A comparison of existing methods for multiple imputation in individual participant data meta-analysis

Abstract

Multiple imputation is a popular method for addressing missing data, but its implementation is difficult when data have a multilevel structure and one or more variables are systematically missing. This systematic missing data pattern may commonly occur in meta-analysis of individual participant data, where some variables are never observed in some studies, but are present in other hierarchical data settings. In these cases, valid imputation must account for both relationships between variables and correlation within studies. Proposed methods for multilevel imputation include specifying a full joint model and multiple imputation with chained equations (MICE). While MICE is attractive for its ease of implementation, there is little existing work describing conditions under which this is a valid alternative to specifying the full joint model. We present results showing that for multilevel normal models, MICE is rarely exactly equivalent to joint model imputation. Through a simulation study and an example using data from a traumatic brain injury study, we found that in spite of theoretical differences, MICE imputations often produce results similar to those obtained using the joint model. We also assess the influence of prior distributions in MICE imputation methods and find that when missingness is high, prior choices in MICE models tend to affect estimation of across-study variability more than compatibility of conditional likelihoods.

Assessment of heterogeneity in an individual participant data meta‐analysis of prediction models: An overview and illustration

“…More advanced multilevel imputation methods have recently proposed to impute missing values in large, clustered data sets such as IPD‐MA and can also be used to impute covariates that are systematically missing for one or more studies . In our case study, we applied the simpler approach as used previously by Steyerberg et al, where the authors imputed missing predictors using the study as a fixed effect in the imputation model.…”

A comparison of one‐stage vs two‐stage individual patient data meta‐analysis methods: A simulation study

“…One‐stage models can directly use multiple imputation techniques since they are standard mixed‐effects models. Although heterogeneity makes the imputation challenging and different approaches have been proposed, missing data at the patient level are common in such investigations and the flexibility of the one‐stage model is invaluable in this context. In addition, evaluating study‐level covariates and their interactions with exposure, although challenging, is only possible through one‐stage models.…”

“…One‐stage analyses rely on widely used mixed‐effects models and are considered more flexible, but they are more challenging in both conducting and communicating the findings, especially regarding visualisation with the hallmark forest plot, although software solutions are available . These challenges with the one‐stage approach drive meta‐analysts to the two‐stage approach, which is more prevalent, although other challenges like power calculations, or multiple imputation, are common across both approaches.…”

Pre-post changes in main outcomes of medical rehabilitation in Germany: protocol of a systematic review and meta-analysis of individual participant and aggregated data

“…Second, any remaining missing data will be addressed via multiple imputation methods. We will create 10 imputed data sets and report the pooled results 29. Third, we will conduct a sensitivity analysis using all available data only.…”

“…For IPD analysis, the method of multiple imputation will be employed to address the missing data values 29. We will report the proportion of the identified missing data and summarise the possible explanations for any missing data.…”