AI based Rapid Evidence Generator with risk-adjusted propensity score matching

BackgroundThe Rapid Evidence Generator (REG) was developed by the AI-based Risk Prediction and Treatment Effect Estimation (AIR_PTE) German-Canadian Project Consortium funded by the Smart Data Economy Program of the German Federal Ministry of Economic Affairs. It utilizes artificial intelligence (AI) methods for risk modelling in order to implement risk-adjusted propensity score matching.Data and methodsThe REG methods have been developed based on regularly available health claims data from Germany and Canada using the example of a treatment effect analysis for venous thromboembolism. The REG was also adapted and successfully transferred to the cost-effectiveness evaluation of the smartCasaplus geriatric coaching program.ResultsThe derived study results indicate that the REG is a fast, effective, and valid alternative to generate real-world evidence based on retrospective cohort studies. It can be applied to replicate results from randomized clinical trials (RCTs) but also to analyze subpopulations, rare indications, and interventions not suitable for RCT study designs.ConclusionREG study reports can be used to focus innovations, to support approval and postmarket surveillance requirements, and to assist medical professionals in making decisions. The REG methods are currently available as a specific form on the eva self-service analysis platform and as a Python module of the DCC Contract Smart Suite.