Bias reduction with an adjustment for participants' intent to dropout of a randomized controlled clinical trial

BACKGROUND: Attrition, which is virtually ubiquitous in randomized controlled clinical trials, introduces problems of increased bias and reduced statistical power. Although likelihood-based statistical models such as mixed-effects models can accommodate incomplete data, the assumption of ignorable attrition is usually required for valid inferences. PURPOSE: In an effort to make the ignorability assumption more plausible, we consider the value of one readily obtained covariate that has been recommended by others, asking participants to rate their Intent to Attend the next assessment session. METHODS: Here we present a simulation study that compares the bias and coverage in mixed-effects outcome analyses that do and do not include Intent to Attend as a covariate. RESULTS: For the simulation specifications that we examined, the results are promising in the sense of reduced bias and greater precision. Specifically, if the time-varying Intent to Attend variable is associated with attrition, outcome and treatment group, bias is substantially reduced by including it in the outcome analyses. LIMITATIONS: Analyses that are adjusted in this way will only yield unbiased estimates of efficacy if attrition is ignorable based on the self-rated intentions. CONCLUSIONS: Accounting for participants' Intent to Attend the next assessment session will reduce attrition bias under conditions examined here. The item adds little burden and can be used both for data analyses and to identify participants at risk of attrition.