How ohdsi/patientlevelprediction works
PatientLevelPrediction is an R package for defining, training, and validating patient-level prediction models with OMOP-format observational healthcare data. The principal workflow runs from prediction-design choices and cohort-based data extraction through preprocessing and model fitting to validation and performance exploration. The README also documents multiple target/outcome combinations, external validation, learning curves, and interactive result viewing. Relationships are based on README capabilities and module names in the tree; source excerpts were unavailable, so detailed call ownership and wiring cannot be verified.
Main entry points: Fit.R Model fitting · ExtractData.R OMOP extraction · EvaluatePlp.R Performance evaluation · SaveLoadPlp.R Saved results
Components 19
Study design and data
- Prediction design
R/PopulationSettings.R - OMOP extraction
R/ExtractData.R - Covariates
R/AdditionalCovariates.R
Model development
- Preprocessing
R/PreprocessingData.R - Data splitting
R/DataSplitting.R - Feature engineering
R/FeatureEngineering.R - Model fitting
R/Fit.R - Model algorithms
R/CyclopsModels.R - Prediction
R/Predict.R
Validation and evaluation
- External validation
R/ExternalValidatePlp.R - Performance evaluation
R/EvaluatePlp.R - Performance summaries
R/EvaluationSummary.R - Learning curves
R/LearningCurve.R
Results and exploration
- Saved results
R/SaveLoadPlp.R - Interactive viewer
R/ViewShinyPlp.R - Performance plots
R/Plotting.R
Optional integrations
- Python ML
R/SklearnClassifier.R
Other components
- Analyst
- OMOP database
Connections 18
- Analyst → Prediction design: defines problem
- Prediction design → OMOP extraction: specifies extraction
- OMOP extraction → OMOP database: reads cohorts and data
- OMOP extraction → Covariates: uses covariate settings
- OMOP extraction → Preprocessing: provides prediction data
- Preprocessing → Data splitting: prepares data
- Data splitting → Model fitting: supplies training data
- Model fitting → Feature engineering: applies feature engineering
- Model fitting → Model algorithms: uses algorithms
- Model fitting → Prediction: produces model
- Prediction → Performance evaluation: provides predictions
- External validation → Performance evaluation: evaluates predictions
- Performance evaluation → Performance summaries: summarizes performance
- Model fitting → Learning curves: supports learning curves
- Performance evaluation → Saved results: saves results
- Saved results → Interactive viewer: supplies results
- Saved results → Performance plots: supplies performance data
- Model algorithms → Python ML: uses when selected