Data preparation
Specify how source records become study variables and analysis data. Document measurement, timing, linkage, derivations, missingness and checks so the resulting dataset can be reconstructed.
Flows for this purpose
- Data source selection and suitability
Research requirements → candidate sources → coverage and quality assessment → selected sources
- Data linkage and harmonization
Source datasets → identifiers and linkage → matched and unmatched records → harmonized data
- Study variable definition
Clinical concept → operational definition → source fields and algorithms → derived variable
- Data quality and query resolution
Incoming data → checks → flagged records → review or query → resolution → analysis-ready data
- Time-at-risk and censoring
Entry or time zero → risk window → outcome or censoring event → analysis time
- Analysis dataset finalization
Prepared data → derivations → reconciliation and review → approved dataset version
- Sources and data preparation
Establish suitable analysis-ready data
- Data preparation and finalization
Resolve queries and derive datasets
- Exposure and covariate ascertainment
Measure comparable prior exposure histories
- Measurements and data preparation
Define variables and resolve missingness
- Data and analysis populations
Finalize data and derive populations
- Participant-data integrity and harmonization
Reconcile participant variables across studies
- Target protocol-to-data mapping
Map trial requirements onto sources
- Emulation exposure and follow-up
Track strategy adherence and censoring
- Risk-set exposure and covariate ascertainment
Anchor measurements to index times
- Self-controlled observation and risk windows
Partition baseline and risk periods
- Time-series data preparation
Align intervals outcomes and denominators
- Comparative change panel preparation
Align repeated measurements and covariates
- Predictor and outcome construction
Align measurements without information leakage
- External data mapping and prediction
Reproduce predictors and generate predictions
- Crossover period data and populations
Track completion missingness and deviations
- Factorial data and analysis populations
Derive combination-specific and pooled populations
- Cluster follow-up and data populations
Track cluster losses and participants
- Stepped-wedge period data and follow-up
Track repeated observations and missingness
- Adaptive follow-up, safety and data integrity
Protect monitoring integrity and participants
- Platform data populations and control pools
Track epochs and comparator eligibility