Dealing with Missing Data in an IPD Meta‐Analysis

This chapter provides an overview of different methods for dealing with missing data in an individual participant data (IPD) meta-analysis. It highlights the specific challenges of dealing with missing data in an IPD meta-analysis context, including how to preserve the clustering of participants within primary studies, whilst allowing for potential between-study heterogeneity. The describes the various types of missing data that can occur in an IPD meta-analysis project, and the strategies, statistical approaches and software to deal with each. It focuses on dealing with missing data in the context of IPD meta-analyses of observational studies, for example for examining prognostic factors or developing prediction models. A number of prognostic factors (‘predictors’) are known to be associated with the incidence of preeclampsia; for example, a woman has a higher risk if she had pre-eclampsia in a previous pregnancy, or if there is a family history of pre-eclampsia, diabetes, or renal disease.