion and changes in hospital volume according to payer. N Engl J Med. 2016;374:196-198. 8. Cunningham P, Rudowitz R, Young K, et al. Understanding Medicaid Hospital Payments and the Impact of Recent Policy Changes. 2016.
Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
9. Dranove D, Garthwaite C, Ody C. Uncompensated care decreased at hospitals in Medicaid expansion states but not at hospitals in nonexpansion states. Health Aff (Millwood). 2016;35:1471-1479. 10. Blavin F. Association between the 2014 Medicaid expansion and US hospital finances. JAMA. 2016;316:1475-1483. 11. Ku L, Jones K, Shin P, et al. The states’ next challenge—securing primary care for expanded Medicaid populations. N Engl J Med. 2011;364:493-495. 12. McClelland M, Asplin B, Epstein SK, et al. The Affordable Care Act and emergency care. Am J Public Health. 2014;104:e8-e10. 13. Chan T, Killeen J, Vilke G, et al. Impact of the Affordable Care Act on the health care coverage of patients seen in the emergency department: initial first quarter findings. Ann Emerg Med. 2014;64:S84-S85. 14. Venkatesh A, Cutter C. Medicaid expansion under the Affordable Care Act: how may it affect emergency department utilization and access? Ann Emerg Med. 2014;64:S3. 15. Pines JM, Zocchi M, Moghtaderi A, et al. Medicaid expansion in 2014 did not increase emergency department use but did change insurance payer mix. Health Aff (Millwood). 2016;35:1480-1486. 16. Medford-Davis LN, Eswaran V, Shah RM, et al. The Patient Protection and Affordable Care Act’s effect on emergency medicine: a synthesis of the data. Ann Emerg Med. 2015;66:496-506. 17. American College of Emergency Physicians. 2015 ACEP Poll: Affordable Care Act Research Results. March 2015. 18. Newhouse JP; Rand Corporation Insurance Experiment Group. Free for All? Lessons From the RAND Health Insurance Experiment. Harvard University Press; 1993. 19. Smulowitz PB, O’Malley J, Yang X, et al. Increased use of the emergency department after health care reform in Massachusetts. Ann Emerg Med. 2014;64:107-115.e103. 20. Miller S. The impact of the Massachusetts health care reform on health care use among children. Am Econ Rev. 2012;102:502-507. 21. Miller S. The effect of insurance on emergency room visits: an analysis of the 2006 Massachusetts health reform. J Public Econ. 2012;96:893-908. 22. Finkelstein AN, Taubman SL, Allen HL, et al. Effect of Medicaid coverage on ED use—further evidence from Oregon’s experiment. N Engl J Med. 2016;375:1505-1507. 23. Taubman SL, Allen HL, Wright BJ, et al. Medicaid increases emergencydepartment use: evidence from Oregon’s health insurance experiment. Science. 2014;343:263-268. 24. DeLeire T, Dague L, Leininger L, et al. Wisconsin experience indicates that expanding public insurance to low-income childless adults has health care impacts. Health Aff (Millwood). 2013;32:1037-1045. 25. Ghosh A, Simon K. The Effect of Medicaid on Adult Hospitalizations: Evidence From Tennessee’s Medicaid Contraction. Cambridge, MA: National Bureau of Economic Research; 2015. 26. Antwi YA, Moriya AS, Simon K, et al. Changes in emergency department use among young adults after the Patient Protection and Affordable Care Act’s dependent coverage provision. Ann Emerg Med. 2015;65:664-672.e662. 27. Hernandez-Boussard T, Morrison D, Goldstein BA, et al. Relationship of Affordable Care Act implementation to emergency department utilization among young adults. Ann Emerg Med. 2016;67:714-720.e711. 28. Hernandez-Boussard T, Burns CS, Wang NE, et al. The Affordable Care Act reduces emergency department use by young adults: evidence from three states. Health Aff (Millwood). 2014;33:1648-1654. 29. Lo N, Roby DH, Padilla J, et al. Increased service use following Medicaid expansion is mostly temporary: evidence from California’s Low Income Health Program. UCLA Center for Health Policy Research Health Policy Brief. 2014. 30. Ghosh A, Simon K, Sommers BD. The Effect of State Medicaid Expansions on Prescription Drug Use: Evidence From the Affordable
Volume
-,
no.
-
:
-
2017
31.
32.
33.
34.
