OBJECTIVES

Data missingness is a common issue in economic evaluations and can lead to substantial bias if not accounted for. Despite this, there is limited guidance on how to address missing data in economic evaluations, and methods are not commonly reported. The objective of this study was to conduct a literature review to identify published literature describing methods to handle missing data in economic evaluations of clinical trials.

METHODS

A targeted literature review designed according to the Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) checklist was undertaken to identify publications describing methods for handling missing data in health economic evaluations. Electronic databases (Medline and Medline In-Process) were searched from inception to October 2020. Full texts were reviewed to determine study eligibility based on predefined criteria, and data from eligible studies were extracted and reported methods for handling missing data were summarised.

RESULTS

The review identified 44 publications meeting the eligibility criteria for the review. The most common methods identified to address missing data were multiple imputation (MI) and complete-case analysis (CCA), used in 27 and 14 studies, respectively. MI-MICE (MI with chained equations) was the most common approach when using MI, and CCA was typically compared against more robust methods in sensitivity analysis. Inverse probability weighting (IPW) was used in four studies, however, was found to be outperformed by MI methods. When data are missing at random (MAR), MI and IPW is more precise than CCA, however, the approaches produce similar results when data are missing completely at random.

CONCLUSIONS

Robust handling of missing data is essential with chosen methods having the potential to directly impact the conclusions made. MI or IPW should be used where data are MAR and compared to results under an assumption of missing not at random, to evaluate the effect of differing assumptions on cost-effectiveness conclusions.