Basic science and statistical understanding

[MrsTerryPratchett]

I'm not a scientist. I do have a BSc. but nothing more impressive. I do wince when I hear some assumptions that people make who don't have basic science (and statistical) knowledge. What basic scientific or statistical principles do you wish the general public (and members thereof) knew about?

I'll start. Causation and correlation. A correlation doesn't prove causation. Particularly when that correlation is a correlation of one. "I had the flu vaccine and got the flu really badly" isn't causation.

@Changechangychange

hamstersarse

1A is the main cause, 1B and 1C are things that caused 1A (ie 1A E. coli septicaemia, 1B ascending cholangitis. You don’t need 1B or C unless they directly caused 1A.

Part 2 is any other contributor, even if it didn’t directly cause death. So in my example above, you could put ischaemic heart disease and diabetes in part 2 if you thought having those made the patient more likely to die of sepsis.

You might put covid in part 2 if you thought having had covid had weakened them. You wouldn't put an asymptomatic positive swab in there. And no, it wouldn’t be counted as a “covid death” if it was in part 2, in the same way that my example patient wouldn’t be classed as a death from heart disease.

OK thanks

So it has to be in part 1 to be counted on the death stats... but doesn't have to be the main cause (1A) to be counted?

Is it also the case that only one doctor is required for certification throughout the pandemic, when previously it has always required 2 doctors?

Ah yes also the fact that "500 deaths today" actually means "500 death reports reached the central gathering spreadsheet today, some of which may have occurred a month ago".

(But it sort of doesn't matter because in future days more deaths from the last 24 hrs will be counted)

I’m not a doctor, but as I understand it 1A isn’t the “main cause”, it’s the “immediate cause”.

There are a couple of questions there, so bear with me and I’ll try to answer them.

Easy one first - it’s always only required one doctor to complete a death certificate. You may need to speak to the coroner in certain circumstances (recent surgery or accident, and anything else that could represent a non-natural death). The coroner may or may not allow you to issue the certificate. There are some circumstances (foul play, misadventure, or where you have no idea what the patient has died of), where you are not allowed to issue a death certificate and there has to be a post-mortem and or inquest. Eg in a hypothetical road traffic accident, it may be obvious what they have died of but you can’t issue a death certificate.

Cremation forms need more information (to exclude foul play before “disposing of the body”), and need two doctors to complete them, one of whom must have “attended to the patient in their final illness”. The second, independent doctor will read the notes, look at the test results, phone up the first doctor, and anyone else they can get a collateral history from (ie ward nurses) before confirming the cause of death sounds right.

It is that part that has changed as part of the covid regulations - see here:

assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/878093/revised-guidance-to-medical-practitioners-completing-form-cremation-4.pdf

Because it was not deemed realistic in a swamped health service. In my trust, our pathologists temporarily took over completion the death certificates and crem forms. It has now reverted to the medical team looking after the patient, as things have calmed down a bit.

So it has to be in part 1 to be counted on the death stats... but doesn't have to be the main cause (1A) to be counted?

Sort of - so if somebody’s immediate cause of death was a pulmonary embolism (blood clot on their lung, well known complication of covid), and part 1B was covid pneumonitis, it would count as a covid death.

If they had a blood clot on their lung, as a complication of their breast cancer (also well-known risk factor), and also had covid a month ago, then I’d put covid in part 2 (or not at all, if I didn’t think it was a factor) and it wouldn’t be counted as a covid death.

There is some judgement involved, but that is always the case with death certificates unless you have a post mortem. And even then, you are going on the pathologist’s judgement (just with a bit more evidence to go on).

The dataset question is slightly further out of my area - I’ve worked with HES data in the past but hopefully somebody else can chip in.

As I understand it, there are two sets of data - the daily updates, which crossmatch deaths and positive covid swabs (I think from HES data). This dataset will include the “asymptomatic positive swab and died in a car crash” patients. But there is no other way of collecting real-time data, and you would hope that even if the exact numbers are not perfect, the relative numbers should still be reliable (because there is no reason to think that the proportion of covid positive patients dying in car crashes would change week on week).

So if numbers double, that should be reliable, even if it is actually a doubling from 200 to 400, not 250 to 500. It’s useful for following trends in real-time, not so much for accurate death statistics. We do need real-time data to judge the impact of interventions such as lockdown, and to respond rapidly to rising rates. And with the existing data sources, it’s probably the best we can do.

