Non-profit hospital community benefit spending based on local sociodemographics

Abstract

Non-profit hospitals are facing greater pressure to address the social determinants of health. Since 2012, with new requirements for greater transparency and community health needs assessments, non-profit tax exemption requirements are believed to incentivize investments in the community, particularly for vulnerable populations. We conducted a cross-sectional analysis of community benefit spending by private, acute care, non-profit hospitals from 2012–2014 to measure if hospitals have begun to address local community needs. We measured total community benefit spending and two subsets of spending—health care-related expenditures and community-directed contributions—as the proportion of their total expenditure. We obtained sociodemographic characteristics for their community, defined by ZIP code. In unadjusted and adjusted analyses using hospital-level and community-level covariates, community benefit spending has not varied and community-directed contribution amounts did not reflect local needs. Stronger incentives—tax-based or otherwise—are needed to steer non-profit hospitals to invest in community health.

Social forces—the conditions in which people live, work and age—are increasingly recognized as important contributors to health.^1,2 Mitigating these forces may significantly affect health outcomes. Over the past decade hospitals have been asked to play a larger role mediating these forces as they face greater accountability for health of populations throughout the care continuum.^3,4 Investing in communities with vulnerable populations has been proposed as one strategy for improving population health.⁵

In particular, non-profit hospitals have the opportunity to play an important role by investing in communities to improve population health. Non-profit hospitals gain their tax-exempt status in part because of the “community benefit” they provide.^6,7 Historically, non-profit hospitals could define community benefit themselves and often defined it loosely, reporting charity care or lost revenue due to lower payment rates from public insurance programs as community benefits. As a result, non-profit hospitals often did not make direct investments in social services or community organizations that address the social forces affecting their patients.⁸

More recently, non-profit hospitals have faced increased scrutiny by Congress and the public about whether they provide sufficient community benefit to justify their favorable and sizeable tax benefit.⁹ In the last decade, tax code reforms have improved transparency about hospitals’ community benefit contributions and moved hospitals towards greater population health expected by the health sector.^8–10 Beginning in 2008, the Internal Revenue Service (IRS) has required that all non-profit hospitals file a Schedule H, reporting each non-profit hospital’s level of a) free or subsidized care for patients, b) investments within the community, and c) educational or research activities for health professionals. With the Affordable Care Act’s implementation (beginning in 2012), non-profit hospitals were further required to conduct and publicly report a Community Health Needs Assessments (CHNA) every three years by convening local public health and community leaders to identify and address local needs.^9,11,12

Given major tax reforms designed to encourage non-profit hospitals to invest in their communities, we sought to describe the early trends in non-profit hospital spending after 2012 and to examine the relationship between the sociodemographic characteristics of a non-profit hospital’s local community and the amount hospitals invest in their communities. We hypothesize that non-profit hospitals located in communities with higher concentrations of vulnerable people invested more in their surrounding communities given higher needs.

Methods

Non-profit hospitals and community benefit data.

Our study evaluated community benefit spending patterns for private, non-profit hospitals that provided general, acute care between 2012–2014. This period aligns with the last major change to the non-profit hospital tax code—a CHNA requirement—and the most recent year of complete tax data filed at the time of our analysis.

For each tax year, community benefit data were obtained from the IRS’s Form 990 and the associated Schedule H. Tax data were obtained from GuideStar, which compiles and digitizes tax-form data from all non-profit organizations who file a Form 990 and their related schedules.¹³ For each tax filing obtained from GuideStar, we confirmed the hospital’s non-profit status by matching the name and address of the hospital with a non-profit hospital listed in the 2012–2014 American Hospital Association (AHA) Annual Survey of Hospitals, and limit our analysis to those organizations with matches in the AHA Annual Survey. Our sample included 1,681 non-profit hospitals in 2012, 1,643 in 2013, 1,674 in 2014.

Form 990 and Schedule H forms require hospitals to report expenditures that provide community benefit within definitions outlined by the IRS (Appendix 1, available from the authors upon request). The definitions for each community benefit category were consistent throughout the 2012–2014 reporting period. Using categories reported within Schedule H, we created three categories of spending for each tax year: total community benefit spending and two subcategories, health care-related spending and community-directed contributions, the sum of which does not equal total community benefit spending because educational and research activities are not captured in these subcategories. We evaluate these two subcategories because hospitals have traditionally provided benefits through free or subsidized care rather than direct community contributions, which may create financial tradeoffs when increased funds are allocated to one over the other.

