Predictors of Payer Mix and Financial Performance Among Safety Net Hospitals Prior to the Affordable Care Act

The objective of this study was to use audited hospital financial statements to identify predictors of payer mix and financial performance in safety net hospitals prior to the Affordable Care Act. We analyzed the 2010 financial statements of 98 large, urban safety net hospital systems in 34 states, supplemented with data on population demographics, hospital features, and state policies. We used multivariate regression to identify independent predictors of three outcomes: 1) Medicaid-reliant payer mix (hospitals for which at least 25% of hospital days are paid for by Medicaid); 2) safety net revenue-to-cost ratio (Medicaid and Medicare Disproportionate Share Hospital payments and local government transfers, divided by charity care costs and Medicaid payment shortfall); and 3) operating margin. Medicaid-reliant payer mix was positively associated with more inclusive state Medicaid eligibility criteria and more minority patients. More inclusive Medicaid eligibility and higher Medicaid reimbursement rates positively predicted safety net revenue-to-cost ratio. University governance was the strongest positive predictor of operating margin. Safety net hospital financial performance varied considerably. Academic hospitals had higher operating margins, while more generous Medicaid eligibility and reimbursement policies improved hospitals’ ability to recoup costs. Institutional and state policies may outweigh patient demographics in the financial health of safety net hospitals.


INTRODUCTION
Safety net hospitals that disproportionately care for low-income populations, often free of charge, face major financial changes in the coming decade due to the Affordable Care Act (ACA). Because safety net institutions predominantly see low-income populations covered by Medicaid or with no coverage at all, costs can be high and revenue inadequate. For example, in 2010, an analysis of hospital survey data revealed that the average financial margin for safety net hospitals was reportedly 2.3 percent, compared to a 7.2 percent for all U.S. hospitals. [1] The 1981 Omnibus Budget Reconciliation Act (OBRA) recognized the high financial burden incurred by hospitals that serve uninsured patients and low-income patients with special needs. [2] Over the following years, what became known as the Medicaid and Medicare Disproportionate Share Hospital (DSH) program provided additional payments to these facilities beyond the normal insurance reimbursement from Medicare and Medicaid. Combined with local and state government subsidies from sales and property taxes, DSH payments were intended to provide at least partial reimbursement for otherwise uncompensated care.
The ACA, which was passed into law in March 2010, brings many key changes to the financial circumstances of these hospitals. On the one hand, most hospitals will likely experience increases in the number of patients with insurance, since one of the law's chief aims was expanding coverage to Americans without health insurance. To accomplish this aim, the ACA included a provision to expand Medicaid eligibility to adults with incomes at or below 138% of the federal poverty level (FPL) beginning in 2014, corresponding to roughly $16,000 in annual income for an individual or $33,000 for a family of four. Additionally, the law made tax credits available to those with incomes up to 400% of the FPL to subsidize premiums for private health insurance.
These two policies have already been linked to an additional 10 million adults gaining health insurance in 2014, thus bringing in needed revenue to safety net hospitals. [3] On the other hand, several sources of safety net funding will decline over the coming decade. Due to predicted coverage increases, the ACA includes a provision to cut Medicare DSH funding by 75% and Medicaid DSH funding by up to 50%. [4] These cuts will occur gradually; for instance, the Medicaid DSH cuts were originally slated to start at 4% in 2014 and increase to 49% in 2020, [4] though recent negotiations have delayed the onset until 2016. [5] Cuts will vary by state, based on the percentage of uninsured in the state and the state's allocation of DSH payments to hospitals that provide more care to Medicaid and uninsured patients. In an attempt to incentivize states to target their DSH funds to the hospitals in greatest need, the biggest DSH cuts will be experienced by states that provide their DSH funding to hospitals regardless of how much uncompensated care they provide. [2] Continued financial viability in the face of such funding cuts presupposes an expansion of Medicaid coverage that can make up the difference. However, in 2012, the Supreme Court effectively ruled that the Medicaid expansion was optional. States could choose to opt-out of expanding eligibility requirements and continue to using their own pre-ACA eligibility cutoffs that were typically much more restrictive. More than 20 states have decided not to expand Medicaid as of October 2014, leaving over 6 million people uninsured and without an option for affordable coverage. [6] Meanwhile, DSH cuts apply to all states, whether they decide to expand Medicaid or not. [7] Furthermore, undocumented immigrants and low-income legal immigrants who have been in the U.S. for less than five years will likely remain uninsured, as the ACA does not grant eligibility for Medicaid coverage to this population. Finally, the perception that the ACA will grant near universal coverage may prompt state or local-level allocation of tax dollars elsewhere, leaving safety net hospitals even more financially vulnerable.
Despite significant policy activity and interest in these issues, the evidence base on the financial performance of safety net hospitals is quite sparse, in large part due to data limitations. Some information on financial performance is available from surveys of hospitals, [8] though the accuracy and uniformity of responses in self-reported data has not been independently confirmed. Others have relied on financial data reported in the Medicare Cost Report -expense reports that must be filed on a yearly basis by all facilities receiving Medicare reimbursements. However, these reports do not disclose important details such as the cost of free care or Medicaid DSH payments. The quality of data in these reports has also been shown to be highly variable, with important factual inconsistencies when compared to hospital financial statements produced by independent auditors as required by outside creditors. [9] Furthermore, financial information from surveys and the Medicare Cost Report generally report on the hospital entity alone and therefore do not reflect the financial condition of the entire organizational entity within which the hospital is owned and operated. Yet most hospitals are in multi-entity systems that often include other hospitals, physician practices, health plans, and other operating entities that can have a material financial impact on the hospital entity.
An alternative approach that overcomes many of these limitations is using audited hospital financial statements, which, as discussed below, are subject to several quality controls and independent validation, unlike survey data or the Schedule G (income statement and balance sheet items) of the Medicare Cost Reports. A recent study by Kane et al. examined the financial performance of large urban publicly-owned safety net health systems using audited financial statements, which demonstrated the feasibility of using this data source for research purposes. Kane et al. found that the sample of safety net systems earned positive operating margins (median of 1.7% in 2010) but hospitals in the sample had considerable variation in financial performance, which was highly contingent on DSH funding and local government tax dollars. [10] It remains unclear what factors are most closely linked to financial distress among safety net hospitals and health systems. Qualitative analyses and case studies have linked better performance among some hospitals to elements of successful strategic management, [11] while others have conceptualized safety net hospitals as operating within a medical arms race, with viability predicated on the adoption of private sector strategies to expand, to renovate, and to attract more generous payers. [12] Often, patient demographic characteristics are assumed to be a key factor in determining the financial Census data, we examine potential risk factors for poor health system financial performance. Our goal was to assess the relative contributions of patient demographics, hospital system features, and state policies to the financial health of safety net hospitals on the eve of the ACA's implementation.

