Data Sources Used to Estimate Medicare Payments
We drew on several data sources to estimate Medicare DSH payments: the Provider of Service (POS) file, the PPS Impact files for FY 1998 and FY2000, and the Provider-specific File (PSF) and HCRIS files. Our starting point for a hospital listing was the CMS Provider of Service (POS) file. According to this file, there were 8,868 providers classified as "hospital" in the United States in FY1998. Of this total, we identified as our initial sample approximately 6,200 acute care hospitals that could be theoretically eligible to receive Medicare DSH payments. We used the information in the PPS Impact files and the PSF files to simulate DSH payments to these hospitals.
PPS Impact Files
The PPS impact files that CMS produces each year as part of the annual update in the hospital prospective payment system contain information that can be used to estimate each hospital's payments for the upcoming federal fiscal year. The variables include current payment parameters (e.g. the applicable wage index for the upcoming fiscal year), hospital characteristics from the most recently filed cost report, and patient characteristics from Medicare claims data from two years prior (e.g. the FY2000 impact file includes data on FY1998 claims received through March 31, 1999). We used a combination of PPS impact files to develop our estimate. That is, the impact file for FY1998 provides the best information on the payment parameters in effect for that fiscal year. The impact file for FY2000 provides the best information on patient characteristics, including SSI percentage, for patients discharged during FY1998. It also provides the best match for the DSH patient percentage that would have been applicable during FY1998.
We used several payment variables from the PSF (and in some case HCRIS) that are not available on the PPS impact file to improve the estimation for capital-related costs. These elements are:
- whether the hospital is paid under the fully prospective methodology or the hold-harmless methodology for capital-related costs. For urban hospitals with 100 or more beds, this information is relevant to determining the federal rate payment eligible for the DSH add-on.
- new capital payment ratio. For hospitals paid under the hold-harmless methodology for old capital costs, the ratio determines the proportion of the federal rate that is payable for new capital.
Summary of Data Sources
We summarize the data sources used to estimate Medicare DSH payments in Table A.1.
|Payment Variable||FY 1998|
|Standard payment rate||FY1998 FR- Table 1|
|MSA for standard payment rate||FY1998 impact|
|Wage indices||Wage index history|
|Cost of living adjustment||FY1998 impact|
|Number of discharges(1)||FY2000 impact|
|Case-mix index 1||FY2000 FR-Table 3C|
|DSH patient percentage(2)||FY2000 impact|
|DSH operating adjustment factor||FY1998 rules|
|DSH capital adjustment factor||FY1998 rules|
|Sole community HSP rate||FY1998 impact|
|Provider type||FY1998 impact|
|Capital federal rate percentage||HCRIS or PSF|
|Eligibility for temporary relief(3)||HCRIS or PSF|
We used these data to determine both Medicare DSH payments and other payments under the prospective payment system (e.g., indirect teaching and outlier) that, while not needed to estimate DSH payments, are need to define classes of hospitals and establish total Medicare payments.
Methodology for Estimating Hospital-specific Medicaid DSH Payments
There are several problems with the Medicaid DSH reports that make it difficult to use them effectively. Although the BBA requires that states submit these reports, compliance with this requirement appears lax and federal enforcement is limited. As a result, CMS does not have a complete set of reports for any fiscal year. The reports that have been submitted contain varying levels of information because CMS gives states considerable latitude in completing them. In a Federal Register Notice dated October 8, 1998, CMS recommended that states file reports that include the name of hospital, type of hospital, ownership (e.g., public or private) and annual payment. However, when states submit reports, they often do not include this minimal level of information. Among the most significant problems are lack of information to adequately identify hospitals, unreliable or missing identification by hospital type, inconsistent identification of mental health or psychiatric DSH payments, and differing payment amounts compared to reporting on Form HCFA-64.
One problem for researchers that want to use the DSH reports is that the only identifier that most states provide for hospitals are their names. In many cases, they use abbreviations or initials. Some reports contain duplicate names with no other identifiers. Others do not clearly identify out-of-state hospitals. Some states include Medicaid numbers or other numerical identification; others provide no numerical identifiers. Poor identification of hospitals makes it difficult to properly group them into hospital types or to conduct more extensive analyses by linking these data with other hospital specific data.
