Hacker News Reader

We have to be extremely cautious when considering any of the death-related statistics for this particular situation.

For example, we know that a dubious methodology was used for counting deaths in Toronto (Canada's most populous city, and the fourth most populous city in North America).

This was publicly admitted to by Toronto Public Health itself, as early as June 2020:

"Individuals who have died with COVID-19, but not as a result of COVID-19 are included in the case counts for COVID-19 deaths in Toronto."

https://twitter.com/TOPublicHealth/status/127588839006028596...

I can't see how that significant city's death stats could be considered reliable or usable for analytic purposes, after considering how the counting was done. The provincial and national stats would also be affected to some extent.

This makes me quite doubtful about the stats coming out of other regions, too.

The question nobody else has been able to answer is "What is a satisfactory analysis for these purposes?"

The closest I've seen to an answer is to remove accidental deaths (eg, hit by a car, firearms fatality, etc) but there's a huge range of arguable cases like stroke which is plausibly covid-affected. Second, you have the problem of comparing stats between locations that used different definitions, so it's harder to do post-hoc correlations.

The decisions seems to have been avoiding false-negative mistakes by increasing false positives in the first wave of analysis: "Let's include everything under a simple rule for covid stats so that we at least have some kind of worst-case baseline modelling with similar datasets, and we can figure out afterwards which are real and which aren't"

Anyway, just wanted to see if you actually have a better answer to the problem than the standard that was used.

The paper those numbers are based on is Assessing the age specificity of infection fatality rates for COVID-19: systematic review, meta-analysis, and public policy implications https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7721859/

The meta-analysis is looking at that question that you are asking.

It also has all the papers that it cites for analysis and an analysis of its own limitations.

> The systematic review encompassed 113 studies, of which 27 studies (covering 34 geographical locations) satisfied the inclusion criteria and were included in the meta-analysis.

You have to look at both sides of the equation. There are excess deaths in every region coinciding with Covid spikes. It takes enormous mental gymnastics to say “some deaths were over counted so deaths as a whole must be over counted” and completely discount the probability of missing deaths too.

The recent Paxlovid study supports this. It reduces deaths days 30-90. Deaths that would not be counted as Covid deaths. But they clearly are if paxlovid reduces them.

Yeah, it is just wacky stuff for people to be "COVID-19 NBD" while the hospitals are full and the morgues can't keep up with refrigeration requirements... it is not worth engaging with. The exact count doesn't matter, the spikes demonstrate the truth of the issue.

I think that's a pretty specious argument though. Other statistics are available (especially excess deaths) that can pretty clearly help show what was going on along with the COVID death numbers (i.e. a significant increase in the number of people dying through the pandemic who most likely wouldn't have died during that time otherwise, in proportions generally tracking more or less to the reported COVID deaths). It's not like the reported numbers of COVID deaths exist in a vacuum to be able to be so sceptical of them!

> This was publicly admitted to by Toronto Public Health itself, as early as June 2020:

> "Individuals who have died with COVID-19, but not as a result of COVID-19 are included in the case counts for COVID-19 deaths in Toronto."

I don't know the particulars of Toronto Public Health, but this wasn't uncommon for immediate reporting vs death record reporting. Since, in many places, comprehensive death records can take a few weeks (or longer!) to go through the pipeline before ending up on finalized reports, it's useful to have more immediate death reporting during a pandemic. Such immediate reporting is necessarily going to be a little rougher around the edges, but it generally gets you pretty close to the real numbers.

But, it's absolutely true that this immediate reporting isn't perfect. Still, it's very useful, and usually close enough for near term needs. Taking the Toronto example, the main tradeoffs are that a) you include some deaths where Covid wasn't actually a contributing factor and b) you miss deaths where the person was never tested for Covid (especially applicable to deaths that occurred at home). Often, these two somewhat balanced themselves out, but only in places with fairly high levels of testing.

Then, in the medium to longer term, you can switch over to relying on more comprehensive death records. This helps to filter out non-causal scenarios (e.g. the "hit by bus, tested positive for covid, died" scenarios). It also helps add in cases that were initially missed (had symptoms of covid, never went to hospital, died at home, etc).

Finally, to validate numbers, we can also look at excess deaths. This helps ensure that we're not wildly off base with reported numbers. Excess deaths isn't a perfect metric to compare to, but it's still a great benchmark.

To wrap up this comment, let me respond to this:

> We have to be extremely cautious when considering any of the death-related statistics for this particular situation.

I fully agree, but.. I would suggest that your comment missed a lot of the accuracy refinement that happens in practice over the medium to longer term. Additionally, it's very much worth pointing out that, with additional analysis (from death records, from excess death analysis, etc), we've found that very few places were likely over-reporting Covid deaths, but that a significant number of places were under-reporting Covid deaths. So, yes, take shorter term Covid death stats with a grain of salt, but know that we've generally ended up reporting too low, not too high.

