External data mapping and prediction

Reproduce predictors and generate predictions

Generate predictions from the original specification before any local updating. Missing required inputs and altered measurements are applicability problems to document, not permission to quietly refit the model.

Flow diagram

External data mapping and predictionF1.S1Map required predictorsF1.S2Align outcome and horizonF1.S3Resolve unavailable inputsF1.S4Generate frozen-model predictionsF1.S5Retain reproducible evaluationinputsF1.N1Validation mapping registerInputOriginaldefinitionAvailabledefinitionMappingDeviationPredictorsOutcomeTime horizonPreprocessingF1.N2Assumptions and rationaleF1.N3Method and protocol evidenceF1.S0 Start flowExternal data mapping and predictionF1.S999 End flowEnd flow

Steps

Map required predictors

Reproduce model inputs, coding and transformations exactly where possible.

Align outcome and horizon

Document ascertainment, follow-up and censoring differences.

Resolve unavailable inputs

Prespecify missingness and mapping decisions without refitting the original model.

Generate frozen-model predictions

Record artifact version, preprocessing and prediction implementation checks.

Retain reproducible evaluation inputs

Separate original predictions from any later recalibrated predictions.

External data mapping and prediction

End flow

Tables and supporting information

Validation mapping register

Validation mapping registerInputOriginaldefinitionAvailabledefinitionMappingDeviationPredictorsOutcomeTime horizonPreprocessing

Trace all deviations; do not silently substitute different predictors.

Assumptions and rationale

Refitting or recalibrating before evaluating changes the model and no longer evaluates the untouched original.

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.

Sources