MedicareMedicaid Linked Analytic File Methodological Summary Centers for Medicare & Medicaid Services

Linked Analytic File Overview ............................................................................................. 1 Previous Analyses of Linked Medicare and Medicaid Data ................................................. 2 Data Source for Linked Analytic File ................................................................................... 2 Linked Analytic File Development ....................................................................................... 3 Chronic Conditions in the CCW ........................................................................................... 3 Defining Medicare-Medicaid Enrollees ................................................................................ 4 Type of Medicare-Medicaid Benefit Status. ..................................................................... 5 Managed Care Enrollment ............................................................................................... 5 Use of Long Term Services and Supports ....................................................................... 5 Reporting of Medicare and Medicaid Spending ............................................................... 6 Data Caveats and Limitations ............................................................................................. 6 Accuracy of Underlying Data............................................................................................ 6 Total Medicaid Spending. ................................................................................................ 8 Managed Care Enrollment. .............................................................................................. 8 Chronic Conditions........................................................................................................... 9 Contact ................................................................................................................................ 9

ii

Linked Analytic File Overview The main data source for the Medicare-Medicaid Enrollee State Profiles and National Summary Report is an analytic file developed by the Centers for Medicare & Medicaid Services (CMS) containing linked calendar year 2007 Medicare and Medicaid administrative and claims data. The linked analytic file includes the following populations for the 50 states and the District of Columbia: 1) all Medicare beneficiaries and 2) all Medicaid beneficiaries who qualify for Medicaid benefits based on age (65 or older), blindness, or disability. These two populations are categorized into four groups based on Medicare-Medicaid coverage status: Table 1. Definitions of Medicare-Medicaid Enrollment Groups in Linked Analytic File Group Name Used in State Profiles and National Summary

Beneficiary Groups

Medicaid Services Provided

Full Benefit

QMB, SLMB, and other Medicare beneficiaries who qualify for full Medicaid coverage based on State eligibility standards (a.k.a., “QMB+”, “SLMB+” and “Other dual eligibles with Medicaid coverage”

Coverage for Medicare Parts A & B premiums, and Medicare deductibles and coinsurance, and full Medicaid benefits

Partial Benefit - QMB

Qualified Medicare Beneficiaries (QMB; a.k.a QMB-only)

Coverage for Medicare Part A & B premiums and Medicare deductibles and coinsurance

Partial Benefit – SLMB/Other

Specified Low Income Medicare Beneficiaries (SLMB; a.k.a. SLMB-only), Qualified Individuals (QI), and Qualified Disabled Working Individuals (QDWI)

SLMB: Coverage for Medicare Part B premiums only

Medicare beneficiaries who are not covered by Medicaid

None

Medicare-only

QI: Coverage for Medicare Part B premiums only QDWI: Coverage for Medicare Part A premiums only

Note: In Wisconsin, the Partial Benefit category included low-income beneficiaries ages 65 and older who participated in the State's SeniorCare pharmaceutical assistance program.

1

This program provided assistance with the cost of prescription drugs and over-the-counter insulin and coordinated with other drug coverage such as Medicare Part D. Previous Analyses of Linked Medicare and Medicaid Data There have been numerous analyses conducted over the past two decades utilizing linked Medicare and Medicaid. These include: •

State-level links of Medicaid and Medicare data conducted in the mid 1990s by several individual states participating in The Robert Wood Johnson Foundation’s Medicare-Medicaid Integration Program; i

•

Linkage of 2003 Medicare Current Beneficiary Survey ii data with the Medicaid Statistical Information System (MSIS) to create a national representative dataset of Medicare-Medicaid enrollees; iii

•

Linkage of 2005 Medicaid Analytic eXtract (MAX) files with Medicare summary spending files (conducted for the Medicare Payment Advisory Commission); iv

•

Linkage of 2004-2006 Medicaid Analytic eXtract (MAX) files with Medicare Beneficiary Annual Summary File (BASF); internal report conducted by CMS staff; and

•

A “cross-payer” study in 2010 based on the experience of Medicare-Medicaid enrollees in Maryland (conducted by The Hilltop Institute of the University of Maryland-Baltimore County). v

