Model Drift
Also known as: data drift, concept drift, performance degradation
Degradation in a deployed model's performance over time caused by changes in the population, the data, or clinical practice, even when the model itself has not been modified.
A model can get worse without anyone touching it. Model drift covers two related effects: data drift, where the distribution of inputs shifts because the patient population, documentation habits, or measurement practices changed, and concept drift, where the relationship between inputs and the outcome itself changes, for example after a new clinical protocol alters how a condition is treated. This is why approval-time validation is a snapshot rather than a guarantee, and why a recurring post-deployment performance review is the governance commitment most likely to catch a real-world failure. Vendor-side model updates produce a similar effect from the opposite direction and should be treated as a trigger for re-evaluation.