Covid-19 update: What is happening now in London? What is (or could be) next?

Case 3: London. Authors: Alan Patrick[1], Maurice Glucksman[2]

YOU CAN download AND make your own COPY of the London model here

This article is reporting in more depth on what has happened in London through the lens of our simulation model. We are now pivoting from “understanding what has happened” to begin to try to answer: “what is next”? or more accurately: “what could be next”?

A short recap of what has happened is repeated here for context and we are showing explicitly where our model of London reveals information that is not, as far as we know, reported anywhere else. Then we use some simple scenarios to ask: what if we do different approaches to easing the lockdown? What are the implications for more what combination of action could be a good way to exit from COVID-19?

This is obviously going to be a contingent answer. Every day we get new information about the nature of COVID-19 and therapies and even possible vaccines. Some of the new information has a material impact on the exit strategy. We will be updating the results and testing continuously to ensure we are adapting rapidly.


The first reported case in The UK was on 23rd January when a person flew into the UK from China. The first UK local case was reported on 19th February. Despite other parts of the UK having recorded infected people first, London fairly soon became — and remains as of today — the UK’s epidemic centre. A large and transient multi-national population, four international airports and one Eurotunnel hub, little screening of arrivals (to this day), a high population density and (still) crowded public transport are most probably the main causes. The UK Government’s originally tried a “test and trace” strategy. It was largely the rapid take-off of the disease in London that forced it to shift to an increasingly severe series of public health strictures, culminating in a strict, formal country-wide lockdown on the 23rd of March.

As of today, London is in its 4th week of lockdown, and the UK government has to now decide the best exit strategy from the lockdown, its choices are essentially whether to continue with the current strictures, tighten them up further, or start to lift them.

To examine possible exit strategies, we have built a system dynamics model, validated it against peer reviewed research and constructed a range exit scenario to see how those options impact our choices. Right now, these are presented crudely as single policy choices. The reality is there are multiple policy options and some of them work powerfully in combination. We are headed there shortly but not in this article. This article is to give you a feel for what is possible.

About the Model

The model is a system dynamics simulation of the end to end flow of a population (London, in his case) from uninfected to infected, and then through the progress of the disease to recovery or death. It is driven by the statistics of the disease, and the assumed response. The London response is shown in Exhibit 1 below:

The red line represents the effect of lockdown in terms of number of people contacted (i.e. possibly infected) per day per person — the sharp drop from Day 75 represents the announcement of the lockdown on March 23rd. It rises depending on the scenario,

The blue line similarly the impact represents social distancing and self-isolating practices by people before lockdown.

The simulation takes whichever is the lower (more-aggressive) line.

We have built the model so that it can be downloaded for free, and users can input their own data and change the core disease assumptions to model a region or various scenarios themselves.

We have used a combination of NHS, Dept of Health and Office of National Statistics data, plus newspaper reports for the purposes of developing this case. As noted above, the reporting for London cases is inconsistent. We are continually updating this as new information becomes available.

Using the Model

Some of the information from the model can be seen directly on the dashboard (Exhibit 2). The available NHS data (red lines in Exhibit 2) show hospital deaths in the model is current up to 17th April 2020 — we have adjusted the data (green lines) in Exhibit 2 to reflect our best estimates where there are conflicting sources. The model tracks our best estimates. The blue lines show how the behaviour develops going forward for variables where we have data. And the model also shows the history and projection for variables where we have no data at all.

