The Impact of Medicaid Disproportionate Share Hospital Payment on Provision of Hospital Uncompensated Care

Abstract

This study examines the association between hospital uncompensated care (UC) and reductions in Medicaid Disproportionate Share Hospital (DSH) payments resulting from the 1997 Balanced Budget Act. Data on California hospitals from 1996 to 2003 were examined using two-stage least squares with a first-differencing model to control for potential feedback effects. Our findings suggest that not-for-profit hospitals did reduce UC provision in response to reductions in Medicaid DSH, but the response was inelastic in value. Policy makers need to continue to monitor how UC changes as sources of support for indigent care change with the Patient Protection and Affordable Care Act (PPACA).

INTRODUCTION

A growing number of individuals in the U.S. do not have health insurance. Census Bureau data indicate the number of uninsured increased dramatically from 31 million in 1987 to 50.7 million in 2009 (US Census Bureau2010). In the U.S. health system, the uninsured often rely on hospitals to provide charity care, or more broadly defined uncompensated care (Bazzoli et al. 2006; Davidoff et al. 2000; Lo Sasso and Seamster 2007). Existing studies estimated the overall amount of hospital uncompensated care costs was about $23.6 billion in 2001(Hadley and Holahan 2003) and $35 billion in 2008 (Hadley et al. 2008). In order to offset the burden from this type of care, hospitals rely on various types of public and private financial support from federal, state, local governments or private philanthropy (Fishman and Bentley 1997; Hadley et al. 2005; Hadley and Holahan 2003).

The Medicaid Disproportionate Share Hospital (DSH) Program is a major funding source for safety net hospitals and provides funding to support free and discounted care for uninsured and Medicaid patients. The Medicaid DSH program accounted for approximately 10.2% of total Medicaid expenditures in 1997 and 4.9% in 2009.¹ The program supported about 36% of total uncompensated care costs for hospitals in 2001 and about 30% in 2008 (Hadley et al. 2005; Hadley and Holahan 2003; Hadley et al. 2008). In the early 1990s, Medicaid DSH payments expanded rapidly. Medicaid DSH spending grew from less than $1 billion in 1990 to more than $17 billion in 1992. In order to limit this dramatic growth, Congress implemented several major reforms to cap the amount of DSH spending.² In 1997, the Balanced Budget Act (BBA) further limited Medicaid DSH payments by reducing state-specific federal allotments³ nationally by $10.4 billion over the 1998 and 2003 period (CBO 1997). Medicaid DSH reductions represented the major sources of federal BBA Medicaid savings, specifically accounting for 61 percent of total Medicaid gross savings over five years. After the BBA, Congress passed the Balanced Budget Refinement Act (BBRA) in 1999. This law eliminated the BBA DSH cuts for Federal Fiscal Year (FFY) 2001 and FFY 2002 and also provided relief by setting 2001 state-specific allotments at 2000 levels adjusted for inflation and setting 2002 allotments at 2001 levels adjusted for inflation. However, the Benefit and Improvement Protection Act (BIPA) of 2000 let the full BBA DSH reductions become effective in FFY 2003 (Mechanic 2004).

Our study examines how changes in Medicaid DSH payments resulting from the BBA affected hospital provision of uncompensated care in California. California is a state with a high rate of uninsured individuals and thus, historically received relatively large amounts of Medicaid DSH payments. Califorina was also substantially affected by the BBA, with its state-specific federal Medicaid DSH allotment declining from $1,085 million in 1998 to $890 million in 2003 (Federal Register 2004). We expand existing knowledge about the effects of payment policy on hospital care provided to uninsured patients. In addition, we consider policy factors other than Medicaid DSH that may affect hosptial provision of uncompensated care. The 2010 Patient Protection and Affordable Care Act (PPACA) calls for much of existing Medicaid and Medicare DSH funding to be redirected to subsidize individual health insurance purchase to expand coverage nationally (CMS 2010; Kaiser 2010a, 2010b; Katz 2010). Insights derived about the effects of past payment policy changes can provide a better understanding of the potential effects of these future policy changes on the health system.

LITERATURE REVIEW

Several studies have examined the effect of changes in different public payment policy on hospital uncompensated care provision. For example, Sheingold and Buchberger (1986) and Campell and Ahern (1993) examined changes in hospital uncompensated care provision in response to the Medicare Prospective Payment System (PPS) and concluded that the new payment system affected hospital resources available to provide uncompensated care. Regarding state policy reform, Thorpe and Phelps (1991) found that increased uncompensated care payment rates in New York State were positively associated with hospital uncompensated care provision. Dunn and Chen (1994) did not find evidence of a relationship between New Jersey State’s Diagnosis-related groups (DRGs) reimbursement reform and hospital uncompensated care, but Gaskin (1997) found that hospital uncompensated care provision was positively affected by the initiation of the New Jersey Uncompensated Care Trust Fund.

