Handling Missing Data in Nursing Research With Multiple Imputation

BackgroundIn the data analysis phase of research, missing values present a challenge to nurse investigators. Common approaches for addressing missing data generally include complete-case analysis, available-case analysis, and single-value imputation methods. These methods have been the subject of increasing criticism with respect to their tendency to underestimate standard errors, overstate statistical significance, and introduce bias. ObjectivesThis article reviews the limitations of standard approaches for handling missing data, and suggests multiple imputation is a useful method for nursing research. MethodSecondary analysis was conducted to examine the effect of a public policy on the health of women using a data set that had a large degree and complex patterns of missing data. DiscussionIn the example, accommodation of the incomplete data was critical to making valid inferences; however, complete-case, available-case, or single imputation could not be defended as an adequate method for dealing with the missing data patterns. Alternative methods for dealing with incomplete data were sought, and a multiple imputation approach was selected given the missing data pattern. Nurse researchers confronting similar complex patterns of missing data may find multiple imputation a useful procedure for conducting data analysis and avoiding the bias associated with other methods of handling missing data.