The Set-up Dashboard in the London model

Some observations and implications from these charts are shown in the blue circles in Exhibit 2,

  1. Virtually all of the data from London is messy. It is very difficult without something to link everything together to make a good estimate of what is likely to be happening. The historical part of simulation allows us to do that because if the pieces of this puzzle do not fit together, the simulation simply will not work. If you will this is enforced rigour by the physics of the model. Because of the physics, the model also helps us make better estimates or select the best source where there are conflicting data sources. We have also in some cases made adjustments to the data where conflicting sources do not track back in history. An example is the recent revision of deaths by the Office of National Statistics stating that about 50% more deaths were happening outside hospitals. Many in care homes. The blue line model tracks our best estimates. The blue lines show how the behaviour develops going forward for variables where we have data. And the model also shows the history and projection for variables where we have no data at all.
  2. Having synchronised the simulation with the historical record allows the model to carry on and show for the charts where we have historical data what is likely to happen next. Going forward there are just two assumptions: a) The epidemiology does not change (the virus does not mutate to a strain with dissimilar characteristics; b) We have assumed that social distancing rigour is gradually relaxed but not more than would cause the active cases and deaths to escalate above levels already experienced
  3. We also can see things that are simply not reported historically and going forward, such as the future track of the disease.
  4. With 3 we have a much clearer foundation for anticipating what is likely to happen and planning for that. We can also vary our assumptions and see what is likely to happen. That is the subject of the scenarios that follow
  5. We can see where London ends up at day 180 (3 million people remain susceptible in our Base Case) and judge whether that is a desirable or undesirable number. It is on the one hand obviously very good that 3 million people have not had to endure the worry and possible suffering to survive and their families have not lost them. On the other hand, they are still susceptible.

From this model set-up we immediately get these benefits.

  1. It helps reconcile a messy data set, so there is just a much higher level of rigour available to cross check the data
  2. The model fills in information that is not published anywhere we know of. This creates a consistent idea about what is going on, instead of having to guess about many things.
  3. The ground work is done to evaluate the impact of actions and develop compound strategies. We believe we can ask much better questions when setting up exit strategy scenarios
  4. As more and better data becomes available, we can rapidly update the model and see where our understanding is impacted and how 3 might change

Modelling Exit Scenarios

To understand what Exit Strategies exist, we model a Base Case (our best model of the existing situation) and 3 other scenarios. These scenarios are shown below in Exhibit 3 below and show the assumed contacts per day before and after the respective lockdowns and lifts. The cases are:

  • The Base Case — what happens if we keep on with the current lockdown until late July, when the simulation ends, c 4 months after the lockdowns start.
  • Scenario One “Stricter Lockdown” — We assume a tighter “Lombardy style” lockdown from at in late April. The aim is to eradicate the disease almost completely.
  • Scenario Two “May Release” — Partial lifting of the current lockdown level from the end of May, with increasing freedoms over time, but some elements of enforced lockdown persist.
  • Scenario Three “Early Release” — New York style lifting of the current lockdown in early May, to a similar lower level of enforced lockdown as in Scenario Two.

For the purposes of this exercise we assume:

  • There will no vaccine available in the time period of this study.
  • Mass testing is not available as a solution in the time period and for a few months after.
  • When c 8m of London’s c 9m population has had the disease, we assume “good enough” Herd Immunity exists with little risk of an unmanageable Wave 2 occurring.

The case studies alter the lockdown policies from c Day 100 (Mid-April). Again, the simulation takes whichever is the lower (better) line. The case studies use our best estimated datasets at the time.

The results in terms of case load, deaths, and remaining uninfected people for each scenario are summarised in Table 1 below.

Table 1 — Impact of Various Exit Scenarios

These simulations give a “solution envelope” to plan from — a range of case and death levels and peak hospital loads and timings, and the range of timings when an effective level of herd immunity can be reached.


The big picture is that the tighter and longer the lockdown the lower the case load and pressure on the health service and the fewer deaths. However, there are 2 downsides:

  • The stricter the lockdown, the less uninfected people — i.e. “herd immunity” — it delivers by Day 180, the period end. The two strict lockdown scenarios (Base Case and Case 4 “Stricter Lockdown”) have not reduced London’s uninfected people to 1m or below by the end of July, necessitating ongoing protective steps until a vaccine is found.
  • Both scenarios require a high level of social and economic inactivity for a long time. Neither of these are likely to be possible for 4 months, never mind longer, without high levels of public unrest and economic destruction, and very possibly both.

