Case study presentation- modelling acute hospital demand, collaboratively and openly

Chris Beeley, Head of data science, Strategy Unit

5 November 2025

Intro

  • Why did we build this model
  • What does it do
  • What are the key design decisions from a policy and technical standpoint

Why?

  • New Hospital Programme came to the Strategy Unit c.2020
  • Predict demand for the future of the hospitals c.2041
  • We built on existing work and knowledge in the SU as well as the literature
  • I was not here!

The landscape

  • Lots of models
  • Lots of consultancy support
  • Lots of repetition / duplication
  • BUT no consistency about definitions
  • Methodological progress is slow
  • Proprietary models means progress is not shared

So what does it do?

'Gif of Homer Simpson pressing a button and saying 'do something cool'

The big picture

  • Demographic change
  • Non-demographic change
  • Types of potentially mitigable activity

The model

  • Sample the parameters (assume normal)
  • Calculate demand at IP, OP, A&E level
  • Do this 256 times and plot the distribution
  • The results are conceptually at row level, but not in practice

The principles

  • Probabilistic vs point estimates
  • Transparent and open source vs black box & paid
  • Collaborative vs top-down (done with vs done to)
  • Reproducible vs unverifiable
  • By the NHS, for the NHS vs taking money (and skills) out of the NHS
  • (And now award winning)

Types of potentially mitigable activity

  • This is a key task for modellers and a key output of the work
  • Definitions
  • Local intelligence from collaborative relationships with schemes and ICBs
  • The National Elicitation Exercise
  • Links with 10 year plan, neighbourhoods agenda, and more

Types of potentially mitigable activity (TPMA)

  • This shows why open source and transparency are so vital- imagine proprietary definitions of these activities!
  • (you actually don’t need to imagine, we already have that…)

Now for the (data) science

Gif of a cat sitting on a keyboard with code on the screen

Big list of technical sounding words coming up…

  • SQL -> databricks
  • Azure compute (Docker)
  • Azure BLOB storage
  • Python for the model
  • R for reporting/ dashboards
  • Quarto for documentation

The future

  • National and regional model runs
  • Bring your own data (FDP?)
  • Understanding more about types of potentially mitigable activity, who thinks what’s possible, and why it matters
  • Increasing understanding of the shift from hospital to the community

We believe that as far as possible…

  • Models should properly account for uncertainty in prediction
  • Modelling results should be verifiably reproducible
  • Concepts and definitions (such as of TPMA) should be open, transparent, and properly documented
  • The NHS should develop, own, and run key models in use inside the NHS