Low levels of missing data suggest that the PEQ-CAMHS is acceptable.
The levels of missing data suggest that the measure is acceptable to patients.
Finally, should missing data patterns and reasons for missing data suggest a bias or confounding, we will conduct sensitivity analyses using pattern mixture models to examine the influence that dropout bias (informative missingness) may have on treatment differences.
Overall, there were very low levels of uncertainty in response to Back-PAQ items, and very little missing data, suggesting that respondents understood the items, and the response options were sufficient for respondents to feel comfortable making a selection.
However, less than one percent of our cohort was lost due to issues of missing data, suggesting that the temporal discontinuity issue was not a significant factor in our study.
Both the questionnaire and software were developed by MEND Australia, however, there were issues with data entry (i.e., no data entry controls) and the high rate of missing data suggests better facilitator training is required to ensure program fidelity in real-world situations.
We did a very limited validation using simulated missing data, which suggested that the scaling of weights was a sensible thing to do.
We addressed this issue through sensitivity analyses after multiple imputation of missing data which suggested that attrition had not introduced any substantial bias; nevertheless, we cannot exclude the possibility of such bias.
Although we were unable to formally examine trends in inequalities due to the periods of missing data, our data suggest that absolute inequalities in exposure declined up to 1 July 2007 (when measured using geometric mean cotinine), but subsequently increased after smokefree legislation was implemented, albeit affecting an ever smaller proportion of the population.