> "Converting this to micromort language, an individual living in New York City has experienced roughly 50 additional micromorts of risk per day because of Covid-19. That means you were roughly twice as likely to die as you would have been if you were serving in the U.S. armed forces in Afghanistan throughout 2010, a particularly deadly year."

This is highly misleading. The average individual may be at such risk, just like humans have ~1.9 legs on average, but that doesn't mean that everyone walks around with their 1.9 legs.

So those statistics are based on the risk to the general population and therefore have approximately zero relevance to any single person.

Almost by definition, the disease will claim those most vulnerable first, so questioning whether vaccination is worth it for those less vulnerable is valid.

What're the risks of vaccination vs covid for a healthy 25 year old who works from home?

This exactly. You would want to see the micromort breakdown for the same populations as covid infection. You might also want to dive deeper and see if there's more specific populations and how they compare.

The 50 micromorts per day is from the excess deaths in NYC during spring of 2020 and comparing that mortality to being a soldier in Afghanistan in 2010.

I don't believe that comparing those two is unreasonable.

However, to your last question that is what micomorts can answer.

The risk of death from a vaccination is 2.9 micromorts. The risk of death from contracting Covid-19 for a 25 year old is 100 micromorts. So 30x greater risk.. and those are directly comparable numbers.

The other point to consider - you're not always going to be 25. I'm much closer to the 55 bracket than the 25 bracket, and that represents a 40x increase in risk... so about 1200x greater risk than vaccination.

Assuming you remain a hermit for some number of years (and Covid-19 remains endemic), is there a point where you would change your mind on the value proposition of a vaccination? And if the answer is yes - why not do it now?

Yea, currently you're comparing 2.9 micromorts vs 100 micromorts. At 25, that comparison doesn't have too much to it. 100 micromorts is not something that I think too much about either (comparable to the risk of death while giving birth for the women out there). I've done road trips that racked up 100 micromorts cumulative (more if you add in the hiking and skiing micromorts on that trip).

I'll also note I'm older now than when I was when I did that roadtrip... and there are things that I would do now to further mitigate the risks that I took back then.

It's not wrong for your age, but it is something that should be said "yes, its a risk and it is a lot riskier."

That said, there's a number even beyond that 100 micromorts to consider.

New data shows long Covid is keeping as many as 4 million people out of work - https://www.brookings.edu/research/new-data-shows-long-covid...

> 1.6 million full-time equivalent workers could be out of work due to long Covid. With 10.6 million unfilled jobs at the time, long Covid potentially accounted for 15% of the labor shortage.

That 4M number is not part that I want to be part of. While I have better than average health insurance, the not-dying also has a significant risk to it too. It is harder to quantify as it isn't an accurate cause of mortality - but it is not something that should be ignored.

Many people here enjoy better than average salaries - consider the risk of going on long term disability for the next 40 to 60 years with the corresponding reduction in compensation. And yes, I know people who are no longer able to work on anything that has a deadline or expectations of attendance in meetings; their lifestyle has changed significantly from three years ago.

> The 50 micromorts per day is from the excess deaths in NYC during spring of 2020 and comparing that mortality to being a soldier in Afghanistan in 2010.

So in other words it's pretty inflated.

1) Putting covid-positive patients in nursing homes instead of isolating them

2) Excess use of ventilators set to maximum

3) Omicron

4) I feel like I'm forgetting something...

The NYC numbers are from the spring of 2020. This was before Omicron and before Delta.

https://www.cdc.gov/mmwr/volumes/69/wr/mm6919e5.htm was a initial report on the excess deaths in NYC in the spring of 2020.

The by age micromorts are based on a paper that was published in December of 2020. Delta was first detected in the US in Feb of 2021 and named in May of 2021 - again, those numbers are before Delta and Omicron (and vaccination).

The numbers may not be current, but they are not inflated. They are an accurate representation of the excess deaths at that time.

> This was before Omicron and before Delta.

> The numbers may not be current, but they are not inflated. They are an accurate representation of the excess deaths at that time.

That is exactly my point for mentioning Omicron: When you introduced them above you were using them for current risk.

Why does vaccine has one micromort where COVID has it broken down by age group? Are you saying vaccine sets the risk to constant regardless of age?

How do those numbers change when you adjust for presence or absence of pre-existing conditions such as obesity and type-2 diabetes?

On the scale that micromorts work, that isn't something that it measures well. Micromorts work best for the acute risk "do X and this is your chance of dying." Adding in things that are chronic risks doesn't factor into micromorts well.

Consider "traveling 230 miles by car is 1 micromort"... but that doesn't account for impaired driving or any of the other factors known to influence likelihood or survival of a car crash.

The numbers would likely change if you further specified them, but they are useful for the broad comparisons and risk analysis.

