How it works | Open Health

How Open Health works

Open Health connects organisations around a shared health challenge, establishes the conditions for collaboration, and enables distributed data, models and services to work together under governed conditions.

A practical example: healthier eating and weight management

Imagine a university research group, a grocery retailer and an NHS network trying to answer a shared question:

Which changes in eating behaviour are most likely to improve health for different people, and which changes are they most likely to sustain?

Each organisation holds only part of the picture.

The university brings research expertise and health outcomes.
The retailer brings insight into everyday food purchasing behaviour.
The NHS network brings a broader patient population.
A GP needs useful, specific advice for an individual patient.
A governed digital assistant can support authorised analysis and recommendations.

Open Health allows these capabilities to work together without requiring every underlying resource to be moved into one organisation.

1. Discover the right partners

Organisations use the Partner Hub to describe what they can contribute and identify complementary partners.

Only information about capabilities is shared at this stage. Sensitive health and commercial records remain with the organisations that hold them.

2. Establish the collaboration and its authority

Partners agree the purpose of the collaboration, the resources involved, responsibilities and the legitimate basis under which each participant may contribute, request or receive something.

These conditions establish the governed context. Open Health does not create that authority; it makes the relevant parts explicit and enforceable.

3. Check each interaction before it happens

When an organisation, person or AI agent requests an operation, governed admission evaluates that request against the agreed conditions.

The decision is specific to the participant, purpose, operation, resource, scope and current collaboration context.

An ALLOW applies only within those limits. It does not create broader rights.

4. Execute within those limits

The authorised computation or service can now run across distributed resources.

Resources can remain under organisational control, but equally importantly, the authority established before execution remains binding during the interaction.

5. Release only what is authorised

A participant receives only the result, model use or derivative that its permissions allow.

Contributing to an activity does not automatically grant rights to its outputs, and an AI agent cannot expand either its own authority or that of the person requesting the result.

For example:

the university may receive research findings;
the retailer may receive evidence relevant to personalised healthy-eating services;
the NHS contributor may strengthen the model without automatically receiving the full model;
a GP may receive an authorised recommendation for a patient.

From prediction to useful services

The first stage can show whether organisations can build a stronger shared prediction model across distributed data.

A later stage can address the more important real-world question:

Which dietary changes actually improve health, for whom, and which changes are sustained?

This is how Open Health moves from distributed data and analysis towards person-centred services built across organisational boundaries.