Overview of Analytical Approach
Using audited financial statements from 2010 and other hospital-reported information, we identified three outcome variables related to financial performance, as defined in the next section. We then collected information on hospital features, demographics of the population living in the area around each hospital, and key state policies related to safety net hospital financing. Our analytical model used multivariate regression to identify which of these institutional, demographic, and policy factors were independent predictors of safety net hospital financial performance.

Hospital Audited Financial Statements and Hospital Characteristics
Our source of financial data was the audited financial statements of the medical systems that contained our sample of safety net hospitals. Audited financial statements, considered the "gold standard" of financial reporting, are required to present a balance sheet, a statement of operations ("income statement"), a statement of changes in net assets (changes in net worth), and a statement of cash flow, as well as provide detailed footnotes disclosing accounting policies and important details such as the amount of free care provided, the amount of Medicaid supplemental payments realized, local government financial support, and other financially material considerations. The financial statements are audited by independent outside auditors, and in order to obtain debt financing, management must attest to the accuracy and fair representation of the financial information.
The Securities and Exchange Commission (SEC) now requires public disclosure of audited financial statements of all health systems issuing tax-exempt debt in a single municipal repository. [14] This recent requirement, enacted in 2009, has generated a record number of public and private non-profit health system filings, as intended. It thus created an unprecedented opportunity to capture the financial health of safety-net hospital systems.
In addition to audited financial statements, we also obtained various hospitalrelated measures such as Medicaid inpatient utilization rates from the American Hospital Directory (AHD), a commercial database derived from the Medicare Cost Report. [15] For multihospital systems these hospital measures were aggregated to the level of the system.
We identified three distinct outcomes of payer-mix and financial performance in these safety net hospitals, which move from the most specific to the broadest:

1.
Medicaid-Reliant Hospitals: From AHD data, we assessed which hospital systems were "Medicaid-reliant hospitals," defined as those with at least 25% of their hospital days coming from Medicaid patients (this represented approximately the top one-third of the sample on this measure).

2.
"Safety Net Ratio": From financial statements and federal DSH data, [16] we constructed a safety net revenue-to-cost ratio, which was defined as the sum of Medicaid and Medicare DSH payments and local government transfers, divided by the total costs of charity care plus Medicaid payment shortfall. Cost of charity care was derived based on dividing the total amount of free care (valued at charges) by the hospital's markup of charges over total operating cost.