In relation to Medicaid DSH payments, Lo Sasso and Seamster (2007) examined the effect of changes in statewide Medicaid DSH spending on uncompensated care between 1990 and 2000, but did not find a relationship between the two. Davidoff et al. (2000) examined the effect of changing state Medicaid payment generosity, which in part reflected increased DSH payments to hosptials, on provision of hospital uncompensated care, and found a positive association between these measures for NFP hospitals. Bazzoli et al. (2006) examined the effects of declining Medicaid payment resulting from the BBA of 1997 and found that core safety net hospitals reduced their uncompensated care in response to this financial pressure. There are several gaps among these existing studies, which are relevant to our study. First, the unit of analysis in Lo Sasso and Seamster (2007) is the state and thus, it is unclear how individual hospitals responded to payment changes. Second, although Davidoff et al. (2000) and Bazzoli et al. (2006) examined Medicaid measures that in part reflected DSH payments, neither explicitly measured DSH payment changes in their analysis. Thus, the specific effect of changing Medicaid DSH payments on hospital uncompensated care is unclear. Our study addresses these gaps.

CONCEPTUAL FRAMEWORK

In modeling private hospital decisions about the provision of uncompensated care, researchers have typically examined non-profit (NFP) and for-profit (FP) hospitals separately. In relation to NFP hospitals, Frank and Salkever (1991) assumed an NFP hospital’s utility function depends on net revenues and the level of unmet indigent care need in the community subject to a financial break-even constraint. Their model suggests that a reduction in hospital net revenues resulting from exogenous price reductions (holding total need for indigent care constant) would lead NFP hospitals to reduce uncompensated care for indigent patients. Specifically, if Medicaid DSH subsidies decline after the BBA, NFP hospitals will adjust their activities to satisfy their break-even constraints. They may substitute care for other types of patients and thus reduce care for low-income patients (i.e., the substitution effect). Additionally, due to the reduction in DSH subsidies, hospitals may receive less financial resources to offset costs of care for low-income patients (i.e., the income effect). Given this, we hypothesize that:

H1: NFP hospitals will reduce their uncompensated care provision in response to Medicaid DSH payment reductions, all other things being equal.

For-profit (FP) hospitals are assumed to maximize their profits and undertake uncompensated care as a means to satisfy community expectations about the role of hospitals in the community. Banks, Paterson, and Wendel (1997) developed a theoretical model to explain uncompensated care provision among FP hospitals. They noted that FP hospitals are subject to community expectations of providing some indigent care and may incur costs if they do not meet these expectations. A number of mechanisms through which FP hospitals would incur these costs exist, including certificate-of-need restrictions and other regulatory sanctions (Davidoff et al. 2000). When net revenue per paying patient declines, FP hospitals may find that the cost of meeting community expectations, in terms of foregone profit, is lower and thus they increase uncompensated care provision (Banks et al. 1997; Bazzoli et al. 2006; Davidoff et al. 2000).

Although researchers have explored FP hospital’s reaction to a payment change for paying patients (i.e., Medicare or commercially insured patients) in determining the provision of uncompensated care, they did not consider FP hospital reaction to declining Medicaid DSH payments (Banks et al. 1997; Bazzoli et al. 2006; Davidoff et al. 2000). Extending the logic of Banks et al. (1997), we argue that FP hospitals will reduce uncompensated care provision as Medicaid DSH declines. Specifically, we conceptualize FP hospitals as using Medicaid DSH payments to offset the costs of uncompensated care. When DSH payments decline, FP hospitals experience higher residual costs they must absorb if they maintain the same level of uncompensated care and this in turn would lower their profits, ceteris paribus. Given this, we expect FP hospitals to reduce their supply of uncompensated care in the face of declining DSH payments. Therefore, we hypothesize that:

H2: FP hospitals will reduce their uncompensated care provision in response to Medicaid DSH payment reductions, all other things being equal.

METHODOLOGY

Data and Sample

Longitudinal data for California hospitals from 1996 to 2003 were studied. Study data were drawn from several databases. First, annual hospital financial data from the Office of Statewide Health Planning and Development (OSHPD) in California were used. This dataset includes state audited financial statements for all California hospitals. Second, the American Hospital Association (AHA) Annual Survey provides hospital structural data. Third, the Area Resource File (ARF) provides extensive information on community demographics, and socioeconomic attributes at the county level. Fourth, HealthLeaders-InterStudy provides data on HMO enrollment at the MSA level. Fifth, Medi-Cal annual statistical reports provide statistical data on the average number of Medi-Cal eligible individuals per month at the county level. Sixth, Medi-Cal Managed Care Annual Statistical Reports provide information about Medicaid managed care programs. Seventh, the Medicare hospital cost reports provide data on hospital Medicare DSH payments.

