Measuring Prescribing Improvements in Pragmatic Trials of Educational Tools for General Practitioners

Abstract

Randomized pragmatic trials of drugs, physician education and drug policies are needed to improve pharmacosurveillance and cost-effectiveness of prescribing. Since 1994, we have developed and tested methods for low-cost education and policy trials to improve prescribing in primary care in Canada. We review methodology for using drug claims and other health services data to evaluate prescribing improvement programs and policies. We apply the lessons to a proposed trial of physician education tools (PET) for quality improvement of prescribing. Design issues for the trial include defining the potential programme in causal terms using counterfactuals, narrowing the denominator to the population affected, excluding noise from the numerator, calculating the prescribing preference, adjusting for baseline differences, controlling for modifiers and confounders, accounting for uncertainty when measuring impacts, and grouping practices for feedback and recognition. Data from a randomized trial of academic detailing illustrate measurement challenges. A decade of progress on methods for evaluating prescribing improvement programs with drug claims data has enabled planning of routine randomized pragmatic trials of education and policies in primary care in Canada.