3.
Operating Margin: From audited financial statements, we measured the overall operating margin for each hospital system (total operating revenue minus total operating expense, divided by total operating revenue).
For potential institutional-level predictors of financial performance, we identified the following hospital-system level features from the AHD database: 1) Ownership status Angeles county is divided into 35 PUMAs. While hospitals may draw patients from areas larger or smaller than PUMAs depending on the circumstance, a population base of roughly 100,000 is more than adequate to capture the key demographic features of a hospital's surroundings while still providing a narrower definition than would be possible using statewide or county-wide measures.
For each hospital, we identified the PUMA containing that hospital using coordinates for each hospital's address mapped against Census PUMA information; geocoding was double-checked by two independent investigators. Next, we assigned each hospital system the survey-weighted percentages of the following three demographic features for all individuals living in the PUMA: 1) percentage of non-citizens, 2) percentage of racial/ethnic minorities (defined as non-whites and/or Latinos), and 3) percentage with incomes below the Census federal poverty threshold. When hospital systems contained multiple hospital locations, we used the weighted average for these statistics, based on the total revenue for each component hospital.

Medicaid Policy Features
We identified measures of several aspects of state policy with major potential implications for safety net hospitals, in the areas of Medicaid and DSH:

Sample Selection
Our target sample for inclusion in the study consisted of large public or non-profit urban safety net hospitals. We identified potential hospital systems for the sample based on hospital ownership, urban location, and bed size greater than 150 using the AHD database. We initially identified 159 large urban publicly-owned hospitals in 30 states, for which we were able to locate 2010 audited financial statements for 95 hospitals in 83 systems. To expand the regional representation of the sample, we added 15 large private nonprofit systems with high Medicaid utilization and uncompensated care, five of which were conversions of public hospitals. This brought our sample to 110 safety net hospitals in 98 systems. The 98 systems also included another 73 hospitals that were not large urban safety net hospitals, but their financial data was consolidated at the system level.
This yielded a sample size of 98 hospital systems across 34 states for our primary analyses. Analyses of the safety net revenue-to-cost ratio were limited to the 62 hospital systems that reported Medicaid DSH amounts in their financial statements.

Statistical Analysis
For each of our three outcomes, we conducted both univariate and multivariate analyses using linear regression models. Univariate analyses considered each set of predictor variables in isolation: hospital ownership, hospital governance, percentage minorities, percentage non-citizens, percentage living below poverty, Census region, Medicaid eligibility index, Medicaid payment-to-cost ratio, DSH allotment (high versus low), and state targeting of DSH allotments to safety net hospitals.
Multivariate models included all these variables simultaneously, with the exception of our analysis of state policies. Three of the state policies considered here were highly correlated with geographical region: Medicaid eligibility, DSH allotment, and DSH targeting were all highest in the Northeast, while eligibility and DSH targeting were lowest in the South. (Medicaid payment-to-cost ratios did not vary significantly by region.) Thus, to estimate the impact of these policies in the multivariate regressions, we excluded regional controls from those analyses, since the regional dummies absorbed most of the meaningful variation in these policies.
All analyses used linear models for ease of comparability of coefficients across the different outcomes, which all refer to percentage-point changes in the outcomes.
Using logistic models for the one binary outcome (high Medicaid payer mix) yielded largely similar results, with a few differences noted below. Results in the next section focus on the multivariate-adjusted regression estimates, unless otherwise indicated, with a p-value for significance p≤0.05. All analyses used Stata 12.0. Table 1 describes the study sample. In our sample of 98 hospital systems, the majority of systems were publically owned by state or local governments and presided over by a publicly-appointed governing board. 59% were located in the South. Linkage with Census data indicates that the hospital systems were located in areas with populations that were disproportionately poor (17%) and belonging to racial or ethnic minorities (39%). Figure 1 presents the distribution of safety net revenue-to-cost ratios and operating margins for our study sample. The median safety net ratio was 1.025, indicating a +2.5% safety net margin, with an interquartile range from -32.4% to +32.6%. The median operating margin was +1.7%, with an interquartile range from -0.3% to +4.5%.  Univariate results for payer-mix were identical when using logistic models instead of linear models. The logistic multivariate model showed the same impact of Medicaid policies and demographics, but also indicated publicly-owned hospitals and those with an elected board of directors were more likely to be Medicaid-reliant. association with the safety net revenue-to-cost ratio. For this outcome, roughly a third of hospital systems did not list any DSH allotments in their financial reports, and we excluded them from our primary analysis. It is unclear whether such hospitals simply did not receive any DSH payments, or received them but chose not to report them. In a sensitivity analysis in which we included these hospital systems and treated them as having $0 in DSH allotments, the results were quite similar as those in Table 3. Table 4 displays results for predictors of hospital system operating margin.

RESULTS
University governance (+0.04, p=0.030) was a positive predictor of operating margin, while location in the Northeast (-0.06, p=0.016) was a negative predictor, and these relationships were similar in both adjusted and unadjusted models. Hospital ownership, hospital area demographics, and state policies showed no significant association with operating margin in either univariate or multivariate models.

