What types of data sets are available to researchers?
Three basic types of data sets are available to researchers: Public Use Files (PUFs), Limited Data Sets (LDSs) and Research Identifiable Files (RIFs). The PUFs have been constructed so that all the beneficiary identifying information contained in the corresponding RIFs (including the Medicare Beneficiary Identifier [MBI], Medicare Health Insurance Claim [HIC] number where available, Social Security Number [SSN] where available, names, address fields, and the plan identifying information) have been removed. In addition, plan identifiers have been removed and some demographic fields such as race and age are aggregated to prevent identification of any individuals.
There are two types of PUFs, baseline and analytic. Analytic PUFs contain a completed cohort of data for all baseline respondents and are constructed to be self-contained with a baseline and follow up component for each beneficiary's record. There is no field that allows identification of a particular individual across the cohorts in the analytic PUFs. Baseline PUFs have been constructed with a unique anonymous ID field that does allow identification of the same individual across multiple baseline cohorts.
LDSs and RIFs are comprised of the entire national sample for a given cohort (including both respondents and non-respondents), and contain all of the HOS survey items, and the physical and mental health summary scores. They also contain protected beneficiary-level health information such as date of birth, gender, race/ethnicity, and county of residence. However, there are differences between the two types of data sets. For example, the specific direct person identifiers (i.e., name, address, MBI, HIC number where available, and SSN where available) are included in the RIFs and allow identification of the same individual across multiple cohorts; however, these identifiers are excluded in the LDSs. Note that SSNs are no longer included in RIFs beginning with Cohort 21. Additionally, the plan identifiers and plan characteristics that are included in the RIFs are blinded, modified, or excluded in the LDSs to prevent identification of specific MAO contracts. For more information, go to the Research Data Files section.
How can I obtain the research files?
The PUFs are available for download on the HOS website. A signed Data Use Agreement with CMS is required to obtain either LDS or RIF data files. A small fee is assessed for each cohort of data. All research requests for LDS data files must be submitted through the CMS Limited Data Set File Process, while the requests for RIF data files will continue to be processed through the Research Data Assistance Center (ResDAC) at the University of Minnesota. ResDAC is a CMS contractor that provides assistance to academic, government and non-profit researchers interested in using Medicare and/or Medicaid data. ResDAC is available to assist in the completion and/or review of data requisition forms for Medicare HOS RIF data files prior to their submission to CMS. For information about how to request either LDS or RIF data files, go to the Research Data Files section on the Data page.