What is the impact of vaccination if we manage to do this in 200 days? 

Last weekend we showed the impact of de UK variant and what would happen if we do not act.

We do have  vaccines and today we would like to show a peak view on a research we did that will be published shortly.

Let´s focus on the impact of vaccination.

 

In our model we simulated the current Dutch strategy, vaccinating from older people towards the younger ones (90+ downwards to 40) in 200 days.

Meaning 10 million people vaccinated in 200 days (50 thousand per day).

 

We  illustrate the impact of the virus on the health care capacity.

To get a clear picture we abstract from the effect of the reproduction factor by assuming it is equal to one, meaning that spreading of the virus remains stable.   

The efficacy of the vaccination assumed to be 90% and vaccinated persons can still infect others.

 

In the following pictures you see the development in units of 16 days,  starting from January the 11th.

The 1 in the graph on the x-axis  illustrates the average number of hospitalizations in the period from January 11th until January 23rd.  

The colors of the bars in the graph illustrate the severity of the level of pressure.

When the level of pressure is above 50% it´s red, when the level is inbetween 25% and 50%, it´s yellow and when it´s below 25% it´s green.   

If we can manage to vaccinate the 10 million in 200 days then we can reduce the impact on the healthcare system massively.

The pressure reduces with more than 80%!

In the above graph we show the development on the infections. The 1 in the graph on the  x-axis, illustrates the same time period as the first graph.

The colors of the bars in the graph illustrate the severity of the level of pressure.

When the level of pressure is above 60% it´s red, when the level is inbetween 50% and 60% it´s yellow and when it´s below 50% it´s green.   

If we can manage to vaccinate the 10 million in 200 days then we can reduce the infections with 44%. 

 

Bottum line it shows that if we manage to vaccinate in 200 days the impact is severe.

The effectiveness of vaccination can be increased by following a different age dependent vaccination strategy and will be be more effective if we assume that a person is no  longer able to infect others when being vaccinated.

 

The  effect of the UK mutation on the number of infected and severe ill

Since November 2020 a new  mutation of  the Sar-Cov-2  virus  started spreading in the Netherlands.  The mutation was first registered in the UK and is therefore called 'the UK variant'.  This UK variant seems to be much more contagious  than the original variant, resulting in a 30% higher reproduction factor  (R) compared to the original variant.

This new variant of the virus made the government of our  country decide to tighten the measures. Currently 10% of the  infections is due to the UK variant, but due to it's higher reproduction factor it is likely that in the coming period this variant will increase it's dominance  and will possibly take over fully.  When this happens and now additonal measures are taken, the effective reproduction factor will go from the current 0.9 to 1.3, resulting in additional hospitalizations. 

 

To give an impression of the development of the infections, we calculated the development  with the current mixture of 10% the original variant  and 90% of the UK variant. The reproduction factors are 0.9 respectively  1.3.  From the graph we can see that the  share of the original (normal) variant  will decrease, where the share of the  UK variant will increase . Assuming no additional measures we can see that by the third week of February  50% of the infections is from the UK variant. At the end of April the infections are only the UK variant, since the original variant is (almost) vanished.  Due to the higher reproduction factor of the UK variant, we see that the number of infections increases again.

 

Connected to the number of infected are the number of severe ill people. After being infected a part of the people will be severe ill and will be hospitalized. Taking into account the development of the infections and the age-distribution, we see that the  number of hopitalizations will first decrease to an minimum of 1.573 (on average) and than will increase to a 5.000 by the beginning of May. It is clear that a number of 5.000 hospitalizations is far too much for the health care system. 

 

With this development in mind, we can understand that the government took additional measures. The latest measure is the intoduction of the curfew. If we assume that a curfew reduces the reproduction factor with 10% the maximum number  of hospitalizations  will be 4.000.  Although this number of hospitalizations is substantially lower  than without curfew, it is still too high for the heath care system.  

 

So we hope that in the coming period the reduction of the  reproduction factor(s) will continu due to the measures and possilbly the effect of the vaccination program. 

Development of hospitalizations from the start of the second wave

The picture above gives some insight in how the Government tried to control the second wave.  The light colored line is the real data wrt hospitalizations and the the other lines are the model predictions we use with are simulator. You can clearly see that the Government tries to control the virus by strengthening or loosen up the conditions of social distancing. We made a simular graph before but now without showing the weight we used. (normally 70/30) The picture above shows us that we are still increasing in hospitalizations but the peak is near. The impact of the lock down will soon start to show. However the solution is near.