35. 36. 37. 38. 39. 40. 41. 42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
Care Act. Cambridge, MA: National Bureau of Economic Research; 2017. Gooptu A, Moriya AS, Simon KI, et al. Medicaid expansion did not result in significant employment changes or job reductions in 2014. Health Aff (Millwood). 2016;35:111-118. Mulcahy A, Harris K, Finegold K, et al. Insurance coverage of emergency care for young adults under health reform. N Engl J Med. 2013;368:2105-2112. Simon K, Soni A, Cawley J. The Impact of Health Insurance on Preventive Care and Health Behaviors: Evidence From the 2014 ACA Medicaid Expansions. Cambridge, MA: National Bureau of Economic Research; 2016. Frean M, Gruber J, Sommers BD. Premium Subsidies, the Mandate, and Medicaid Expansion: Coverage Effects of the Affordable Care Act. Cambridge, MA: National Bureau of Economic Research; 2016. Agency for Healthcare Research and Quality. Healthcare Cost and Utilization Project Fast Stats Data. 2016. Kaiser Family Foundation. Status of State Action on the Medicaid Expansion Decision. Menlo Park, CA: Kaiser Family Foundation; 2015. Kaiser Family Foundation. Total Monthly Medicaid and CHIP Enrollment. Menlo Park, CA: Kaiser Family Foundation; 2016. Ruggles S, Genadek K, Goeken R, et al. Integrated Public Use Microdata Series: Version 6.0. 2015. Bureau of Economic Analysis. Regional Economic Accounts. Washington, DC: US Department of Commerce; 2016. Cameron AC, Miller DL. A practitioner’s guide to cluster-robust inference. J Human Resources. 2015;50:317-372. Glied S, Ma S. How Will the Affordable Care Act Affect the Use of Health Care Services? New York, NY: The Commonwealth Fund; 2015. Sun BC, Hsia RY, Weiss RE, et al. Effect of emergency department crowding on outcomes of admitted patients. Ann Emerg Med. 2013;61:605-611.e606. Healthcare Cost and Utilization Project. Statistical Brief #143: Emergency Department Visits for Dental-Related Conditions, 2009. Washington, DC: Agency for Healthcare Research and Quality; 2012. Hinton E, Paradise J. Access to Dental Care in Medicaid: Spotlight on Nonelderly Adults. Menlo Park, CA: The Henry J. Kaiser Family Foundation; 2016. Shartzer A, Kenney GM. QuickTake: The Forgotten Health Care Need: Gaps in Dental Care for Insured Adults Remain Under ACA. Washington, DC: Urban Institute Health Policy Center; 2015. Gallup. Strong relationship between income and mental health. Gallupcom. 2007. Available at: http://www.gallup.com/poll/102883/ strong-relationship-between-income-mental-health.aspx. Accessed June 15, 2016. Jones CM, Logan J, Gladden RM, et al. Vital signs: demographic and substance use trends among heroin users—United States, 2002-2013. MMWR Morb Mortal Wkly Rep. 2015;64:719-725. Compton WM, Jones CM, Baldwin GT. Relationship between nonmedical prescription-opioid use and heroin use. N Engl J Med. 2016;374:154-163. Colorado Hospital Association. Impact of Medicaid Expansion on Hospitals: Updated for Second-Quarter 2014. Greenwood Village, CO: Colorado Hospital Association; 2014. Galarraga JE, Pines JM. Anticipated changes in reimbursements for US outpatient emergency department encounters after health reform. Ann Emerg Med. 2014;63:412-417.e412. Melnick GA, Fonkych K. Hospital pricing and the uninsured: do the uninsured pay higher prices? Health Aff (Millwood). 2008;27:w116-w122. Medicaid and CHIP Payment and Access Commission (MACPAC). Report to Congress on Medicaid Disproportionate Share Hospital Payments. 2016.
Annals of Emergency Medicine 11
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits Table E1. ICD-9 and clinical classification codes that go into the specific categories of visits. Visit Type
ICD-9 Code (%)
Clinical Classification Code (%)
Abdominal pain Back or neck pain Dental Headache Injury Mental health/substance abuse Skin infection Total
251 (4.2) 205 (2.9) 520.0–523.9 (0.7) 84 (2.2) 660–661, 800–909.2, 909.4, 909.9 910–994.9, 995.5–995.59, 995.80–995.85, 980.0, 965.00–965.02, 965.09 (20.2) 650–659, 662, 663, 670 (0.1) 197 (2.5) (33)
35
Source : Fraction of total visits from HCUP-Net for 2013 in parentheses. Total ED visits in 2013 were 134,739,675.