While the findings related to Medicare-Medicaid enrollees presented in the State Profiles and the National Summary Report are generally consistent with prior research and data sources, there are differences in enrollment counts and expenditures which are attributable to differences in the data sources and methodologies. These are described in the Data Caveats and Limitations section. In addition to data caveats and limitations, the following sections present key information about the data sources used in creating the linked analytic file, and definitions of terms. Data Source for Linked Analytic File The linked analytic file is primarily based on the CMS Chronic Condition Data Warehouse (CCW). The CCW is a set of linkable research databases created in response to the Medicare Modernization Act of 2003, Section 723, which outlines a plan to improve quality of care and reduce cost of care for chronically ill Medicare beneficiaries. vi At the time the linked analytic file was created, the CCW contained 100% of Medicare claims for calendar years 1999 through 2010 and Medicaid Analytic eXtract (MAX) files for calendar years 1999 through 2007. To supplement the Medicare spending information in the CCW, which is limited to fee-for-service (FFS) claims, CMS included data on beneficiary-level capitation

2

payments to Medicare Advantage plans. vii Medicaid managed care capitation payments are already included in the MAX Other Therapy Claims file. Four broad types of beneficiary information are available in the CCW: 1) enrollment and eligibility; 2) chronic condition flags; 3) home health and skilled nursing facility assessments; and 4) claims. Whereas the enrollment, eligibility, and claims data are available for both the Medicare and Medicaid programs, the information on chronic conditions and the home health and nursing home assessments are limited to Medicare beneficiaries (including those eligible for both Medicare and Medicaid). The home health and skilled nursing facility assessment information was not included in the 2007 linked analytic file. Linked Analytic File Development Each beneficiary in the CCW is assigned a unique, unidentifiable identification number which can be used to links across the research databases. For the purposes of the Medicare-Medicaid Enrollee State Profiles and National Summary Report, a linked Medicare-Medicaid beneficiary file was created through the merging of Medicare and MAX enrollment files: •

Medicare Beneficiary Summary - annual, person-level summary containing enrollment, eligibility (including dual eligibility for Medicaid), date of death (if applicable), state of residence, age, and other demographic and geographic information

•

Chronic Condition Summary - yearly indicator and first occurrence date for the set of conditions tracked in the CCW (see below for more detail)

•

MAX Person Summary - annual, person-level summary containing Medicaid enrollment, eligibility (including dual eligibility for Medicare), demographic, summarized utilization of select health services such as institutional and inpatient hospital care, and claims payment summaries.

Variables of interest were developed based on the enrollment, eligibility, demographic, and chronic condition data. The linked beneficiary file was then merged with Medicare and MAX monthly claims files. Finally, cost and utilization variables were created. Additional detail is available on request via the MMCO mailbox ([email protected]). Chronic Conditions in the CCW At the time the linked analytic file was created, the CCW included information on whether the beneficiary had ever had any of 21 chronic conditions. viii Eleven of these conditions

3

were retained for the linked analytic file: •

Alzheimer’s Disease, related disorders or senile dementia

•

Chronic kidney disease

•

Chronic obstructive pulmonary disease

•

Congestive heart failure

•

Depression

•

Diabetes

•

Ischemic heart disease

•

Osteoporosis

•

Rheumatoid arthritis/osteoarthritis

•

Stroke

•

Cancer (a combined flag based on ever having female breast, colorectal, prostate, lung, or endometrial cancer).

The following CCW conditions were excluded from the linked analytic file: acute myocardial infarction; Alzheimer’s disease; atrial fibrillation, cataract, glaucoma, and hip/pelvic fracture. Alzheimer’s disease was omitted because it is duplicative with the “Alzheimer’s disease, related disorders or senile dementia” category, and the other five because they were considered to be less disabling on a long-term basis than the 11 conditions listed above. Additional information about the CCW and criteria used in identifying the conditions can be found at http://www.ccwdata.org/index.htm. Limitations of using these CCW chronic conditions to understand the Medicare-Medicaid enrollee population are described in Data Caveats and Limitations. Defining Medicare-Medicaid Enrollees Medicare-Medicaid enrollees were identified based on an algorithm which takes into account the Medicare eligibility code in the CCW file and the Medicaid and Dual eligibility codes from the MAX file. Enrollment counts in the linked analytic file represent beneficiaries who "ever-enrolled" during calendar year 2007. Beneficiaries were categorized by Medicaid benefit status (see Table 1), age group, managed care enrollment status, and use of Medicaid-funded long term services and supports (LTSS). To avoid double-counting beneficiaries who were in more than one group within these categories during the calendar year, rules were used to assign beneficiaries to single categories. These rules are described below.