Vaccinations have been started already. Until then stay safe! 

 

The picture below gives some insight in the development of the infections over time. Het you can see that the number of persons that can infect someone else is at this moment around 120k. (the RIVM dashboard shows 122 k based based on December the 10th).

 

New Crown Dashboard for January 4th 2020.

The expected number of hospital and ICU beds has increased compared to our last forecasts . It this prediction we assume 88 thousand persons that can be infecting others and the weight is determined between the R= 1 and R= 0.9. If the R increases during the Christmas days, this will effect the amount of hospitalizations.

With an R = 1.5 this would result in almost  300 thousand infections and over 8000 hospitalizations.  

Of course a scenario nobody want to happen, but it shows how thin the line really is.

No wonder the RIVM is stressed out about the days to come!

New Crown Dashboard for 10 December 2020.

The expected number of hospital and ICU beds has decreased significantly compared to our last forcasts due to a change in the view of the older people that become servere sick. We now assume that this group is in the hopital or on the ICU for a short while. We think using the adjusted assumptions will lead to a more accurate prediction (of course as long as the reproduction factor remains 0.9).

From November 1st.

Here the new Barometer based on information from 27/10

Our prediction for November 26th based on data from October 27th.

The number of hospital and ICU beds is expected to stabilize. In the coming week we hope to see the number of infections decreasing and the expected number of hospital and ICU beds to decrease as well. So far we have been predicting every week for 28 days in advance. We did this by using our Crown simulation model.

Last week me made a few minor updates based on new available data from the RIVM. 

 

How well is our model performing?

We would like to show how the model would run from August 12th until November 16th and how well we the outcomes matches the current situation.

Looking at the graph ahead four lines are calculated. Starting with a few thousand  infections on August 12th the model simulates until November the 16th how the admissions (number of hospitalizations) would evolve. The dark green(lower) line is the calculation based on R= 0.9 and the upper red line is a based on R= 1.15 and the middle light green line is the weighted outcome of both scenarios (let it go and full control). Taking a weight of 30/70 as we did before the outcome reaches a result that comes very close to reality. The blue line the real numbers!

At this moment (November 2nd) the amount of hospitals beds amount to 2545 and outcome in the model 2519!

This is calculated from the start (12th of August) so it looks that the model is doing exactly what in reality happens.

 

But..

A crucial week for the Netherlands is coming. Last weeks many people were infected. So there is still a lot in the pipeline that has to come out.

Depending how fast the and how severe the infected persons are hit the amount of hospitaliszations will continue to rise.

We expect in the week of November 4th the number will surpass the 3000 beds.

 We hope we are wrong, but so far our model predicted the right outcome. 

 

What was predcited on October the 13th in a press conference by Ernst Kuipers (chairman of the LNAZ)  

Below we show what was predicted on October the 13th.

Ernst Kuipers showed us three scenarios. What if:

1) The R  gets to 0.9

2) The R gets to 1

3) Keerps going as it us, so te current trend continues

 

On October the  8th Erns Kuipers warned that the number of beds would rise towards 1250.

In reality this was 1390  which was within our 95% Confidence level.

Analysing the picture above around 900 COVID patiens using upto x 1.2  more bed capacity where estimated.

Bringing the R to 0.9 would amount to almost 3000 beds for the end of November. With an R = 1 this would be 4000 and if the current trend continues: 5415 beds

Currently (november 3nd) this would be:  2400 (R= 1 and 0.9) and 2700 if the trend continues. We expect that on November the 3rd the amount of beds would reach 2600.

 

If we compare the analyses with our model outcomes  we can conclude that most of the estimations done by Erns Kuipers are under estimated.

Why could this be the case?

 

A reason for this underestimating could be the lack of granularity.  In our Crown model we integrate granularity like age, but also cross infections and a probability that you end up in the hospital related to the age in our model.

 

Not doing that will in the beginning over estimate the outcome, since more young people are infected and when time passes under estimate because more and more older people get infected and therefor hospitalized.

 

This is also what you can see in the analysis done above.

One of the reasons we made an simulation model is to add these factors which are otherwise very difficult to project analytically.

From October 23th.