We included all short-term, non-federal general acute care hospitals in California. Kaiser hospitals, which constitute 25 hospitals in each study year, are excluded because they do not report data to the state. In addition, we excluded California hospitals that experienced ownership changes (i.e., hospital closure or ownership conversion) during the study years because these changes could affect hospital charity care missions and the provision of uncompensated care. About 8% of hospitals were excluded due to this restriction. There were a total of 2,287 hospital-year observations in California, representing 318 hospitals that reported data at least one year. California hospitals that provide high volume of care to low-income and uninsured patients are eligible for receiving Medicaid DSH payments and were defined as DSH hospitals. Between 1996 and 2003, 130 hospitals received Medicaid DSH payments in one or more years, and about 52 percent of those hospitals received Medicaid DSH continuously for more than six years.

Variable Definition

Hospital Uncompensated Care Provision

We measured hospital uncompensated care provision as total uncompensated care charges divided by the hospital’s average charge per admission, which is then divided by the number of hospital beds. We called this variable: uncompensated care admissions per bed. This variable is similar to the one developed by Banks et al. (1997), Gaskin (1997) and Hsieh et al. (2010). By standardizing this measure relative to hospital admissions and hospital beds, the provision of uncompensated care across different sized hospitals and multiple years can be compared.

Medicaid DSH Payments

Many researchers have noted that the Medicaid Disproportionate Share Hospital payment program is a complicated financing system (Coughlin et al. 1994; Coughlin and Liska 1997; Mechanic 2004). It is necessary to know how the financing mechanism works in order to construct accurate measures of the Medicaid DSH payments that individual hospitals received.

The majority of state governments use intergovernmental transfers to generate matching federal funds for DSH payments. In the California Medicaid DSH program (also called the SB855 DSH program), the state collects funds through intergovernmental transfers (IGTs) from public hospitals. Then, the federal government matches the state’s funds at the California federal Medicaid matching rate of 51.55%. After that, the state retains a portion of the combined funds in its state General Fund and distributes the other portion of the combined funds to both public and private hospitals (Coughlin et al. 1994; Coughlin and Liska 1997; McCue and Draper 2004; Mechanic 2004). A hospital is eligible for a Medicaid DSH payment when it meets one of the two criteria: (1) the hospital’s number of Medi-Cal inpatient days must be at least one standard deviation above the statewide mean; (2) the hospital’s revenues from low-income patient utilization (including Medi-Cal and uncompensated care) must account for 25% or more of its total revenues (Fonkych and Melnick 2010).

McCue and Draper (2004) and Baicker and Staiger (2005) have argued that the Medicaid DSH payments received by public hospitals should be reduced by the portion of intergovernmental transfers used to obtain matching federal funds. Thus, it is necessary to calculate net Medicaid DSH as the amount of Medicaid DSH payments that a hospital received less the amount of funds collected by the state through IGTs. Similar to uncompensated care provision, we rescaled the net Medicaid DSH payment divided by the hospital’s average charge per admission and then divided by the number of hospital beds. In this way, we can have the same scale between uncompensated care admissions per bed and the net Medicaid DSH variable.

Other Governmental Subsidies

To assess the impact of the decline in Medicaid DSH payments on uncompensated care, it is important to control for other governmental subsidies that hospitals receive to offset the unreimbursed costs of care for Medicaid and uninsured patients. Specifically, a change in Medicaid DSH, even it looks large, may not have much effect on hospital uncompensated care because Medicaid DSH may be relatively small in magnitude relative to other financial support received by the hospital. The variable of other governmental subsidies consisted of two components. The first one was the state and local governmental subsides, which national studies indicate covered approximately 31% of total uncompensated care costs in 2001 and 2008 (Hadley and Holahan 2003; Hadley et al. 2008). This type of funding is highly variable, typically based on available state and local budgets. In California, state, county, and district hospitals most commonly receive this governmental funding. The second component was Medicare DSH payments, which national studies indicate covered approximately 20% of total uncompensated care costs in 2001 and 2008 (Hadley and Holahan 2003; Hadley et al. 2008). The amount of Medicare DSH that a hospital received is determined by the proportion of all Medicare days that are attributable to beneficiaries of Supplemental Security Income (SSI) and the proportion of all patient days for which Medicaid is the primary payer. We measured other governmental support as the sum of state and county tax appropriations, restricted donations and subsidies for indigent care, subsidies for district hospitals and Medicare DSH payments. This sum was divided by the hospital’s average charge per admission and then divided by the number of hospital beds so that the scaling for the variable was also the same as the uncompensated care dependent variable.

Market Characteristics

Several market factors were also controlled. We used data from 1996-2003 Medi-Cal Annual Statistical Reports to construct Medicaid eligibility variable as the ratio of the number of average monthly Medi-Cal eligible individuals to the total population at the county level.⁴ Medicaid managed care is measured as the ratio of the number of Medicaid managed care enrollees to the total population at the county level. This latter variable is based on 1996-2003 Medi-Cal Managed Care Annual Statistical Reports.⁵

A Herfindahl-Hirschman Index (HHI) is calculated on the basis of hospital admissions to measure market competition at the county level. In calculating HHI, hospital admissions for those hospitals in the same multi-hospital system within a county were combined and treated as if the system were one organization. The prevalence of public hospital beds in a county is measured as the percentage of the sum of county and district hospital beds to total hospitals beds. Likewise, the presence of teaching hospitals is measured as the percentage of teaching hospital beds to total hospital beds within the county. The reason for including these two variables is because public hospitals have obligations locally to provide a broad range of care for all types of patients and teaching hospitals may need to provide a full range of services to train new physicians. Thus, hospitals located in a community with relatively more public hospital or teaching hospital beds may have less need to provide unprofitable services (Bazzoli et al. 2006; Davidoff et al. 2000).