A second problem is that DSH expenditures reported on Form HCFA-64 and the state reports often do not match. In some cases, these differences are attributable to inconsistent time frames--Form HCFA-64 data represent federal fiscal years and several states submitted reports for state fiscal years or calendar years. In other cases, states reported only the state share of DSH expenditures. There are also several instances in which total DSH expenditures reported on Form HCFA-64 differ from those reported on hospital specific reports for the same time period. For example, Pennsylvania's hospital specific report for FY 1998 showed a total of $41 million in DSH payments whereas the state claimed $546 million in DSH payments on Form HCFA-64. Without standardized reporting by federal fiscal year and method of accounting between the hospital specific reports and Form HCFA-64, it is difficult to determine the accuracy of the information reported on the hospital specific reports.
A third problem is that information concerning ownership and type of hospital is not uniform and is sometimes inconsistent with other sources. Many states do not identify whether hospitals are state-owned, other public, or private. When reported, this information is sometimes at odds with other available sources. For example, New Hampshire's DSH reports identify several hospitals as public that the American Hospital Association (AHA) indicates are private, non-profit. Information on the type of hospital (e.g., acute, children's, psychiatric, rehabilitation) is also often missing or inconsistent. For example, several hospitals in Iowa are identified as both acute and psychiatric in the Medicaid DSH report.
A fourth problem is that identification of DSH payments to psychiatric facilities is not consistent. For example, California, Iowa, Mississippi, and New Mexico all had limits of $0 on DSH payments to institutions for mental diseases (IMD) in FY 1998, yet all four states report payments to hospitals that are identified as psychiatric facilities by the state's DSH report, the AHA, or Medicare files. Several states' reported expenditures to psychiatric hospitals from hospital specific reports are also not consistent with what they report on Form HCFA-64 as IMD DSH payments.
CMS staff have compiled states' DSH reports into spreadsheets that include each hospital's name, the total annual DSH payments paid to that hospital by the state, and information about the hospital including ownership (public or private) and type (acute, children's, teaching, or psychiatric). We combined the data for each of the states into a single spreadsheet for the nation. To be consistent with the Medicare DSH estimates, and to take advantage of the largest group of available Medicaid reports, we used FY 1998 as the benchmark year.
In order to merge the Medicaid DSH payment data with our estimated Medicare DSH payments, we had to identify hospitals in the Medicaid reports using their Medicare provider numbers. Only two states, Michigan and North Carolina, put Medicare provider numbers on their hospital specific Medicaid DSH reports. Project staff used the CMS On-line Survey and Certification Reporting System (OSCAR) and Provider of Service (POS) files, as well as AHA data from the on-line American Hospital Directory (www.ahd.com) to match hospital names in the Medicaid reports with Medicare provider numbers. This task had to be done by hand, and was further complicated because of the lack of detailed information about the hospitals. In general, discrepancies in hospital ownership or type were resolved by retaining the classifications from the Medicare or AHA data.
Several states included a few individual hospitals in their Medicaid DSH reports that we could not identify with sufficient confidence to match them with their Medicare provider numbers. We created dummy Medicare provider numbers for these hospitals to keep them in the data set, but they could not be linked up with Medicare DSH payment information. In addition, eight states lacked hospital specific payment information for a much larger share of their Medicaid DSH payments; they were Alabama, Colorado, Georgia, Illinois, Indiana, Minnesota, New Jersey, and Pennsylvania. For these states, we created one dummy variable to account for the missing DSH payments to acute care facilities and a separate dummy variable to account for missing IMD DSH payments. Unidentified hospitals account for 12 percent of total Medicaid DSH expenditures in the completed analysis, although they account for a very small share of reported expenditures in all but a few states.
In a few cases, we could not get hospital specific reports for FY 1998 but received reports for other time periods. We did not get FY 1998 reports for Louisiana, Minnesota, New York, or South Carolina. In Louisiana, total payments in the state fiscal year 1998 report that we received were virtually identical to total payments in federal fiscal year 1998, so we used the state fiscal year report. The adjustments to the other three states are noted in the state specific notes, below.
Once all 49 states with DSH programs were part of the data set, we compared the total amounts of Medicaid DSH payments reported on the state's hospital specific reports with aggregate totals reported on Form HCFA-64 for FY 1998. The HCFA-64 is widely considered the most accurate record of Medicaid expenditures available. In general, most states reported total expenditures on their hospital specific reports that were extremely close if not identical to total DSH expenditures reported on Form HCFA-64. Based on this comparison, we made changes to Alaska, Delaware, and Wyoming as noted below.