The measures within Schedule H used to calculate health care-related spending included charity care (i.e., providing free or discounted services for patients who qualify under a hospital’s financial assistance policy), unreimbursed costs incurred by participating in means-tested government programs (e.g., Medicaid), and subsidized health services (i.e., clinical services provided despite financial loss to the organization not covered by the previous two categories). It does not include bad debt or unpaid medical bills.⁹ Community-directed contributions included community health improvement services (e.g., activities or funds for the direct purposes of improving community health, such as community-based anti-smoking campaigns), cash and in-kind contributions for community benefit (e.g., monetary contributions to local organizations or health programming) and community-building activities that promote the health of the community they serve (e.g., housing, economic development, environmental improvements, and leadership development for community members). We standardized each measure—total spending, health care-related, and community-directed contributions—to account for differences in hospital size and thus spending, by dividing each by the total hospital expenditure reported. Previous studies have validated these data by comparing Schedule H data with independent data sources.¹⁴

Community sociodemographic data.

We approximated the geographic reach of a non-profit hospital’s community benefit to the immediate ZIP code. We believe this to be a conservative approximation of a hospital’s community. We used 2010–2014 American Community Survey data to calculate six sociodemographic characteristics believed to affect the health of populations: the percent of the total ZIP code population that was uninsured, in poverty (i.e., household income below the federal poverty level), Black, Hispanic, unemployed, and having a high school degree or less.¹ Because ZIP code-level data are only available for five-year American Community Survey estimates, the same population-level sociodemographic characteristics were applied to all three tax years.

Statistical analysis.

We first examined mean spending within the three categories of community benefit spending as a percent of total hospital expenditure across all three years. We found no significant changes in spending across all three years, therefore the remaining analyses focus on 2014 community benefit spending—the most recent year of complete tax data. For each of the six ZIP-code characteristics, we divided the characteristic into quintiles and examined mean community benefit spending in 2014 within each quintile for each of the three spending categories. We tested significant differences across the years and across quintiles of sociodemographic characteristics using ANOVA.

We then used multivariable linear regression to further explore the association between sociodemographic characteristics and community benefit spending in 2014. For the regression models, the dependent variable was the three hospital expenditure variables defined above as a percentage of a hospital’s total expenditures. The independent variable of interest was a continuous measure of the hospital’s community’s sociodemographic characteristic (e.g., percent uninsured or percent in poverty). Each sociodemographic characteristic was analyzed separately in a model containing covariates believed to confound the relationship between community benefit spending and the sociodemographic characteristics of the local community. These included hospital-level characteristics (i.e., church affiliation and teaching hospital), community-level characteristics (i.e., being the sole community provider and location in an urban environment), market-level (i.e., percent of a ZIP code’s total publicly available hospital beds and the Herfindahl-Hirschman index based on the hospital’s share of the market for acute care services within each ZIP code), and the capacity to spend money towards community benefit (i.e., the margin of profit or loss for each hospital).

Human subject protections.

This study protocol was reviewed and was deemed exempt by the Institutional Review Board at the University of Pennsylvania.

Results

The percentage of total community benefit spending, health care-related spending, and community-directed contribution did not vary significantly between 2012–2014 (Table 1). In 2014, total community benefit spending averaged 8.1% of total hospital expenditures (standard deviation [SD] 8.2%; equaling $17 million), 6.7% on health care-related (SD 7.4%; $11 million), and 0.7% on community-directed contributions (SD 2.3%; $1.2 million).

Table 1.

CHARACTERISTICS OF NON-PROFIT HOSPITALS, 2012–2014

	2012 n=1681	2013 n=1643	2014 n=1674	p-value
Characteristics
Church affiliated, %	19.2	19.2	19	.98
Teaching hospital, %	5.4	4.8	5	.70
Sole community provider,%	91.7	91.8	92.1	.93
Urban hospital, %	69.4	69.2	69.8	.94
Percent of ZIP’s public hospital beds, %	75.4	75.2	75.1	.91
Herfindhal-Hirschman Index, mean (sd)	1,791 (1,429)	1,813 (1,447)	1,783 (1,434)	.84
Profit margin in $ millions, mean (sd)	9.54 (26.7)	10.7 (35.9)	12.1 (39.8)	.09
Community benefit spending
Total community benefit
mean % of total hospital expenditure (sd)	8.1 (8.2)	8.2 (8.2)	8.1 (7.3)	.97
$ in millions	16.5 (44.8)	16.4 (42.5)	17.0 (40.5)
Healthcare-related
mean % of total hospital expenditure (sd)	6.7 (7.4)	6.9 (7.5)	6.8 (6.5)	.88
$ in millions	10.7 (19.5)	11.0 (20.6)	12.1 (23.8)
Community-directed
mean % of total hospital expenditure (sd)	0.7 (2.3)	0.7 (2.4)	0.7 (2.8)	.91
$ in millions	1.2 (3.4)	1.2 (3.6)	1.3 (4.3)

Notes

^ap-value is for statistically significant difference in spending across the years 2012–2014 based on ANOVA.

