A frequent conundrum faced by nurses when implementing research findings and developing new interventions is that their patients, staff and sociocultural environment may be quite different from those in the original studies. Therefore, implementing and evaluating new interventions may need a carefully thought out and creative approach. One such approach is realist evaluation (RE). This paper will give an overview of RE and how this methodology can be used to support nurses when evaluating the implementation of evidence-based care.
RE is a research methodology, otherwise known as a system of enquiry, that provides a framework for conducting an evaluation of a healthcare intervention.1 An RE approach enables researchers to develop theories about how their intervention works by asking ‘What works for whom, in what circumstances, in what respects, and how?’ (Pawson and Tilley, p2).2 This contrasts with traditional approaches to evaluations which simply measure and report the results without considering the environment, culture or what the human behaviours and emotional reactions between the intervention and outcome may be. This style of evaluation has been described as a ‘black box’, meaning that the inner workings are hidden from view.3 The RE approach is a ‘clear box’ evaluation where in addition to the outcomes, the processes of how the intervention has worked, in terms of peoples’ actions and reactions, and in which sociocultural context, are reported transparently (Kazi, p204 and Scriven3).
The theories that describe how the interventions work are constructed by recognising that Outcomes are the result of Mechanisms triggered in a specific Context and are written as Context, Mechanism, Outcome configurations (CMOc) (Pawson and Tilley, p60).1 An example of a CMOc is as follows:
Some people in prison are rebellious (context), so will refuse to engage in non-mandated activity (mechanism), so hepatitis C test uptake rates will not improve (outcome).
The process of undertaking RE research involves following a cyclical model (figure 1).