The death certificate data comes in weeks later, once the death certificate has been completed, the death has been registered, and that registration has been reported to ONS. It is much more reliable, but far too slow to be any use reacting to outbreaks.

Can we please acknowledge that RT-PCR is a subset of PCR? Or is that being too pedantic?

@Changechangychange

what's being recorded is any death where the person also tested positive for covid (even if the covid presents no symptoms at all) or any death where the doctor reckons covid was present (don't give a shit about actual cause)

we have absolutely no idea of the real extent to which it causes death (rather than happens to be present at the same time as death

This is what I am taking issue with. Death certificates do not have “Covid pneumonitis” as main cause of death (part 1A diagnosis) if you had a positive swab but no symptoms, or if you tested positive but clearly died of something else. But you know that very well.

I'm not sure where the tone of this has come from, especially as you say in a later post, exactly as I'm saying, that the doctor's own judgement comes into it. Data based on judgement is not good data. If policy decisions are being made based on data then it should be verified data not data based on a judgement by one person, with no clear evidence to back it up.

Are you genuinely suggesting that doctors are always correct when they certify a death as being from covid? Because that would make covid 100% unique as the only ever COD of that was always recorded accurately. No other cause of death produces anything like accurate data.

I've found my tribe!

To be clear, just because you seem to have implied that I think the deaths of cancer patients, where cancer is the obvious cause, are being recorded as being from covid, I am not suggesting that. What I'm talking about, as I think you know, are the many many deaths, especially in elderly patients, where the cause is very hard to determine. I mean 'advancing years/old age' is an 'acceptable' COD for someone over 80! That's not even a cause! To say that doctors are reliably and accurately recording covid deaths for those types of patients is just not true.

"Died with covid" confuses me. Let's say Mr Smith has a slight cough and therefore gets a test. He isolates for 14 days. On Day 15, he goes for a walk and gets knocked over by a car. Sadly he dies. Does he count in the covid death stats? Even though his death was nothing to do with Covid?

TheDailyCarbuncle but people do not start campaigns against heathy heart campaigns because “we have absolutely no idea of the number of people dying of heart disease in the UK”. People do campaign to get rid of lockdown and other public health restrictions, on the grounds that death certificates are now all being filled in fraudulently.

There is a clear agenda here, in implying that all of these covid deaths are fake news.

@Changechangychange

TheDailyCarbuncle but people do not start campaigns against heathy heart campaigns because “we have absolutely no idea of the number of people dying of heart disease in the UK”. People do campaign to get rid of lockdown and other public health restrictions, on the grounds that death certificates are now all being filled in fraudulently.

There is a clear agenda here, in implying that all of these covid deaths are fake news.

Are you saying I'm implying that covid deaths are fake news? Because if you are then you must be reading something other than what I'm posting.

I don't think it's the slightest bit controversial to say that in a pandemic it's important to know that actual number of deaths being caused by the virus. It could well be a lot, lot higher than the number being recorded, given that there is no system in place to check and verify COD - it's definitely the case that in March a significant proportion of deaths outside hospital failed to be recorded as covid deaths. What I'm advocating for, in the spirit of the thread is accurate data - whatever that data might be, because it is a basic principle that if you're basing decisions on data, that data should be correct.

Incidentally there have been campaigns around the role of cholesterol and statins in heart disease because there was legitimate concern that a false connection was being made between cholesterol in food and blood cholesterol, combined with concern about the actual effectiveness of statins vs their side effects. Concerns about data meant that people refused to take statins and were unsure/concerned about advice regarding diet, possibly leading to poorer health outcomes. So what you gave there is actually a very good example of situations in which poor reporting of health data can impact engagement with and the impact of public health measures. Accurate data is actually not too much to ask for.

OK, so the old car crash thing keeps coming up. do we actually know if this has ever happened, even once?

"Absence of evidence is not evidence of absence".

An inability of people to think about important issues which may have been overlooked and then dismiss people raising the concern as 'anecdotes rather than considering this example as a very concern that may warrant investigation which may cast doubt on whether existing 'knowledge' may have a massive blind spot and change our perceptions of the entire idea and concept.