Adjusted regression models included the following additional controls:
APPENDIX E1 Regression specification Unadjusted regressions models were performed with ordinary least squares with heteroscedasticity-robust, state-clustered standard errors, using the following regression equation: Ysq ¼ a0 þ a1POSTq þ a2EXPANSIONs þ a3POST EXPANSIONsq þ usq
Ysq ¼ b0 þ b1POST EXPANSIONsq
(1)
Y is either total ED visits per 1,000 state population or the share of non-Medicare visits covered by Medicaid, without a source of coverage, or private insurance. Post is an indicator variable set to 1 for visits occurring on or after the fourth quarter of 2013, with all other periods set to zero. Expansion is an indicator variable set to 1 for visits occurring in states that expanded Medicaid in 2014 (Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont). All other states in the data set (Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin) were set to zero. The interaction of these variables gives the difference in the change in visits over time between expansion and nonexpansion states, or the difference-in-difference effect.
11.e1 Annals of Emergency Medicine
(2)
þ fs þ gq þ Xsq y þ usq
The regression is the same as in equation (1), but adds state fixed effects (fs) and quarter-year fixed effects (gq). The former controls for time-invariant characteristics between states, and the latter flexibly controls for common changes in visits occurring in both expansion and nonexpansion states over time. We also include time-varying, state-level controls for unemployment, high school degree or less, income categories, fraction of male patients, marriage rates, and the fraction of the population in 9 categories of age (Xsq). Unemployment is the fraction of the labor force (employed peopleþunemployed people) who were unemployed in each state and each quarter. Income is a set of variables describing the fraction of the population that fell into the following income categories: <$5,000, $5,000 to $24,999; $25 to $74,999; $75 to $100,000; and >$100,000. The age categories are <5 years (omitted) and 6 to 17, 18 to 29, 30 to 39, 40 to 49, 50 to 59, 60 to 64, 65 to 74, 75 to 84, and 85 years or older. All time-varying controls were created with the monthly current population survey analyzed at the quarterly level.
Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Table E2. Full regression output. Total Visits per 1,000 Population Variables
Share Medicaid
Share Uninsured
Share Private
Coefficient
95% CI
Coefficient
95% CI
Coefficient
95% CI
Coefficient
95% CI
2.265 0.867 2.401 –0.890 –1.725 –0.548 0.994 –3.916 –1.761 0.818 –2.020 20.54 4.505 –1.433
0.752 to 3.778 0.003 to 1.731 1.448 to 3.354 –1.860 to 0.079 –2.620 to –0.831 –1.679 to 0.583 –0.269 to 2.256 –4.881 to –2.950 –3.629 to 0.106 –1.024 to 2.659 –3.659 to –0.382 –25.30 to 66.39 –23.50 to 32.51 –23.84 to 20.98
0.091 –0.006 –0.007 –0.005 0.006 0.001 0.002 –0.018 0.001 0.017 0.024 0.449 0.198 0.457
0.054 to 0.128 –0.015 to 0.003 –0.018 to 0.004 –0.014 to 0.004 –0.004 to 0.015 –0.009 to 0.010 –0.011 to 0.014 –0.039 to 0.003 –0.022 to 0.024 –0.003 to 0.037 0.003 to 0.045 –0.136 to 1.033 –0.156 to 0.552 0.059 to 0.856
–0.055 0.008 0.013 0.007 –0.003 0.005 0.009 0.000 –0.024 –0.036 –0.045 –0.313 –0.189 –0.447
–0.089 to –0.022 –0.000 to 0.015 0.002 to 0.023 –0.006 to 0.020 –0.015 to 0.008 –0.008 to 0.018 –0.007 to 0.025 –0.023 to 0.023 –0.046 to –0.002 –0.058 to –0.015 –0.069 to –0.020 –0.978 to 0.351 –0.534 to 0.155 –0.820 to –0.075
