Multilevel regression modelling to investigate variation in disease prevalence across locations.

In this article, we show how to investigate the role of individual (personal) risk factors in outcome prevalence in multicentre studies with multilevel modelling. The variation in outcome prevalence is modelled by introducing a random intercept. In the next step, the empty model is compared with the model containing the risk factor(s). Because the outcome is dichotomous, this comparison can only be carried out after having rescaled the models' parameter values to the variance of an underlying continuous variable. We illustrate this approach with data from Phase Two of the International Study of Asthma and Allergies in Childhood (ISAAC) and provide a corresponding Stata do-file.