Cambridge research team estimates 12% of England has been infected already | Hacker News Reader

Cambridge research team estimates 12% of England has been infected already

mrc-bsu.cam.ac.uk

48 pointsgwd69 comments

Other highlights:

Death rate for under-45 is under 0.024%

Look at the "Deaths incidence" tab of the graphs at the bottom to see actual deaths with the curve-fit model.

EDIT: Meant to include in my initial "highlights" comment:

Estimated "peak" daily infections was 23 March, with 500k infections. Estimate for daily infections as of 10 May closer to 10k; estimated country-wide Rt at 0.75, still well below 1.

Bonus highlights: Method seems to be from this 2009 paper, "Bayesian modeling to unmask and predict influenza A/H1N1pdm dynamics in London":

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3215054/

Not equipped to evaluate their method, really; but if it's reasonably accurate, then the UK is not currently really headed towards "natural herd immunity".

death is easy to measure quickly but the QALY impacts are not.

As a 55 and over with a dad who's 85 and over one can excuse me for not being on the herd immunity train.

I also have friends who've gotten sick and suffered impaired health. Figure the 80/20 rule. 80 people suffer impaired health for every 20 that die. We're talking about 1 in 50 having damaged health.

The "herd immunity train" doesn't mean we should get people in high-risk categories (even 55+, let alone 85+) infected. It means we should get the lowest risk people infected, which blankets the high-risk people in coverage. Regardless of where you stand on that plan, it's worth calling out.

Not obvious this works though. 70% is herd immunity threshold when there is no epidemic. With an active epidemic it may be 95%.

Here are some epidemiologists arguing that: https://www.nytimes.com/2020/05/01/opinion/sunday/coronaviru...

538 made a model to play with: https://fivethirtyeight.com/features/without-a-vaccine-herd-...

So, the plan may be wrong. Proponents may have sincere but incorrect beliefs.

Basically herd immunity overshoot is a thing and shows up both in the historical data and in all models.

There is also the problem that as a virus takes years to burn through clusters of susceptible people. Currently it would take over a 100 years to achieve herd immunity in California.

Note that in order for this kind of overshoot to happen, you basically have to have no social distancing whatsoever - not even voluntary changes to behaviour. If there is any then it decreases the reproductive number R further, meaning that it drops below one and the epidemic peaks before enough people have been infected to provide herd immunity after everything returns to normal, and then herd immunity is only reached during the overshoot past the peak.

(As I understand it, the UK government's original plans couldn't have lead to herd immunity in the first wave without this, since they were planning on substantial measures to reduce spread even under those plans.)

How do you achieve that? It's already ripping through care homes because the people who work in them are unable to shield themselves and simultaneously actually go to work and do their job.

Honestly? And I'm genuinely surprised that this wasn't the first course of action: keep the staff locked in with the residents. You're either locked in at work, or you're locked in at the nursing home so it doesn't much matter does it?

For what it's worth there are less drastic options. Kingston, Ontario for instance was able to avoid having even a single infection within a nursing home. [1]

[1] https://www.theglobeandmail.com/canada/article-with-an-early...

Because no-ones going to pay for them to do that. And most of the staff would quit.

They have families, children, partners. You think they'll give up 12 weeks of their lives locked in with their wards?

You really sound like a dangerous idiot, glad you're not in power.

That’s not as crazy an idea as you make it sound. Furlough those who can’t and hire those who were let go and can. The government could help support costs.

> Because no-ones going to pay for them to do that. And most of the staff would quit.

Seriously? Cost is now the factor? I think we can find it in our 3 trillion dollar stimulus budget.

> They have families, children, partners. You think they'll give up 12 weeks of their lives locked in with their wards?

Is this a pandemic or is this stay and home tea party time?

> You really sound like a dangerous idiot, glad you're not in power.

Thanks man, regardless of whether you agree with me or not, I'd point you at the HN post guidelines.

You seem incapable of understanding that people in demographic groups cluster. The problem is most of my dads friends and associates are over 60. You can only get herd immunity by infecting 60-70 percent of THEM. A high infection rate in the under 45 set does NOTHING to protect them.

That's... not how herd immunity works. If it was, all your dad's friends would have Polio right now :P

It's literally how vaccines work. The proposal (again, without opining on its validity) is the same as vaccination, but comes with much greater risk.

