Association Between State Medicaid Expansion and Emergency Access to Acute Care Hospitals in the United States

Key Points Question Are states that chose not to expand Medicaid under the Patient Protection and Affordable Care Act associated with reduced emergency access to acute care hospitals? Findings In this cross-sectional study of acute care hospital availability in all 50 US states and the District of Columbia, states that did not expand Medicaid experienced worsened emergency access to acute care hospitals compared with states that expanded Medicaid. Meaning This study found reduced emergency access to acute care hospitals in states that did not expand Medicare, which could negatively impact the quality of care for time-sensitive conditions such as acute myocardial infarction, stroke, sepsis, and trauma.


Introduction
Timely access to acute care services can be lifesaving in medical emergencies such as acute myocardial infarction, 1-3 stroke, 4-6 sepsis, [7][8][9] and trauma. [10][11][12][13] For decades, the primary approach to maintain or improve access to acute care hospitals for patients with medical emergencies has been to implement organizational approaches such as regionalization of acute care. [14][15][16] In contrast, relatively little attention has been paid to the role of insurance reform, sources of hospital revenue, and the financial sustainability of hospitals providing emergency services. Signed into law in 2010, the Patient Protection and Affordable Care Act (ACA) included provisions for states to receive enhanced matching federal funds to expand eligibility for Medicaid up to 138% of the federal poverty level for adults. As of December 2017, 19 states had not expanded coverage to those newly eligible under the ACA. Evidence suggests that the decision not to adopt Medicaid expansion has contributed to hospital closures in those states. 17 However, the extent to which hospital closures have affected access to care is not known. As hospital closures could occur in areas with duplication of services or in areas with declining populations, fewer hospitals does not necessarily translate to decreased population access. At the same time, closures of safety-net hospitals specifically may constitute a practical loss of access for some patients, even if other nearby hospitals remain open, as underinsured persons may be dissuaded from accessing services because of the potential for high out-of-pocket expenses.
To address this knowledge gap, we evaluated the association of Medicaid expansion under the ACA with changes in emergency access to acute care hospitals in the overall and low-income US population. We examined access both to short-term acute care hospitals overall and to safety-net hospitals, as safety-net hospitals are potentially more sensitive to changes in uncompensated care. [18][19][20][21] We evaluated both overall and low-income population access, as the low-income population was specifically targeted for coverage expansions under the proposed Medicaid eligibility changes.

Methods
Our analyses involved 3 linked steps: (1) identifying and geolocating all short-term acute care hospitals in the United States; (2) estimating populations without emergency access to acute care hospitals, which we defined as living outside a 30-minute driving distance of any hospital, and (3) using a difference-in-differences approach to compare changes in population access to acute care hospitals in states that expanded Medicaid with those that did not. The study included all 50 US states and the District of Columbia. As all analyses used aggregated population and hospital-level data, the project did not meet criteria for human subjects research and informed consent requirements according to the University of Pittsburgh Human Research Protection Office. We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.
Bureau's American Community Survey 5-year samples to determine the population earning less than the federal poverty line in the prior 12 months, localized to the zip code level. As American Community Survey population poverty results are not available for the years 2008 through 2010, we   created zip code-level estimates for those years based on linear growth population trajectories and   absolute counts from 2011 through 2017. Fifth, for geolocation and rural location classification we used topologically integrated geographic encoding and referencing cartographic files from the US Census Bureau. 24 To estimate driving times we used a 2012 Environmental Systems Research Institute road atlas and applied standard driving regulations. 25

Hospital Geolocations and Operational Status
We identified hospitals directly from HCRIS. We geolocated each hospital using their reported street address and categorized each hospital in each year either as new, existing, or closed relative to the hospital's reporting in the prior and subsequent years. Hospitals that changed facility status from short-term acute care to any other facility type (eg, rehabilitation hospital) were considered closed, as under that change they would no longer provide essential services for time-sensitive medical emergencies. Hospitals that changed ownership but remained open in the same location without a gap in services were considered to be continuously open. Safety-net hospitals were defined as those in the highest quartile of hospitals in 2008 according to their percentage of patients eligible for Supplemental Security Income 26 at the state level, as measured in HCRIS. 27 We manually verified the operational status of all annual changes using a variety of methods,

State Medicaid Expansion Definition
We defined our primary exposure, state Medicaid expansion status, as either expansion or nonexpansion using the first complete calendar year of expansion in each state. We determined the status and timing of state decisions to adopt ACA Medicaid expansion between January 2014 and December 2017 using public reporting. 29

