A support study uses a stepped-wedge design: check the calendar

Identify the cluster and the switching sequence

A cluster might be a service, workplace, or another group defined by the study. Read which unit was randomized and how its switch time was determined. Individuals receiving support later are not necessarily individually randomized to it. A phased implementation can resemble the diagram without having the same allocation design.

Cochrane describes stepped-wedge cluster trials and highlights adjustment for time trends. The exercise here explains a design label, not a conclusion about any current support program. It does not establish that a randomized sequence was implemented correctly or that a particular quitting app participated in such a trial.

Cochrane: Chapter 23: Including variants on randomized trials

Map support status against calendar time

For a hypothetical example, four adult services introduce a new conversation format at different scheduled periods. Each service contributes observations before and after its switch, while the number using the format grows over the calendar. Write the service’s condition and period together rather than labeling all earlier observations untreated and all later observations treated without context.

A general change in staffing, public information, season, or access could coincide with later periods. Those possibilities do not prove that the trial is invalid; they explain why a simple early-versus-late average is inadequate. Read the actual model and authors’ account of relevant changes instead of inventing a calendar explanation for the results.

Check whether the reported method addresses both dimensions

Look for the analysis’s handling of the group structure and calendar period. Repeated observations within a cluster are not automatically independent. The report should explain how these design features were considered, but identifying a relevant term in the methods does not certify that the entire analysis was appropriate.

Also check the outcome definition and population at each period. A later service sample may differ from an earlier one. Avoid claiming that every person improved because a group average changed. If the report does not establish whether the same individuals were followed, keep that participant-continuity question open.

Retain rollout and effect as separate facts

Your note might say: “The format was introduced in this randomized sequence; the reported effect used this handling of calendar time and clusters.” Availability to all groups by the end is an implementation fact. Whether the measured outcome changed because of the intervention requires the actual comparison and its limitations.

For clinical decisions, seek qualified interpretation of the evidence. You can use the design map to identify a missing explanation without performing a reanalysis. A personal record taken before and after you start an app is not a stepped-wedge trial: it lacks randomized clusters and the study’s comparison structure. The label should describe the research design, not decorate a personal timeline.