An investigation of the representativity of neuropsychiatrists and their patients in a drug utilization observation study on fluoxetine.

This study presents a methodological approach to an expost facto investigation of sample bias in drug utilization observation (DUO) studies using the example of a DUO with the nontricyclic antidepressant fluoxetine. A total of 479 psychiatrists and neurologists and 2,401 patients were investigated. The purpose of the study was to judge the representativeness of our DUO sample for two populations: first, for all psychiatrics and neurologists prescribing fluoxetine or all patients being treated with fluoxetine in Germany and, second, for all psychiatrists and neurologists prescribing antidepressants or all patients being treated with antidepressants in Germany. Criteria for the representativeness test were physician variables (gender, size of community where practicing, federal state, age, volume of prescriptions) and patient variables (gender, age, prescription-related diagnosis, concurrent illnesses, concomitant medications). The study shows that the DUO sample can rightfully claim representativeness in the majority of parameters for the psychiatrists and neurologists prescribing fluoxetine and for the patients being treated with fluoxetine. There are more noticeable discrepancies with regard to the psychiatrists and neurologists in general and to the patients being treated with antidepressants in general. The methodological problems of pharmacoepidemiological investigation of representativeness are discussed.