The final market measures are related to demand for uncompensated care. Ideally, it is best to measure uninsured demand by using the number of uninsured and low-income individuals in the county. However, there are no publicly available data that allow measuring this variable over time.⁶ Therefore, we used median household income and unemployment rate for each county as proxy variables to capture uninsured demand. These demographic data were from the Area Resource File for 1996 to 2003.

Hospital Characteristics

Several hospital characteristics are also included. Hospital ownership type consists of three dummy variables to identify for-profit hospitals, county hospitals, and district hospitals (not-for-profit hospital is the omitted category). Ownership data are from OSHPD hospital annual financial data. We included a system affiliation variable that identified whether a hospital is a member of a multi-hospital system. One dummy variable is used to identify hospital teaching status. Hospital nurse labor force was measured as the sum of the number of full time equivalent (FTE) registered nurses, divided by total staffed hospital beds. Medicare share of inpatient days was also used as a control variable. These hospital characteristics were constructed from the AHA annual survey. Additionally, annual dummy variables were included to identify the study years.

Empirical Model

The basic reduced form empirical model is presented in equation (1):

$$U{C}_{\text{it}}={\delta }_0+{\delta }_{1}DS{H}_{\text{it}}+{\delta }_2{Z}_{\text{it}}+{\delta }_3{M}_{\text{it}}+{\delta }_4{H}_{\text{it}}+{\delta }_5(F{P}_{\text{it}}\times DS{H}_{\text{it}})+{\delta }_6(CNT{Y}_{\text{it}}\times DS{H}_{\text{it}})+{\delta }_7(DISTRI{C}_{\text{it}}\times DS{H}_{\text{it}}){\delta }_{t}Yr+{\tau }_i+{\nu }_{\text{it}}$$

where i= an individual hospital; t=year; UC_it is the provision of hospital uncompensated care for hospital i in year t; DSH_it represents net Medicaid DSH payments; Z_it is a variable for other governmental financial support for indigent care (including Medicare DSH and state and local governmental supports). M_it is a vector of market characteristics. H_it is a vector of hospital characteristics, including teaching status, multi-hospital system status, and Medicare share. These variables can vary from year-to-year because hospitals may change their involvement in teaching new physicians, may be acquired or dropped by multi-hospital systems, and may experience fluctuations in the number of Medicare patients they treat. Yr are year dummy variables. FP_it is a dummy variable for for-profit hospitals. CNTY_it is a dummy variable for county hospitals. DISTRIC_it is a dummy variable for district hospitals (not-for-profit hospitals are the omitted category). Because we are interested in the effect of Medicaid DSH payments (DSH_it) among different ownerships, we interacted ownership dummy variables and DSH payment (DSH_it) in the model. τ_i is a hospital-specific error component. ν_it is a random error term.

This empirical model allows us to test study Hypotheses 1 and 2. The coefficient estimate for DSH (i.e., Δ₁) captures the effect of changes in Medicaid DSH payments on NFP hospital provision of uncompensated care. The coefficient signs are expected to be positive and significant because NFP hospitals may reduce the provision of uncompensated care provided to low-income patients in response to reductions in Medicaid DSH payment. In addition, as proposed in Hypothesis 2, FP hospitals may reduce uncompensated care provision when Medicaid DSH payments decline. Thus, the total effect of DSH payments for FP hospitals, which is the sum of the coefficients Δ₁ and Δ₅, is expected positive.

Analytical Strategy

Our estimation approach takes advantage of eight-year panel to control for unobserved time invariant hospital and market characteristics. The empirical model was estimated with two-stage least square (2SLS) and a first differencing transformation. Two-stage least squares was used because there is a potential “feedback effect” or sequential exogeneity" between hospital uncompensated care provision and Medicaid DSH payment. Specifically, the Medicaid DSH payment that a hospital receives in time t depends on low-income patient utilization lagged one year. We used a first differencing transformation to both remove the hospital-specific error component τ_i and to correct for arbitrary autocorrelation because this transformation will make highly persistent time process become weakly dependent (Woolridge 2006: pp.397-398). We then used the lagged-difference in net Medicaid DSH variable (ΔDSH_i,t-1) as an instrumental variable for the first-differenced regressor of Medicaid DSH effect (ΔDSH_i,t), following the approach described by Wooldridge (2002). Similarly, we used the lagged-difference variable in the interaction term between ownership (e.g. for-profit binary variable) and net Medicaid DSH variable as an instrumental variable. We also used lagged-differenced in other governmental subsidies (ΔZ_i,t-1) as an instrumental variable for the first-differenced regressor of other governmental support (ΔZ_i,t). A Durbin-Wu-Hausman test was conducted to test for endogeneity. The results suggested that estimators in the 2SLS model were consistent. In addition, we implemented first-stage diagnostic test for the relevance of instruments, and obtained a Cragg-Donald Wald F statistic of 11.88, which indicates that the instrumental variable does not appear to be weak.

