Estimates of Dual and Full Medicaid Benefit Dual Enrollees, 1999

INTRODUCTION

There is a clear need to develop better estimates of dual (Medicare and Medicaid) enrollees and the subpopulation of dual enrollees who receive full Medicaid benefits. Dual enrollees that may receive full Medicaid benefits include: qualified Medicare beneficiaries (QMBs), specified

* A need to improve coordination of public funds from Medicare and Medicaid to meet the service needs of these vulnerable populations.

* Continuing increases in utilization and program spending for these vulnerable populations, especially dual disabled enrollees. These spending increases are straining Medicaid budgets in times that States are in fiscal crisis.

* A need for baseline estimates of State spending amounts for prescription drugs provided to dual enrollees by Medicaid to support cost estimates for these populations once drug coverage for these groups begins in 2006 under Medicare.

* A need to monitor changes in utilization and spending levels for dual enrollees under Medicaid.

The estimates shown in the tables are not official CMS estimates and should not be construed to represent data used for purposes of implementing the provisions of Section 103 of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (Public Law 108-173) relating to the Federal assumption of Medicaid prescription drug costs for dual enrollees.

NEED TO LINK MEDICARE AND MEDICAID DATA

Neither the Medicare nor the Medicaid systems, by themselves, permit complete and accurate reporting of dual enrollees.

Medicare

The Medicare system maintains data on persons enrolled in Medicaid and for whom Medicaid has paid the Medicare Part A and B insurance premiums in the enrollment database (EDB). Historically, these third-party liability data were housed in a Medicare data set commonly known as the TPEarth file (or the third party buy-in). Data from these Medicare systems have traditionally represented an undercount of all dual enrollees because States do not necessarily pay Medicare premiums for all dual enrollees.

Medicaid

The Medicaid analytic extract (MAX) data include two possible data elements that may identify dual enrollees. The first is the "dual eligibility flag." In its current form, this data element was first required of State Medicaid agencies beginning with fiscal year (FY) 1999 reporting under the Medicaid statistical information system (MSIS), the source for MAX data. Data quality may vary substantially from State-to-State for this data element. The second is a pair of data elements that report Medicare deductible and coinsurance payment amounts paid by Medicaid for a dual enrollee on an individual claim. Again, data quality is uncertain because reporting of these amounts was also required of State Medicaid agencies for the first time, beginning with FY 1999.

MEDICARE AND MEDICAID LINK

The source data for the most recent link are the Medicaid MAX data for calendar year (CY) 1999 and the Medicare EDB for the 50 States and Washington, DC. In order to maximize the quality of the linking process, the Medicare health insurance claim (HIC) was not used as a primary linking variable. Instead, the linking criteria use the Medicaid enrollees' Social Security Number (SSN), date of birth (DOB), and sex. The link effort begins with Medicaid MAX data and consists of two steps:

* The first step has different criteria for aged versus disabled Medicaid enrollees. For aged Medicaid enrollees, SSN, and sex must match exactly. For disabled Medicaid beneficiaries, either the enrollees' SSN or the DOB must match exactly, or SSN and sex must match exactly, and two of the three elements in DOB must match exactly.

* In the second step, there is an attempt to link the Medicaid SSN to a claim account number (CAN) from the HIC in the EDB for records that were not linked in the first step. This is done because some enrollees incorrectly report the CAN from an account on which they receive auxiliary benefits (as a spouse, widow, child, etc.) as their own SSN. For example, a spouse will report her husband's SSN as though it were her SSN. A check on sex and DOB assures that a correct link is made.

Once it is determined that the enrollee appears in both the MSIS and EDB data sets, it is necessary to determine if the enrollee was eligible for both programs at the same time.

* For each MAX eligibility record, monthby-month Medicaid enrollment is compared to repeating segments of Medicare enrollment A dual indicator is set whenever an overlap occurs. An annual (CY) dual indicator is set if the dual indicator for any month is set. The result is an enhanced MAX eligibility data set that includes information about the results of the EDB link.

