Charity Care and Community Benefit in Non-Profit Hospitals: Definition and Requirements

Abstract

Policymakers are using different ways to measure the community benefit provided by non-profit hospitals because different policy makers have different policy objectives. We compare 3 commonly used measures of community benefit; examine the correlation across the 3 measures; examine how the distribution of community benefits varies across non-profit hospitals; and compare the factors associated with the level of community benefit for each definition. The main dataset for this study is the Schedule H of IRS Form 990 data for 2017. We merged this data with the 2017 American Hospital Association (AHA), the 2017 CMS Hospital Cost Report, and the 2018 American Community Survey data. The final sample consists of 1904 non-profit hospitals. We define 3 measures of community benefit: (1) Total community benefits: combining all 17 possible measures in the 990 data; (2) Total community benefits less unreimbursed Medicaid care because it reflects a policy choice made by the state; and (3) only charity care. We also subdivided the community benefits into individual and service-based benefit. Gini Coefficients and descriptive analysis show the distribution of 3 types of community benefit measures. On average, hospitals spent 8.1% of their expenses on all community benefits; 4.3% on community benefits less unreimbursed Medicaid; and 1.7% on charity care. The provision of charity care showed more variation (Gini coefficient) than the other 2 measures. Different hospital and geographic characteristics were associated with each definition, suggesting that different types of hospitals place emphasis on different community benefits. When policy makers choose among different definitions of community benefit, they should consider what incentives they want to instill.

Keywords: IRS schedule H; charity care; community benefit; non-profit hospital; policy choice.

Figure 1.

Distribution of different community benefits in non-profit hospitals.