The information on all health plan members in the pharmacy data base is limited.  For some studies36-38 we have carried out external adjustment using the method of Schneeweiss40.  We obtained information about the prevalence of possible confounding variables such as cigarette smoking from samples of members who have responded to a mailed questionnaire called the Member Health Survey.  This also permitted us to estimate the association between the confounder and use of the drug being studied.  We obtained the confounder/outcome association from the literature.