Model Drift
Also known as: Data Drift
The gradual decline in an AI model's accuracy as real-world data changes from what it was trained on.
Model drift describes the gradual decline in an AI model's accuracy over time as real-world data, such as patient population, imaging equipment, or clinical workflow, shifts away from the data the model was originally trained and validated on.
In a clinical setting, model drift is a key reason vendors and health systems are expected to monitor deployed AI tools on an ongoing basis rather than treating validation as a one-time event at go-live.