DISCUSSION
Financial circumstances of safety net hospitals will change significantly as a consequence of the ACA. Understanding the factors driving these hospitals' financial performance prior to the ACA is critical for guiding policy in this area. Our paper contributes to an ongoing effort to strengthen what has been a research literature plagued by sparse evidence and non-validated data sources. Using audited financial statements and focusing on the broader health systems within which safety net hospitals operate, we present suggestive findings on the underlying risk factors for poor financial performance in large urban safety net hospitals.
In assessing the predictors of financial performance, our findings suggest that one piece of conventional wisdom -that patient demographics determine a hospital's financial destiny [20] -may not hold. Being located in an area with high shares of minorities, low-income individuals, and non-citizens was significantly associated with a high Medicaid payer mix, as was the generosity of each state's pre-ACA Medicaid eligibility criteria. However, patient demographics did not significantly predict either of our financial performance measures -safety net revenue-to-cost ratio or operating margin.
Instead, we found that several hospital system-level features and state policies were stronger predictors of financial performance. Systems in states with higher Medicaid reimbursement rates and more generous Medicaid eligibility had higher safetynet ratios, as would be expected. Meanwhile, state DSH funding level and state DSH targeting were not significant predictors of financial performance. This finding may be due in part to the sample size reduction (from 98 hospital systems to 62 hospital systems) required to run the analysis on the safety net ratio, since one-third of hospital systems did not report DSH funding in their financial statements. However, it also may simply be a logical reflection of the relative magnitude of total federal and state Medicaid spending (about $415 billion in 2012) versus DSH spending (about $17 billion in combined funding in 2012). [21,22] Overall, one of our key findings is that state policy decisionsin particular those that govern Medicaid reimbursement and eligibility -may outweigh demographic characteristics in their ability to shape the financial health of safety-net hospital systems.
In terms of the operating margin, which is a broader measure of hospital performance than looking at just one component of costs and revenues (as in the safetynet ratio), we did not find any significant association with state policy decisions. Two other parameters were the strongest predictors. Hospital systems in the Northeast fared worse on average than systems in the South, West, and Midwest, a finding that may be due in part to financial repercussions from Massachusetts' 2006 health reform, which redirected much of the state's uncompensated care fund away from several large safety net hospitals; [13] however, our findings were persistent even when excluding Massachusetts. Other regional market trends may be to blame, such as high overall levels of health care spending in several Northeastern cities, [23] though we do not have the sample size or variation over time that would be useful to further assess this issue.
Meanwhile, we found that academic medical centers were at a significant advantage compared to other safety net hospital systems in terms of operating margins, even after accounting for state policy decisions and patient case mix. This is consistent with previous research on indirect medical education (IME) funding, which provides academic hospitals with additional funding ostensibly tied to the higher costs of care in teaching facilities. One recent study indicated that current IME levels are two to three times as large as what would be an "empirically justified" level. [24,25] Given that the ACA does not include any IME cuts, university ownership may continue to provide some safety net hospital systems with a financial buffer as Medicaid DSH funds decline substantially over the next five years.
Our paper has several important limitations. First, by virtue of using audited financial statements, which is a time-intensive data source to collect, we were limited to a single year (2010) of pre-ACA financial statements, and our overall sample size was small. The SEC requirement for hospitals to file these financial statements only took effect in 2009, precluding any longer-term historical analyses. In designing our sample, we aimed to target hospital systems that provide the bulk of hospital care in the U.S.
health care safety net, by identifying large urban-based systems. However, the fact that we had just a single year and a sample of 98 hospital systems means that our regression analyses may have been underpowered to detect significant associations between financial performance and the demographic, hospital, and policy features we studied.
Nonetheless, large effects should have been evident even in this small sample, and we did identify several factors associated with changes in operating margins on the order of 4-6 percentage points, which was roughly the gap between the 1 st and 3 rd quartiles in our sample.
Our results may not generalize to hospitals involved in safety net care that did not meet our inclusion criteria. More specifically, our observed pattern of predictors of financial performance in large urban safety net hospitals may not apply to rural hospitals (including critical access hospitals), small safety net hospitals, and for-profit hospital systems that provide large amounts of care to low-income populations.
Finally, our analysis is by definition a cross-sectional descriptive study. None of the associations identified here can be attributed to a causal relationship without further study and, ideally, policy variation over time. For better or for worse, the ACA is likely to provide just such an experiment for safety net hospitals over the coming decade.
In conclusion, in this exploratory analysis of large urban safety net hospitals, we find that audited financial statements can be used to examine financial performance and identify risk factors for distress. Overall, population demographics appear to play less of a role than state policy decisions such as Medicaid eligibility and payment levels, and hospital governance features -particularly academic affiliation -may be protective for hospitals against the risk of financial distress. Future research will be critical to monitor these patterns over time and identify policy approaches for protecting the hospital safety net in the coming years of immense change.