For many decades, the prevailing view has been that the randomized controlled trial (RCT) is the criterion standard research design. The RCT is a strong and highly desirable research design as it enables researchers to generate the highest level of evidence and allows clinicians to be confident to use research findings in practice. At the same time, the limitations of the RCT design in addressing questions of relevance to nursing are well known. In some cases, it is not feasible or ethical to randomize participants to control and intervention conditions. More and more of our research requires testing of multicomponent interventions, which are quite complex to do within an RCT design. Take cancer-related symptoms as an example. Advances in nursing science mean we now have a much deeper understanding of the multifactorial etiologies, intricate pathophysiologic mechanisms, and multiple environmental and social influences that contribute to symptom experiences. Nursing interventions are becoming increasingly sophisticated to deal with this symptom complexity. Randomized controlled trials can take years to complete, by which time knowledge has advanced and practice has moved on. The time and resources required to conduct an RCT are increasingly being questioned as we face diminishing resources for research and increasing pressure to achieve rapid translation of knowledge into practice. Pragmatic approaches to trialling healthcare interventions have been described in the literature for some decades.1 The popularity of these approaches is growing as we respond to the challenges facing modern healthcare systems.
My team has conducted a number of RCTs during my 20 years as a nurse scientist. We have successfully completed trials of psychoeducational interventions to manage pain, fatigue, and dyspnea. I am a strong advocate of well-conducted RCTs wherever possible. However, I am increasingly frustrated with the limitations of the RCT and growing more impatient with how long it takes to generate findings which improve outcomes for patients. These concerns have led our team to explore how modifications to the traditional RCT can be used to more efficiently achieve high-quality research outcomes.
One example of an innovative approach to research design is the stepped-wedge approach. The stepped-wedge cluster randomized trial involves random and sequential crossover of clusters from control to intervention until all clusters are exposed to the intervention.2 Proponents argue that this design has a number of advantages over a standard cluster design. First, all participants eventually receive the intervention as the intervention is rolled out sequentially to all clusters (epochs), with each site receiving the intervention in a random sequence over time.3 Some argue that stepped-wedge designs require fewer clusters compared with a standard RCT; however, others note that this is due to the large number of measurements of the dependent variables rather than being an inherent characteristic of the design.3 A further advantage of the stepped-wedge design is that it allows an intervention to be improved during the life of the study, although such changes may not be desirable as they may influence interpretation of the findings.3
Our team has recently completed analysis of data from the PROMPT study, a trial using a stepped-wedge design. The study involved testing of a nurse-led psychosocial support intervention across 5 sites in Australia. We published the protocol for this study4 to share our approach and promote debate and critique, and we expect to publish results soon. Our experience with this design confirms that a major strength of the stepped-wedge design is that it enables engagement across all participating sites. The design was appealing to participating services who knew they would at some time receive the intervention. This approach also enabled us to test the intervention in a range of care contexts (rural and metropolitan, private and public, large and small) without the need to ensure equivalency of settings as assumed in a traditional cluster design. While the sample size estimates and data analysis approach presented new challenges for our statisticians, these challenges were not insurmountable.
More recently, I was introduced to the SMART design, a sequential, multiple-assignment randomized trial. The key features of this design include sequencing and tailoring of decisions regarding patient care, as various components of interventions can be introduced one stage at a time, depending on patient response.5 These features potentially allow a researcher to test different components over time while simultaneously changing direction if the effectiveness of the intervention wanes over time or if the patient does not respond.5 While this design is quite new, a number of trials of various interventions across diverse fields are currently underway (see https://methodology.psu.edu/ra/adap-inter). This pragmatic approach is appealing as it reflects the reality of clinical practice and allows us to be more responsive to patient needs.
I offer these 2 examples of innovative yet pragmatic research designs, but there are many more, including n of 1 trials and noninferiority trials. The CONSORT group has recently acknowledged the growing use of pragmatic trials and has offered their statements to give guidance on reporting of these newer and increasingly popular research designs (see http://www.consort-statement.org/extensions). I acknowledge the limitations of these designs, but I am excited by the possibilities of research designs that better reflect the world in which we live. As a researcher and educator, I am keen to understand the strengths and limitations of these approaches. I encourage all nurse researchers to be innovative in the way in which we approach our research to ensure our work contributes to better outcomes for our patients.
References