Nutrition is difficult to measure accurately and thus liable to measurement error. Although validation studies of dietary assessment instruments are regularly carried out, most studies stop at describing the reproducibility and correlation with the true intake.
Studies looking at the quantitative effects of measurement error on the conclusions of the studies carried out are relatively rare, and also mostly focus on health effects of nutrients. For public health purposes, studies on foods rather than of nutrients are important, as it are foods that are eaten, rather than nutrients. This project will develop and evaluate new methods to handle intake data that contain zeros in a regression calibration context, evaluate the performance of these methods, quantify an effect that is handled mostly qualitatively and explore the impact of ignoring measurement error on the burden of disease estimates among others. The lessons learned will have wider impact over and above the case of nutrition and cancer.
In order to evaluate the impact of policies or public health interventions, modelling of effects - like with the DYNAMO-HIA model -l is required. To date these models do not take effects of measurement error into account, and therefore underestimate potential impacts. This research will contribute to a more realistic estimation of the impact of policies.