Study Designer · RWEdnesdays

Design aid, not a protocol approval. Do not enter patient-level, confidential, or regulated information. This prototype records design decisions but does not establish data fitness, causal identification, regulatory acceptability, or statistical validity. A qualified methodologist must review the exported brief.

Stage 01 · question

Frame the decision before choosing a method

A precise question determines whether a comparative, descriptive, diagnostic, or policy design is appropriate. Start with the decision the study should inform.

Working title Use a neutral title. Avoid implying benefit or harm before the design is specified.

Primary objective State one population, one contrast, one outcome, and one decision-relevant horizon.

Design intent

Primary effect or summary measure

Population

Exposure or strategy

Comparator Required for causal comparative questions; optional for descriptive and some validation studies.

Primary outcome

REC

Choose a design intent to receive a starting recommendation.

The recommendation is a prompt for review, not an automatic selection.

Design architecture

Override the recommendation when the scientific question, data, or identification assumptions require a different design.

Selected study design

Eligibility criteria

Time zero / index event Eligibility, treatment assignment, and follow-up should align at this moment.

Treatment or exposure strategy

Exposure washout / clean period

Baseline covariate window

Start, end, and censoring of follow-up

Causal contrast / estimand

Competing event likely?

Data fitness and observability

Name what the source can and cannot observe before selecting an analysis model.

Primary data source

Study period and data vintage

Observable person-time rule

Outcome ascertainment and validation

Bias control plan

Document the bias domains first. The eventual model should follow from these decisions.

Confounder domains and measurement timing

Primary adjustment strategy

Missing-data strategy

Planned sensitivity and falsification analyses

Assess overlap and positivityInspect propensity distributions, tails, and effective sample size. Pre-specify balance diagnosticsDefine standardized-difference thresholds before outcome modeling. Include a falsification strategyUse negative controls or another residual-bias diagnostic where feasible. Version code lists and algorithmsRecord vocabulary version, date, provenance, and validation evidence.

Governance and decision context

The same design may need different documentation, validation, and review depending on how the evidence will be used.

Intended use

Review owner or accountable role

Known open decisions

Method spine used by this prototype

Target trial emulation Protocol and SAP elements Time-zero alignment Active-comparator new-user design Observable person-time Outcome algorithms Target trial framework StaRT-RWE template RECORD-PE