–0.036 –0.002 –0.006 –0.002 –0.002 –0.006 –0.011 0.018 0.023 0.020 0.021 –0.135 –0.009 –0.010
–0.052 to –0.020 –0.006 to 0.002 –0.010 to –0.001 –0.007 to 0.004 –0.008 to 0.004 –0.015 to 0.003 –0.019 to –0.003 0.007 to 0.028 0.007 to 0.038 0.006 to 0.033 0.007 to 0.034 –0.357 to 0.086 –0.093 to 0.076 –0.165 to 0.145
21.88 –43.65 –9.688 14.53 –1.114 –11.52 8.926 4.963 –25.76
–30.84 to 74.61 –146.1 to 58.76 –117.6 to 98.21 –122.2 to 151.2 –139.5 to 137.3 –124.7 to 101.7 –97.75 to 115.6 –109.1 to 119.0 –157.7 to 106.2
0.299 0.788 1.316 1.191 2.203 1.131 1.526 1.504 1.552
–0.320 to 0.917 –0.815 to 2.390 –0.124 to 2.757 –0.481 to 2.862 0.390 to 4.017 –0.357 to 2.619 –0.117 to 3.168 –0.306 to 3.315 –0.729 to 3.833
–0.330 –0.436 –0.901 –0.473 –1.403 –0.319 –0.936 –0.545 –0.695
–0.993 to 0.333 –1.692 to 0.820 –2.038 to 0.236 –1.606 to 0.659 –3.043 to 0.236 –1.484 to 0.846 –2.345 to 0.473 –2.001 to 0.910 –2.666 to 1.276
0.032 –0.351 –0.416 –0.717 –0.800 –0.812 –0.590 –0.959 –0.857
–0.302 to 0.365 –1.250 to 0.547 –1.262 to 0.431 –1.755 to 0.321 –1.847 to 0.247 –1.839 to 0.215 –1.574 to 0.394 –2.129 to 0.212 –2.189 to 0.475
–16.41 –14.20 –22.98 –21.83 –37.86 10.55 275 0.659
–64.80 to 31.99 –41.32 to 12.93 –43.33 to –2.633 –37.33 to –6.328 –81.40 to 5.676 –37.67 to 58.76
–0.049 0.514 0.236 0.242 –0.784 –0.202 275 0.802
–1.062 to 0.965 –0.010 to 1.038 –0.031 to 0.504 –0.041 to 0.526 –1.386 to –0.182 –1.299 to 0.895
–0.258 –0.600 –0.292 –0.328 0.698 0.454 275 0.777
–1.420 to 0.904 –1.100 to –0.101 –0.551 to –0.034 –0.590 to –0.066 0.199 to 1.196 –0.714 to 1.622
0.306 0.086 0.056 0.085 0.086 0.748 275 0.509
0.016 to 0.597 –0.114 to 0.287 –0.057 to 0.168 –0.040 to 0.211 –0.242 to 0.414 0.396 to 1.100
Expansionpost 2012q2 2012q3 2012q4 2013q1 2013q2 2013q3 2014q1 2014q2 2014q3 2014q4 Female patient High school or less Married Age categories, y 6–17 18–29 30–39 40–49 50–59 60–64 65–74 75–84 85 Income categories, $ <5,000 5–24,999 25–74,999 75–100,000 Unemployment rate Constant No. state quarters R2
This table presents the full regression output for our main results (Table 2).
APPENDIX E2 Because ED visits within a state were correlated over time, we had to cluster our analysis at the state level. However, our sample contains 25 states, which some researchers argue is too few clusters to use typical cluster-robust standard errors estimation techniques.40 To address this problem, we performed a t-percentile
Volume
-,
no.
-
:
-
2017
wild-cluster bootstrap procedure proposed by Cameron and Miller.40 We found that the interpretation of the results did not change. Adjusted P values for the results from Table 1, Figure 3, and Table 3 are presented in appendix Table E3.
Annals of Emergency Medicine 11.e2
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits Table E3. Corrected P values generated by the t-percentile wild-cluster bootstrap. Table 1 Total visits, all states ED visits per 1,000 population Medicaid share Uninsured share Private share Total visits, top tercile states ED visits per 1,000 population Medicaid share Uninsured share Private share Total visits, bottom tercile states ED visits per 1,000 population Medicaid share Uninsured share Private share
Table 3
0.008 0.004 0.032 0.010 0.008 0.018 0.192 0.034 0.020 0.018 0.503 0.018
The table presents adjusted P values from a wild-cluster bootstrap t-percentile procedure.
