TL28: A novel cluster randomized pragmatic research study design for evaluating interventions
*U Vijapurkar, Janssen 

Keywords: stepped-wedge cluster randomized design, intervention trials, design effect, power and size of study

The stepped-wedge cluster randomized (SWCR) design is a new research study design that is becoming increasingly popular in evaluating interventions in many areas from healthcare, education, social policy, and quality improvement. The SWCR design is a unique study design that can help mitigate the ethical dilemma of randomizing subjects to a control group and withholding interventions to that group when not in equipoise. That is, it may be unethical to deny benefits to the control group when interventions designed to bring about a positive change or are believed to do more good than harm are introduced. In this novel SWCR pragmatic research study design, clusters are essentially randomized to a time order for receiving the intervention (i.e. when and not if they receive the intervention). Under this design all clusters ultimately receive the intervention over time.

At this roundtable, we will give a brief overview of what an SWCR study design is, discuss and illustrate the typical features, nuances of key design considerations, similarities and dissimilarities with the conventional parallel cluster designs and potential benefits where SWCR designs can be implemented. In addition, discuss statistical considerations behind determining the design effect, sample size and power of the SWCR study and the methodology for analyzing the data. Any ideas or experiences, current practices, and challenges faced in designing, conducting and analyzing data from SWCR studies will be welcomed at this roundtable.

The following topics and questions may be of interest to the group:

1. Discuss opportunities and advantages of an SWCR design. 2. What are the key logistical and practical considerations in implementing the SWCR design (number of time steps, duration of step, cohort vs cross sectional, dropouts/missing data)? How do we size an SWCR trial, determine the design effect and plan the consequent statistical analysis? 3. What opportunities and guidelines exist to incorporate such novel study designs (eg, CONSORT guidelines) towards regulatory considerations? 4. Discuss some examples of SWCR implementation and analysis of trial data.