In addition, the use of hospital-specific data may be a source of heteroskedasticity. To account for unequal error variances across hospitals, we used a heteroskedasticity-robust standard error adjustment. Furthermore, we recognized that intra-county variations in county-level indicators for the Medicaid program or demand characteristics and intra-hospital variations in Medicaid DSH payments may exist that could bias downward estimated standard errors (Davidoff et al. 2000) As such, the estimated variance-covariance matrix for estimated coefficients was adjusted with the Huber-White correction by using two-way clustering approach developed by Cameron, Gelbach and Miller (2006).⁷

RESULTS

Table 1 provides descriptive data on uncompensated care admissions and uncompensated care admissions per bed for two hospital groups (DSH hospitals and non-DSH hospitals) and for all hospitals across the study years. Looking first at DSH and non-DSH hospitals, DSH hospitals provide about twice the annual uncompensated care admissions and uncompensated care admissions per hospital bed when compared to non-DSH hospitals. For DSH hospitals overall, annual uncompensated care admissions increased 91% from 851 average admissions in 1996 to 1,630 in 2003. When we adjust for hospital beds, the annual uncompensated care admissions per hospital bed increased over 122% from 3.51 in 1996 to 7.78 in 2003. Non-DSH hospitals also experienced an increase in annual uncompensated care admissions and also uncompensated care admissions per hospital bed.

Table 1.

Descriptive statistics for Dependent Variable, 1996-2003

		DSH Hospitals╩			Non-DSH Hospitals			All Hospitals
Variable	Year	N	Mean	SD	N	Mean	SD	N╬	Mean	SD
Uncompensated Care Admissions	1996	68	851.88	1,040.72	216	469.89	480.78	284	561.35	677.27
	1997	75	828.72	1,079.09	215	495.91	527.17	290	581.98	724.74
	1998	75	1,061.15	1,296.13	214	546.88	613.90	289	680.34	872.58
	1999	77	1,210.30	1,395.83	189	552.74	549.32	266	743.09	928.50
	2000	84	1,400.30	1,474.87	198	684.75	779.96	282	897.89	1,084.65
	2001	96	1,258.92	1,311.71	197	686.19	668.44	293	873.84	965.51
	2002	99	1,309.40	1,475.01	194	744.24	744.66	293	935.20	1,080.93
	2003	101	1,630.51	1,875.70	189	897.87	947.02	290	1,153.03	1,386.75
	1996-2003 Average		1,223.19	1,432.69		629.32	686.18		804.60	1,005.19
Uncompensated Care Admissions per Hospital Bed	1996	68	3.51	3.12	216	2.63	1.91	284	2.84	2.28
	1997	75	3.58	3.35	215	2.96	3.04	290	3.18	3.37
	1998	75	4.54	3.38	214	3.18	3.01	289	3.57	3.20
	1999	77	5.33	4.14	189	3.10	2.42	266	3.81	3.35
	2000	84	6.07	4.98	198	3.65	2.89	282	4.37	3.77
	2001	96	5.60	4.38	197	3.58	2.76	293	4.25	3.46
	2002	99	5.93	4.83	194	4.07	3.25	293	4.68	3.92
	2003	101	7.78	7.98	189	4.87	4.40	290	5.83	5.97
	1996-2003 Average		5.45	5.07		3.48	3.08		4.06	3.88

^╩Note: DSH hospitals are defined as those hospitals that received Medicaid DSH payments.

^╬Those hospitals experienced any ownership change or closures during study period are excluded in the study hospital sample.

In contrast to the trend of increased annual uncompensated care admissions per bed, we observed decreased net Medicaid DSH financial support when measured relative to admissions and hospital beds among DSH hospitals. The mean value of net Medicaid DSH (scaled) trended downward beginning in 1999, as illustrated in Figure 1.

Figure1.

Trend for the Mean Value of Uncompensated Care Admission per Bed and Net Medicaid DSH payment (Scaled) among DSH Hospitals, from 1996 to 2003

Note: Net Medicaid DSH (scaled) equals Medicaid DSH payments less IGTs divided by hospital average charges per admission and then divided by hospital beds. This adjustment was used so that the DSH data are scaled in a similar manner to the uncompensated care data.