Here the new Barometer based on information from 22/10

Big warnings from the Crown Actuaries. A black scenario is in the making!

Our prediction for November 19th based on data from October 22th.

Since last week it became clear that the Government has lost more and more control in the fight against COVID. So, we have adapted our assumptions slightly.

Where we in our former models one month ago we used a weight of 70/30 ( The weight between the Full control and Let it go model) we adjusted this towards 50/50. 

This will be the case if the current measures will have its effect.

 

Knowing that data always comes late, and that measure cannot be having an immediate effect a 70/30 weight so far worked well. However, since there are hardly any analyses towards the source (so we do not know where a positive tested person got it and who might be infected too) the weight must be changed. And we did that!

 

With that knowledge in our head the estimated the amount of hospital beds for the next 28 days will grow towards 3266. The percentage of ICU has slightly increased from 20% towards 25% due to the case that more elderly people are taken into the hospitals, leading to 816 ICUs.

 

We would like to put a big warning though.

So far, the measures taken have not shown us any effect, so more strict measures like a full lock down, evening clock must be considered. 

 

If the current measures do not work, our model expects the amount of hospitalizations within the next 28 days to rise to more than 5200 beds, with 1300 ICU beds. That means that within a month we could enter the situation that healthcare is no longer guaranteed. 

This is new and must be prevented at all cost.

 

Let´s go into the deep and see why we think this could happen!

The first graph is an outcome of 100 simulations where we assume full control.

The second graphs show what could happen of the situation develops like it did before, 

so that the new measures have no impact.

 

If we calculate that a 50/50 scenario, the amount of beds will be around 3200. This is also our prediction for our Crown Barometer.

 

However, if we are not able to reverse the current trend, the amount of beds will grow towards 5190 in 28 days. That also means that the amount of ICU will grow towards 1300.

 

Erst Kuipers made clear that the amount of ICU beds could grow towards 1350 beds by the end of next week. So, if the measures does not work all those beds will be filled with COVID patients. That is a black scenario.

 

Why we think this could be the case.?

In our last blogs we showed a shift from the younger generation towards the middle one. Now we also see it is moving even further and that is when the ICU start getting fuller. Underlying graph shows the movement in time.

First the movement wrt to infected persons. 

The majority of infected persons are from the younger generation, but a you see the first 7 periods  a movement towards the middle-aged groups,  and finally, it slips through to towards the older people and that is the moment when the amount of hospital beds start to rise faster. After 9 periods (36 days) the process repeats itself. Since more older persons get into the hospitals a drop in the amount of infections is expected. This is caused by the fact that less persons will be asymptomatic. So, the second ´wave´ will be less than the first and so on.

 

Analyzing the second graphs we notice that the increase in the slope of the line of the Hospital Capacity is explained by the fact that more elderly people are occupying the hospital beds. The Green, Dark Blue, And Brown areas are getting more weight in time. If we cannot stop that we will reach eventually a peak of 7000 beds by the end of November. (period 11)

 

So 

A dark black scenario is in the making!

 

Let´s hope the table turn and the government measures will have its effect.

We keep you informed!

 

Crown Actuaries.

Kees, Rob en Servaas

 

From October 15th 2020:

This is the Crown Barometer for 12 november 2020.

Based on the current status (15 October) of number of infected, reproduction factor and measures taken by the governement we expect around 3.000 people to be hospitalized (including ICU) at 12 November. Taking into account the uncertainty in the estimates around 3.465 are expected with 95% certaintly.

 

From October 8th 2020:

This week  (8 October) the RIVM did not make an update for the number of infections and the reproduction factor.

However, 

Ernst Kuipers, who is the chairman of the LNAZ which controls the distributions of the IC and hospital beds in the Netherlands, mentioned that the amount of beds will reach 1250 in the beginning of next week.

 

This in line with our estimation we did on the data of September 15th, where we estimated 1212 on average for 13 October.  We will follow the development of the number of beds closely in the coming period.

 

Due to the absence of the RIVM update, we will focus on the impact of more elderly people getting infected this week

So far in our model most of the infected are younger people that infect their own generation. But also they will infect their parents for a part. That parents went to work and infected  other parents as well. The parents visit their mothers and fathers, and they get infected too. And that is when a problem arises, because the older people have a high chance of getting severe ill from the virus. This effect is happening at this very moment. In underlying graph, we try to illustrate this:

The graph shows how the currently estimated 175 thousand infected are divided by age. The three colors (blue, orange & red), represent the three generations we use in our model (young, middle and older). In the graph you can see that 87% are from the young, 12% are middle and 1% from the older generation.

