Building on The Hammer and The Dance Covid -19 Action Plan: It Has To Be Local

Viral infections happen in clusters. A forecasting tool that can be localised is essential: e.g. where should the next shipment of respirators be sent?

On 19 March 2020 Thomas Pueyo published this follow up article that examines in detail at a country level how to manage the Covid-19 response. He called it ‘The Hammer and Dance” to provide a metaphor for hitting hard like a hammer and then reacting in a dance with the outbreak to keep it at bay until a vaccine is available.

In response to this article a week earlier by Thomas Pueyo: a sobering call to action; we were by coincidence working on a tool that can be part of implementing the Hammer and The Dance. We believe this is powerful idea and until something better comes along its important to support it.

Decisions are made in different countries at different levels whether at nation state, county, town or village. Viral infections happen in local clusters that do not respect these organisational structures. Viral outbreaks must be managed locally around the cluster.

This article is a summary of work over the past week. It is just a start to answer a multitude of questions about how the Covid-19 virus and load on the healthcare system is evolving. We are using dynamic modelling to look ahead and anticipate bottlenecks and stresses. Our first focus is the stress on the healthcare system.

After you read it you will know the following:

1. National actions to halt to slow or contain Covid 19 are critical but must be localised. Transmission happens in local clusters and each one can be different
2. To do this effectively it is necessary to forecast and understand what actions have leverage for change in each cluster
3. We have developed a dynamic model tool (as of 28 March — you must use any web brower but not Microsoft EDGE — it does not work) that can be localised to a region, a city, a town, a hospital catchment area, even a building. The tool sees ahead and calculates leverage
4. The tool is distributed free with some details at the end of this article and we are asking for feedback and ideas for how you would like to use the tool

Progress so far:

Thomas Pueyo made abundantly clear urgent action is needed to slow the spread of Covid -19 at a country level — which is excellent and has strongly influenced us — but the actions we take must be heavily influenced by local conditions. Since last week In the neighbourhood of New Rochelle New York there is a complete lockdown with nobody in or out. Similar measures have been put in place in South Korea, Singapore and Hong Kong. The precise actions are all different but share one thing in common: they are at a neighbourhood level. And to implement them well it is important to know what is happening not just now but what is expected in the weeks and months ahead. It is not just for the virus but also the healthcare system’s capacity.

It is complex, and it is personal and local. It requires answers to questions like these. :

1. What is happening in my neighbourhood: how many people are infected but with no symptoms and contagious? What if someone walks out their door and walks past someone: how likely are they to be infected without knowing? What is the probability that chance encounter spreads the infection?
2. What if medical help is needed? It can take up to two weeks before symptoms show and the local hospital may already be struggling to cope. What is the chance of getting adequate help two weeks from now?
3. In the hospital, will there be enough supplies and doctors and nurses? Will that impact the chances of recovery? What if all the doctors are getting sick? Will there be enough housekeeping staff to keep the hospital clean?
4. Where should the logistics team send the next shipment of respirators? How many lives will the shipment save?

The answers cannot be delivered with a single discipline. It is necessary to draw on expertise from epidemiologists, doctors, virologists, healthcare process management, economists, supply chain management and more.

All of the answers are changing daily. It is a rapidly evolving dynamic complex, interconnected web of interactions.

Over the past week we formed a team of 24 people each with different domain knowledge and expertise and individually we have all reached out to many more. Our aim is to deliver a tool that anyone can use for free to answer these questions.

Many of our team have been working with dynamic models to solve complex problems for decades so our collective instinct was to build a dynamic model see what it says. We did that and benchmarked it using the Imperial College study published on March 16th to validate the model.

The Imperial College study has been criticised for not modelling the spread of the virus the right way: some of the criticism revolves around taking local measures to contain the virus and possibly misrepresenting the scale of subsequent outbreaks because of the modelling methodology. Others are much more qualified to comment on that that I am, some of them are in our team. While the benchmark may not be perfect and can be improved upon we believe it is directionally correct and accurate for our purpose: to assess the impact of today’s actions on the longer term outcomes, in this case survival rate from Covid -19.

Most of what we have read including Thomas Pueyo’s excellent articles examine the challenges from a country or regional level. We want to address the need to see a little bit ahead in local areas: quantify as best as possible how fast the outbreak is growing and what will happen at local hospitals. The dynamic model we have developed is for the whole of the UK, but it can be very rapidly adapted for local areas with a few small changes. To give you an idea of what it produces, here are some examples for the whole of the UK.

We are at day 57 in this simulation:

A live copy of THE MODEL CAN BE ACCESSED HERE: warning this is work in progress but you can make a copy and do what you like if you make your own copy: its free. For the latest version please contact the author.

For some help see

************Sheetless tips and references below**************

Next Steps:

We will now work to clean up and make the tool more accessible and useful and adaptable to local conditions.

We plan to provide a roadmap for how things are likely to play out over the coming weeks in anyone’s neighbourhood. It will provide a foundation for preparations to be extra vigilant when the infections in a local area are at a peak. It will also help us better understand when the local healthcare system will be in a position to provide adequate medical care.

In the coming days we will do our best to explain how to use it and adapt it so it is possible to localise it and personalise it. We hope others will find it useful and build on it or try to personalise it, share that with us and make suggestions.

If you have any suggestions please send them through Linked-in here

Acknowledgements: This was a rapidly assembled team effort. Special thanks to Dr Hans Schepers and Dr Kim Warren for modelling help, and thinking quality checks. Thanks also for all the timely inputs and advice from: Stephen Allott, Dr David Bennett, Partha Bose, Alvin Cheng, Dr Andreas Coutras, Dr Glenn Cornett, Dr Larrie Ferriero, Hans Chistopher Fuchs, Dr Myron Glucksman, Dr Pantelis Katharios, Barbara Mester, Doug Mester, Dr Romi Navaratnam, David Partridge and Alan Patrick. Thanks to other members of the modelling team who learned how to use the software in 1 day: Markos Glucksman and Argyaputra Sakti. And we are really grateful to John Hill and Chris Spencer for creating the Sheetless modelling platform: we have learned how to use so rapidly and now we can seek feedback from anyone, anywhere, for free, on our live model.

**************Sheetless software tips and references***************

• The model can be opened, changed and explored in a web browser immediately, with no need to install anything
• A Sheetless community account can be opened for free at sheetless.io to save a copy of the model for yourself
• The basics of how to use Sheetless are at sdl.re/sheetlessstart
• If you want more in depth training to understand what ‘Stocks’ in the model are all about and reference material this Linked-In article sdl.re/LIPstock is useful.