Led by Aarhus University Hospital in Denmark, this use case aims to test the adaptability of an advanced AI/ML model for predicting outcomes in patients with CKD across multiple European healthcare data environments. Originally developed in Canada and previously tested in Scotland and Denmark, this AI model uses a “super learner” approach, which integrates multiple machine learning models to select the best-performing one for each dataset. By using this approach, SHAIPED seeks to accurately predict the risk of kidney failure and mortality for CKD patients in different countries, particularly those with CKD stages 3b or 4. The super learner strategy addresses a common limitation in previous models, which often overlooked mortality risks when estimating non-fatal outcomes.
In this study, the model will be deployed within secure data environments at Health Data Access Bodies (HDABs) in Denmark, Finland, and France, using patient data from laboratory and nephrology records. To enable comparisons, a common data model will standardize variables, allowing the super learner algorithm to evaluate model performance across these datasets. This case study is crucial in demonstrating AI model “transportability,” providing HDABs with validated, predictive AI models that support chronic disease management across diverse populations in Europe.