4

Type of Medicare-Medicaid Benefit Status. A Medicare-Medicaid enrollee was defined as a Full Benefit Medicare-Medicaid enrollee if the beneficiary had at least one month of Full Benefit Medicare-Medicaid enrollment during the year. The next group assigned was Partial Benefit - QMB, then Partial Benefit - SLMB. Remaining Medicare beneficiaries were categorized as “Medicare-only,” meaning that the beneficiary did not have Medicaid in addition to Medicare at any time over the course of the calendar year (private insurance was not taken into account). Age Group. MAX age group codes were used to assign enrollees into two broad age groups representing people ages 65 and older and those between the ages of 18 and 64. Individuals who originally qualified for Medicaid based on disability but were coded as age 65 or older in MAX were categorized as age 65 or older in the linked analytic file. Managed Care Enrollment. For Medicare managed care enrollment, any beneficiary with at least one month of enrollment in Medicare Advantage was categorized as a Medicare Advantage enrollee. The remaining beneficiaries were categorized as being in Medicare FFS. For Medicaid managed care, those with at least one month in an LTSS Medicaid managed care program were categorized as a managed LTSS enrollee. Those who were not managed LTSS enrollees, but who had at least one month in a comprehensive managed care program (PACE and programs which include the full range of Medicaid acute care services) were categorized as comprehensive managed care. Those who were neither LTSS nor comprehensive, but who had at least one month in a limited managed care program (such as a managed mental health and substance abuse, a non-emergency transportation, or a dental program) were categorized as limited managed care. Those who were not categorized into any managed care category, but who had at least one month in a primary care case management (PCCM) program were categorized as PCCM. The rest were categorized as Medicaid FFS. Use of Long Term Services and Supports. Use of Medicaid-funded LTSS and the type of LTSS were determined by Medicaid claims payments. The LTSS categories were mutually exclusive in that a beneficiary could not be assigned to more than one type of LTSS. Any beneficiary with a paid amount greater than $0 for an institutional service was categorized as using institutional LTSS. Within this category, beneficiaries were further categorized into two types of institutional service: nursing facility or other type of institutional service which included Intermediate Care Facility for the Mentally Retarded (ICF-MR) and inpatient psychiatric facility. Recipients of institutional care were also categorized by length of stay: short-stay (< 90 days) and long-stay (90+ days). Service dates on claims were used to ensure that the days included in the length of stay count were unique days. The days were not required to be consecutive to be included in the count.