The other key aspect is deaths — the total deaths range from c 15,000 to c 45,000 depending on the scenario.

The tighter lockdowns cost far fewer lives in the period simulated (15,000 for Stricter Lockdown and 27,000 for the Base Case respectively) but as noted above leave many millions (6m and 3m respectively) people are uninfected, with a high and potential for further deaths without ongoing lockdowns.

The strictest scenario (Case Four “Stricter Lockdown”) leaves about 2/3 of the population still not infected, leaving a high risk of a large “Wave 2” infection that could easily swamp any “test and trace regime” if the lockdown is released and could require another major lockdown

The two lightened lockdown cases (Case 2 “Early Release” and Case 3 “Partial Release” incur more deaths (from 35,000 to 45,000) in the period, but after that effective herd immunity has been achieved and there is no need for further vigilance.

London at the crossroads

Lombardy, Hong Kong, Athens and Australia have made a choice to limit the deaths as much as possible, but to follow through with that choice they are effectively living in lockdowns now until a vaccine is available or a breakthrough in recovery therapy is found.

At the other extreme, New York appears to be in a sprint to herd immunity by end of May. Sweden is on the same track but perhaps with geographic advantages that have allowed testing and isolation to be a mitigation strategy that the Swedes have judged an acceptable risk.

Where next: London?

We know that the cases from the initial sources of infection may have peaked, and the model predicts (and latest data confirms) the death rate is also declining. This is in some cases being mistakenly taken as evidence that we are past the point of maximum danger. But because the social distancing measures have been successful (it may be they are more like the Stricter Case”, the base case shows there is still a huge proportion of London’s population that remains uninfected, so removing the lockdown can only work if the post-lockdown measures are sufficiently effective to stop the disease from blowing up again. This will be true for some time to come.

These are London’s main considerations now:

· How long, realistically, can a severe lockdown be enforced before large numbers of people (mostly those either already infected or with little to fear from Covid) refuse to obey.

· Similarly, how long can London (and London is the engine of the UK’s economy, so it’s a UK problem) afford to stop its economy, education system and other major functions operating.

· Are there other policy options that can work in tandem with any type of sudden or gradual lockdown release that can help London start to function more normally without risking so many lives?

It may seem like a let-down to leave it here. But we are not leaving it here, we have to get more information about London’s capacity to manage an exit and then work to get better answers. But we are leaving it here for this article.

How to manage the next phase depends very much on estimates about what medical breakthroughs may be on the way and how well prepared the Hospitals and healthcare workers are, and how many (and how healthy) they are.

That is the end of this update on London. We will be working to develop additional scenarios and incorporate more data as it becomes available.

You can help by giving us your feedback and ideas so we can try to improve on what may seem to be a choice between a few very unpleasant options.

We believe we can all do much better than that and we are aiming to contribute to the solution.


[1] Alan Patrick Alan studied Engineering and IT systems, and has lectured on simulation and systems dynamics. He has consulted for Deloitte and McKinsey, and worked at senior levels in Telecoms, Internet and Marketing businesses. Recently his team competed in the IARPA Geo-political challenge and won prize for the DataSwarm predictive systems. Alan is the leader on this London case and is working on localisation for Sweden.

[2] Maurice Glucksman is an Investor and former Equity Analyst now focused on Disruption. He is a published author on business and policy analysis using Simulation methodologies and was an Associate Principal with McKinsey & Co. He is now also an advisor to businesses in Aquaculture Health, Artificial Intelligence, Marine Shipping, Pharmaceuticals and others. He is one of the architects of this project, a developer of the simulation model and co-author of a number of the localisation cases

Investor and analyst focused on disruption. Former equity analyst and management consultant with Engineering and Management degrees from MIT and U of Michigan