I think he does have a point. Thorough the social measures you take and depending what age groups you expose the most, you can (should) be able to skew the numbers in having more infections in a age group than he other. If you skew the numbers towards the younger population, the end result is _better_ than the normal, presumably uniform, distribution.

EDIT: I misunderstood GP's point. I'll still leave my reply as-is since i still think the general idea I was trying to convey is sound.

The comment above is overstated, but more or less correct. All the naive herd immunity calculations assume a well-mixed population. A large fraction of recovered and immune young people will protect the elderly from contacts with their children, with store clerks, etc., but not from contacts with other elderly. If we successfully protect the elderly, then herd immunity among young people will decrease the probability that the virus gets introduced into a nursing home; but it does nothing to slow the virus's spread once it does.

That heterogeneity also goes the good way though, since people in jobs most prone to spreading the virus (doctors, nurses, store clerks, etc.) tend to get infected first and therefore become immune first. There are models suggesting we might reach overall herd immunity with only ~30% infected (https://news.ycombinator.com/item?id=23149725), though this is obviously speculative.

Herd immunity is backwards... You can say "hey, look, the epidemic has died because of herd immunity", but you can't reasonably desire herd immunity if you can avoid it.

Think of Ebola and herd immunity for more that a second, and you'll understand what it means: herd immunity is not the only way. Korea proves it too. SARS 2003 died without herd immunity either. What about HIV? Fancy trying herd immunity?

Herd immunity shouldn't be a goal.

> Herd immunity shouldn't be a goal.

Herd immunity is how vaccines work. It's the reason not everyone needs to get the measles vaccine so long as most people do.

> Think of Ebola and herd immunity for more that a second, and you'll understand what it means: herd immunity is not the only way. Korea proves it too. SARS 2003 died without herd immunity either. What about HIV? Fancy trying herd immunity?

If you can obtain a slice of the population that is not particularly strongly affected by a disease large enough to achieve the same result as you would with a vaccine, it's an avenue. That's true of folks under 45, broadly speaking. You may not like it -- I may not like it -- but it is an effective avenue.

It is the goal. This conversation is about different ways to achieve that goal.

Ah, I see, you're of the opinion there's some way to eradicate this disease without immunization -- by continued lockdowns? Given the R0 could be 5 and millions infected, that ship has long since sailed IMO, to the extent that I don't really even consider that an option.

That's true. The terminology has gotten a bit confused; most people who say they want to follow a "herd immunity" plan really mean "I don't think we can avoid it". Country A achieving containment doesn't necessarily mean Country B can achieve it, at least not without impossible or unacceptable costs.

> you can't reasonably desire herd immunity if you can avoid it.

And that is the crux of it. Can you avoid it? If you can avoid it, by all means do it. If you can't you might as well make sure you reach herd immunity by infecting the least susceptible age groups.

The ability of the population to stay indoors is limited, _especially_ if they don't see the effects of the disease (because they are indoors, right?) so you can't be able to keep up the pace for 2 years.

This is basically a bet on whether the resolution will be reached through herd immunity or some other means, some drug or quick vaccine.

If you do get the resolution by herd immunity anyway, you want to infect the least susceptible population and sustain the highest economic activity and standard of life for the population. If this bet pays off, you're least affected. If it doesn't and some magic solution shows up, you will have already bore the brunt of your approach and more people will have died than they should have.

If you get some sort of technical, quicker solution to the problem, the you will have the least overall number of deaths and will have paid a comparatively small cost by doing a lock down for a couple of months only. If this doesn't pan out and you get my herd immunity anyway, your overall number of victims will be higher than the herd immunity folks because statistically the spread will be uniform through age groups and your economy will be in worse shape because more people had to stay at home for a longer time.

It's all in what you think is the most likely resolution to the crisis and how well you follow through with your plan. I am though of the opinion that if you're to commit to a strategy, commit to it fully.

That's not the 80/20 rule. There must be empirical estimates of that number.

I'm with you on this one, and the downvotes are peculiar to me. The death rate in the under-45 bracket doesn't matter to me and I don't know why the thread root author focused on it except as reassurance for that age range. What I was trying to call out is that even in that age range that observed death rate shouldn't be reassurance to justify going out-and-about.

