Opportunities for Prevention: Assessing Where Low-Income Patients Seek Care for Preventable Coronary Artery Disease

Abstract

The Affordable Care Act has many aspects that are aimed at improving health care for all Americans, including mandated insurance coverage for individuals, as well as required community health needs assessments (CHNAs), and reporting of investments in community benefit by nonprofit hospitals in order to maintain tax exemptions. Although millions of Americans have gained access to health insurance, many—often the most vulnerable—remain uninsured, and will continue to depend on hospital community benefits for care. Understanding where patients go for care can assist hospitals and communities to develop their CHNA and implementation plans in order to focus resources where the need for prevention is greatest. This study evaluated patient care-seeking behavior among patients with coronary artery disease (CAD) in Florida in 2008—analyzed in 2013—to assess whether low-income patients accessed specific safety net hospitals for treatment or received care from hospitals that were geographically closer to their residence. This study found evidence that low-income patients went to hospitals that treated more low-income patients, regardless of where they lived. The findings demonstrate that hospitals—especially public safety net hospitals with a tradition of treating low-income patients suffering from CAD—should focus prevention activities where low-income patients reside. (Population Health Management 2015;18:337–341)

Introduction

The Affordable Care Act (ACA) has many aspects that are aimed at improving health care for all Americans. The most reported focus of the ACA has been on mandated insurance coverage that aims to increase health care security for individuals who are currently uninsured or underinsured. Another obligation of the ACA is for nonprofit hospitals to increase their community benefit by conducting community health needs assessments (CHNAs) and reporting investments in community benefit in order to maintain tax exemptions. Investments in community benefit have been the obligation of nonprofit hospitals as a condition of their federal tax-exempt status for decades, and most states impose similar requirements; however, the level of investment has not been studied until recently.¹

Even though the ACA was designed to increase the number of insured patients, how affordable insurance will be for individuals with low socioeconomic status—especially if they are sicker—remains an important policy question. Katz² reports that 23 million will remain uninsured even under the ACA and that these individuals likely will require hospital care services via trauma/emergency department admissions. Therefore, it may be assumed that a vulnerable population will still exist and “safety net” hospitals will still be needed to provide services to low-income communities.

Hospital community benefit

In 2002, the financial advantage nonprofit status brought to hospitals was an estimated $12.6 billion.³ Although the Internal Revenue Service (IRS) legally established community benefit in 1969, to date, no minimum expenditure level for valuing community benefit has been established.⁴ Under increased Congressional scrutiny, in 2009 the IRS began requiring all community benefit expenditures be reported on a “Schedule H” work sheet.⁵ The work sheet—attached to the Form 990 that all tax-exempt organizations must complete—effectively creates a nationwide, publicly available reporting system that enables assessment of how individual hospital community benefit investments are linked to community needs.¹ The clarity of the Schedule H definition was a vast improvement over state community benefit laws that vary widely in their variability and reporting requirements, prohibiting meaningful comparisons.⁴

Since its implementation, Schedule H has identified community benefits as activities including: the provision of financial assistance such as free or discounted health services provided to indigent patients, hospital costs associated with hospital participation in Medicaid and other entitlement programs, and other “community health improvement” services.⁵ The IRS clearly delineates what it considers community benefits as opposed to bad debt and losses associated with hospitals' Medicare participation, neither of which is included.

Measuring community benefit expenditures has been mired in controversy.⁴ An American Hospital Association report estimated that hospitals reported 11.3% of their 2009 total annual expenditures as “benefits to the community.”⁶ However, the report included expenditures specifically excluded from the IRS definition of community benefit: Medicare losses and bad debts. A more accurate analysis, excluding the disallowed expenditures, estimated tax-exempt hospitals spent 7.5% of their operating expenses on community benefit in 2009, with most (more than 45%) going to offset Medicaid losses and only 2.2% spent on community health improvement.⁷ Young et al⁷ also found that the level of community benefit expenditures is highly variable, and only the presence of at least some level of state reporting requirement explained the variability.

Community health needs assessment

The implementation of the ACA may lead to increasing investments in community health improvement as it requires the addition of a CHNA process that includes an evidence-based planning and prioritization process and is completed with an implementation strategy to guide hospital investment.¹ The CHNA process requires partnerships between hospitals and local public health agencies with a clear mandate to improve preventive services to address population health issues. CHNAs must include a description of the community served in order to assist community partners in reaching those with the greatest need.¹ If hospitals treating more low-income, vulnerable patients can identify where these patients come from, the implementation of preventive programs may be more efficient and effective.

