|
| 1 | +--- |
| 2 | +title: "Case study presentation- modelling acute hospital demand, collaboratively and openly" |
| 3 | +author: |
| 4 | + - "[Chris Beeley, Head of data science, Strategy Unit](mailto:[email protected])" |
| 5 | +date: 2025-11-05 |
| 6 | +date-format: "D MMMM YYYY" |
| 7 | +format: |
| 8 | + revealjs: |
| 9 | + theme: [default, ../su_presentation.scss] |
| 10 | + transition: none |
| 11 | + chalkboard: |
| 12 | + buttons: false |
| 13 | + preview-links: auto |
| 14 | + slide-number: false |
| 15 | + auto-animate: true |
| 16 | + footer: | |
| 17 | + Learn more about [The Strategy Unit](https://www.strategyunitwm.nhs.uk/) |
| 18 | +--- |
| 19 | + |
| 20 | +## Intro |
| 21 | + |
| 22 | +* **Why** did we build this [model](https://www.strategyunitwm.nhs.uk/new-hospital-programme-demand-model) |
| 23 | +* **What** does it do |
| 24 | +* **What** are the key design decisions from a policy and technical standpoint |
| 25 | + |
| 26 | +## Why? |
| 27 | + |
| 28 | +* New Hospital Programme came to the Strategy Unit c.2020 |
| 29 | +* Predict demand for the future of the hospitals c.2041 |
| 30 | +* We built on existing work and knowledge in the SU as well as the literature |
| 31 | +* I was not here! |
| 32 | + |
| 33 | +## The landscape |
| 34 | + |
| 35 | +* Lots of models |
| 36 | +* Lots of consultancy support |
| 37 | +* Lots of repetition / duplication |
| 38 | +* BUT no consistency about definitions |
| 39 | +* Methodological progress is slow |
| 40 | +* Proprietary models means progress is not shared |
| 41 | + |
| 42 | +## So what does it do? |
| 43 | + |
| 44 | +<img src="https://media1.giphy.com/media/v1.Y2lkPTc5MGI3NjExMTc3dXMyanF0cjJoem16bjhieDZ5djRqZ2RoNDd2eG43aDg3aXkweSZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw/l2JdSlA1a1zKVAyze/giphy.gif" alt="'Gif of Homer Simpson pressing a button and saying 'do something cool'" height="500"> |
| 45 | + |
| 46 | +## The big picture |
| 47 | + |
| 48 | +* Demographic change |
| 49 | +* Non-demographic change |
| 50 | +* Types of potentially mitigable activity |
| 51 | + |
| 52 | +## The model |
| 53 | + |
| 54 | +* Sample the parameters (assume normal) |
| 55 | +* Calculate demand at IP, OP, A&E level |
| 56 | +* Do this 256 times and plot the distribution |
| 57 | +* The results are conceptually at row level, but not in practice |
| 58 | + |
| 59 | +## The principles |
| 60 | + |
| 61 | +* Probabilistic vs point estimates |
| 62 | +* [Transparent and open source](https://www.strategyunitwm.nhs.uk/news/transforming-hospital-planning-open-source-demand-and-capacity-model) vs black box & paid |
| 63 | +* Collaborative vs top-down (done with vs done to) |
| 64 | +* Reproducible vs unverifiable |
| 65 | +* By the NHS, for the NHS vs taking money (and skills) out of the NHS |
| 66 | +* [(And now award winning)](https://www.strategyunitwm.nhs.uk/news/strategy-unit-demand-model-wins-prestigious-florence-nightingale-award) |
| 67 | + |
| 68 | +## Types of potentially mitigable activity |
| 69 | + |
| 70 | +* This is a key task for modellers and a key output of the work |
| 71 | +* Definitions |
| 72 | +* Local intelligence from collaborative relationships with schemes and ICBs |
| 73 | +* The National Elicitation Exercise |
| 74 | +* Links with 10 year plan, neighbourhoods agenda, and more |
| 75 | + |
| 76 | +## Types of potentially mitigable activity (TPMA) |
| 77 | + |
| 78 | +* This shows why open source and transparency are so vital- imagine proprietary definitions of these activities! |
| 79 | +* (you actually don't need to imagine, we already have that...) |
| 80 | + |
| 81 | +## Now for the (data) science |
| 82 | + |
| 83 | +<img src="https://media3.giphy.com/media/v1.Y2lkPTc5MGI3NjExMjVkNTFoMXNrb3kya3c1aHMzc3VwNGtmbmNkOWJmeHhhemE3NG8wbiZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw/3oKIPnAiaMCws8nOsE/giphy.gif" alt="Gif of a cat sitting on a keyboard with code on the screen" height="500"> |
| 84 | + |
| 85 | +## Big list of technical sounding words coming up... |
| 86 | + |
| 87 | +* SQL -> databricks |
| 88 | +* Azure compute (Docker) |
| 89 | +* Azure BLOB storage |
| 90 | +* Python for the model |
| 91 | +* R for reporting/ dashboards |
| 92 | +* Quarto for documentation |
| 93 | + |
| 94 | +## The future |
| 95 | + |
| 96 | +* National and regional model runs |
| 97 | +* Bring your own data (FDP?) |
| 98 | +* Understanding more about types of potentially mitigable activity, who thinks what's possible, and why it matters |
| 99 | +* Increasing understanding of the shift from hospital to the community |
| 100 | + |
| 101 | +## We believe that as far as possible... |
| 102 | + |
| 103 | +* Models should properly account for *uncertainty in prediction* |
| 104 | +* Modelling results should be *verifiably reproducible* |
| 105 | +* Concepts and definitions (such as of TPMA) should be *open*, *transparent*, and *properly documented* |
| 106 | +* The NHS should *develop*, *own*, and *run* key models in use inside the NHS |
0 commit comments