Insights
Articles on clinical AI
Practical articles on building and adopting AI in healthcare, each with references you can check. Clinical statements are awaiting review by a named clinician, as marked on each article.
AI scribes for clinical notes
Ambient AI scribes promise to give clinicians their attention back. Early studies are encouraging but mixed, and design details decide whether the promise holds.
Evaluating clinical AI before use
Good test-set performance is only the first of four checks. A practical guide to the others, using the reporting guidelines the field already agrees on.
Protecting patient data in AI systems
Health data protection depends less on policy documents than on where data goes, who can reach it, and what is kept. Six architecture choices that make privacy hold.
From spoken family history to pedigree
What it takes for software to turn a patient’s spoken family history into a standard pedigree that a genetics team can trust and correct.
Making clinician review work
“A clinician reviews every output” is only a safeguard if the review is designed to work. What research on automation bias says about building it well.
AI for tumour board preparation
Multidisciplinary team meetings change how patients are assessed and treated, and they take a lot of specialist time. Where multi-agent AI can help with preparation, and where it must stay out.
Showing uncertainty in clinical AI
A prediction without a sense of how sure it is invites both over-trust and dismissal. Why calibration, case-level uncertainty, and the option to abstain belong in clinical AI.
Home rehabilitation after surgery
Most recovery after surgery happens at home. What the evidence says about telerehabilitation, and what software must get right to support it safely.
Connecting AI tools to health records
Most clinical AI projects that stall do so at the interface, not the model. Why integration should be designed first, and what FHIR and SMART on FHIR do and don’t solve.
Measuring time saved by AI tools
Claims that AI saves hours are easy to make and hard to check. A practical approach to measuring what actually changes before, during, and after rollout.
