Missing data continue to be the chest wall mechanics, abdominal status, and IAP [ 60].
To compensate for missing data, recruitment continued until 186 participants returned their questionnaires.
In the most-likely scenario it was assumed that, in the absence of any intervention, participants with missing data would continue to fall at the baseline falls rate.
In this case, the complete records form a random subsample of the target sample, so any parameters which would have been estimated without bias in the absence of missing data will continue to be estimated without bias.
However, review of follow-up rates in centers suggests that the problem was related to the difficulty of collecting audit data in routine practice rather than default, and it is likely that most individuals with missing data are continuing to use DAFNE principles and attending some follow-up.
Missing outcome data continues to be a common problem in RCTs.
For the smoking cessation outcome variables, intention-to-treat analyses [ 23] were utilised with missing data classified as either continuing smoking or not reducing the amount smoked by 50%.
There have been attempts to refine these correction methods (e.g., Le & Schmidt, 2006), albeit with further assumptions and approximations for missing data, and, therefore, the debate continues.
For these analyses, missing data were classified either as continuing smoking or as not achieving a 50% reduction in tobacco use.
This review highlights the inconsistent reporting of missing data in cohort studies and the continuing use of inappropriate methods to handle missing data in the analysis.