In 2014, communities with a higher percentage of members who were Hispanic and unemployed had higher total community benefit spending (Table 2). Communities with a higher percentage of members who were uninsured, Hispanic, and unemployed had higher health care-related spending. Hospitals’ community-direct contributions were not different across quintiles of sociodemographic characteristics in their communities.

Table 2.

ASSOCIATION BETWEEN COMMUNITY BENEFIT SPENDING AND QUINTILES OF SOCIODEMOGRAPHIC CHARACTERISTICS OF NON-PROFIT HOSPITALS’ SURROUNDING COMMUNITIES^a

	Average unadjusted spending as a percent of total hospital expenditure
	Quintiles
Percent of population that:	1 (lowest)	2	3	4	5 (highest)
Was living in poverty
Total community benefit	7.7	8.3	7.8	8.3	8.8
Health care-related spending	6.4	7.0	6.7	7.4	6.8
Community-directed contributions	0.7	0.9	0.6	0.5	0.7
Was uninsured
Total community benefit	7.7	8.5	8.0	7.8	8.7
Health care-related spending*	6.1	7.2	6.7	6.9	7.3
Community-directed contributions	0.6	0.7	0.9	0.5	0.7
Was Black
Total community benefit	7.7	7.8	8.6	8.3	8.4
Health care-related spending	7.0	6.9	7.1	6.9	6.4
Community-directed contributions	0.6	0.6	0.6	0.7	0.9
Was Hispanic
Total community benefit**	8.1	7.8	7.8	8.1	9.0
Health care-related spending**	7.4	6.9	6.2	6.4	7.5
Community-directed contributions	0.5	0.6	1.0	0.7	0.7
Was unemployed
Total community benefit*	7.3	8.4	8.2	8.4	8.5
Health care-related spending**	5.9	7.2	7.0	7.2	6.9
Community-directed contributions	0.8	0.5	0.6	0.7	0.8
Had a high school degree or less
Total community benefit	8.3	7.9	8.0	8.5	8.3
Health care-related spending	6.3	6.9	6.7	7.4	6.9
Community-directed contributions	0.9	0.6	0.9	0.6	0.5

Notes

^aTable displays the average unadjusted spending as a percentage of total hospital expenditure within each quintile

^*p<.05,

^**p<.01,

^***p<.001

In our adjusted models (Table 3), no local sociodemographic characteristic was associated with total community benefit spending, health care-related spending, or community-directed contributions, regardless of the tax filing year.

Table 3.

PERCENTAGE POINT DIFFERENCE IN 2014 COMMUNITY BENEFIT SPENDING ASSOCIATED WITH A 10-PERCENTAGE POINT DIFFERENCE IN LOCAL SOCIODEMOGRAPHIC CHARACTERISTICS, ADJUSTED FOR HOSPITAL, COMMUNITY, AND MARKET LEVEL CHARACTERISTICS

Ten-percentage point difference in the percent of population that:	Total community benefits	Health care	Community- directed
Was living in poverty (95% CI)	0.34 (−0.08 to 0.76)	0.23 (−0.17 to 0.62)	0.02 (−0.11 to 0.15)
Was uninsured (95% CI)	0.58 (−0.01 to 1.2)	0.47 (−0.08 to 1.0)	0.12 (−0.06 to 0.30)
Was Black (95% CI)	0.06 (−0.15 to 0.28)	−0.08 (−0.29 to 0.11)	0.07 0.00 to 0.13)
Was Hispanic (95% CI)	0.13 (−0.11 to 0.37)	0.05 (−0.17 to 0.28)	0.01 (−0.07 to 0.08)
Was unemployed (95% CI)	0.58 (−0.24 to 1.4)	0.54 (−0.23 to 1.3)	−0.04 (−0.30 to 0.21)
Had a high school degree or less (95% CI)	0.36 (−0.10 to 0.82)	0.35 (−0.08 to 0.78)	−0.07 (−0.22 to 0.06)

