If the system can say "this person needs a blood pressure check," something or someone made that call. Who, and on what authority?
Why it matters
This is where clinical accountability either holds or quietly dissolves. Get it wrong in one direction and you have an automated system making clinical judgements nobody signed off. Get it wrong in the other and you have a coordination layer so cautious it cannot coordinate anything.
The questions inside the question
- Can a clinical guideline generate something automatically, without a clinician in the loop?
- Can an AI system propose one? Under what constraints, and visible to whom?
- When must a clinician approve before anything reaches the person?
- How are conflicting guidelines resolved, given they frequently disagree?
- When does something stop being relevant, and who decides that?
- Who is accountable when something important goes unresolved for a long time?
What would count as an answer
A clear map of what may be generated by rule, what requires clinical review, what an AI may suggest but never assert, and where accountability sits in each case. Ideally something a medical council would recognise as reasonable.
Related worry
We use the word "obligation" internally for the thing being created, and we are aware the word is doing heavy lifting. It is meant as an operational primitive, not a legal duty and not something anyone is forced to do. But words drift, especially once they become fields in a database. See IHMR-Q-003.
Help us with this
Clinicians and people who have worked on clinical governance or medical regulation: this is yours. We would particularly like to hear where analogous systems have got this wrong.