5

For the remaining beneficiaries (not assigned to an institutional category), those with a paid amount greater than $0 for §1915(c) waiver programs or for home health or personal care services were categorized as community-based LTSS. Within the community-based LTSS category, beneficiaries were identified as either using waiver services or state plan services (home health or personal care). Beneficiaries who had no FFS Medicaid expenditures for any of the LTSS services were presumed not to be users of LTSS. It is important to note that Medicare-Medicaid enrollees who received LTSS through capitated managed care arrangements would not be identified as users of LTSS because there were no encounter data in the linked analytic file to document the service use. Reporting of Medicare and Medicaid Spending Variables were constructed to summarize monthly Medicare and Medicaid spending by the beneficiary's type of enrollment. These variables could be used to summarize spending specific to those enrollment months when a beneficiary was a Full Benefit MedicareMedicaid enrollee. This was useful in developing average monthly spending given that a significant share of Medicare-Medicaid enrollees were in more than one Medicaid benefit category during the year (e.g., transitioned from "Medicare-only" to "Full Benefit MedicareMedicaid enrollee"). On average, 76% of beneficiaries in the 2007 linked analytic file who were defined as Full Benefit Medicare-Medicaid enrollees by virtue of having at least one month of Full Benefit enrollment were Full Benefit for all months in the year. ix This ranged from a low of 60% in Ohio to a high of 88% in Tennessee. Thus, the spending figures in Figure 10 of the State Profiles and Figure 5.1 of the National Summary Report represent the average monthly spending of Full Benefit Medicare-Medicaid enrollees and Medicareonly beneficiaries for those months they were enrolled in these respective categories. Data Caveats and Limitations Accuracy of Underlying Data. With any data source, there are known, suspected, and unknown data gaps and inaccuracies. The CCW and MAX data were not adjusted to account for these issues. In general, the findings related to Medicare such as MedicareMedicaid enrollees as share of Medicare enrollment vs. Medicare spending are consistent with published sources such as MedPAC's 2011 Medicare Data Book. x The average perenrollee spending figures for Medicare-Medicaid enrollees and for Medicare-only beneficiaries are somewhat higher in the linked analytic file as compared to MedPAC's figures, but differences are minimal. And, the ratio of Medicare-Medicaid enrollees' annual Medicare spending to that of Medicare-only beneficiaries is the same as found by MedPAC (with Medicare-Medicaid enrollees' Medicare spending over twice as high as the Medicareonly population's). MedPAC's analysis is based on the Medicare Current Beneficiary Survey, which, by virtue of being a point-in-time survey, would be expected to produce different findings. MedPAC also uses a slightly different definition of a Medicare-Medicaid

6

enrollee in that the months the beneficiary qualifies for Medicaid must exceed the months he or she qualifies for supplemental insurance. Given the state variation in the Medicaid program, the MAX file data are particularly susceptible to anomalies. Documentation of anomalies in MAX person summary and claims files can be found on the CMS Website at: http://www.cms.gov/MedicaidDataSourcesGenInfo/07_MAXGeneralInformation.asp. Anomalies in the underlying Medicaid Statistical Information System data are documented at: http://www.cms.gov/MedicaidDataSourcesGenInfo/MSIS/list.asp. The data anomalies most relevant to analysis of Medicare-Medicaid enrollees are summarized below. Known MAX data gaps and inaccuracies include: instances of missing FFS claims data and miscoding of services on FFS claims, and a lack of encounter data for some states with Medicaid managed care programs. The approach used to report findings for states with significant managed care enrollment of Full Benefit Medicare-Medicaid enrollees is discussed further under Managed Care Enrollment. Researchers have described MAX data gaps and inaccuracies related to home and community-based services (HCBS) waivers. xi These issues result in inaccurate HCBS waiver spending estimates and participant counts for some states. Further, there has been considerable state variation in the coding of services received through HCBS state plan services and waivers in MSIS and MAX. While CMS is working with states to implement a standard set of HCBS service codes, the MAX 2007 files would not reflect these efforts. For this reason, the State Profiles and National Summary report do not include analysis of utilization of specific services provided through HCBS waivers. There are also potential inaccuracies in MAX in the reporting of Medicare-Medicaid enrollment: both the total number of enrollees and the share with partial Medicaid benefits. xii In 2007, five states were identified as anomalous in the Medicaid Analytic Extract State Anomaly Tables because there was a discrepancy of more than 5% between the number of Medicare-Medicaid enrollees identified in MAX and the number of Medicaid enrollees in MAX identified as having Medicare through a link to the Medicare Enrollment Data Base. In addition six states were identified as anomalous because the share of Medicare-Medicaid enrollees reported as having "restricted" (partial) Medicaid benefits exceeded 40%. The Medicare-Medicaid enrollment data reported in the linked analytic file differ somewhat from ever-enrolled statistics published on the Kaiser Family Foundation’s www.statehealthfacts.org website: the linked analytic file counts are 3% higher, on average, for all Medicare-Medicaid enrollees and 6% higher, on average, for Full Benefit Medicare-Medicaid enrollees. There are two main differences in methodology that could explain these discrepancies: definition of a Full Benefit Medicare-Medicaid enrollee and time period. The main reason for the difference is that the linked file used a hierarchical method of assigning Medicare-Medicaid enrollees to categories, with Full Benefit being used to classify persons with any Full Benefit eligibility in the calendar year, while the 7