APPENDIX E3 Robustness tests To determine whether our results were attributable to temporary changes in a particular quarter, to states that expanded Medicaid under the ACA before 2014, or to differences in trends by region, we conducted several sensitivity tests and found that our results were robust to a number of sensitivity analyses, including omitting the first and second quarter of 2014, excluding states that expanded
Medicaid under the ACA before 2014, and controlling for regional time trends. To determine the robustness of our results to the choice of preperiod, we re-estimated the results, including data from 2011. Finally, to determine whether our results were due to the particular set of states in our data set, we also and re-estimated the results, limiting to states that overlapped with the sample from 2 recent articles.
Table E4. Test of differential pretrends. Main Sample, 2012 to 2013 Preperiod
Total visits per 1,000 population Medicaid share of visit Uninsured share of visits Private share of visits
Expansion3Trend
P Value
–0.17 0.001 <0.001 –0.001
0.32 0.346 0.783 0.148
Source: AHRQ Fast Stats ED data. The table presents the results of a regression of our dependent variables on an interaction between expansion status and a linear quarterly trend, controlling for state fixed effects, quarterly dummies, and control variables. Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and nonexpansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Total ED visits per capita are quarterly visits per 1,000 state population, and payer mix is the share of non-Medicare ED visits covered by Medicaid, with no source of coverage, and covered by private insurance. Results are weighted by using 2014 state population. Standard errors are heteroscedasticity robust and clustered at the state level. The results are also robust to using a wild-cluster bootstrap to estimate P values corrected for a small number of clusters.
11.e3 Annals of Emergency Medicine
Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Table E5. Robustness tests.
Total visits per 1,000 population Share Medicaid Share uninsured Share private Number of observations
Main Estimates
Exclude 2014 q1
95%CI
95%CI
Exclude 2014 q1 and 2014 q2
Region3Quarter FE
95%CI
95%CI
2.47
1.06 to 3.88
2.04
0.62 to 3.45
1.75
0.19 to 3.31
2.62
1.16 to 4.08
0.088 –0.053 –0.035 275
0.050 to 0.126 –0.089 to –0.017 –0.050 to –0.020
0.097 –0.056 –0.041 250
0.053 to 0.141 –0.098 to –0.014 –0.055 to –0.026
0.106 –0.063 –0.043 225
0.061 to 0.151 –0.104 to –0.021 –0.060 to –0.027
0.086 –0.052 –0.034 275
0.048 to 0.123 –0.088 to –0.016 –0.049 to –0.019
Drop early expanders
Overlap with Hempstead sample7
95%CI Total visits per 1,000 1.74 0.79 to 2.69 population Share Medicaid 0.078 0.033 to 0.122 Share uninsured –0.049 –0.102 to 0.005 Share private –0.029 –0.044 to –0.014 Number of observations 242
Overlap with Pines sample16
95%CI
Begin in 2011
95%CI
1.64
0.42 to 2.85
2.39
0.88 to 3.89
0.047 –0.028 –0.019 98
0.015 to 0.080 –0.074 to 0.018 –0.039 to –0.000
0.094 –0.058 –0.037 209
0.051 to 0.138 –0.099 to –0.017 –0.053 to –0.020
95%CI 1.74
0.23 to 3.24
0.089 0.051 to 0.127 –0.054 –0.090 to –0.018 –0.035 –0.051 to –0.019 375
Figure E1. Medicare ED visits per capita by 2014 Medicaid expansion status. Source: AHRQ Fast Stats ED data. The figure plots mean seasonality-adjusted total Medicare ED visits per 1,000 state population in each calendar quarter, with 95% CIs. Data are weighted by 2014 state population. The solid line represents the best-fit line in the pre-expansion period (2012 Q1 to 2013 Q3) and the dashed line represents the projection of the best-fit line into the postperiod (2013 Q4 to 2014 Q4). Expansion states include Arizona, California, Hawaii, Iowa, Illinois, Kentucky, Maryland, Minnesota, North Dakota, New Jersey, Nevada, New York, Rhode Island, and Vermont, and non-expansion states include Florida, Georgia, Indiana, Kansas, Missouri, North Carolina, Nebraska, South Carolina, South Dakota, Tennessee, and Wisconsin. Volume
-,
no.
-
:
-
2017
Annals of Emergency Medicine 11.e4
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Figure E2. ED visits per capita by state.
11.e5 Annals of Emergency Medicine
Volume
-,
no.
-
:
-
2017
Nikpay et al
Effect of the ACA Medicaid Expansion on Emergency Department Visits
Figure E3. ED payer mix by state.
Volume
-,
no.
-
:
-
2017
Annals of Emergency Medicine 11.e6