Predictor and outcome construction
Align measurements without information leakage
Information recorded after prediction time can leak the outcome into the model. Define measurement availability and a feasible missing-input strategy at deployment. All learned preprocessing belongs inside resampling.
Flow diagram
Steps
Define predictors at prediction time
Record measurement availability, units, transformations and clinical accessibility.
Define outcomes and horizon
Specify ascertainment, censoring and competing events where relevant.
Plan missing-predictor handling
Specify missing-input handling during development and at deployment. Fit learned imputation within the resampling training data; exclude information unavailable at prediction time.
Place preprocessing inside resampling
Fit imputation, scaling and feature selection using training folds only.
Freeze reproducible feature definitions
Version transformations and ensure outcome information cannot leak into predictors.
Predictor and outcome construction
End flow
Tables and supporting information
Prediction variable dictionary
Retain measurement timing and deployment availability.
Assumptions and rationale
Preprocessing learned on all data before validation can leak information and inflate measured performance.
Method and protocol evidence
Attach the protocol, methodological sources and versioned software specification supporting these decisions. Record citations and provenance; do not enter participant identifiers.