enrollment statistics on statehealthfacts.org categorized concomitant Medicare-Medicaid enrollment based on beneficiaries’ last month of enrollment during the year. xiii In addition, the enrollment is based on the Federal Fiscal Year, which may be slightly lower than calendar year due to a trend toward growth in the Medicaid enrolled population. xiv In addition, the enrollment counts from these State Profiles are expected to be higher than enrollment reports based off States’ monthly Medicare Modernization Act (MMA) file submissions to CMS. This is because enrollment defined as “ever-enrolled over the course of a calendar year” will always be higher than a simple snapshot of enrollment at any given time during the year. Future analytic efforts will use the best elements from the three methods discussed above as follows: State-reported monthly MMA data will be the source data used to develop “ever-enrolled” Medicare-Medicaid enrollee counts, which will then be classified by most recent Medicare-Medicaid enrollee status. Total Medicaid Spending. As MAX data are limited to claims, spending estimates based on these data do not include supplemental Medicaid payments to states and providers which are made outside of the claims system, such as upper payment limit, disproportionate-share hospital payments, and intergovernmental transfers. The total Medicare-Medicaid enrollee spending amounts by state reported in the State profiles and National Summary Report tend to be significantly lower than those published on the Kaiser Family Foundation statehealthfacts.org web site because Kaiser Family Foundation adjusted the latter to include these additional payments based on States' quarterly reporting on the CMS Form 64. The Form 64 is a broader reporting of state expenses eligible for federal Medicaid reimbursement in that is based on other sources in addition to claims. Managed Care Enrollment. As noted above, managed care enrollment, both in Medicare and Medicaid, affects the accuracy of the findings for any analyses which are based on FFS claims experience. The linked analytic file does not contain encounter data to provide information on service use in managed care. Thus, the higher the rate of participation in managed care for a beneficiary group, the less representative are the remaining FFS claims of that group. For the purpose of assessing the impact of Medicaid managed care on Medicaid spending and utilization, only the enrollment in managed care programs that included LTSS as a capitated service were considered since Medicaid would be the primary payer for these services (but not for most primary and acute medical services). The states with the highest Medicaid managed care participation of Full Benefit Medicare-Medicaid enrollees in such programs in 2007 were Arizona (79%) and Minnesota (36%). xv Medicare Advantage enrollment also impacts findings for Full Benefit Medicare-Medicaid enrollees on spending and use of services for which Medicare is the primary payer. The

8

same two states, Arizona and Minnesota, had very high Medicare Advantage enrollment for Full Benefit Medicare-Medicaid enrollees (45% and 36% respectively), with Oregon and Pennsylvania also having very high participation (45% and 37% respectively). High participation in Medicare Advantage may also attenuate findings related to the number and prevalence of chronic conditions as these are derived from coding on Medicare FFS claims. As a rule of thumb, analyses were not considered sufficiently representative if the managed care participation rate for Full Benefit Medicare-Medicaid enrollees was 35% or higher. For states with managed care participation rates below 35% but above 19%, a note was added to the relevant analyses to inform the reader of the potential impact of managed care enrollment on the findings. Such notes were also added with respect to Medicare Advantage enrollment of Medicare-only beneficiaries, although this group's enrollment rate did not preclude the analysis from being in the Profile. Chronic Conditions. Only a limited number of chronic conditions were investigated for this State Profiles project; unfortunately these did not include the range of mental health or developmental conditions. A recent analysis conducted by the Kaiser Commission on Medicaid and the Uninsured based on 2003 data from the Medicare Current Beneficiary Survey found that nearly half of the Medicare-Medicaid enrollee population between the ages of 18 and 64 had a mental or cognitive impairment, as did 44% of Medicare-Medicaid enrollees of all ages. xvi Beneficiaries with depression represented just over half of the group. CMS is currently vetting a list of proposed additional chronic conditions to better describe the experience of the Medicare-Medicaid enrollee population. These include additional mental health disorders (e.g., bipolar, anxiety, post-traumatic stress syndrome, schizophrenia and other psychotic disorders), intellectual and developmental disabilities, HIV/AIDS, substance abuse, alcohol abuse, and others. Contact For questions about the State Profiles, National Summary Report, or linked analytic file and its development, please contact [email protected]. i

