Quasi-Experimental designs for quality improvement research

Abstract

Presentation Quality Improvement (QI) research may be defined as “the design, development and evaluation of complex interventions aimed at the re-design of health care systems to produce improved outcomes”. The challenge of QI lies in bridging the gap between knowing what needs to happen at an individual patient level and implementing this at a systems level. The inherent complexity of systems poses challenges in terms of implementation, but also presents the researcher with circumstances for which conventional research methods may not prove useful. Explanatory trials are designed to answer the question “does this intervention work under ideal circumstances?” Patient and system variability are typically rigorously controlled. Pragmatic trials seek to answer how well an intervention works in usual practice [1]. It is important to contend with variation (e.g., in patient volume or complexity) and not control for it. Consider the analogy of water sampled from a pond versus a river [2]. If one takes random samples of water from a still pool of water one can draw inference about the pond as aw hole, as it is relatively static and unchanging. This is the principle we are using in attempting to extrapolate the findings of a randomized controlled trial to a population. The real world however, behaves far more like a river where the water changes from second to second, influenced by innumerable complex interacting factors such as the season, rain, construction. In QI research it is important to understand the changing nature of the river (i.e. causes of system variation) in order to be able to predict how to make an intervention work under all the conditions in which it will be expected to perform. QI research should focus therefore on robust, sequential experimentation. Too often, quality improvement investigators seek to proceed to clinical trials before sufficient exploration, investigation, and understanding of the complex system and its interactions have been achieved. Campbell et al present a trajectory for QI research required to build requisite knowledge [3]. The design and testing of complex interventions in care delivery proceeds through a series of planned stages. One begins by developing a concept or theory and then progresses to designing a prototype. Next, an intervention is piloted on a small scale before performing a detailed test and finally disseminating the ideas generated. A variety of study designs may be used as learning proceeds across this trajectory of understanding. Research methods that address issues of internal validity without randomization of individuals are referred to as “quasi-experimental” designs and include time-series, equivalent time series, multiple baseline and factorial design.