Table 2 provides descriptive statistics on study variables. These statistics combine data across eight years from 1996 to 2003 for DSH hospitals, non-DSH hospitals, and hospitals overall. Among hospitals that received Medicaid DSH payments in any of the study years, about 12% were district hospitals, 21% were county hospitals, 37% were NFP hospitals and 30% were FP hospitals. The data also suggest that hospitals receiving Medicaid DSH payments tend to receive greater amounts of other types of governmental financial support (p<0.01), have larger bed size(p<0.01), and are more involved in teaching than are hospitals that do not receive DSH (p<0.01). DSH hospitals also tend to be located in counties with lower household income (p <0.01), greater unemployment rate (p<0.01), higher ratios of Medicaid eligible to total population (p <0.01), higher Medicaid managed care enrollees to total population (p<0.01), and greater presence of public (p<0.01) and teaching hospitals (p<0.01).

Table 2.

Descriptive Statistics for Explanatory Variables by DSH Hospital Type, 1996-2003

	All Hospitals		DSH Hospitals		Non-DSH Hospitals		P-value
Variables	Mean	SD	Mean	SD	Mean	SD
Governmental Financial Supports
Net Medicaid DSH payments (Scaled)	0.84	2.49	2.86	3.91	-	-	-
Other Governmental Supports (Scaled)	1.77	0.06	3.53	0.16	1.03	0.03	0.000***
State and Local Governmental Subsidies (Scaled)	0.54	2.31	1.55	4.00	0.12	0.53	0.000***
Medicare DSH subsidies (Scaled)	1.23	0.03	1.98	0.08	0.91	0.03	0.000***
Market Characteristics (at county level)
Ratio of Medicaid eligible to total population	16.44	5.77	17.84	5.58	15.86	5.74	0.000***
Ratio of Medicaid managed care enrollees to total population	6.39	5.20	7.48	5.25	5.93	5.11	0.000***
Ratio of private HMO enrollees to total population at MSA Level	41.60	19.84	42.69	18.66	41.14	20.30	0.087*
Herfindahl-Hirschman Index	0.30	0.26	0.24	0.23	0.32	0.27	0.000***
Presence of public hospitals in county (%)	18.97	21.24	20.90	22.83	18.16	20.49	0.005***
Presence of teaching hospitals in county (%)	28.33	21.74	31.98	20.19	26.80	22.19	0.000***
Median household income (in 1000s)	37.62	7.70	36.48	7.38	38.10	7.79	0.000***
Unemployment rate (%)	6.78	3.44	7.15	3.67	6.63	3.33	0.001***
Hospital Characteristics
FTE RNs to staffed bed	1.35	0.62	1.22	0.58	1.40	0.63	0.000***
Hospital staffed bed size	195.42	155.84	213.16	162.75	187.99	152.29	0.004***
Hospital Medicare share (%)	42.39	15.29	34.72	16.01	45.60	13.77	0.000***
Proportion system affiliated (%)╠	0.64	0.48	0.57	0.50	0.67	0.47	0.000***
Proportion teaching hospital (%)╠	0.17	0.38	0.31	0.46	0.12	0.32	0.000***
Proportion not-for-profit hospital (%)╠	0.55	0.50	0.37	0.48	0.62	0.48	0.000***
Proportion for-profit hospital (%)╠	0.25	0.43	0.30	0.46	0.23	0.42	0.001**
Proportion county hospital (%)╠	0.07	0.25	0.21	0.41	0.01	0.07	0.000***
Proportion district hospital (%)╠	0.13	0.34	0.12	0.33	0.14	0.34	0.245
N (Study Hospital Observations, 1996-2003)	2,287		675		1,653

¹Note: DSH hospitals are defined as those hospitals that received Medicaid DSH payments.

²Net Medicaid DSH payments (scaled) equals Medicaid DSH payments less IGTs divided by hospital average charges per admission and then divided by hospital beds. Similarly, other governmental support (scaled) equals the sum of other financial supports received by the hospital divided by hospital average charges per admission and then divided by hospital beds. This adjustment was used so that the Medicaid DSH and other governmental support data are scaled in a similar manner to the uncompensated care data.

³P-value presented here is to compare DSH and non-DSH hospitals for explanatory variables. T-test was used to compare DSH and nonDSH hospitals when explanatory variable is continuous variable; Chi-square test was used for categorical variables with superscript ╠.

^***p<0.01

^**p<0.05

^*plt;0.1

Table 3 reports the detailed results from the 2SLS with first-differencing estimation. Based on the results in Table 3, we further calculated Medicaid DSH total effects on hospitals by ownership types as presented in Table 4. Looking first at the net Medicaid DSH (scaled) variable in Table 3, the coefficient reflects the impact of these payments on the behavior of NFP hospitals, a positive and significant association (p<0.01) was found between net Medicaid DSH payments and hospital uncompensated care admissions per hospital bed. These results suggest that NFP hospitals significantly reduced their uncompensated care provision when reductions in Medicaid DSH payments occurred during the study years. Other things being equal, the marginal effect of the net Medicaid DSH (scaled) variable indicates that a one unit decrease in the amount of DSH payment is associated with decreased uncompensated care admissions per hospital bed by 0.944 (p<0.01) by NFP hospitals, which supports our hypothesis H1. To examine hypothesis 2, we summed the main coefficient for DSH in the model plus its interaction with hospital for-profit status as shown in Table 4. The total effect of a one unit change in the Medicaid DSH variable for for-profit hospitals was not significant, suggesting that for-profit hospitals did not substantially change uncompensated care provision as DSH declined. We further examined the Medicaid DSH total effects on uncompensated care provision among county and district hospitals. The coefficients did not show significant effect. This may be because county and district DSH hospitals have other sources of support to offset their uncompensated care costs or because the number of these facilities in our sample is small and thus power may be an issue.

