Glossary
Retrospective study
A study that looks back at data already collected, such as past medical records or scans.
A retrospective study analyzes information gathered in the past. Many AI medical studies are retrospective: researchers run a model on stored scans or records and compare its output with what was later known about the patients.
These studies are relatively fast and cheap, and are useful for early evidence. But the data may be cleaner or more selected than what a tool would face in a real clinic, and the study cannot show how doctors and patients would respond to the tool's output.
Strong performance in a retrospective study is a reason to run a prospective one, not proof that the tool works in practice.
Read more: our explainer