BACKGROUND: Since the introduction of prescription drug coverage in 2006, United States Medicare data are increasingly used in pharmacoepidemiology research. The data capture a large proportion of those aged 65 years and older, often allow follow-up until death, and can be linked to other data sources. Using supplemental data sources can increase validity and should be considered when designing pharmacoepidemiology studies to evaluate and mitigate bias due to measurement error in exposures, outcomes, and covariates.
OBJECTIVES: This symposium aims to 1) provide an overview of Medicare data available for research; 2) illustrate novel methods of using supplemental data sources to increase validity of exposure, confounder, and outcome definitions; 3) provide examples of current validation studies for Medicare data; and 4) discuss remaining gaps and the feasibility of future linkages to other data sources.
DESCRIPTION: The symposium will focus on sharing best practices for validation studies using Medicare data. The co-chairs of the session, Til Stürmer and Elizabeth Andrews, will provide an introduction to the challenges of identifying exposures, confounders, and outcomes using claims data in an older population. Five speakers will present examples using different methods to validate Medicare data, highlighting best practices for implementing each method and operational challenges: 1) Introduction (6') 2) Overview of Medicare data available for research: Invited, CMS (10') 3) Exposure validation: Using linked interview data to assess prevalent drug use prior to claims-based drug initiation: Jonsson Funk (15') 4) Covariate validation: Using cohort data to validate claims-based measures of physical function in older adults: Lund (12') 5) Outcome validation: a. Chart abstraction to obtain clinical information: Johannes (15') b. Linkage with state cancer registries: Gilsenan (12') 6) Interactive discussion with the audience about conducting validation using Medicare data, including CMS perspectives on data linkage (20').