 

Running a simulation based on these numbers shows and decrease in the hospital capacity.

Assuming we just let the virus go and using the 175 thousand infected we get the following result: 

How would this picture look if the 1% of infectors from the Older generation would in fact be zero? See the following picture:

How would this picture look if the 1% of infectors from the Older generation would in fact be zero? See the following picture:

From September 29th 2020:

This barometer shows the expected hospital capacity for Covid-19 patients per October 27th for the Netherlands

The RIVM director Jaap van Dissel made another statement yesterday:  

" Around Mid-October the amount of hospitalizations will grow towards 2250 beds and the number of ICU´s  to 400!"

Could he be right? 

 

Let´s Find out!

Our model last week predicted 1500 hospitalizations and 250 ICU´s,  3 weeks from now with a R of 1.33 and 100 thousand people that could infect others.

Assuming 1 out of 6 get into the ICU. The percentage of ICU will probably go more towards 20% instead of 16%.

 

What changed since last week?

The R dropped this week slightly to 1.27 and the number of infectors rose towards 146 thousand.

Since the R stayed stable, this makes it is possible to estimate the amount of beds from our last prediction.

Taking 9 periods instead of 7 (1 week further) shows us that the amount of expected hospitalizations.

 

Looking at the graphs from last week (scroll down to see) this results in:

1) Control the virus. The expected amount of beds will be 900

2) Let the virus go. The expected amount of beds will be 4500

 

Calculating the same formula we used last week the total expected amount of beds will be:

70% x 900 + 30% x 4500 = 1980. Taking tinto acount the model uncertainties this would be approximately 2000 hospitalizations.

 

Current Crown Model outcome:

The current model calculations amount to 2115 which are very close to the prediction based on last week's model.

A few minor differences plus considering its uncertainty explains the differences. 

 

So, 2250 in 2-3 weeks' time looks like too much. Why this sudden increase? 

It it possible that the Government has lost more control. This could indeed be the case since the source investigation are not as strong compared to last week.

The GGDs announced that tracking down sources would now rely more on peoples own responsibility and that could cause slight lost of control.

Looking 2-3 weeks ahead the following can be concluded:

Looking 3- 4 periods ahead the amount of beds in both viewpoint amount to:

1) 2400 beds en 2) 1200 beds. 

Using the formula results in:

70%* 1200 + 2400*30% = 1560 beds. If we assume 20% will go to the ICU instead of 1/6th this leads to 312 ICU beds. 

This is lower that then predicted by the RIVM. 

 

Conclusion:

Based on the model that the current prediction of the RIVM might be a little to high.

We expect the amount of hospilitazations rise towards 1600.

 

The coming weeks wil make clear if our assumptions were correct or not.

One thing is clear. Covid is still around and once again it shows it highly unpredictable behavior.

Only for that reason the current measures taken by the Government are needed. 

 

Crown Actuaries

Kees, Rob en Servaas

 

From September 22th 2020:

This barometer shows the expected hospital capacity  for Covid-19 patients per october 20th for the Netherlands

The RIVM director Jaap van Dissel made a clear statement yesterday:  

" If the trend continues we can expect the amount of hospitalisations to tripple towards 1500 beds and the number of ICU´s  to 250!"

Could he be right? We think he could!

Underneath we explain in more detail why we think this is plausible.

 

First we go back to our latest dashboard outcome last week

It was already close to the current expectation, but since then we made a few improvements.

1) It now also considers the current amount of hospital beds, that influences the capacity in the future

2) We increased the model-population we simulate tenfold. (from 1000 to 10 thousand)

3) The amount of simulations increased from 25 to 100

 

The current model we use shows our outcomes are simular with the current "Prognoses" of the RIVM

How do we get to this outcome.  Let´s get into more detail fo find out.

 

 Crown Simulation model:

In order to understand the behavior of the virus we use two points of views:

1) Control the virus. With a certain amount of infections and or hospital capacity to take full control to lower the R back to 1. Meaning that when a test is found positive its source is found and all the connections around the source are tested. This scenario is unfortunately not realistic in the Netherlands. We can only deduct 50% of the positive outcomes to its origine. However, in our model it gives us a minimum value of expected hospiral beds.