Table 3.

Two Stage Least Square Regression Result: First Differencing with Instrumental Variable Estimators

VARIABLES: Uncompensated care Admission per Hospital Bed	Coefficient	StdErr	p-value
Net Medicaid DSH Payment (Scaled)	0.944	(0.304)	0.002***
Interaction Term of For-Profit Hospital and Net Medicaid DSH (Scaled)	−0.541	(0.524)	0.302
Interaction term of county hospital and Net Medicaid DSH (Scaled)	−0.936	(0.387)	0.016*
Interaction term of district hospital and Net Medicaid DSH (Scaled)	−3.621	(1.923)	0.060*
Other Governmental Subsidies (Scaled)	0.250	(0.187)	0.182
Ratio of Medicaid Eligibles to Total Population	0.082	(0.072)	0.255
Ratio of Medicaid Managed Care Enrollees to Total Population	0.040	(0.033)	0.221
HMO Penetration at MSA level (%)	−0.039	(0.020)	0.047**
Herfindahl-Hirschman Index	−1.265	(0.677)	0.061*
Presence of Public Hospitals in County (%)	−0.008	(0.012)	0.538
Presence of Teaching Hospitals in County (%)	−0.013	(0.011)	0.247
Median Household Income (in 1000s)	0.076	(0.054)	0.157
Unemployment Rate (%)	0.041	(0.074)	0.583
FTE RNs to Hospital Staffed Bed	0.919	(0.162)	0.000***
System Affiliated	−0.722	(0.376)	0.055*
Teaching Hospital	1.065	(0.353)	0.003***
Hospital Medicare Share (%)	−0.005	(0.011)	0.664
Hospital Medicare Share (%) at Post BBA Period	0.006	(0.010)	0.514
Year 1999	0.276	(0.371)	0.456
Year 2000	0.724	(0.763)	0.342
Year 2001	0.747	(1.341)	0.577
Year 2002	0.952	(1.883)	0.613
Year 2003	1.879	(2.184)	0.389
Constant	0.108	(0.455)	0.812
Hospital Year Observations	1,596
R-squared	0.12

Note: Robust standard errors in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Net Medicaid DSH payments (scaled) equals Medicaid DSH payments less IGTs divided by hospital average charges per admission and then divided by hospital beds. Similarly, other governmental support (scaled) equals the sum of other financial supports received by the hospital divided by hospital average charges per admission and then divided by hospital beds. This adjustment was used so that the Medicaid DSH and other governmental support data are scaled in a similar manner to the uncompensated care data.

Table 4.

Summary of Medicaid DSH Total Effects on Hospitals by Ownership Types

	Coefficient	StdErr	P-Value
Medicaid DSH total effect on NFP hospitals (δ1)	0.944	(0.304)	0.002***
Medicaid DSH total effect on FP hospitals (δ1+δ5)	0.403	(0.370)	0.277
Medicaid DSH total effect on County hospitals (δ1+δ6)	0.008	(0.185)	0.966
Medicaid DSH total effect on District hospitals (δ1+δ7)	−2.677	(1.859)	0.150

Note: Robust standard errors in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

In addition, the findings indicate that hospitals located in markets with growing HMO market penetration (p<0.05) and growing market concentration as measured by the Herfindahl-Hirschman Index (p<0.1) had declining uncompensated care admissions per hospital beds. With respect to other hospital characteristics, the results indicate that hospitals that have increasing numbers of nurse staffing to bed provided more uncompensated care (p<0.01). Hospitals that became affiliated with a multi-hospital system experienced a reduction in uncompensated care (p<0.1). Hospitals that became teaching hospitals significantly increased their annual uncompensated care admissions per hospital bed (p<0.01).