> Death rate for under-45 is under 0.024%

The rate of death for people under 45 is very low. It's going to be more than 0.024% because:

> Deaths which have COVID-19 on the death certificate are included in ONS figures but not here; we are looking into their inclusion.

That isn’t likely to change the deaths under 45’s by much if it all as the ONS figures only a additionally include out of hospital deaths primarily form care homes and hospices in which the average age will be well above 45.

If you remove comorbidities in the under 45 pool the rate drops even further for the most part if you are under 45 and healthy your chances of dying from covid-19 are exceedingly small.

> if you are under 45 and healthy your chances of dying from covid-19 are exceedingly small.

Yes, I did say that.

However,

> as the ONS figures only a additionally include out of hospital deaths primarily form care homes and hospices in which the average age will be well above 45

The ONS figures include all people where coronavirus was mentioned on the death certificate no matter where they died.

Care homes isn't just elderly people, although that's a common misconception. Care homes here are using the Care Quality Commission service specification definition. This includes residential services for people with learning disability and or autism, but also other settings for children and working age adults.

  Care home services with nursing (CHN)
  Care home services without nursing (CHS)
  Specialist college services (SPC)

NHSE+I have only just started including this table in their daily death count reports. https://twitter.com/chrishattoncedr/status/12609675671555153...

That would be a valid conjecture if we didn’t had the ONS data already there are virtually no additional reported deaths for that age group above what PHE reports:

https://www.ons.gov.uk/peoplepopulationandcommunity/birthsde...

In fact even deaths under 65 only make up 11% of all COVID-19 deaths.

So again the likelihood of the death rate changing in any significant manner for that specific age group due to the inclusion of the ONS data is currently about zero, for other age groups especially those above the age of 65 there would likely be some changes.

> if we didn’t had the ONS data already there are virtually no additional reported deaths for that age group above what PHE reports:

You've neglected to account for the data lag. From your link:

> he information used to produce these statistics is based on details collected when certified deaths are registered with the local registration office. In England and Wales, deaths should be registered within five days of the death occurring, but there are some situations that result in the registration of the death being delayed. For example, when a death needs to be investigated by a coroner. Therefore, there may be some deaths involving COVID-19 that occurred in March and April but are yet to be registered, meaning they will not be included in this analysis.

I'm talking about deaths of young people living in care and nursing homes. All of these deaths need to be investigated by the coroner.

I didn’t neglect anything you can match PHE England data for those dates yourself, I’ve also included the deaths that were reported in early May but were recorded in March for PHE.

Your assumption is that young people die in care homes is not substantiated by any data we have.

It doesn’t mean that young people don’t die they do even if it’s in extremely rare cases they simply die in hospitals since the current triage grants them priority and that they are less likely to turn to the worse before they can be hospitalized or rehospitalized.

Also unlike (much) older patients they are also not released back to the care homes once they are admitted because they can still be effectively saved.

Overall nothing in the data we have suggests that the death rate will increase for the younger population if anything it will likely decrease slightly as the ONS data skews even higher for the older populations than the PHE data.

So sorry but this is yet again you making assumptions that go directly against the data we have.

0.024% is 9,100 people.

https://www.ons.gov.uk/peoplepopulationandcommunity/populati...

Figure 8 you can highlight the age group (0-45) and get a summary figure: 37,980,275 people aged 45 or less * 0.00024 = ~9,100

I'm not sure I call something like 10,000 dead a highlight.

> I'm not sure I call something like 10,000 dead a highlight.

Not on an individual level of course, but on a population scale, it can be. That level would be comparable with driving, with the flu and so on for that demographic.

The world will never be risk-free, and "exceeds expectations" can be a highlight even if the expectations were low.

The 10,000 are likely to be mostly additional deaths, over and above driving casualties, 'flu sufferers etc.

We are already getting "additional death" scores here which is a sort of measure by proxy of the effect of SARS-CoV-2. There is a standard way of counting this, at least across Europe (and pre-Brexit UK!) There is of course a fair amount of variation but having COVID-19 is something to avoid if you can.

