O3-1.5 Using data linkage to explore use of GP services by smokers

Abstract

Introduction It is understood how confounding of mediator-outcome associations resulting in collider biases may cause error when estimating direct and indirect effects. Here we assess the impact of non-differential and independent measurement error of the exposure and mediator. Methods We used as a motivating example the association of childhood socioeconomic position (exposure) with adult psychological distress (outcome), and possible mediation by adult socioeconomic position (mediator). We use quantitative bias analysis methods to quantify the impact of misclassification of exposure and mediator on three target parameters: the total effect of exposure on outcome; the direct effect (by conditioning on the mediator); and the indirect effect (identified by the percentage reduction in the excess OR upon adjusting for the mediator). Results ORs before and after adjustment for mediators are both biased to the null by non-differential misclassification of the exposure, but the percentage reduction in the excess OR is not affected by measurement error of the exposure. Conversely, measurement error of the mediator rapidly biases the percentage reduction the excess OR downwards. Conclusions If the research objective is to quantify the proportion of the total association that is due to mediation (ie, indirect effect), then minimising (or adjusting for) non-differential misclassification bias of the mediator is much more important than that for the exposure (or outcome). Considering the relative importance of collider bias and measurement error when estimating direct and indirect effects, measurement error (of mediators in particular) will often be a (much) greater source of systematic error.