Public hospitals (state and local) have been the traditional safety net venue for low-income and uninsured patients.^8

–11 However, the number of these hospitals has decreased by at least 27%.⁶ If these safety net hospitals no longer treat low-income patients, and if the number of insured is not guaranteed via the ACA, then there is a need for CHNAs, especially for private not-for-profit hospitals that will be responsible for treating this patient population.

Coronary artery disease (CAD)

Although CAD is a common problem, with 36.9% of the US population exhibiting some form of heart disease, the costs for treatment of CAD are predicted to triple by 2030.¹² Many hospitalizations for CAD are preventable; however, patients with low socioeconomic status experience disparities in mortality rates from CAD as a result of a lack of social capital as well as the higher density of alcohol outlets and fewer opportunities to purchase healthy foods (ie, food deserts) in their communities.¹³ In a review of socioeconomic factors and cardiovascular disease, Kaplan and Keil¹⁴ also found that controlling for all well-known risk factors, education, income, occupation, employment status, and social class contribute to the incidence and prevalence of cardiovascular disease. Public outreach should be focused on disease risk factors but also social determinants of health and access to preventive services.^14,15 Kanjilal et al¹⁶ found that preventive care led to a reduction of risk levels for cardiovascular disease among all income groups, especially for lowering cholesterol and high blood pressure. Pearson et al¹⁷ also endorsed prevention in a statement made jointly by the American Heart Association and the Centers for Disease Control and Prevention.

This paper describes how patients suffering from nonacute CAD gravitate to hospitals for treatment. An evaluation of patient care-seeking behavior was conducted to assess whether low-income patients access specific safety net hospitals for treatment or receive care from those that are geographically closer to their residence. Understanding where patients go for care can assist hospitals and communities to develop their CHNA and implementation plans in order to focus prevention resources where the need is greatest.

Market share approach

This study applies a market share approach to understand low-income care-seeking behavior. Identifying where high-risk patient populations go for care can help define hospital market share and tailor programs and services to market needs. The specific research question of this study is: in order to facilitate hospital community outreach programs—particularly for low-income vulnerable populations—is there a natural clustering either by (a) patients' home zip codes, (b) patients' payment type, or (c) source of admission? If a geographic concentration exists for hospitals, this would be a natural first step for community outreach programs. Patient payment type would guide more community outreach to patients who are either elderly (Medicare) or have low income (Medicaid plus the uninsured). Although not as convenient as a geographical focus, outreach could be conducted in partnership with other community-based organizations that work with these population groups. Finally, source of admission could aid outreach programs by coordinating health planning with a frequent admitting physician or clinic. This also could serve as a preliminary step in developing medical homes, another objective of the ACA.¹⁸

Methods

Three separate data sets were merged in order to perform this study: the Florida Agency for Health Care Administration (data on hospital capacity), the Hospital Inpatient Discharge Data (patient characteristics), and lastly the Hospital Financial Data for Florida for specific service capabilities. All data were from 2008 to avoid possible outlier effects attributable to the financial crisis (analyses were conducted in 2013). General, short-term stay hospitals were included and specialty or long-term care facilities were excluded. Institutional Review Board approval was not required as all data were de-identified secondary data.

The following procedures were implemented to derive the final data set measuring market share based on patient-hospital pairs. In the first step, patient data by insurance status, admission source, and admission type were consolidated by hospital for all patients admitted with a diagnosis of CAD (International Classification of Diseases, Ninth Revision codes 412–414, and 429.2). This included angina pectoris, other forms of chronic ischemic heart disease, and coronary heart disease. In step 2, admissions were consolidated by zip code. In the third step of this process, the hospitals from the hospital admission profile in step 1 were linked to patients by zip code in step 2. Patients from zip codes where the population fell below the federal poverty level (FPL) were the focus of the analysis. The FPL was chosen rather than the 133% of the FPL typically used because Florida uses the former to determine Medicaid eligibility. Descriptive statistics for the dependent and independent variables used in this analysis are provided in Table 1.

Table 1.

