The instinct with any system like this is to collect everything. We think that instinct is wrong, and we would like to know how much less would actually do.
Why it matters
More data is not free. It is more risk if it leaks, more burden to collect, more to keep current, and more that can be wrong. A system holding everything about everyone is also a very different thing politically from one holding the minimum it needs.
Our working principle: keep the minimum sufficient model of a person needed to make good decisions. We do not yet know what that is.
The questions inside the question
- What genuinely changes a decision, and what merely feels relevant?
- How much can be inferred rather than collected?
- What has to be current, and what is fine being years old?
- How does this work for someone with almost no recorded history, which is a great many people?
- What does the system do under partial or intermittent connectivity?
A distinction we keep returning to
A health record answers what happened. Something that maintains health has to also answer what should happen next. Those need different information, and the second may need surprisingly little.
There is a real risk of this project drifting into simply centralising more records. We would consider that a failure even if the records were excellent.
What would count as an answer
A defensible minimum set, with reasoning for each element and an honest account of what is lost by excluding things. Ideally tested in simulation against what fuller data would have concluded.
Help us with this
Clinicians: what do you actually need to know before deciding what someone needs next? Privacy researchers: what would you refuse to hold?