Concordance check improves trust in human digital twins predictions
Concordance Plausibility Check for Intervention-Conditioned Human Digital Twins
Artificial Intelligence
Summary
Human digital twins are computer models that simulate how a patient might respond to different treatments that haven't been tried on them yet. Since these alternative treatments can't be directly observed in the same patient, it's hard to know if the model's predictions are believable. The authors propose a Concordance Plausibility Check (CPC), a method that cross-examines real-world data, simulations, and clinical knowledge to see if the predictions make sense. CPC then reports whether the predictions pass the test, fail it, or are uncertain, helping users decide how much to trust the model's advice.
What this means in practice
- •For clinical data teams: Evaluate and validate synthetic treatment predictions to ensure they align with known clinical evidence and causal relationships.
- •For healthcare ai developers: Incorporate the concordance check to improve reliability of AI systems simulating patient-specific treatment outcomes.
Authors
Saeed Farzi, Christopher Kipps, John McBeth, Gabriela Czanner, Sofia Michopoulou, Benjamin Macarthur, Sarah Ennis, Paul Elkington, Nicholas C Harvey, Rahman Attar
Abstract
Intervention-conditioned human digital twins predict treatment alternatives that cannot be observed in the same patient. Ordinary predictive accuracy therefore cannot establish counterfactual credibility. We propose the Concordance Plausibility Check (CPC), a claim-centered framework combining real-world evidence, qualified simulation and clinical knowledge. CPC matches evidence to the causal claim, audits contradictions and shared dependencies, and reports Pass, Fail or reason-coded Indeterminate.