Total expenditures for inpatient DSH and IMD DSH from the hospital specific reports were also compared to FY 1998 DSH limits from the BBA. Based on these comparisons, we feel that the hospital specific Medicaid DSH payments used in our analyses are a good reflection of actual payments made in FY 1998.
Most hospitals in Alabama take part in a managed care initiative called the Partnership Hospital Program (PHP). No hospital-specific data are reported for the 112 hospitals that participate in the PHP, so we created a single dummy variable to account for the $346 million in acute care DSH payments paid to these hospitals.
We increased Alaska's payment to its single DSH hospital from the $12.7 million figure reported on the hospital specific report to $13.8 million to match the HCFA-64.
Delaware reported only the state's share of total payments to its single DSH hospital. The state's FMAP is 50%, so we doubled the reported amount.
The data we received from Georgia were for state fiscal year 2001 and state officials noted that the amounts paid to specific hospitals may have been significantly different from those paid in FFY 1998 due to recent program changes. We created a single dummy variable to account for the $413 million in DSH payments that Georgia made to acute care facilities.
The state reported $123 million in DSH payments on the HCFA-64, but included only $116 million in total payments on its hospital specific report. The total IMD DSH payments claimed on both reports were identical, so we added a dummy acute care DSH payment of $6.8 million to make up the difference.
The state reported $235 million in DSH payments on the HCFA-64, but included only $154 million in total payments on its hospital specific report. All of the missing payments appeared to be IMD DSH payments, so we added a dummy IMD DSH payment of $81 million to make up the difference.
The lone hospital specific report submitted by Minnesota was for calendar year 1997. The total payments in this report were comparable to the total payments claimed on Form HCFA-64 for FY 1998, but the distribution among acute care and IMD facilities from the hospital specific report was inconsistent with the reported distribution from Form HCFA-64 and the state's IMD DSH limit from the BBA. We replaced the hospital specific data with separate dummy values for the total acute care DSH payments and total IMD DSH payments reported on the HCFA-64.
New Hampshire's hospital specific report for FY 1998 did not include New Hampshire Hospital, a psychiatric facility that received an IMD DSH payment of $25 million. We added this hospital to our data set to make the state's total DSH payments match what was reported on the HCFA-64.
The state reported $1.058 billion in DSH payments on Form HCFA-64, but included only $876 million in total payments on its hospital specific report. The total IMD DSH payments claimed on both reports were identical, so we added a dummy acute care DSH payment of $182 million to make up the difference.
New York did not submit a hospital specific report to HCFA for FY 1998. We started with the distribution of payments to specific hospitals from the state's FY 1999 report, then adjusted the payments to all the hospitals so that total DSH payments equaled the FY 1998 amount (statewide). Payments to acute care facilities were multiplied by 0.9043; payments to IMD facilities were multiplied by 1.0237 (The 1999 report included only the federal share).
The hospital specific report submitted by Pennsylvania for FY 1998 included only $41 million in total DSH payments, while the state claimed $546 million in total DSH payments on Form HCFA-64. We replaced the hospital specific data with separate dummy values that match the total acute care DSH payments ($216 million) and total IMD DSH payments ($330 million) reported on the HCFA-64.
South Carolina did not submit a hospital specific report to HCFA for FY 1998. We started with payments to specific hospitals from the state's FY 1999 report, then adjusted the payments to all the hospitals so that total DSH payments equaled the FY 1998 amount (statewide). Payments to acute care facilities were multiplied by 1.0285; payments to IMD facilities were multiplied by 1.0162.
Wyoming reported only the federal share of total payments to its DSH hospitals in its hospital specific report. The state's FMAP is 63.02%, so we multiplied the reported payments by 1.5868 (=1.0/0.6302) to get the combined federal and state total payments of $106,315.
1. Unlike the cost report, the CMI and discharges are adjusted for short-stay transfer cases.
2. The DSH patient percentage is based on the percentage of Medicare patients who are entitled to SSI and the percentage of all patients who are eligible for Medicaid (and not Medicare). The SSI percentage is based on FY1998 claims. The Medicaid percentage is from the most recently settled cost report.
3. During FY1998 and FY1999, certain hospitals qualified for a higher update than other hospitals ("temporary relief" hospitals). To qualify, the hospital could not receive DSH or IME payments and needed to be located in a state where such hospitals had a Medicare negative operating margin in FY1995. The provision provided eligible hospitals with negative operating margins with a .5 percent higher update in FY1998 and .3 percent higher update in FY1999.