Quasi experimental designs in pharmacist intervention research.

Background In the field of pharmacist intervention research it is often difficult to conform to the rigorous requirements of the "true experimental" models, especially the requirement of randomization. When randomization is not feasible, a practice based researcher can choose from a range of "quasi-experimental designs" i.e., non-randomised and at time non controlled. Objective The aim of this article was to provide an overview of quasi-experimental designs, discuss their strengths and weaknesses and to investigate their application in pharmacist intervention research over the previous decade. Results In the literature quasi experimental studies may be classified into five broad categories: quasi-experimental design without control groups; quasi-experimental design that use control groups with no pre-test; quasi-experimental design that use control groups and pre-tests; interrupted time series and stepped wedge designs. Quasi-experimental study design has consistently featured in the evolution of pharmacist intervention research. The most commonly applied of all quasi experimental designs in the practice based research literature are the one group pre-post-test design and the non-equivalent control group design i.e., (untreated control group with dependent pre-tests and post-tests) and have been used to test the impact of pharmacist interventions in general medications management as well as in specific disease states. Conclusion Quasi experimental studies have a role to play as proof of concept, in the pilot phases of interventions when testing different intervention components, especially in complex interventions. They serve to develop an understanding of possible intervention effects: while in isolation they yield weak evidence of clinical efficacy, taken collectively, they help build a body of evidence in support of the value of pharmacist interventions across different practice settings and countries. However, when a traditional RCT is not feasible for logistical and/or ethical reasons researchers should endeavour to use the more robust of the quasi experimental designs.