* For persons identified as dual enrollees, selected data elements from the EDB are added to the Medicaid enrollment data. Because this is a Medicaid database, all MAX records are retained. However, information on dual enrollment status is not retained if the EDB contains an indication of dual enrollment status, but there is no record in the MAX file for the enrolled person.

COUNTING DUAL ENROLLEES USING MAX DATA

Following the EDB link, the MAX data provides counts of confirmed dual enrollees, by State. There is the potential for bias both in terms of undercounting and overcounting. The potential for undercounting may be caused by one or more of several factors: (1) the record for a dual enrollee may have been missing from either the EDB or the MAX file, (2) SSN may have been missing in the MAX file, or (3) there may have been errors or number transpositions in the recorded SSN. The possibility of overcounting is not as likely, but could be caused if an enrollee moved to a different State during the year because the MAX data are State-specific data sets. Because of this, there has been no attempt to unduplicate persons across States.

Estimates include adjustments for under-counting persons reported as dual by Medicaid, but not linked with an SSN or with incorrect/non-matching SSNs. However, estimates do not include adjustments for undercounting of persons reported as dual enrollee by Medicare, but not linked to Medicaid (e.g. persons on Medicare TPEarth). The estimates do not adjust for over-counting that may occur if the Medicaid person was enrolled in more than one State or if more than one person was identified with the same SSN in Medicaid. In both cases where adjustments were not made, the extent of overcounting and/or undercounting should be extremely minor and offsetting.

DUAL ELIGIBLE COUNTS-ADJUSTING FOR BIAS

Two sets of State-specific estimates are produced in Table 1. The first set is known as the "best estimate." It consists of enrollees confirmed to be dual enrollees as a result of the EDB link and selected Medicaid enrollees not linked to EDB (those identified as dual enrollees by Medicaid and having at least one claim in the year where Medicare copayment and/or deductible was paid by Medicaid in 1999). The second set of estimates is known as the "upper bound estimate." It consists of enrollees confirmed to be dual enrollees as a result of the EDB link and selected Medicaid enrollees not linked to EDB (those identified as dual enrollees by Medicaid or having at least one claim in the year where Medicare copayment and/or deductible was paid by Medicaid. Because of data inconsistencies for several States, these estimates are adjusted to not exceed the total number of aged and disabled enrollees in each State.

Estimating Full Medicaid Benefit

Currently it is not possible to estimate full Medicaid benefit dual enrollees using Medicare data alone. However, there are two Medicaid data elements that are used to increase the accuracy of these estimates.

The first of these data elements is the dual eligible flag. This data element was first required in MSIS reporting for FY 1999. While MSIS has established a 2-percent error tolerance for this data element; reporting remains inconsistent. One State (Pennsylvania) did not report dual enrollment status. Five other States (Georgia, Ohio, Rhode Island, Tennessee, and West Virginia) reported no full Medicaid dual enrollees. Findings for these six States are inconsistent with national estimates that about 90 percent of all dual enrollees are full Medicaid dual enrollees. However, the most pervasive data reporting problem for this data element was that many States reported dual eligibility status of unknown for a high percentage of their dual enrollees. Based on MAX data for 1999, 21 States reported greater than 20 percent of dual enrollment status of unknown. Among those States, 11 reported greater than 50 percent unknown.

There are two estimates of full Medicaid benefit dual enrollees that are produced using this data element (Table 2). The first estimate, known as the "lower bound estimate", assumes that dual enrollees of unknown type are distributed according to the same percentages as those for whom type is known. This assumption becomes questionable as the percentage of dual enrollees of unknown type grows, but it does establish a lower bound for the number of full Medicaid benefit dual enrollees. The second estimate, known as the "best estimate", assumes that all dual enrollees of unknown type are full Medicaid benefit dual enrollees. This is a reasonable assumption because, as noted previously in the national estimates, about 90 percent of all dual enrollees are full Medicaid benefit dual enrollees. Also, it is likely that States would have correctly identified dual enrollees who do not receive full Medicaid benefits because of the need they have to coordinate coverage and reimbursement with Medicare.

