Our work
HFA's priorities are set by its Steering Committee and carried in a work plan that is revised as circumstances change. What follows describes how the work is done, the components it is organised into, and the threats it is currently applied to.
The obstacle, and the way around it
The obstacle to collaborative forecasting has rarely been method. It has been that the data cannot move — for reasons of law, of sovereignty, and of patient privacy that are entirely legitimate. HFA is built so that the data does not have to move.
Models travel to the data
There is no central repository of primary national data. Models are trained and run locally, on infrastructure the data holder controls. What is shared between partners is anonymised model parameters and the resulting insights, using established privacy-preserving methods. A member is never required to share data as a condition of taking part.
The cycle of work
Each stage belongs to someone. Priorities are set together at the start; the data never leaves the authority that holds it; the decision at the end belongs to the government taking it; and what is learned returns to the next round of consultation rather than ending there.
- Consultation on priorities Jointly determined.
- Data Held under national or regional sovereignty, and staying there.
- Analysis and modelling Carried out at hub and national level.
- Intelligence Made transparent and actionable.
- Decision and response Authority rests with each government.
- Reflection Joint, published, and carried forward.
- Reflection carries into the next round of consultation. The cycle is continuous, and HFA's accompaniment does not stop at the point of decision.
Outputs that can be interrogated
A forecast that a health official cannot question is a forecast they cannot responsibly act on. HFA works with explainable methods, so that the reasoning and the data behind a prediction can be examined by the people who have to use it. HFA also publishes how its forecasts performed: a body that asks authorities to act on predictions has to show what those predictions were worth.
Built on shared public infrastructure
The technical work follows digital public infrastructure principles for health data exchange, so that what is built in one region can be adopted in another instead of being rebuilt from nothing.
One Health throughout
Human, animal and environmental health are studied together rather than separately, an approach generally described as One Health. The threats HFA is most concerned with cross species and borders, and they cannot be forecast well by looking at people alone.
Five key components
The work is organised into five components. They describe work rather than standing bodies, and each one covers the stages of the cycle it serves rather than owning a stage of its own.
| Component | Stages served |
|---|---|
| Community of Practice The space where reflection and consultation happen | Every stage |
| Federated Analytics and Model Exchange Standards, architecture, model exchange | Data, analysis, intelligence |
| Advanced Computational Capacity Shared platforms and access | Data, analysis, intelligence |
| Support to Decision- and Policy-Making Translation, uncertainty, briefs | Intelligence, decision, reflection |
| Capacity Building and Knowledge Integration Continuous accompaniment | Every stage |
Access is not determined by ability to pay
Computing capacity is where equity in this field most often fails in practice. Access to the infrastructure HFA provides or brokers is not determined by a region's ability to pay for it.
Current use cases
Work is organised around threats the hubs have prioritised together. Each is taken forward by a team drawn from more than one region.
What is developed is shared
Methods, models and tools developed through HFA are held for the benefit of its members and the wider public health community, and are openly licensed by default. Anything a member brings with it remains the member's own.