In a stepped-wedge trial every cluster starts as control and crosses to the intervention at a scheduled step, so every record needs a cluster, a period and an exposure status. Capture holds those in one audited study. Build free, no credit card.
Free sandbox · No credit card · 21 CFR Part 11 aligned
months
What a stepped-wedge trial asks of its data capture
The design
In a parallel cluster trial, half the clusters receive the intervention throughout and half do not. In a stepped-wedge trial, all clusters begin as control and cross to the intervention one sequence at a time, until all are exposed. It suits interventions that will be rolled out anyway, that cannot be withdrawn, or that stakeholders insist everyone should eventually receive. For the parallel design, read EDC for cluster-randomised trials; this page is about what changes when time is part of the design.
Time changes the data in three ways. First, the same cluster contributes control and intervention observations, so each record needs a period and an exposure label. Second, there are different ways to collect data: the same participants followed over time (a closed cohort), a continuous flow of new participants, or repeated cross-sectional samples taken in each period. The choice affects consent, identifiers and the number of forms per person. Third, some designs include a transition period around each switch when the intervention is being implemented and outcomes are excluded or analysed separately.
Analysis for these designs is the domain of your statistician, and sample-size methods differ from parallel designs: the free cluster sample size calculator explicitly covers only a two-arm parallel design. Capture does not do the sample size or the analysis. It supplies the structured, time-stamped dataset the analysis needs.
Stepped wedges are common in health-service and nursing research, where wards or practices roll out a care bundle in stages. See EDC for nursing research for a ward example with falls and pressure-injury outcomes.
Sequence
SEQExposed to intervention this period
EXPOSEDDate cluster switched
SWDTScheduled switch date
SCHDTDays late
Calculated, read-only
14 CalculatedSwitched later than scheduled
Observations this period
NOBSStudy elements
| Design element | What the data looks like | Where it lives in Capture |
|---|---|---|
| Cluster | Ward, clinic, school, practice or village | A site with its own participant numbering and by-site filtering |
| Sequence | Order of switchover assigned to the cluster | A site-level form field set once from the statistician's allocation list |
| Period | Calendar block in which an observation is made | A period field on each record; study visits or epochs, tested in the sandbox |
| Exposure status | Control or intervention at the time of observation | A required field or derived from the switch date; edit check if inconsistent |
| Switch date | Actual date the cluster crossed over | Site form with an audit trail, compared with the scheduled date |
| Participant outcomes | Clinical measures, events or questionnaires | eCRFs and ePRO on the participant's phone, linked to site and period |
| Implementation fidelity | Training done, intervention delivered as planned | Repeating site form per period |
| Consent | Individual consent, cluster-level waiver or opt-out | eConsent where individuals consent; record the waiver approval otherwise |
Capture does not generate the stepped-wedge sequence or perform the analysis.
Create your clusters as sites, add a period summary form and an exposure check, and enter demo data across the steps. Free sandbox, no credit card.
Data quality
A wrong period or exposure label is hard to find later. Use edit checks to flag impossible combinations as the data is entered, for example a record dated before the cluster's switch date but labelled as exposed. Each flagged entry becomes an auto-query for the site, and every correction keeps its reason in the audit trail.
Edit checks / auto-queries
2Type
Range High
Operator
Greater than
Value
180
Priority: High
Type
Range Low
Operator
Less than
Value
80
Priority: Normal
Query raised automatically
Value 192 violates limit (180). Please verify.
Design choices
In a closed cohort, the same participants are followed across periods, so each person needs a stable identifier and follow-up visits; Capture's visit schedule and ePRO tasks fit this well. With continuous recruitment, new participants enter throughout and each contributes to the period they enrolled in, so enrolment date and period assignment are critical and site-level numbering keeps clusters distinct. With cross-sectional sampling, different people are sampled in each period, often with a short survey or record extract and sometimes with no individual identifiers at all; here a lightweight form per record, rather than a full visit schedule, may be enough.
Decide which you are running before building forms, because it determines the number of forms, the identifiers, and what the consent covers. Write the choice in your data management plan; the data management plan template is a starting point. If the unit is a hospital or clinic network, multi-site clinical trial management covers site set-up.
Build sequence
Clusters, sequences, steps, periods and exposure, defined once and used on every form.
Give each its own numbering, and record its sequence from the statistician's allocation list.
Upload the protocol (PDF, DOCX or DOC) and let the AI study builder propose forms. Nothing is saved until a person reviews it.
A period summary form and edit checks that flag inconsistent exposure labels.
Enter demo data before and after a switch and confirm the export has cluster, period and exposure columns.
Approved forms lock for live use. Real participants start the paid live phase.
Before first step
Sequence for each cluster supplied by the statistician and stored before the first switch.
Calendar boundaries and step lengths written once.
Closed cohort, continuous recruitment or cross-sectional, with matching consent.
Who records the actual switch date, and how late switches are handled.
Whether outcomes in implementation periods are excluded or flagged.
CSV or Excel with the data dictionary opened in the statistics package, with cluster, period and exposure columns present.
Your statistician produces the allocation order and Capture stores it against each cluster. This page does not claim an automatic stepped-wedge randomiser; for random lists see the randomisation pages and test your design in the sandbox.
Use a site-level form with the scheduled and actual switch dates. Every change is kept in the field-level audit trail with the reason, and a calculated field can show days late.
No. The cluster-randomised page covers parallel designs where clusters stay in one arm. This page covers the extra data needs of stepped-wedge designs: sequences, periods, exposure status and switch dates.
No. It supplies the structured dataset with cluster, period and exposure for your statistician. Stepped-wedge sample size needs methods beyond the free parallel-cluster calculator.
Yes for both as data capture: cohorts use visit schedules and ePRO tasks, cross-sectional designs can use a light form per record. Consent and identifiers differ, so settle them with your ethics committee.
Yes. The free sandbox has every feature and you pay only once you go live. Check data hosting with your ethics office; Capture hosts in the EU (Frankfurt) and the USA (N. Virginia).
The sandbox is free with every feature, no credit card and no time limit. You pay only once you go live with real participants; see the pricing page.
Keep exploring
EDC for cluster-randomised trials
Parallel cluster designs.
EDC for nursing research
A ward-level stepped-wedge example.
Cluster sample size calculator
Parallel-design calculator.
Pragmatic clinical trial software
Embedded, real-world trials.
Edit checks software
Flag inconsistent entries.
Pricing
Free sandbox, pay at go-live.
Free sandbox with every feature. No credit card, and you pay only when you go live.