Using multiple imputation for analysis of incomplete data in clinical research

Abstract

Background: Sample loss and missing data are inevitable in multivariate and longitudinal research. Ad hoc approaches such as analysis of incomplete data or substituting the group mean for missing data, while common, may unnecessarily reduce statistical power and threaten study validity. Multiple imputation for missing data is a newly accessible, methodologically rigorous approach to dealing with the problem of missing data.

Approach: To (a) discuss the problem of missing data in clinical research, and (b) describe the technique of multiple imputation. A case of analysis of multivariate psychosocial data is presented to illustrate the practice of multiple imputation.

Results: The advantages of multiple imputation are it (a) results in unbiased estimates, providing more validity than ad hoc approaches to missing data; (b) uses all available data, preserving sample size and statistical power; (c) may be used with standard statistical software; and, (d) results are readily interpreted.

Discussion: Accessible, user-friendly computer programs are available to perform multiple imputation for missing data making ad hoc approaches to missing data obsolete.