Analysis of the Joint Distribution of Disproportionate Share Hospital Payments

7. Exploratory Analyses Using National HCUP Data

[ Main page of Report | Contents of Report ]

Contents

• Data.
• Methods
• Results
• Conclusions

In this chapter, we report the results of our analyses using the HCUP national inpatient sample. We used the HCUP data to explore the relationship between various inpatient measures of low-income patient care that rely on inpatient utilization or gross revenue data by payer class and the implications this might have for using the different measures to allocate DSH funds.

Data

HCUP National Inpatient Sample

The HCUP national inpatient sample (NIS) is comprised of a 100 percent claims for inpatient discharges occurring during 1998 in a nationally representative sample of about 20 percent of the hospitals in 24 states. In total, there are 984 community hospitals stratified on five characteristics: ownership/control, bed size, teaching status, urban/rural location, and region. Not all states allow individual hospital identifiers to be released. We were able to link the NIS to only 640 hospitals located in states for which we have Medicare cost reports and hospital-specific information on Medicaid DSH payments. These hospitals are located in 15 states: AZ, CA, CO, CN, FL, IL, IO, MA, MD, MO, NJ, NY, UT, WA and WI. We dropped Pennsylvania hospitals because of the lack of hospital-identifiable Medicaid DSH data. We estimated what Medicare DSH payments would be for Maryland hospitals in the absence of the waiver and retained them in our simulation.

The NIS includes uniform categories for the expected primary payer:

• Medicare (both fee-for-service and managed care patients)
• Medicaid (both fee-for-service and managed care patients)
• Private insurance
• Self-pay
• No-charge (includes charity care)
• Other

The "other" payer category includes patients covered by CHAMPUS, Workmen's Compensation, Title V and other government programs. These categories were previously broken out but discontinued in the 1998 data because of problems in coding the data uniformly across states. As explained below, we use the 1997 data to develop an estimate of patients covered by local indigent care programs. Compared to the national estimates (Table 7.1), our HCUP/DSH analysis file has a lower percentage of discharges and days where Medicaid is the expected primary payer (Table 7.1). Additionally, the percentage of Medicare days is considerably higher. The statistics for the uninsured (self-pay) are quite similar to the national average.

HCRIS

We used the HCRIS data for FY1998 to develop each hospital's cost-to-charge ratio for inpatient hospital services. We also used the total margin and total margin net of DSH measures used in our evaluation of measures of financial viability (see Chapter 5).

AHA Survey

Consistent with other analyses, we used the adjusted discharges and adjusted days from the 1998 AHA survey. As previous noted, these measures include outpatient volume by adjusting the inpatient statistic by the ratio of gross patient revenues to gross inpatient revenues.

[Go To Contents]

Methods

We used the HCUP data to establish the proportion of total inpatient days, total discharges, total charges, and case-mix index (the average DRG relative weight) attributable to each payer. We estimated the Medicare SSI statistics by applying the hospital's SSI ratio to the Medicare data. Doing so assumes that the Medicare SSI patients have the same length of stay, case mix and charges as other Medicare patients. We determined each hospital's proportion of patients that were classified as "other government" in 1997 and applied the relevant ratio to the FY1998 "other" category to obtain an estimate of proportion of the hospital's patients covered by local indigent care programs. We used the data on patients covered by Medicare SSI, Medicaid, self-pay, no-charge and local indigent care programs to establish the claims-based measures of care provided to low-income patients that we discussed in Chapter 6.

We used correlation analysis to examine the relationship between key low-income patient measures that could be used in an allocation formula (e.g. proportion of days, discharges, and revenues and case-mix index including/excluding Medicare SSI patients). The degree of correlation between the measures can be used to indicate whether the choice of the measure (utilization or revenue) used to describe hospital's low-income patients is likely to have a significant effect on the distribution of funds. The HCUP data is for inpatient services only and allowed us to test only inpatient utilization and gross inpatient revenue measures. HCUP does not have the outpatient data and uncompensated care data that would allow us to evaluate measures using financial risk.

We also simulated potential DSH allocation policies and compared the results to the current distribution of DSH. We used the correlation between the DSH payment and the hospital's net income as a comparative measure of how well the payments target financially vulnerable safety net hospitals.