^*p<.05,

^**p<.01,

^***p<.001

Discussion

From 2012 to 2014, non-profit hospitals’ spending on community benefits was concentrated in health care-related contributions, with no significant change in spending during this period. In unadjusted analyses of 2014 data, spending levels on total and health care-related contributions varied with some sociodemographic characteristics, but community-directed contributions did not. After controlling for hospital and market-level features, community benefit spending was not associated with community sociodemographic characteristics. Despite intentions to use the non-profit hospital tax code to increase community-directed contributions over the past decade, these contributions are a very small fraction of non-profit hospital expenditures, have changed little in the early years after reforms to the tax code, and were unrelated to the potential needs of their community.

Previous studies of non-profit hospitals found similar trends. In an earlier analysis of 2009 tax data, when Schedule H first became a requirement for non-profit hospital tax filings, spending on total community benefit was, on average, 7.5% of total hospital expenditure and less than 1% of total hospital expenditure was devoted to community-directed contributions.¹⁴ Another analysis pairing 2009 tax data with County Health Ratings showed community-directed contributions were not affected by the needs within a county, but county needs did affect health care-related spending.¹² These differences are likely related to the use of county-level variables and 2009 tax data. We present more contemporary data and believe the use of ZIP code as a definition for a community is potentially a more realistic measure of community than the county definition. An analysis of 2011 tax data similarly found low community-directed contributions and no shifts in spending based on community need.¹⁰ Our study adds to this literature by showing the pattern of spending across years continued for the subsequent three years after the requirements for Schedule H and CHNA were actually enacted.

Our work has several limitations. First, ZIP code may be too narrow an approximation of a hospital’s community. However, in the absence of a defined consensus surrounding what geographic marker to use for defining a hospital’s community, we feel this is the best available estimate for a hospital’s community in that it balances the need to have a wide enough geographic region with the need to not be too broad. Second, state and local tax policies and benefits may influence how hospitals decide to allocate community benefit dollars towards the community. Non-profit hospitals, though, derive more than 50% of their tax benefit from the federal government, making it a significant driver of community benefit spending patterns.¹⁰ Third, we did not collect data related to bad debt or measured shortfalls from Medicare reimbursement. However, these data points are excluded from the definition of community benefit data by the IRS. Fourth, because CHNAs or any community-based efforts may require additional time before implementation, more recent data—once available—may provide a clearer picture of the spending trajectory. However, there have not to our knowledge been recent changes to policy or tax regulation that are expected to significantly change hospitals’ community benefit spending. Our findings from the first three years do not signal significant changes underway. Fifth, we used sociodemographics as indicators of community health needs. Measures such as the prevalence of chronic diseases, negative health behaviors (e.g., smoking rates or unsafe sex practices), or indicators of the physical environment (e.g., air and water quality) may better represent needs. However, the chosen sociodemographic characteristics likely correlate with more accurate indicators of need and were universally measured and available for all ZIP codes. Finally, direct community investments as measured by Schedule H may be imprecise measures of community investments and the impact a hospital may have on the sociopolitical climate within its local area, which may significantly affect the community-level characteristics and resources within a hospital’s local area.

Nonetheless, our findings provide clues concerning how hospitals allocate funds for their local communities and these findings may have significant policy implications. Community benefit spending has not altered since the requirement to file Schedule H and a CHNA. As a result, these tax policies may have little impact on the social determinants of health. While these tax reforms were implemented as mechanisms to increase transparency regarding the allocation of resources in efforts to drive hospitals towards increasing spending within their community, based on our analysis, hospitals have not changed their spending. Though CHNAs stimulate multisector conversations about strategies to improve community health, acting on the identified needs is not required and mandating CNHAs may not have been effective in stimulating hospitals to address the needs identified.¹⁵ Requirements of CHNAs could go further by mandating implementation of evidence-based initiatives aimed at addressing the identified community needs, such as housing-first initiatives, funding for mental health resources, or improving transportation.^5,16 Finally, hospitals that benefit the community should address the evolving needs of their community. Hospitals could be held accountable for adjusting their community-directed contributions based on local needs identified by independent data collection mechanisms, such as Census data reported in the American Community Survey.

Ultimately, understanding and addressing the influence of social and economic factors on health outcomes is central as the health sector continues to face increasing pressure to address the out-of-hospital social forces affecting patient outcomes. Stronger incentives—tax-based or otherwise—are needed to steer non-profit hospitals towards greater investments in community health.