It is going to take some really funky stats to work out the real effect because there is less 'flu already in say Aus due to less travel and N hemisphere residents bringing it down south as they head into autumn/winter. Air quality is up nearly everywhere so COPD might be mitigated somewhat. Less road travel means less accidents. However suicides might be up and at least here in the UK there is a fairly well documented lack of people going to A&E (ER) when they should out, of fear or being stoic. Many other treatments have been put on hold.

SARS-CoV-2 has turned the world upside down. I do look for bright sides eg the huge social WFH experiment seems to be a huge success and opportunity. However, additional people are dying in rather unpleasant ways that they would not have done less than six months ago.

isn't the all ages symptomatic ifr for flu about 0.02? and the flu is worse for the elderly. it sounds like taking the young, total infections including asymptomatic cases fatality rate and comparing that to the elderly, symptomatic ifr and saying it's good is playing with stats

what if the measures you take because of COVID kill more than 0.024% of the population should you still take those measures?

What deadly measures would those be?

all hospital visits are way down.

No general health checkups, no doctor visits are allowed

No screening for cancer, no screening for TB

No "elective" surgeries may be performed (note how everything other what would kill you tomorrow is called "elective")

Depression,

Food security,

Reduced exercise, reduced stimuli

Tens of millions of kids not vaccinated, shall I continue?

Those are not measures they are some of the awful, additional side effects we see due to COVID-19. I'm sure we all try to reduce them as best we can.

Here, the NHS repeatedly asks people with problems that would normally require a visit to A&E (ER) to do so. They really do have capacity to deal with cancer, heart attacks, stroke, etc in a safe way. There are entire wards and buildings that are segregated and "clean" of COVID-19 for this purpose.

Elective is a funny old word and is a bit more nuanced than you imply. Some elective surgeries here are still going ahead.

The list you provide is a starter for 10 of things that are and will go wrong for a long, long time but it is what seems to happen when a massive world wide pandemic strikes. This is a world war where everyone is near enough on the same side.

These are not side effects of the disease.

These are side effects of the "safety" measures such as "shelter in place" and lockdowns.

I feel that you are side tracking the discussion. The original topic here is that a disease that causes 0.024% IFR should not be addressed with measures that cause higher deaths in a different way.

I would up vote you more if I could. There is a weird group think happening on HN where you can't even question the quarantine measures. It's knee jerk virtue signalling and not data driven.

> The original topic here is that a disease that causes 0.024% IFR

You misread the report because it puts the ifr at 0.63% (and that's going to be higher because they omit some deaths).

The 0.024% was for the age group under 50, that includes perhaps two thirds of the population.

no dentist btw. which reduces life expectancy.

This has has always been my concern.

How many people have died because they couldn't get life saving elective procedures?

How many deaths will be attributed to the economic slow down?

As a counterpoint though I must also ask:

How many people have NOT died because of reduce automobile travel? Or from COVID19?

How many people will not die due to the cleaner air?

You got downvoted probably because it sounds flip and you didn’t provide any data...but you’re right. Some of the air quality improvements have been astounding, such as seeing the Himalayas from New Delhi 125mi away possibly for the first time in decades

https://www.bbc.com/news/world-asia-india-52313972

Herd immunity will kick in around 70%, so the actual people that would get it is not 100% of the age group.

> I'm not sure I call something like 10,000 dead a highlight.

Given that people before were saying the death rate was 2-5%, yes, it is a good thing.

I find it really strange that people in this thread are trying to be mad about a low death rate...

Looks optimistic when large scale serological testing in Spain, which has a higher per capita death rate[1] suggest prevalence under 5%, and under 12% in the epicentre of the disease in Madrid.

[1]yes, caveats apply

Potentially, but it's not unheard of for a disease to be differently severe in different places for inscrutable reasons. (Of course, that also does suggest those of us outside of England shouldn't generalize the findings too far.)

It's not necessarily all that inscrutable either. We know that a lot of deaths are in places like nursing homes, and that measures which reduce spread amongst the general population aren't necessarily going to be as effective at preventing spread in nursing homes since those obviously have to be staffed, etc.

If you take the low end of the prevailing IFR estimates (~0.5%), the 5700-odd deaths would mean an infection rate of ~12% in London (population ~9M). If you assume 40% of deaths are not properly reported as COVID-19, that gets you to around 20%. But NYC, with a similar population, has had three times as many confirmed deaths, and antibody tests there suggest a total infection rate of 20%. LA antibody tests indicated something like 4%. I'm betting that when antibody tests are done, London will be lower than 20%, and the other areas are likely at the low end, or below, this estimate of 5-14%.