As with any study, several limitations of the analysis must be acknowledged. First, due to the lack of comparable data on key study variables in other states, we only examined hospitals in California. The results may not be generalizable to other states. Second, we conducted a pre-and-post design that covered the study period between 1996 and 2003 in order to capture the specific impact of Medicaid DSH payment cuts resulting from the 1997 Balanced Budget Act. However, other policy reforms (i.e., the Omnibus Budget Reconciliation Act of 1993, which was effective in 1995) that capped the expansion of Medicaid DSH expenditures may have had some residual historical effects on hospital outcomes for early study years (i.e., 1996 and 1997).² Third, our analytical approach distinguishes between hospitals receiving DSH and those that do not, with the assumption that the former are primary providers of hospital safety net care in a community. Although several researchers have defined safety net hospitals using criteria similar to that used in California to identify DSH recipients (Bazzoli et al. 2012; Dorn and Buettgens 2010; Gaskin, Hadley, and Freeman 2001; Hadley and Cunningham 2004), we recognize that this approach may not correctly identify all safety net institutions serving a community. Fourth, our analysis excluded hospitals that closed during the study period and if DSH payments reductions were a primary factor in these closures, we may be underestimating the effects of DSH reductions on uncompensated care. Fifth, following work of Newhouse (1970), Frank and Salkever (1991), and Hoerger (1991), we assumed that a hospital funds care that is uncompensated in time period t with the pool of resources available to it in time period t. To some degree, though, current provision of uncompensated care may be affected by hospital consideration of what future DSH resources it will derive, especially for those hospitals that perceive they are near the threshold for receiving DSH in a subsequent year based on the California criteria for DSH determination. This is an important consideration that might best be explored in future research using a mixed method qualitative and empirical approach.

SUMMARY AND DISCUSSION

In response to the growing number of uninsured, policy makers have sought to reduce the number of uninsured individuals by either expanding public insurance coverage or to subsidize the cost of uncompensated care for health care providers (Weissman 2005). Medicaid DSH payments are one of the major funds that traditionally have supported health care providers offsetting a large portion of their costs for providing care to uninsured patients. However, these financial subsidies were reduced as a result of the BBA of 1997. As Figure 1 shows for California hospitals, there was a growing trend of uncompensated care among DSH hospitals between 1996 and 2003. However, the financial support from Medicaid DSH payments declined over the years and the gap between this support and hospital financial burden for uncompensated care has become larger. Hospital industry leaders and policy makers have concerns that hospitals may reduce medical care provided to uninsured patients when faced with reductions in Medicaid DSH payments. To address this concern, we examined the association between changes in Medicaid DSH payments resulting from the BBA policy changes and hospital uncompensated care provision.

This study applied economic theory of hospital behavior as a framework to examine the association between the hospital provision of uncompensated care and Medicaid DSH payments. Data for California hospitals from 1996 to 2003 were examined. The study findings suggest that not-for-profit hospitals reduced their provision of uncompensated care in response to reductions in Medicaid DSH payments. To put the magnitude of these responses into perspective, it is worthwhile to examine the elasticity of hospital responses, namely the percent change in uncompensated care provision associated with a percent change in the net Medicaid DSH variable. We calculated these using mean values for NFP hospitals and the marginal effect estimates based on the results in Table 4. For NFP hospitals, our study findings imply that a one unit change in the scaled net Medicaid DSH variable equaled 32.9% reduction in net Medicaid DSH (i.e., 1 divided by 3.04) and this was associated with a 15.47% (i.e., 1 divided by 6.1 and then multiplied by 0.944) decline in hospital uncompensated care. This implies that the elasticity of UC provision to a net Medicaid DSH change was 0.47.⁸ Overall, for NFP hospitals, these results suggest an inelastic but still sizable response to declining Medicaid DSH.

With respect to the recently enacted U.S. health care reform law – the Patient Protection and Affordable Care Act (PPACA) of 2010, the law intends to redirect a substantial amount of Medicaid DSH payments to provide subsidies to aid individual purchase of health insurance (Berenson et al. 2009; Hall 2011; Katz 2010). Specifically, under PPACA, federal allotments for Medicaid Disproportionate Share Hospital (DSH) payments will be reduced by $18.1 billion for 2014-2020 (CMS 2010; Kaiser 2010a, 2010b). Although safety net hospitals will have fewer uninsured patients, PPACA will not eliminate all uninsured and all uncompensated care. In particular, CMS (2010) estimates that 23 million individuals will remain uninsured after full enactment of PPACA, and Holohan and Garrett (2010) estimated that the costs of hospital uncompensated care will likely equal $46.6 billion in 2019. The net effect of reduced uncompensated care burden due to fewer uninsured and reduced DSH dollar support is unclear, especially because about 52% of the newly covered will be Medicaid beneficiaries (CMS 2010), and Medicaid traditionally pays hospitals less than their costs of care (Katz 2010; Ku et al. 2011).

Existing evidence from the State of Massachusetts, which implemented similar health reform to PPACA, suggests that safety net hospitals continue to serve an important role in providing care for the uninsured and underinsured, and in fact, that patients’ demand for care at these institutions remains strong. For example, the National Association of Public Hospitals (2009) reported that the two safety net hospital systems in Massachusetts cared for a growing number of low-income individuals after state-wide health reform was implemented, with the majority of the increase attributable to Medicaid patients. Ku et al. (2011) also found a similar effect that the demand for care at safety net facilities continues to rise in Massachusetts that occurred with health care reform. In light of this, federal and state officials need to carefully monitor how care for the residual uninsured changes after national health reform is implemented and whether residual DSH funding is appropriately directed to those facilities providing the bulk of this care so that the financial condition of these hospitals is not adversely affected and they can continue to provide needed access to services for the remaining uninsured and growing underinsured.