Bias Correction and Sensitivity Analysis for Meta-Analysis of Studies with Zero-Inflated Outcomes

• Activity Number: 542 - New Research Synthesis Methods in Data Science
• Type: Invited
• Date/Time: Wednesday, July 31, 2019 : 2:00 PM to 3:50 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Zhengyang Zhou and Minge Xie* and Thomas Trikalinos and Eun-Young Mun
• Companies: University of North Texas Health Science Center and Rutgers University and Brown University and University of North Texas Health Science Center
• Keywords: Bias correction; inference synthesizing; meta-analysis; wrong model; zero-inflated outcomes

Abstract

In clinical fields, zero-inflated outcomes are quite common. However, the zero-inflated nature of such outcomes is often ignored in many studies, leading to biased estimations. When we perform a meta-analysis to synthesize data from relevant studies, the aggregated outcome that includes such studies is biased. In this talk, we will quantify the estimation bias caused by using a wrong model and show results from a sensitivity study to demonstrate the impact of ignoring zero-inflation in the outcomes. We will also provide a general correction method, with supporting theories, to mitigate the impact of biased estimations (caused by ignoring zero-inflation) on meta-analysis. This correction method minimally requires the summary information from the wrong model and an estimated amount of zero outcomes in the study without the need to access any individual participant data, and thus can be broadly applied. This work was motivated by Project INTEGRATE, a synthesis study of aggregate data and individual participant data in the field of brief alcohol interventions for young adults. Our method development will be demonstrated using both simulation studies and the real data example.