Descriptive Statistics (N=2063)

Variable	Mean	Std. Deviation	Minimum	Maximum
Market Concentration—the combination of patients who come from a specific zip code “i” going to a specific hospital “j”	52.40	52.17	12	526.00
Percent private pay patients treated by hospital j	26.00	0.08	4.00	55.00
Percent Medicare patients treated by hospital j (omitted variable)¹	62.00	0.09	22.00	85.00
Percent low-income (Medicaid, charity, bad debt) patients treated by hospital j	10.00	0.05	2.00	65.00
Percent of patients by low-income zip codes	12.00	0.08	0.00	50.00
Admission source private physician (omitted variable)	35.00	17.00	1.00	83.33
Admission source other (including nursing homes, direct admission clinics)	65.00	0.17	16.00	99.00

Certain categories were omitted from the regression analysis in order to maintain variability in the remaining categories.

Market concentration was defined as the measure of the total number of patients whose home zip code, denoted by “i,” go to a specific hospital, denoted by “j.” In other words, market share is defined by patient-hospital combinations.¹⁹ Admission source indicated if the patient was admitted by a physician or from another source including public health clinics, nursing homes, and the emergency department (this percentage was below 1% because of the focus on nonacute CAD), or another hospital (again this percentage was well below 1% of total admissions).

Nonparametric rank tests were used to assess whether hospital ownership had a statistically significant relationship with the number of low-income patients treated, as well as the percent of their patient load from low-income zip codes.²⁰ The results are presented in Table 2.

Table 2.

Nonparametric Tests on Ownership, Care for Low-Income Patients, and Patients' Zip Codes (For-Profit Omitted) ^*

Hospital Ownership	Percent of Low-Income Patients Treated	Poverty Level of Zip Codes by Patients	Admission Source: Private Physician	Admission Source: Other	Private Insurance	Medicare
Government/Public (n=232)	20.0	16.0	41.00	59.00	24.6	51.00
Private/Not-For- Profit (n=1967)	9.0	11.0	35.00	64.00	25.9	63.00
F test	38.70	46.80	19.33	19.33	5.88	287.00
Pr <	0.0001	0.0001	0.0001	0.0001	0.01	0.0001

Unit of analysis is the hospitals. Only government/public and private not-for-profit are included as they are the hospital groups used in the regression analysis.

Pr, probability.

Finally, to account for payer mix, the research team derived uncompensated care (charity plus bad debt plus Medicaid patients), percent of patients covered by Medicare, and percent of private pay patients. The study included a total of 205 hospitals comprised of 2063 pairs of patient (count by zip code)–hospital combinations.

After assessing univariate relationships among the 3 measures described (zip codes, payer type, and admission source), logistic regression was used to discern whether patients clustered to hospitals because they were from lower income neighborhoods or low-income patients tended to go to hospitals treating more low-income patients irrespective of address. The dependent variable was market share, but because it was not linearly related to hospital features, and because hospitals want to focus their outreach mostly on their patient base, the dependent variable was divided into high market share (more than 33 patients per zip code), which was the median of market share per zip code, and low market share (fewer than 33 patients per zip code). The research team used the split at the median rather than at 1 standard deviation to avoid losing observations. Parameters are interpreted as log odds as well as in terms of odds ratios.

Results

As noted in Table 1, approximately one quarter of patients in the sample hospitals were private pay patients compared to more than 60% Medicare patients and 10% low-income patients. Twelve percent of patients were from low-income zip codes, and the vast majority were admitted via another source as compared to admission via a physician.

Hospital-based descriptive statistics are presented in Table 2. Only government/public and private not-for-profit hospitals were included because they were the hospital groups used in the regression analysis (described later). As expected, public hospitals in the sample treated significantly more low-income patients as compared to private not-for-profit hospitals (P<.001). Public hospitals also treated more patients from lower income neighborhoods (P<.001) and had more patients admitted via a private physician (P<.001). Although this latter result seems contradictory, private physician may include those physicians practicing in public health clinics that provide medical care services to lower-income patients. ²¹ Private not-for-profit hospitals treated more patients who were admitted from other sources as well as more patients covered by private pay and Medicare; however, a majority of admissions for both types of hospitals came from sources other than physicians. The sample did not include emergency cases of CAD, explaining the low number of emergency department admissions. Compared to the for-profit sector hospitals in the sample, the not-for-profit hospitals treated significantly more low-income patients from lower income neighborhoods (P<.001).