Saucier, P., Bezanson, L., Booth, M., Bratesman, S., Fralich, J., Gilden, D., Goldstein, E. K., O'Connor, D., Perrone, C. V., and Willrich, K. K. (1998). Linked Data Analysis of Dually Eligible Beneficiaries in New England. Health Care Financing Review, 20, 91-108. ii The Medicare Current Beneficiary Survey collects information on use, cost, and access, chronic conditions, health insurance status, and other topics for a nationally representative sample of Medicare beneficiaries. iii Kasper, J., O'Malley Watts, M. and Lyons, B. (2010). Chronic Disease and Co-Morbidity Among Dual Eligibles: Implications for Patterns of Medicaid and Medicare Service Use and Spending. Kaiser Commission on Medicaid and the Uninsured. iv Report to the Congress: Aligning Incentives in Medicare, Chapter 5. (2010). Medicare Payment Advisory Commission.

9

v

Tucker, A., Johnson, K., Rubin, A., and Fogler, S. (2008, September). A Framework for State-Level Analysis of Duals: Interleaving Medicare and Medicaid Data. Baltimore, MD: The Hilltop Institute, UMBC. vi About Chronic Condition Data Warehouse, Chronic Condition Data Warehouse Website, accessed on January 23, 2012 at http://www.ccwdata.org/about/index.htm. vii The Medicare Advantage payment information comes from CMS' Medicare Advantage Prescription Drug System. viii Since that time, six more conditions were added to the CCW and newly proposed conditions targeted for use with Medicare-Medicaid Enrollees are being finalized. ix The total enrollment months in the year for each beneficiary were limited to those months during which they were alive. x A Data Book: Health care spending and the Medicare program. Section 3. Dual-eligible beneficiaries. MedPAC (June, 2011). Accessed on March 29, 2012 at http://www.medpac.gov/chapters/Jun11DataBookSec3.pdf. xi March 2012 Communication with Pamela Doty and John Drabek (Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services) on a forthcoming study on LTSS use and expenditures based on MAX 2006 and 2007. xii "Table 5. Reporting of Dual Enrollment in Medicaid and Medicare in MAX 2007" in Medicaid Analytic Extract State Anomaly Tables, 2007. (March, 2011). Prepared by Mathematica Policy Research for the Centers for Medicare & Medicaid Services (ORDI) under CMS Contract: HHSM-500-2005-00025I, Task Order 0003. The states cited in the first example are Alaska, Arkansas, Illinois, Minnesota, and Ohio; the second, Alaska, California, the District of Columbia, Maryland, Massachusetts, and New York; and the third, Alabama, Delaware, Florida, Georgia, Mississippi, and Nevada. xiii Future publications of the Medicare-Medicaid Coordination Office will define Medicare-Medicaid Enrollees according to the "Dual Status Code" reported by states in the Medicare Modernization Act Files they submit at least monthly to CMS. Full versus partial status will be assigned based on the most recent month's dual eligibility status. xiv Centers for Medicare & Medicaid Services Website, Medicaid Analytic eXtract Frequent Asked Questions, Answer ID 9230. xv The assessment is based on the share of Full Benefit Medicare-Medicaid enrollment months in these types of programs. This is a more accurate measure than the share of Full Benefit Medicare-Medicaid Enrollees given the share of this group with more than one benefit status during the year. xvi Kasper, J., O'Malley Watts, M. and Lyons, B. (2010, July.) Chronic Disease and Co-Morbidity Among Dual Eligibles: Implications for Patterns of Medicaid and Medicare Service Use and Spending. Kaiser Commission on Medicaid and the Uninsured. Accessed on January 31, 2012 at http://www.kff.org/medicaid/upload/8081.pdf. Even these percentages are likely understated because cognitive impairment in the Medicare Current Beneficiary Survey was limited to intellectual disability, thus not reflecting a wider range of developmental disabilities.

10