Also see poor comparison making eg comparing data for vaginal births with all csections, when in reality this isn't how life works. Instead you should be looking at a planned vb which includes emergency csections as a possible outcome v planned Csections (which also includes cases where someone has unexpectedly ended up with a vb).

And averages in general. The average person has less than 2 legs. Averages need context and meaning. You also need to think about outliers and whether they have common traits which shouldn't be ignored or overlooked as 'unimportant' because on average they aren't at risk.

people do not start campaigns against heathy heart campaigns because “we have absolutely no idea of the number of people dying of heart disease in the UK”.

Well, no because the healthy heart campaigns don't do any harm afaik.
Lockodwns do terrible harm so we have to be sure they're necessary.

Once, in the US, piggy and it was then taken out of the stats.

www.snopes.com/fact-check/florida-motorcyclist-covid-death/

@Greenandcabbagelooking

"Died with covid" confuses me. Let's say Mr Smith has a slight cough and therefore gets a test. He isolates for 14 days. On Day 15, he goes for a walk and gets knocked over by a car. Sadly he dies. Does he count in the covid death stats? Even though his death was nothing to do with Covid?

Also, what if Mr Smith gets positive for covid, is quite ill for 2 weeks, then gets much worse and is admitted to hospital-deteriorating slowly over another couple of weeks. He dies 29 days after his positive test.

Does he not get counted in the covid death count?

The 'Mr Smith' stories are just silly. I can't understand whether they're a willful misinterpretation of the situation or people are just stupid.

It's not the case that people tested positive then died in a car crash and were recorded as a covid death. In that situation, the doctor (who incidentally is a highly trained person and likely not an idiot) wouldn't even consider that person's positive result - there's every chance they wouldn't even know about it, given that there's no need to record positive results on a person's health record unless they're given treatment. What the restrictions in reporting were trying to address is the situation I'm talking about - a good example is a case I have direct experience of, which is a patient with severe ill-health for years, who's been steadily declining for the last year. The patient then died, with no identifiable cause of death (they were malnourished and extremely thin, but those aren't acceptable causes). They had a negative covid swab but the doctor recorded the death as from covid due to findings on the pre mortem x-ray. That patient may well have had covid some time back, but there was no positive result within the last 28 days so a covid death should not have been recorded. It was, however, recorded. There are many more like that. The 28 day thing is arbitrary - it's essentially to prevent lazy certification like this case and to stress the point that doctors should have reliable evidence for a covid certification. It hasn't really worked unfortunately.

That made me think of the dear old DfE/Ofqual algorithm red! The crucial issue of overlooked outliers there!

It's worth remembering too that a person can get more than one covid test so the 28 days thing isn't set in stone. It's not a hard and fast rule, it's just a guideline to attempt to get clearer death data. It's not every effective but it's better than nothing.

One other complicating factor is that all infections can lead to sepsis. That usually happens some time after the initial infection and is a complication of infection rather than a feature of the infection itself. That's part of the reason why it makes sense to have a cut-off date. If I have a strep infection, get very ill, take antibiotics, end up in hospital, go into septic shock and eventually die a month after initial infection, that's recorded as a death from sepsis, caused by strep. It's not a death from strep - if such deaths were recorded as strep deaths then there'd be massive panic about strep and infections in general given that 11 million people die of sepsis caused by those infections every year. Recording deaths in that way would make covid seem like a minor annoyance by comparison.

Exponential growth. The old thing about folding a piece of paper 42 times and it becomes as thick as the distance to the moon is a good illustration.

Not science, just basic maths.

It ties in with people thinking that we can just carry on and 'shield the vulnerable'... if we let spread run exponentially this really isn't possible.

Just BSc here too....primary school teacher actually considering moving to secondary maths/science to help do something about the appallingly low levels of understanding of both in society.

Oh this might have been covered but PCR 'amplifying' the genetic material as if that's some kind of fiddle....that's how it works!

Just BSc here too....primary school teacher actually considering moving to secondary maths/science to help do something about the appallingly low levels of understanding of both in society.

Mmmmmm I think you should.

primary school teacher actually considering moving to secondary maths/science to help do something about the appallingly low levels of understanding of both in society.

No! Stay in primary! We need as many strong mathematicians and scientists in primary as possible. I teach scientific literacy and number sense right from reception.