Data governance
The rules and roles that decide who may use which patient data, for what purpose, with what protections, and who answers for it when something goes wrong. AI raises the stakes in three places. Training a model is a new use of data collected for care, so it has to be checked against the legal basis and approvals the data was collected under. Sharing data with a vendor needs a contract that states whether the vendor may keep it or train other products on it. And outputs such as risk scores written back into the record become patient data themselves. When a vendor asks for a data extract, data governance is the process that asks what for, how much, and for how long.
A radiology department shares ten years of de-identified CT studies with a startup in exchange for discounted licenses. Two years later the same data shows up in the training set of a product the startup sells to other hospitals. Whether that was allowed depends on contract wording nobody in radiology read. The question to settle before any extract leaves the building is what the recipient may do with it once the original project ends.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →