Approaches to Improve Slovak Government COVID-19 Model
The Slovak Government introduced COVID-19 Model for the country which, based on my own data & assumptions has room for improvement and therefore better decision making.
This part is a sequel to Why the Government Model Is Wrong. In this post I will try to go further into the details.
No virologists or epidemiologists can estimate the secondary or tertiary effects on the economy. Also — “zero deaths” as public policy at war — is very suboptimal choice, to say at least. Under estimating seriousness of both, can be fatal for the country.
Would you save lives for any cost or would you save economy for any cost? How do you choose the right equilibrium?
How many deaths are “optimal” for 1% decrease in GDP? How many deaths are “optimal” for 10% decrease in GDP? Would you save 1,000 people, if you have to give up 20% GDP? 5,000 for 50%? What are the “right” numbers?
How many deaths would accept for 10,000 unemployed? 10? How many deaths for 100,000 unemployed? 100? Why not less, why not more? How do you choose the right equilibrium?
Is the — “as few deaths as possible” — the strategy? Is the least GDP decrease the strategy? Is the least unemployed the strategy? What is the optimal strategy here?
Who is responsible and owns this exact equations? What are the variables? What is the decision process? Who are the experts on these ideologically opposing sides? Are both sides at the table? Do they understand, they both need to find the right and weighted consensus?
What is the strategy? Who owns it? What are guiding principles & targets?
What is the cost of protecting one life?
What is the cost for protecting 200,000 jobs?
There are no experts on COVID-19. There were none before COVID-19 and there will be too many after COVID-19. Every great expert is flexible in light of new & incoming facts, data and adjusting on the go.
Slovak Government COVID 19 model output introduced 28-Mar-20:
Red — peak of total cases, no intervention.
Green —peak of total cases, including interventions.
Here are the key points for model’s improvements:
- The model itself may be correct, I do not challenge the model itself — I am unclear on that topic yet (see Ferguson’s UK NHS and US capacity models compared later, as I have doubts). Also the authors correctly point out other improvements to the model itself, in future (e.g. foreign travel, temperature, mobility, etc.). However it needs much, much more!
- The room for improvement I see in both: inputs, variables & assumptions — that go into the model. Which then translate to much more improved output (or destruction of the old output, choice of naming is to each her own).
Other Countries & Sample Sizing
- The comparison to other counties data: Coronavirus Daily News Cases are peaking in Europe. The USA, which may be hit the most in next days — daily new cases are peaking in April as well. Slovak Republic wants to toughen the measures, while our neighbour Czech Republic is already thinking of slowly reopening the economy. This would suggest limited sample sizing, some wrong assumptions and variables, even the whole model as such.
Europe, top 5 countries (cases & deaths), peaking:
USA (Beds, ICUs, ventilators — capacity requirements), peaking in April:
Czech Republic (new cases per day), peaking:
- Sample size: Slovakia did only approximately 3000 tests per million inhabitants. Limited sample size (as input) also skews the model output. On top, the more tests you do — the faster you go after the virus outbreaks and limit the future spread. Each country is different in size, however very similar in virus spread & curves. How do you know it’s not just quietly spreading & what’s the plan to make the invisible — visible? Making the known unknown — known known..
Testing Comparison to other countries:
Reproduction of the Virus (R0 value).
- R0 is estimated today (3.4.2020) at 1.65. That value, is too high, too conservative. Estimates from other countries suggest R0 to be around 1.0; 2–4 weeks after “quarantine”, depending on the country. R<1.0 means you are wining in the war on virus — one person infects less than one person.
- If R0 =< +/-1.0 already today (3.4.2020), that would also mean the best decision was to act very early & very decisively — to both enforce quarantine & close borders early. The relative plus could also be limited openness of the country (limited number of foreigners, business travellers, country’s density, etc.). Respecting government rules plays a key role as well. The trust in governance.
- As an addition to the first point (incorrect R0) is — the more Slovak Republic tests, the less % wise infected we find (e.g. 300 tests/20 positives, 1000/30, 2000/25), which would suggest — that despite well targeted testing — we cannot find proof of increased exponential growth! Again R0 = +/- 1.0.
Example from Czech Republic: the more tests the lower the percentage COVID-19 cases = proof of R0 decrease to proximity of +/-1.0, which goes even more down in time.
- Or if the country may go the other way — under-test — which lets the virus silently spread even further. And that will be detected by real numbers later, with critical cases in hospitals & death tolls.
- Zero Testing Assumption: even if the country did 0 tests, the critical cases & deaths won’t be hard to find, they would be very real.
- R0 assumptions are incorrectly modelled across time — from March through April, May, June, etc. — till end, It’s R0 = 1.65 -> 1.4 -> 1.2, in the study. Even the government study suggest this short coming, see model from Japan, as example from the study. Rapid R0 decrease in short timeframe. However, you can use any other country as example.
- If we assume R0 = 1.1, 1.0 or <1.0, weeks after quarantine, the model’s output would blow up. Downwards & bring future events — much, much faster to the present. It could even show peak new cases /day— is already happening +/- week.
R0>1 = Strategic priority is fighting COVID-19.
R0<1 = Strategic priority is fighting for economy restart.
For the whole government, for the whole nation.
- There is also a claim in the model literature — that the more we delay the peak, the longer it will take. This is wrong as the R0 is not static: the number of tests must increase multiple fold, the hunt for positives will bring the peak forward in time and R0<1. Not lowering & delaying it, but hammering it. It’s like saying — if weed multiplies, there is lot of it. No it is not, you see it — you kill further spread.
- The “linear growth theory” — is baseless — either you cannot properly test the citizens (severely slow & low testing) or the exact opposite is true. The slow & low testing can give you false sense of security (real cases of COVID-19 do not equal measured cases). The critical cases & deaths do show up eventually, as both end up in hospital. Even with zero testing.
Mobility Assumption & Reality
- Mobility of the nation is estimated for 100%, 70%, 30–35%. The baseline number is 35%. This number is very conservative based on data from other data providers (google, tomtom, foursquare, tens of them, etc.). The authors of the study correctly points this out, and expect real data from mobile operators.
- During the Quarantine — some data suggest even 2x less mobility than 35%, which again pushes down R0 (as mobility is correlated to spread). Some people work from home, kids are locked, fun, sport, retail is almost closed, people may move — but alone or only for work (essentials), and in “distancing mode”.
Example from Google: Bratislava, the Capital:
- Let’s not kid ourselves, we’re being voluntary self-prisoned in our houses. Also you may not move — but no one is coming to visit you either. Which pushes R0 related to mobility way down.
- For economy re-opening — the mobility will go upwards, therefor you need to plan bio surveillance system, eventually for millions. The model is too static, you would need much more dynamic model, more decisive actions & blitz speed execution.
- You cannot counterbalance the spread by limited mobility for months. You need a vaccine. The vaccine is both — testing & daily bio surveillance system.
Your strategy cannot be trying to lower the R0 — only — by in-prisoning people at home for months. That’s a bad strategy, wrong governance and lack of vision.
- If you really believe the Gov Model output is right, while you will re-open the economy (mobility goes upward) -> that would be a suicide. The R0 would increase — the model does not show this causation, it’s like a gas on fire — if the model’s output is right — which of course I know it is not, based on evidence.
Day Zero Date
- Day 0 is considered 28.3.2020, and from that date — the exponential curve is modelled from. (If I understood the assumption correctly). Why? Why 28.3.2020? Why not much sooner?
- However: you need to go back in time to first case caught (6.3.2020), assume there were more cases (80% asymptomatic) + spread already pre 6.3.2020 + back-model the critical cases & deaths we already have + back-model all potential asymptomatic cases to those; and redo the exponential curve again. And state as Day 0 = say 1.3.2020. Afterwards you will see — the new output does not correspond to previous model. And it would suggest peak new cases/day +/- week in time. This would be an eye opener!
- Modelling the previous — you may easily find the R0 < 1. +/- two cents.
- This is also reason, why the model output from 17.3.2020 (500,000 total cases at peak in July 2020) needed to be adjusted, and the model output from 28.3.2020 (170,000, July 2020) will be also adjusted lower, much lower in the end. Maybe < 10,000 & April; who knows, I am not prophet.
- Another reason why — if the model assumptions are left unchanged, as is — the output will be adjusted downward with each new iterations based on real world data (real inputs=reality check) as you won’t see the exponential growth. Again R0 ≤1.
Capacity Models vs Re-opening Economy toolkits
- Tale of two capacity models: Ferguson, UK NHS model vs US Government model. The Slovak Gov model literature list the UK NHS as example. Honestly, no idea how close these are related (input, assumptions, variables & modelling wise).
- One of the models is wrong. They cannot be both correct! And US Government capacity model beats all others (vast sample size & testing, most data, large resources & land, most complex, 350+ mil. citizens, transparency and co-lead & supply chain managed by the US army).
Ferguson model, NHS capacity model
The UK NHS model was debunked already..
US Healthcare system capacity for COVID 19
- Back to Slovak Gov model: This is a capacity model for healthcare preparedness, not a model for opening the economy. This is the major drawback besides incorrect assumptions.
You cannot reopen the economy with 5000 tests a day. That’s a suicide. Think 1000x! Daily..
How do you get people go back to retail, to travel, to safely work or to drop all fears?
5 million people..
- The Government model also does not count with exponential increase in testing, while you hunt for positives and kill fires on the backend of opening the economy, which pushing R0 even lower.
To really open the economy, you as a state — need to ensure — one person cannot infect the whole factory, a mall, a school, a shop. While one person could be negative one day and positive next. For the economy to get off the knees — you already need to start to work on a daily bio surveillance system, for millions of employees and consumers.
Got it? Good!
This is what the Germans & US is doing on the backend, and will lead and reopen first, and then help its allies (I would assume).
While Slovak Government talks China’s Wuhan or Singapore non sense — these guys won’t help you, the allies eventually will, while some capability could have been already developed at home. Yesterday.
The Government rhetoric and solutions from its study
- Several countries were continuously mentioned in both, government rhetoric and the study: China, South Korea, Singapore or Taiwan.
- China’s data are false by a long-shot (5x-10x+who knows?), this has been debunked at the time of reporting (Jan & Feb 2020) and its falsehood is already a consensus in the community now; from health experts to secret services. No idea why anyone would model — any scenario based on false data sets & curves. Let’s stop using those data, it’s misleading at least & frankly — pure lies.
- Also Slovakia is not Wuhan, just look at the density. Wuhan metro is 20 mil, Slovakia as a whole country is 5.5 mil people. You can lock down a city or village, but why to take the whole country & its economy as hostage — at this stage?
- South Korea has a very unstable neighbour on the north border — North Korea. The country was expecting a bio weapon attack for decades. The citizens were ready to quarantine, with face masks, and obey the rules; immediately. The government infrastructure & response was in place. Long, long before COVID-19.
- Singapore & Hong Kong are not countries, these are “city states”; territory & COVID-19 wise. You lock the borders, quarantine & deploy testing.
So, you do blackout, fine. You destroy the economy.
One infected person can start this madness all over again & you destroy the economy forever.
- The experience, data & common sense suggest — social distancing works, increase targeted testing X-fold, get real time bio control and start to plan reopening. You do not have the capability? Fine, call the allies — we are at war!
- In peace time everyone thinks he or she is a Steve Jobs. Where are all the creative geniuses at war time?
Strategic Priorities for the Country Are Backwards
This means, that based on the output of the Government Model, Slovak Republic has two very opposing top strategic priorities at the moment (3.4.2020). Two ideologically opposing camps — quasi blackout or continued long term quarantine (economy destruction) versus economy restart.
The opposition to “blackout” talk is easy & cheap when lives are at stake, yet there is no single entity in the world challenging the Slovak Government model output, per se. No reporter, no politician, no expert. Yet. None.
There has never been any crucial strategic priority of the country that is as important as the reopening of the economy.
Never, in the history of the country; time to act- is Now.
You cannot simply get this wrong, it would be fatal for the whole nation. This would be destruction. For families, employees, companies, savings, supply chains, consumer confidence, debt, capital, and future generations. If any country makes a strategic mistakes — this stories end up in its history books..
We are at War. The enemy is decentralised & invisible. Nonetheless, it’s a war. Are you willing to die for your Model? I am, for my assumptions. Let’s not let the enemy be ourselves and our economy.
What is the plan to re-open the country?
What’s the strategic toolkit?
When is the time to shift gears? How do you know?
However I can be wrong on one point: we tested very few people and thus the virus is spreading without control. This would suggest that the government’s ability to test targeted individuals or groups correctly — failed. And it could cost us dearly. Let’s hope not, but hope is not a strategy to pursue. Especially now.
Cowards act from place of fear and lead by consensus. Leaders acts from place of strength, are decisive and accept full responsibility at hand. We are at war!
The government can create public — private partnerships or even nationalise. This is war. Why wait & beg some entity for anything?
Aren’t you over reacting if your unemployment rises the same way as countries with thousands of deaths; and tens of thousands of infected?
In the end, the rising unemployment claims would bring all politicians to their senses. But, at what costs & why so late..
Being contrarian is a lonely place. However when all biases are washed away by time — the hidden truth eventually reveals itself — and becomes consensus. And the joy that follows is priceless. Pure joy for everyone..
Good luck, this post is dedicated to all & everyone really trying hard on the frontlines to help their nations and citizens. We are grateful & you are our heroes!
She was exponential..
You can find updates here, if any.