Missing data and long-term outcomes from nutrition research in the critically ill

Abstract

Purpose of reviewThe use of functional outcomes in critical care nutrition research is increasingly advocated; however, this inevitably gives rise to missing data. Consequently there is a need to adopt modern approaches to the foreseeable problem of missing functional and survival outcomes in research trials.Recent findingsAnalyses that ignore unobserved or missing data will often return biased effect estimates. An improved approach is to routinely anticipate the types and extent of missing data, and consider the likely mechanisms of that missingness. The researcher and their statistical advisor may then choose from a number of modern strategies to assess the sensitivity of the research conclusions to the patterns of missingness contained in these research data. Methods widely employed include multiple imputation of missing observations, mixed regression models, use of composite outcome variables with patients who die being attributed a value reflecting the lack of ability to function, and selected Bayesian methodology.SummaryConclusions from clinical research in critical care nutrition will become more clinically interpretable and generalizable with the adoption of modern methods for the statistical handling of missing data.