Dual Medicare-Medicaid Enrollment and Integrated Plan Identification: T-MSIS Analytic Files Data Quality

In this data quality report, we investigate the quality of 2016 data on dual MedicareMedicaid enrollment in the Transformed Medicaid Statistical Information System (TMSIS) Analytic Files, hereafter called the TAF. We also explore the TAF’s capacity to identify people dually enrolled in Medicare and Medicaid in integrated plans. For both exercises, we compare TAF data with enrollment information from the Medicare Master Beneficiary Summary File (MBSF). Our analysis produced several key findings:

• Agreement across the TAF and MBSF on which individuals were dually enrolled in Medicare and Medicaid in a given month during calendar year 2016 was approximately 84 percent.
• Concordance in dual enrollment status across the TAF and MBSF varies considerably across states; some states’ data are unusable (e.g., Arkansas), whereas very little evidence of data issues related to enrollment status exists in other states (e.g., Nebraska).
• Medicare Savings Program classifications (e.g., Qualified Medicare Beneficiary) for people identified as dual enrollees in both the TAF and MBSF seem consistent across data sources for most states. However, a few states have low consistency (e.g., California).
• The Program of All-Inclusive Care for the Elderly (PACE) is the only integrated plan identifiable in the TAF, and agreement in PACE enrollment is high across the TAF and MBSF.
• The TAF’s data on generic integrated plan enrollment (non-PACE) identify approximately 44 percent of Medicare-Medicaid Plan enrollees, 17 percent of Dual Eligible Special Needs Plan enrollees, and no Fully Integrated Dual Eligible Special Needs Plan enrollees in the MBSF.
• The TAF identifies most Medicare-Medicaid Plan enrollees in Ohio, Massachusetts, New York, Texas, and Virginia, or in half of the states with capitated Financial Alignment Initiative demonstrations.

The high degree of agreement between TAF and MBSF data on dual enrollees is encouraging, suggesting the TAF enrollment data quality is good for most states. Still, some states’ enrollment data are unusable. Consequently, researchers need to incorporate state-specific considerations into their research designs. Thus, one overarching implication is that nationwide analyses are infeasible using the TAF; some state exclusions are necessary. Another broad implication is that the TAF alone is not equipped to identify and study those enrolled in integrated plans, except PACE, and additional data sources, such as the MBSF, are necessary to identify integrated plan enrollees in the TAF.