RiskStratifiedEstimation R package
This tool helps researchers see if a treatment works differently for patients based on their individual risk levels. It uses real-world health data to compare how well treatments work across different patient groups.
At a glance
Use when
Assessing how treatment effects vary by patient risk level in real-world data; conducting comparative effectiveness research across diverse populations; validating findings in multiple databases
Avoid when
Data quality is poor or OMOP mapping is incomplete; outcome prediction model performs poorly; unmeasured confounding is suspected to be substantial
Inputs
Observational health data in OMOP Common Data Model format; definitions for treatment, comparator, and outcome cohorts; prediction models for outcomes and propensity scores
Outputs
Stratified estimates of relative and absolute treatment effects; diagnostic plots for model calibration, ROC curves, propensity score distributions, and covariate balance
How it works
An R package for evaluating treatment effect heterogeneity using a risk-based approach in observational databases structured under the OMOP Common Data Model. It integrates functionality from PatientLevelPrediction and CohortMethod packages, enabling risk stratification through prediction modeling, propensity score adjustment, and estimation of relative and absolute treatment effects across risk strata. The framework supports large-scale regularized regression, diagnostics for model performance, and visualization of results across multiple databases.
- Project
- EHDEN
- Funding
- IMI
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Categories
- RWEHeterogeneity
- Technology
- Non-specific
- Assumptions
- Treatment effects can be meaningfully stratified by predicted baseline risk; sufficient data exists to build reliable prediction models; confounding can be adjusted for using observed covariates and propensity scores
- Strengths
- Enables assessment of treatment effect heterogeneity across risk strata; leverages standardized OMOP CDM data for generalizability; integrates comprehensive diagnostics for prediction and estimation steps; supports multi-database analysis
- Limitations
- Relies on quality of observational data and correct model specification; risk stratification depends on predictive performance of the outcome model; does not account for unmeasured confounding
- Also known as
- RiskStratifiedEstimation
Questions this answers
- › Does this treatment work better for sicker patients compared to healthier ones?
- › How does treatment effectiveness vary across different levels of patient risk?
- › Can we identify subgroups of patients who benefit most (or least) from a treatment?
- › Is the treatment effect consistent across different risk groups?
- › How well does the prediction model identify patients at high risk of the outcome?
- › Are results similar across multiple databases or healthcare settings?
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.