Izurieta HS, Wu X, Lu Y, Chillarige Y, Wernecke M, Lindaas A, Pratt D, MaCurdy TE, Chu S, Kelman J, Forshee R. Zostavax vaccine effectiveness among US elderly using real-world evidence: Addressing unmeasured confounders by using multiple imputation after linking beneficiary surveys with Medicare claims.
Pharmacoepidemiol Drug Saf 2019;
28:993-1001. [PMID:
31168897 DOI:
10.1002/pds.4801]
[Citation(s) in RCA: 9] [Impact Index Per Article: 1.8] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/08/2019] [Revised: 04/12/2019] [Accepted: 04/17/2019] [Indexed: 02/06/2023]
Abstract
PURPOSE
Medicare claims can provide real-world evidence (RWE) to support the Food and Drug Administration's ability to conduct postapproval studies to validate products' safety and effectiveness. However, Medicare claims do not contain comprehensive information on some important sources of bias. Thus, we piloted an approach using the Medicare Current Beneficiary Survey (MCBS), a nationally representative survey of the Medicare population, to (a) assess cohort balance with respect to unmeasured confounders in a herpes zoster vaccine (HZV) effectiveness claims-based study and (b) augment Medicare claims with MCBS data to include unmeasured covariates.
METHODS
We reanalyzed data from our published HZV effectiveness Medicare analysis, using linkages to MCBS to obtain information on impaired mobility, education, and health-seeking behavior. We assessed survey variable balance between the matched cohorts and selected imbalanced variables for model adjustment, applying multiple imputation by chained equations (MICE) to impute these potential unmeasured confounders.
RESULTS
The original HZV effectiveness study cohorts appeared well balanced with respect to variables we selected from the MCBS. Our imputed results showed slight shifts in HZV effectiveness point estimates with wider confidence intervals, but indicated no statistically significant differences from the original study estimates.
CONCLUSIONS
Our innovative use of linked survey data to assess cohort balance and our imputation approach to augment Medicare claims with MCBS data to include unmeasured covariates provide potential solutions for addressing bias related to unmeasured confounding in large database studies, thus adding new tools for RWE studies.
Collapse