If you've got someone going skydiving every weekend, what does that do to their life insurance rates? Well, skydiving is 8 micromorts per jump. So that's 400 micromorts per year. And that goes into some insurance premium cost calculations.

From Wikipedia:

> Micromorts for future activities can only be rough assessments, as specific circumstances will always have an impact. However, past historical rates of events can be used to provide a ball park, average figure.

Another part of this is a study about risk acceptance.

> An application of micromorts is measuring the value that humans place on risk. For example, a person can consider the amount of money they would be willing to pay to avoid a one-in-a-million chance of death (or conversely, the amount of money they would receive to accept a one-in-a-million chance of death). When offered this situation, people claim a high number. However, when looking at their day-to-day actions (e.g., how much they are willing to pay for safety features on cars), a typical value for a micromort is around $50 (in 2009). This is not to say the $50 valuation should be taken to mean that a human life (1 million micromorts) is valued at $50,000,000. Rather, people are less inclined to spend money after a certain point to increase their safety. This means that analyzing risk using the micromort is more useful when using small risks, not necessarily large ones.

That last sentence is an important one when looking at this.

Would you be willing to be paid $150 to get vaccinated?

Please don't. What you are doing is called "P-hacking". Yes, if you keep selecting for subsets of a dataset you'll eventually find one that, just due to random chance, looks like a statistically significant result that proves whatever it is you want to see. That's not the way you do science.

As always, xkcd explains it best: https://www.explainxkcd.com/wiki/index.php/882:_Significant

I would not call "P-Hacking" wishing to know the stats of the group you're in, especially if there are significant differences between groups.

No that's not P-hacking. The majority of US residents have been infected at this point. The sizes of those population groups both with and without serious pre-existing conditions are so huge that there is minimal chance of an incorrect random result.

Your link is not relevant to this issue so I don't know what point you're trying to make.

To be clear I support vaccination. But patients should make their own decisions free of coercion based on informed consent with full knowledge of the risks on both sides. Don't presume to patronize them with spurious, bad faith comments about "P-hacking". Let's see the actual data.

Right, you have a hypothesis. But you don't have data. Trying to tease out evidence for your hypothesis from existing data by slicing the data set up is the very definition of p-hacking. If you refuse to believe existing work on the basis of an unmeasured hypothesis, well, that's a different sort of logical fallacy on top of the bad statistics.

And FWIW, if you want a more complete (but much less amusing) explainer on the subject, wikipedia always comes through: https://en.wikipedia.org/wiki/Data_dredging

This is not at all correct. By your definition looking at anything in greater detail is P hacking.

For example, peanut allergies don't exist because most people don't have reactions to them. You absolutely can ask the question for if peanuts are good or bad for a subpopulation.

The alternative is just sticking your head in the sand.

The difference is in the motivations for asking the question and what's done with the answer (so agenda). People are assuming and agenda of others in this thread, and are crafting responses to head of that perceived agenda without actually answering the question, which is rarely productive, IMO. Better to ask than assume, or if the answer is known, provide it with a caveat that it doesn't actually support the specific cases you're worried it might be used for erroneously.

Wrong, I do not have a hypothesis (or any sort of preconceived notion). I am just curious to see what the data shows about risk stratification across age groups with and without common risk factors. Do you have any references to recent peer reviewed studies on that topic?

I'm not sure why you're trying to derail the discussion by bringing up data dredging. That would only potentially be relevant as a possible criticism of the methodology in a particular study. So what's your point?

The global risk of death from childbirth is X.

It isn’t p-hacking for a dude to inquire what the chances are for the male subgroup.

You didn’t stratify vaccination risk by age which immediately calls into question your comparative data. Indeed the myocarditis risk appears to be inversely correlated with age (exactly the opposite of the positive correlation of risk COVID itself has with age).

These numbers are useless going forward if they are based on the original Covid-19 strains and not the Omicron strains, which now account for virtually every case and are tens of times less lethal.

That's not true. Death rates are now much lower, but that seems to be mostly due to people having (partial) immunity due to vaccination and/or previous infection, and better treatment. Omicron variants are less intrinsically severe than delta, but delta was more intrinsically severe than the earlier variants, so it's not clear omicron's actually any lower than the original strain, alpha, etc.

For example, according to "Challenges in Inferring Intrinsic Severity of the SARS-CoV-2 Omicron Variant" (https://www.nejm.org/doi/full/10.1056/NEJMp2119682?query=fea...): "This meaningful but fairly small difference [vs Delta] implies that omicron, alpha, and wild-type SARS-CoV-2 have similar intrinsic severity."

The argument does not hold water. As Omikron replaced Delta, case mortality reduced immediately, despite skyrocketing cases. Population immunity does not explain this. The figure of 75% is only arrived at by correcting for underascertainment, which is an arbitrary estimate.