[Go To Contents]

Results

Low-Income Patient Shares

We examined the relationship between the proportion of low-income patients by payer to address the question of whether one of the more readily available statistics, e.g., the percentage of Medicaid days, is an appropriate proxy for a hospital's percentage of low-income days. We combined Medicare SSI and Medicaid days into a measure of "joint" days and we also computed the patient percentage used in the current Medicare DSH formula (Table 7.2). We found the percentage of Medicaid days is highly correlated with the percentage of low-income days (0.920). The correlation using "joint" days is lower (0.862). Self-pay and no-charge days are poorly correlated with either Medicaid days alone or the "joint" days. The correlation between the DSH patient percentage and the proportion of low-income days is 0.838.

We also explored the relationship between the different types of measures of the hospital's care to low-income inpatients. In Table 7.3, we report the correlations between the hospital's percentage of Medicaid and all low-income patient days, discharges, and inpatient charges. The hospital's proportion of Medicaid inpatient charges is highly correlated with its Medicaid discharges (0.955) and Medicaid days (.950). Compared to its average proportion of inpatient days (.129), Medicaid has on average a higher proportion of discharges (.147) and lower proportion of inpatient charges (.110). The large number of Medicaid maternity cases probably accounts for the differences. The same pattern is seen in the overall measures of care provided to low-income patients.

Case-mix

One objective of the DSH allocation policy is to use an indirect measure of a hospital's costs of providing care to low-income patients. When discharges are used as an allocation statistic, differences in a hospital's case mix need to be taken into account. Since data on a hospital's overall case mix are not readily available, we examined measures that might be used as a proxy such as the Medicare case mix index. We found that there is only a moderate correlation between the Medicare case mix index and the Medicaid case mix index (0.504) and the case mix index for all low-income patients (0.624). An alternative to using a case mix would be to use the proportion of days or charges as the low-income patient measure.

Comparison of Current and Alternative DSH Policies

Table 7.4 summarizes the results of our analysis of the current distribution of DSH payments to the hospitals in the HCUP file. Most of the hospital categories in the first column are self-explanatory. The grouping of hospitals by the percentage of low-income patient days is based on the proportion of low-income patient days attributable to Medicare SSI, Medicaid, local indigent care programs, self-pay and no-charge patients. The safety net hospitals category includes all hospitals with at least 20 percent of their inpatient days attributable to low-income patients. The 20% threshold derives from the 15% threshold used in the Medicare program based solely on Medicare SSI and Medicaid inpatient ratio plus 5% for uninsured patients. These hospitals are grouped by their total margin net of DSH payments.

We eliminated 8 hospitals from our simulations because we did not have margin data for them. In total, 492 of the remaining 632 hospitals received DSH payments from Medicare and/or Medicaid. The total margin net of DSH (Medicare plus the federal share of Medicaid) was 2.0% and the total margin with DSH was 5.7%. The total margin with DSH is about 1% higher than the margin we found for all hospitals in our analysis file in our Chapter 5 analysis (4.7%). In particular, the county-owned hospitals in the full analysis file had a total margin with DSH of 5.0% compared to 12.9% in the HCUP analysis file. This may be indicative of the uniqueness of each state's DSH program and the danger of using selected states to draw national conclusions about DSH payments. Safety net hospitals accounted for 45% of adjusted inpatient days and received 88% of the DSH funds. Safety net hospitals with total margins net of DSH of 5% or higher received 20% of DSH funds. DSH payments were made to 73 hospitals with less than 10% low-income patient days (which resulted from the Medicare DSH formula).

We examined alternative DSH allocation policies using different combinations of the measures discussed in Chapter 6. For illustrative purposes, we present the results from a basic simulation that allocated DSH based on the hospital's percentage of low-income days other than Medicare SSI days (Table 7.5). The allocation factor was based on the formula:

(Medicaid+ local indigent care+ self + no-charge patients) -.15 x WI x adjusted inpatient days
————————————————————————————————————
total inpatient days

Because Medicare SSI patients were not included in the allocation policy, the patient percentages used to allocate funds are not the same as those used to establish each of the low-income patient categories in the first column. This explains why 23 safety net hospitals did not receive any DSH funds under the simulation. Overall, the number of hospitals that would receive DSH funds decreases from 492 to 276.

