Research Note: Using effectiveness studies for prescribing research, Part 2

Trials that consider the effects of interventions on prescribing behaviour amongst clinicians often have complex design implications resulting in data that has inherent hierarchical structure. It follows that both experimental design and analysis plans must account for this structure and thus results should be considered in terms of clinician behaviour rather than individual patient response. We describe this change in perspective and the necessity for using statistical techniques that allow incorporation of potential confounding effects. We also discuss the appropriateness of some specific outcomes in relation to these trials.