Yes, it would be one thing if this was an antibody/serotological testing result and we could add it to the mix of similar studies, like the ones suggesting ~1.1% rate in Spain (with 5% infected) or ~0.8-1.0% rate in NYC (with 25% infected). But this is not empirical data; it's pure modeling.

It's worth pointing out that the overall IFR is estimated at between 0.49% and 0.83%, with 0.63% being the median. That's only a bit lower than your NYC quote, and still about twice the IFR from the study from Germany (0.34%).

This is interesting ...

The positive rate for those tested in the Bay Area started at 10% 2 months ago, but new test in past week are under 2%:

https://www.sccgov.org/sites/covid19/Pages/dashboard.aspx#ca...

Without having a source to back it up, I'm guessing that a month ago tests were more scarce. So you had to have a strong clinical indication to get tested.

As evidenced in this epidemic, epidemiological modeling is either not very good or overconfident. I mean, all models are wrong, but theirs have also been useless. Why trust a model if you can do a serology study instead?

> Why trust a model if you can do a serology study instead?

Because studies require a lot more time and money, while a model is just on a computer and can be constantly and quickly updated as things change?

> Why trust a model if you can do a serology study instead?

I think the earlier tests were pretty bad, and the newer tests are expensive, but also this kind of modelling is used for other respiratory illness deaths (flu, flu and pneumonia, rhinovirus, adenovirus) so they want to start developing the models to be used in those annual respiratory illness death reports.

Sounds more plausible than the 50% estimate put out two months ago by Oxford.

It would be really cool if we could confirm some of these estimates. Like maybe if we used tests? The government could run a program to test people. Then they’d know. And they wouldn't need to keep guessing. It wouldn't even need to be everyone, just an appropriate sample.

Nah, fuck it, let's just keep testing MPs and guessing about what to do next.

Currently around 500 deaths/ million. If you need ~60% for "herd immunity", that implies they'd get around 2500/1m, or 0.25% of the population. Pretty gruesome.

0.25% IFR is among the lowest we've seen yet, on the plus side. Lower than Gangelt @ 0.36%.

If I'm reading it right, they report an IFR from 0.49% to 0.81% (95% credible interval). Ballpark population of England is 50 million. Herd immunity at ~70% means over 200,000 deaths at median IFR of 0.63%. Extrapolating to US population of ~350 million is around 1.5 million deaths.

Overall IFR was estimated at 0.63%, because IFR in the 75+ bracket is [EDIT] 16%. This is definitely a deadly disease for the elderly.

Seriously, click the link and look around -- there's a lot of interesting stuff to see.

Their normal death rate is ~1%, right? Are we sure that there would be little to no overlap between these groups and they'd end up with a death rate of 1.25%?

There's a ton of information to poke around at the link; but here's their [EDIT] median estimated IFR by age range:

15-24: 0.0032%

25-44: 0.018%

45-64: 0.28%

65-74: 1.8%

75+: 16%

The last one looks similar to the measured death rates by age group in Germany -- i.e., the COVID-19 deaths of 80+ in Germany divided by people tested positive in Germany was around 18% when I looked before, IIRC.

So nearly the entirety of the "1-3% fatality" comes from the fact that it's so incredibly deadly to the elderly.

The overall death rate for 2020 will probably be even below 1%. The old will still die, but the young are no longer driving, smoking, etc and social distancing is preventing other diseases like the flu.

Meant to include in my initial "highlights" comment:

Estimated "peak" daily infections was 23 March, with 500k infections. Estimate for daily infections as of 10 May closer to 10k; estimated country-wide Rt at 0.75, still well below 1.

Method seems to be from this 2009 paper, "Bayesian modeling to unmask and predict influenza A/H1N1pdm dynamics in London":

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3215054/

Not equipped to evaluate their method, really; but if it's reasonably accurate, then the UK is not currently really headed towards "natural herd immunity".

Shame we can't prove it until they start randomised testing.

I was part of a randomised testing survey in Oxfordshire which involved an antibody serology test and swab tests. Hopefully the results come out soon (I took part last week).

Reply on news.ycombinator.com