Meta-Analysis With a Continuous Covariate That Is Differentially Categorized Across Studies.

Citation:

Perin, J., Fischer Walker, C.L., Black, R.E. & Aryee, M.J. Meta-Analysis With a Continuous Covariate That Is Differentially Categorized Across Studies. Am J Epidemiol 183, 5, 507-14 (2016).

Date Published:

2016 Mar 1

Abstract:

We propose taking advantage of methodology for missing data to estimate relationships and adjust outcomes in a meta-analysis where a continuous covariate is differentially categorized across studies. The proposed method incorporates all available data in an implementation of the expectation-maximization algorithm. We use simulations to demonstrate that the proposed method eliminates bias that would arise by ignoring a covariate and generalizes the meta-analytical approach for incorporating covariates that are not uniformly categorized. The proposed method is illustrated in an application for estimating diarrhea incidence in children aged ≤59 months.