- Published
- Updated
In a stepped wedge cluster design, units switch to an intervention at different times. Outcomes can also change because of general care, season or recording practices. Calendar time therefore needs explicit consideration in design and analysis rather than being treated as an incidental date.
Define the question and assumptions
Define sequences, periods, measurement arrangements and the expected onset of effect after transition. State whether the same people or different samples are observed in successive periods. Learning after implementation and general period effects are distinct mechanisms; operational evidence and specialist statistical assessment help support appropriate models and interpretation.
Hypothetical study example
In a hypothetical study, all units adopt a new practice during a year when a hospital-wide improvement programme also begins. The team does not attribute the falling event rate solely to the study intervention. It evaluates transition timing, concurrent changes and observation processes together when interpreting the comparison.
Design review checks
- Record planned and actual transition dates for each randomised cluster.
- Track changes in care and documentation practices across the study periods.
- Define treatment of observations during transition or learning intervals prospectively.
Record the decision and its limits
Accepted reporting makes the calendar and cluster structure understandable. Saying everyone eventually receives the intervention does not settle every scientific or ethical issue. Discuss assumptions about period effects, implementation delays and missing observations. Assess comparability of people observed in different periods separately, and keep causal certainty within what the design and data support.
Frequently asked questions
- Is a before-and-after difference sufficient?
- No. Calendar changes and concurrent activities can also affect outcomes.
- Does the transition decision establish implementation?
- Planned and actual start dates may differ; track both.