Measurement of Population Access to Hospital-Based Emergency Services
We defined the population without emergency access to an acute care hospital as the count and percentage of the state population that lived outside a 30-minute drive of any short-term acute care hospital. A 30-minute threshold was used because medical care delivered within this time frame is associated with improved outcomes for many emergencies through early stabilization and initiation of definitive care. 1,5,12,13 Calculating drive times from home zip codes is appropriate given that a majority of acute myocardial infarctions, 30 traumas, 31 strokes, 32 and general medical emergencies 33 occur near or at home. We performed access calculations using denominators of all and low-income state residents (defined as those reporting incomes below the federal poverty line). We included all ages for both the numerator and denominator because short-term acute care hospitals are frequently the first point of hospital contact for time-sensitive emergencies for both adult and pediatric patients. We converted zip code regions into geometric centroids, using the center of each

JAMA Network Open | Health Policy
State Medicaid Expansion and Emergency Access to Acute Care Hospitals zip code to measure population access, and performed calculations using Network Analyst in ArcGIS 10.6 software. For safety-net hospital population access calculations we performed the same steps using the subset of safety-net hospitals.

Statistical Analysis
To determine the association between Medicaid expansion and population access we used a difference-in-differences approach with state-year as the unit of analysis. We fit a series of linear regression models using cluster-robust variances. We fit a total of 4 models with 4 different dependent variables defined at the state-year level: total population access to any acute care hospital, low-income population access to any acute care hospital, total population access to a safety-net hospital, and low-income population access to a safety-net hospital. Each model included a relative term for year (with time zero being the year before Medicaid expansion in states that expanded eligibility and 2013 in states that did not), a term for Medicaid expansion status, an interaction term for relative year and Medicaid expansion status, and a state-year level error term.
Models accounted for stable state-level characteristics before and after the year 2013 (or year of Medicaid expansion, for states that expanded later). This approach allows us to control for other factors that could be associated with changing population access by evaluating changes within the same state relative to the year of Medicaid expansion for that state.
To better understand the population impacted by changes in emergency acute care hospital access, we further estimated the change in access in 2017 that was potentially associated with Medicaid nonexpansion. We did this by projecting the differential change in population access from

Population and Hospital Distribution Characteristics in 2008
In total, there were 4601 hospitals, including 1118 safety-net hospitals, serving 291.   (Figure 1). Both expansion and nonexpansion states experienced net hospital closures in most years (Figure 2).

Difference-in-Differences Analyses of Population Access to All Hospitals
In the difference-in-differences analysis, states that did not expand Medicaid experienced an increase in the population without emergency access to an acute care hospital (6.76% to 6.79%

Difference-in-Differences Analyses of Population Access to Safety-Net Hospitals
In the difference-in-differences analysis, states that did not expand Medicaid experienced an increase in the population without emergency access to a safety-net hospital (

Discussion
Timely access to an acute care hospital is a key determinant of improved clinical outcomes for conditions such as acute myocardial infarction, 2,3 stroke, 4,6 sepsis, 8,9 and traumatic injuries. 10,11 We found that states that did not expand Medicaid under the ACA experienced more acute care hospital closures, which was associated with a loss of timely access to care for an estimated 421 000 total persons, of whom 48 000 had low incomes. In states that did not expand Medicaid, an estimated additional 2.2 million persons overall and 364 thousand persons with low incomes lost timely access to safety-net acute care in association with safety-net hospital closures.  Our analysis contributes to the understanding of how state-level Medicaid expansion decisions impact public health for time-sensitive emergencies, both for the low-income and the overall population of the United States. While recent work has demonstrated an association between Medicaid expansion and hospital closures, 17,34 summaries of closures alone cannot determine whether the shifting hospital landscape is associated with worsened, unchanged, or even potentially improved public health access for time-sensitive conditions. 35 To answer this question, this drivetime analysis for all acute care hospitals in the United States before and after the ACA on an annual basis accounted for births, deaths, population migration, and rural-to-urban population changes to assess whether closures and new hospitals were associated with emergency access to acute care hospitals. Without such an analysis, it would not be possible to determine if emergency access was associated with changes in hospital service locations. This study found that changes in Medicaid eligibility at a state level may have negative repercussions for emergency access to acute care hospitals for persons at any income level.
Our analysis found a potential spillover effect from national health policy reform on changes in the local availability of services, with unanticipated and undesirable repercussions at patient,   Medicaid expansion year Short-term acute care hospitals Annual change in hospital count, No.

Medicaid expansion year
Short-term acute care safety-net hospitals Net hospital counts for states that expanded Medicaid Net change in hospital numbers for states that did not expand Medicaid

Limitations
Our analysis has several limitations. We did not include hospital capacity in our models, and so true availability of services could have been overestimated or underestimated in some geographic regions. Similarly, while we limited inclusion to short-term acute care hospitals, we did not verify that all facilities had an operational emergency department or evaluate prehospital emergency services. We