• The percentage of DSH funds distributed to county-owned hospitals would decrease from 33% to 27 %. The average payment per adjusted day (for all county-owned hospitals) would decline from $246 to $207 but these hospitals would continue to have higher than average total margins with DSH (10.5%). Non-profit hospitals would be the primary winners ($123.5 million).
• Among teaching hospitals, those with 10-100 residents would lose $103.5 million in DSH funds. The gains are spread across the other teaching categories.
• The hospitals with at least 60 percent low-income patient days (including Medicare SSI patients) would increase their share of DSH funds from 16 to 29%.
• Hospitals defined as safety net hospitals (at least 20 percent low-income patient days) would receive 99% of the DSH funds. Most of the redistribution would benefit those hospitals with total margins net of DSH in the -5.0% to +5.0% range. However, hospitals with margins in excess of 5.0% would increase their share of DSH funds from 19 to 24%.

We examined the relationship between the alternative DSH allocation factors and net income (Table 7.6). The purpose was two-fold: to determine if there are significant differences in the DSH funds that would be allocated to safety net hospitals under the alternative polices and to determine the extent to which the policies direct funds towards safety net hospitals that are financially vulnerable. Our measure of financial vulnerability was net income per day. We used the allocation factors without a threshold for DSH payments so that we could examine the relationship between each allocation factor (e.g. % low-income patients exclusive of Medicare SSI patients x wage index) and net income per day. Ideally, we would have found a strong negative correlation between total income net of DSH and the allocation factor.

The variables used to establish the allocation factors that we report in Table 7.6 are:

• Simulation A: Proportion of inpatient days attributable to low-income patients.
• Simulation B: Proportion of inpatient days attributable to non-Medicare low-income patients.
• Simulation C: Proportion of inpatient days attributable to non-Medicare low-income days x wage index. This is the simulation reported in Table 7.5.
• Simulation D: Proportion of discharges attributable to non-Medicare low-income patients.
• Simulation E: Proportion of inpatient charges attributable to non-Medicare low-income patients.

There is no strong correlation between any of the allocation statistics and net income; however, the current Medicare and Medicaid DSH policies are more correlated with net income than any of the measures used in the alternative DSH policies (e.g. -0.48 for the joint DSH payment per day compared to -.17 for the Simulation C allocation policy). The allocation alternatives are strongly correlated with each other and moderately correlated with the current DSH policies.

[Go To Contents]

Conclusions

There are several findings from the HCUP analysis that have import in designing a DSH allocation policy. First, it appears that the patient population (e.g., with or without Medicare SSI beneficiaries) included in the allocation statistic is more important than how the care provided to those patients is quantified. Ideally, the allocation statistic would take into consideration all low-income patients. If this is not administratively feasible, using Medicaid patients only is preferable to "joint" days or the Medicare DSH patient percentage, both of which are less correlated with low-income patients.

The different measures of the amount of care provided to a low-income population (days, discharges, or charges) are highly correlated. However, the choice could have implications for certain hospitals. Those which have a high volume of Medicaid maternity cases or shorter than average length of stay (e.g. California hospitals) would benefit if discharges were used instead of days as the measure of the proportion of care provided to low-income patients.

The Medicare case mix index is not a good proxy for the hospital's low-income patient case mix. In the absence of data on the case mix of low-income patients, days or charges should be used instead of discharges as the allocation statistic.

Neither the current DSH allocation policies nor the alternatives that we examined in the analysis target DSH payments in a way that is strongly correlated with net income. This is an issue that warrants further investigation and understanding. The different Medicare formulae and the Medicaid DSH program's flexibility may provide mechanisms to target financially vulnerable hospitals in a way that a single formula-driven allocation may not. Targeting financially vulnerable safety net hospitals may require taking into consideration more factors than the amount of care a hospital provides to low-income patients.

Finally, the diversity of the Medicaid DSH programs makes it difficult to draw conclusions from an analysis of selected states. The lack of information on "new" Medicaid DSH funds further compounds the problem. Since we have data from only selected states- and not even all the hospitals in those states- we have not presented information on the redistributions of DSH payments that might occur across states under the alternative allocation policies. Nevertheless, it is important to not lose sight of the differences in DSH expenditures across states and the likelihood that a national allocation policy would result in substantial state-level redistributions as well as redistributions across classes of hospitals.

Where to?

Top of Page | Contents

Main Page of Report | Contents of Report

Home Pages:
Office of Health Policy (HP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)