Workshop on Data Linkages to Improve Health Outcomes

John Drabek, OASPE, HHS

Several years ago, staff from four agencies NCHS, CMS, SSA, and OASPE met to discuss how the respondents in the NHIS might be linked to their Medicare and Social Security records. An agreement was developed that safeguarded privacy, yet enabled the database to be constructed. This spring the database was made available to researchers at the NCHS Data Center. Extensive documentation is available on the NCHS website: http://www.cdc.gov/nchs/r&d/nchs_datalinkage/data_linkage_activities.htm.

A large amount of data has been made available. The 1994, 1995, 1997, and 1998 National Health Interview Surveys were matched with National Death Index data through 2000, with Social Security data (1963-2003), and with Medicare data (1991-2000). The data are particularly useful to disability researchers since the 1994 and 1995 NHIS contain an extensive disability supplement. These data have been merged with records from the SSDI, SSI, and Medicare programs. Similarly, since the 1994 NHIS formed the initial sample for the LSOA, the merged data can be combined with the followup interviews in the LSOA to form a rich database on people as they age. Finally, since the NHIS started providing the initial sample for MEPS in 1996, it is possible to combine some of the linked data with MEPS respondents.

This project has taken a great deal of effort on the part of many persons. Completing the match, verifying the results, and preparing the analytical files and documentation has been a huge effort. Although the NHIS is well documented for users, SSA and Medicare files are designed primarily for program operations, not for research purposes. Although the SSA and Medicare files contain a wealth of information, it requires substantial effort to understand how to properly use that information.

The four agencies have devoted a considerable amount of staff time to this project. In addition, OASPE provided funds for preparing the analytical files and documentation that are on the NCHS website. OASPE is providing additional funds to NCHS to further support this project and related projects.

Jerry Riley of CMS has already used the linked database to study the characteristics of individuals who are receiving SSDI, but are waiting to become eligible for Medicare (after the mandatory 2 year waiting period for those under age 65). The results of this study will soon be published in the journal Inquiry. Other investigators have submitted, or will be submitting, proposals to NCHS to use the linked database at the Data Center.

OASPE is in the process of awarding a research contract to demonstrate the usefulness of the linked database for policy research.  Three areas will be studied: (1) understanding the interaction of the population with disabilities with the SSDI and SSI programs; (2) using administrative data to follow respondents in the NHIS, especially in terms of later application to SSDI and SSI; and (3) understanding family support and caregiving for SSDI and SSI beneficiaries.

OASPE has supported this project because it is essential that we have good data for policy research and program evaluation. Linking databases is a relatively new undertaking, and the design and use of these products is not well understood. Only by conducting analyses will we understand how useful the data are, and what further improvements need to be made. In particular, we need to figure out how to make optimal use of a database that has various parts covering different time intervals. Survey data collected at a point in time are merged with program histories often covering many years. Figuring out how to meaningfully summarize the program history data for analysis is a new research task.