Logistic regression was used to identify how the variables used in the singular tests interacted in determining how patients gravitate to a certain hospital. The dependent variable is defined by low versus high market share—below or above the median—of a patient by zip code–hospital combination. The findings from the logistic regression are used to discern how market size related not only to these singular tests, but also whether larger hospital market share comes from low-income neighborhoods, or whether patients, irrespective of lower-income zip codes, gravitate to public or private not-for-profit hospitals (for-profit hospitals were omitted because of the significantly lower rates of low-income patient admissions). The findings from the logistic regression model are presented in Table 3. Hospitals that treated more low-income patients increased the log odds of a higher market share (3.380). Conversely, there was no statistically significant relationship between the income level of the patients' home zip codes and market share (−0.69). Patients who were admitted from another source of care had increased log odds of a higher market share than patients admitted to the hospital via a physician. Finally, the odds of more concentration of patients were statistically higher for public versus private not-for-profit hospitals.

Table 3.

Logistic Regression Results (N=2063; Low Market Share [n=1012]; High Market Share [n=1051])

Variable	Coefficient Model 1	Odds Ratio	95% Confidence Interval	t Statistic
Intercept	−.1.05^**
Private insurance patients	0.56	1.75	0.50	6.08
Low-income patients (all zip codes)	3.38^**	29.34	7.04	122.21
Patients from low-income zip code	−0.69^NS	0.50	0.156	1.62
Admission source other	0.84^*	2.31	1.29	4.16
Public ownership	0.58^**	1.79	1.30	2.45
Likelihood	45.67^**
score	44.98^**
Wald test	43.24^**
concordant %	60.00

P<0.01.

P<0.0001.

NS, not significant.

Converting the parameters to their natural logarithms, the odds of hospitals having a larger market share were assessed. For example, the odds of a hospital treating a high market share were 24 times greater for every 1% increase in the percentage of low-income patients. Conversely, a 1-unit change in admissions from another source of care increased the odds of a high market share by greater than 9%. Finally, the odds of having a higher market share for public hospitals were about 2 times higher than for a private not-for-profit hospital.

Discussion

This study found evidence that low-income patients go to hospitals that treat more low-income patients, regardless of where they live. The increase in the odds of a higher market share for hospitals admitting patients from another source as opposed to a physician's office offers evidence for partnering with community agencies that admit high numbers of patients with preventable health concerns. There also is evidence that there is more market concentration for public ownership hospitals than for not-for-profit hospitals, strengthening the case for the importance of CHNAs and community prevention activities, particularly for hospitals that treat lower income populations.

Public and not-for-profit hospitals that serve low-income patients can use this evidence to increase the number of preventive programs, which have the potential to reduce the uncompensated care burden, at least for conditions that are largely determined by environmental and personal health habits such as CAD. Here, empirical evidence is offered indicating how traditional measures of market share can be used by hospitals to identify vulnerable patients for community outreach purposes, as deigned by the provisions of the ACA.

It is important to note some limitations of this work. Low-income patients are defined as those living at or below the FPL; therefore, low-income patients who live above the poverty level (such as those above 133% of the FPL as guaranteed in the ACA), would increase the potential market share for public and not-for-profit hospitals and change some of the parameter estimates. However, as long as Florida does not accept expansion of Medicaid, the research team cannot make predictions about the impact of Medicaid expansion on hospital systems. Second, because of ambiguity in the data, bad debt and uncompensated charity care had to be combined. Although the IRS forbade using bad debt as part of the community service requirement for nonprofit status, doing so was necessary for this analysis.

From the hospitals' and patients' perspectives, mandating insurance for patients may not be a cure-all for states such as Florida, which has a population consisting of individuals who are ineligible for private insurance subsidies because of undocumented worker status, and state administrations/legislations that do not wish to expand Medicaid.²² This study demonstrates that focusing community outreach from hospitals, especially public hospitals with a tradition of treating low-income patients suffering from CAD, would be a worthy endeavor benefiting patients who could avoid the more serious complications of the disease, as well as reducing the number of low-income patients who may become ill. This conclusion is supported by other literature that suggests providing preventive care that addresses cholesterol and high blood pressure in a community translates into lower health care costs,²³ which is corroborated by the findings that low-income groups respond positively to these prevention endeavors.¹⁶ As for where the public hospitals should focus these outreach endeavors, the study findings suggest focusing on low-income communities that are most likely to access services from these safety net hospitals.

