Missing data in surveys: Key concepts, approaches, and applications

A recent review of missing data in pharmacy literature has highlighted that a low proportion of studies reported how missing data was handled. In this paper we discuss the concept of missing data in survey research, how missing data is classified, common techniques to account for missingness and how to report on missing data. The paper provides guidance to mitigate the occurrence of missing data through planning. Considerations include estimating expected missing data, intended vs unintended missing data, survey length, working with electronic surveys, choosing between standard and filtered form questions, forced responses and straight-lining, as well as responses that can generate missingness like “I don't know” and “Not Applicable”. We introduce methods for analysing data with missing values, such as deletion, imputation and likelihood methods. The manuscript provides a framework and flow chart for choosing the appropriate analysis method based on how much missing data is observed and the type of missingness. Special circumstances involving missing data have been discussed, such as in studies with repeated or cohort measures, factor analysis or as part of data integration. Finally, a checklist of questions are provided for researchers to guide the reporting of the missing data when conducting future research.