Accounting for missing data in statistical analyses: multiple imputation is not always the answer

Choice of method for dealing with missing data is crucial for validity of conclusions, and should be based on careful consideration of the reasons for the missing data, missing data patterns and the availability of auxiliary information.