2) Let the virus go. Use the current R value and its amount of infections and run the model without further interventions. The current R = 1.33 and the amount of persons that can infect is currently roughly 100 thousand. This viewpoint is not realistic in the Netherlands, but it give us a maximum value.

 

 With the two viewpoints we can determine the minimum and maximum impact on the hospital capacity.

Simulating our model based on these two viewpoints gives the following result:

Control the virus                                                                                                                                                  and                                                Let the virus go

The scenarios we ran show a wide variation in the amount of needed hospital bed and the red lines show the averages. 

When we translate this outcomes into the current situation in the Netherlands (so from a selected population of 10 thousand towards 17.5 million) 

it shows the following pictures.

In the two pictures above we show the average and the 5% en 95% percentile. It immediately shows that "blowing up " the population with a factor 1750 decreases the bandwith of the percentiles., due to the "law of large numbers".

Using this technique the minimum amount of beds would be 800 and the maximum 3300.

Leveling both viewpoints leaves an average of around 2000 beds.  

Of course we expect, especially with the current rise of the infections, that the actions would more tends towards getting full control of the virus. For that reason a weight of 70% on the first outcome and 30% on the second would be very likely. In that case the average amount of expected beds would amount to 1549 beds at October 21. We assume 1/6 of this beds to be ICU (258).

 

Why is it plausible that the amount of hospitalisation would increase?

To answer this question we need to go into more depth of our model. High level the reason is while for starters relatively young people get infected, in time it shifts towards the older generations and they get hospilitalized. Underneath pictures show what happens if we run the simulation if we do not try to control the virus.

 

Looking at the movement in time on both infections and the capacity of the hospital beds you see the following shifts:

In the beginning the amount of infection is still low (around 700 in 10 thousand ), but while it is increasing a shift takes place from younger people (in the 30´s and 40´s) towards the older generation. After the first 4 days the younger generation amounts to 80%, but it lowers within a month to 30% and makes a shift to the older generation.

And this translates in the increase of hospital beds. After 4 timesteps (16 days) an increase in the hopspital beds starts.

This makes it that the amount of hospital beds needed after a month grows to 19 beds per 10 thousands civilians, which comes down to 3300 beds on the total population. The maximum capacity is still not reached, but there is a reason to be concerned.

The key take away is therefore: yes, the RIVM is right, but we also need to consider that we are better prepared, need less capacity since the treatments improvement and the time COVOD-19 patients spend in the hospital decreased from 22 days to 8 days. So we are further in the process and lets  hope we continue to do so.

 

Conclusion

If we are able to stabilize and control  the virus 1000 beds would be needed at October 21. However for Dutch situation, this is would entail that we are able to find the source if an infection occurs.  Last measurements showed that we can only do this for about 50% of the cases so we are more heading towards an " let the virus go" situation.  Our Prime Minister was quite clear on that: " Currently it feels if we are driving through a big fog" .

When there are no obstacles on the road we are safe. But its still possible if superspread events occur.  Lets hope that it will not be the case.

 

Crown Actuaries

Kees, Rob en Servaas

From 15 September 2020:

This barometer shows the expected hospital capacity  for Covid-19 patients per october 13th for the Netherlands.

Our first chart had a few diferent assumptions that lead to much higher estimate. That of course needed to investigated more.

Further analysis made clear a few changes were needed:

• Our starting point now (30 thousand infected) is set up differenty.
  Using the last RIVM report (epidemiological situation Covid-19 in the Netherlands)
• In this report we looked at the distribtion of positive tested from june onwards
• We conclude that more young people could be possibly infected
  ( We use Young, Middle and Older as generation) where 2/3 is now Young, 1/4 Middle and the rest comes from the Oldest generation.
• Furthermore we made more model updates that are more in line what could be expected (assympotimatic behaviour programmed differently and set to 75%, the time spent in hosptitals went down with 1/3, the way generations can infect each other and others is slightly changed, we changed the possibility of infected people (more young) in getting in to he hospital. This is now slightly lower. Now overall there is a 4% chance that one infected ones can end up in the hospital instead of 5%. 

Conclusion is that the amount capacity that is expected droppped considerably because of these changes and are now more in line what intuitively is expected.