Footnotes

Author Disclosure Statement

Drs. Klaiman, Valdemanis, Bernet, and Moises declared no conflicts of interest with respect to the research, authorship, and/or publication of this paper. The authors received no financial support for the research, authorship, and/or publication of this paper.

References

Rosenbaum

, Rieke

, Byrnes

Hospital community benefit expenditures: looking behind the numbers. June 2013. http://healthaffairs.org/blog/2013/06/11/hospital-community-benefit-expenditures-looking-behind-the-numbers/. Accessed December 10, 2014 .

Katz

Future of safety net under health reform. JAMA., 2010; 304:664–670.

Congressional Budget Office. Nonprofit hospitals and tax arbitrage.. 2006. http://www.cbo.gov/sites/default/files/cbofiles/ftpdocs/76xx/doc7696/12-06-hospitaltax.pdf. Accessed December 4, 2013 .

Rosenbaum

, Byrnes

, Rieke

. Hospital tax-exempt policy: a comparison of schedule H and state community benefit reporting systems. Frontiers in Public Health Services and Systems Research, 2013; 2:Article 3.

Internal Revenue Service (IRS).. Instructions for Schedule H (Form 990). http://www.irs.gov/pub/irs-pdf/i990sh.pdf. Accessed December 4, 2013 .

Ernst & Young. Results from 2011 Tax-exempt Hospitals' Schedule H Community Benefit Reporting.. 2013. www.aha.org/content/14/schedhreport.pdf. Accessed December 4, 2013 .

Young

, Chou

, Alexander

, Lee

SYD

, Raver

. Provision of community benefits by tax-exempt U.S. hospitals. N Engl J Med., 2013; 368:1519–1527.

Ayanian

, Weissman

, Schneider

, Ginsburg

, Zaslavsky

. Unmet health needs of uninsured adults in the united states. JAMA., 2000; 284:2061–2069.

Thorpe

, Brecher

. Improved access to care for the uninsured poor in large cities: do public hospitals make a difference?. J Health Polit Policy Law., 1987; 12:313–324.

10.

Davis

, Rowland

. Uninsured and underserved: inequities in health care in the United States. Milbank Mem Fund Q Health Soc., 1983; 61:149–176.

11.

Duggan

Hospital ownership and public medical spending. Quarterly Journal of Economics., 2000; 115:1343–1373.

12.

Bufalino

, Masoudi

, Stranne

, et al. The American Heart Association's recommendations for expanding the applications of existing and future clinical registries: a policy statement from the American Heart Association. Circulation., 2011; 123:2167–2179.

13.

Cohen

, Farley

, Mason

. Why is poverty unhealthy? Social and physical mediators. Soc Sci Med., 2003; 57:1631–1641.

14.

Kaplan

, Keil

. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation., 1993; 88:1973–1988.

15.

Wilper

, Woolhandler

, Lasser

, McCormick

, Bor

, Himmelstein

. A national study of chronic disease prevalence and access to care in uninsured US adults. Ann Intern Med., 2008; 149:170–176.

16.

Kanjilal

, Gregg

, Cheng

, et al. Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971-2002. Arch Intern Med., 2006; 166:2348–2355.

17.

Pearson

, Mensah

, Alexander

, et al. Markers of inflammation and cardiovascular disease: application to clinical and public health practice. A statement for healthcare professionals from the Centers for Disease Control and Prevention and the American Heart Association. Circulation., 2003; 107:499–511.

18.

Koh

, Sebelius

. Promoting prevention through the Affordable Care Act. N Engl J Med., 2010; 363:1296–1299.

19.

Garnick

, Luft

, Robinson

. Appropriate measures of hospital market areas. Health Serv Res., 1987; 22:69–89.

20.

Norton

, Staiger

. How hospital ownership affects access to care for the uninsured. Rand J Econ., 1994; 25:171–185.

21.

Valdmanis

, DeNicola

, Bernet

. Public health capacity in the provision of health care services. Health Care Manag Sci. Epub ahead of print April 1, 2014.

22.

Kaiser Family Foundation.. The Florida health care landscape. http://kff.org/health-reform/fact-sheet/the-florida-health-care-landscape/. Accessed December 4, 2014 .

23.

Murray

, Lauer

, Hutubessy

, et al. Effectiveness and costs of interventions to lower systolic blood pressure and cholesterol: a global and regional analysis on reduction of cardiovascular-disease risk. Lancet., 2003; 361(9359):717–725.