The second data element is the "restricted benefits flag." As with the dual eligible flag, this data element was first reported by States, in MSIS for FY 1999. While this data element has a 5-percent error tolerance for States, it is reported that data quality is questionable (Ellwood, 2004). A code value of 3 for this data element indicates that the person is enrolled in Medicaid, but only entitled to restricted benefits based on Medicare dual-eligibility status (e.g. QMB only, SLMB only, qualified disabled and working individuals-QDWIs or qualifying individuals-QI1s or QI2s) (Centers for Medicare & Medicaid Services, 2004a). An estimate of full Medicaid benefit dual enrollees is made using this data element to subtract numbers of dual enrollees with restricted benefits from the total numbers of dual enrollees. These estimates are also shown in Table 2.

CONCLUSION

As a best estimate, there were about 6.881 million dual enrollees, nationally, ever enrolled in both Medicare and Medicaid during 1999. This represented about 16.2 percent of all Medicaid enrollees. An upper bound estimate was 7.288 million dual enrollees.

Because the quality of reporting was uncertain for data elements used to estimate full Medicaid benefit dual enrollees, the reliability of those estimates is less certain than the estimates of all dual enrollees. However, the estimates of full Medicaid dual enrollees ranged from a lower bound estimate of 5.916 million (86.0 percent of all dual enrollees) to a best estimate of 6.091 million (88.5 percent of all dual enrollees).

DISCUSSION

These estimates of dual enrollees compare favorably with estimates from other sources:

* An estimate for FY 1999 is 6.982 million duals, using an actuarial rules of thumb regarding the percentage of aged and disabled who are dual enrollees (95 percent of Medicaid aged and 40 percent of Medicaid disabled beneficiaries) on reported FY 1999 MSIS summary statistics (Klemm, 2004; Centers for Medicare & Medicaid Services, 2004b). The data reported in this article are quite close to this estimate because both estimates are counts of enrollees ever enrolled in a year. The primary difference is that one estimate is for CY 1999 and the other is for FY 1999.

* The Kaiser Commission on Medicaid and the Uninsured (2003) reported 5.84 million full Medicaid dual enrollees for FY 2000. Colleagues Bruen and Holahan (2004) reported 7.2 million dual enrollees and 6.13 million full Medicaid dual enrollees for 2002. These estimates are also counts of persons ever enrolled in a year.

* The Henry J. Kaiser Family Foundation (2004) reported 5.8 million dual enrollees as of the August 2002 billing cycle, reflecting enrollment as of June 2002. Estimates of dual enrollees for the first quarter of FY 1999 were 5.46 million (Ellwood, 2002). Using a similar methodology, Ku (2003) estimated 5.4 million full Medicaid dual enrollees in 1999. The Medicaid Chart Book reports an average number of 6.4 million dual enrollees during CY 2000 (Centers for Medicare & Medicaid Services, 2003). Data from the Medicare Current Beneficiary Survey in 1999 show 6.277 million persons with health insurance coverage through Medicaid (either as Medicare buy-in individuals or as reported by survey respondents). Clark and Hulbert (1998) reported between 6.4 and 6.7 million dual enrollees for 1997, using (form) HCFA-2082 reports that were actuarially adjusted to represent person years of enrollment and to approximate average monthly enrollment. It is reasonable that estimates reported here should be higher than these quarterly, monthly, or point-in-time estimates because of enrollment turnover through the year.

* Finally, Dale and Verdier (2003) estimated that there were 6 million dual enrollees in 2002.

ACKNOWLEDGMENTS

The author would like to thank William Clark, Suzanne Dodds, Marilyn Ellwood, David Gibson, and John Klemm for their technical input and review. The author would also like to thank Matt Gillingham and Yifei Hu for providing background data for the estimates. Finally, the author would like to thank Chuck Brinker for his work on the linking methodology between Medicaid (MAX) and Medicare (EDB) data.