10 ms·

Epidemiologist here, a few thoughts reading other comments: -Death data is preferable to new cases in some ways, since it doesn't rely on testing, which ramps

Epidemiologist here, a few thoughts reading other comments: -Death data is preferable to new cases in some ways, since it doesn't rely on testing, which ramps up/down from locality to locality and over time. -To compare deaths from the Spring with deaths recently isn't apples to apples: we know now remdesivir and dexamethasone are helpful treatments, especially the latter, and these may help reduce mortality rate. -Mortality now may be lower partly due to: if the virus killed off the more vulnerable populations (nursing home) early on, it has less frail individuals to infect now. These communities got ravaged, we had something like 40% of deaths in my area in nursing homes in our 6-week case peak. -Trying to tease out immunity is very tough as well, early on there were few good tests and their specificity and sensitivity were less well known. At my agency, there were multiple tests we were collating, with different false positive and negative %s, and different reporting times (which all can influence R0, the variable number our behavioral interventions seek to tamp down). -Comparing different countries is very tricky. Sweden is a favorite example for non-lockdown approach, but there we see it's very tricky to tease out the benefits. On the one hand, they didn't do much better epidemiologically than their neighbors. On the other hand, their economy was surrounded by locked down ones, so that may have tamped down any benefits their non-lockdown would've garnered due to neighbors' activities. Did it pay off? The verdict is still out, IMO. Things are very complex in infectious diseases epidemiology. It's extremely hard to know anything for certain with this virus, it's behavioral tangled up with comorbidity with genetics and demographics and evolving treatment and viral dosage and strain of virus...it will take a few years of unpacking the mountains of data before we truly have a grasp of what happened here.

asah 6y ago
q for you: early on, there was a big variance (region vs region, one region over time, etc) in how mortality was reported, e.g. heart attack vs COVID, single cause vs multi-cause, etc. Is mortality more standardized now?
- craftinator 6y ago
  I've been wondering this myself; last I heard (2 months ago, fairly credible investigative journalism), there was a massive disconnect, state to state, how COVID-19 deaths were assessed and recorded. IIRC, Florida was on one end of the spectrum, tagging deaths as from any comorbidity present (only tagging as COVID if they couldn't find anything else), to California, which was tagging as COVID without test confirmation, if enough of the symptoms were present.
  - username90 6y ago
    California and Florida have about the same gap between excess mortality and registered covid deaths, with Florida being slightly better. You can see it here: https://www.nytimes.com/interactive/2020/05/05/us/coronavirus-death-toll-us.html https://www.nytimes.com/interactive/2020/05/05/us/coronaviru...
  - sunsetMurk 6y ago
    Do you have any links/sources for this?
  - spfzero 6y ago
    I think the CDC numbers say that, at least nationwide, 6% of deaths list only Cov 19 as the cause of death. 94% list other causes. This data is so messy, I don't think there is much of a take away. Of the 6%, could there have been other reasons which were missed? Certainly some. Of the 94%, would they have died anyway sans Cov 19? Certainly not all, and not then. As others have pointed out, we should pay attention to the delta in deaths over some other year with an average flu season, and a year with a bad flu season. Someone will run these numbers at the end of 2020, I hope, and then we'll have a better idea of the effect of this particular pandemic.
    - chrisco255 6y ago
      They've already been running those numbers and they're available on CDC's website: https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm (Scroll way down for the chart) We had a bad 4-6 weeks a couple months ago but we're already down in the range of a normal flu season as far as deaths go.
- beagle3 6y ago
  Not GP. In my opinion, the only reliable measure is excess death (if you believe death count, which I trust for most countries, but not all countries), because it is invariant to these (often political) classifications. See e.g.[0] (it is up-to-date for 11-sep, despite the link date saying April 21). In fact, these graphs would look the same even if you didn't have access to any testing/diagnostics/causes-of-death. It is easily visible which countries have gone back to baseline, which countries haven't, and which didn't even have a "first wave", which likely means that they are in great danger of one happening. [0] https://www.nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html https://www.nytimes.com/interactive/2020/04/21/world/coronav...
xorfish 6y ago
Would't any preexisting immunity be accounted for by any decent estimate for R0 and the IFR?
- bluGill 6y ago
  Not really. If there is existing immunity then the measured R0 will be lower than the actual R0, but we don't know how to measure the true R0 in absence of the current population with their immunity so we can never know what the true R0 might actually be. Note that true R0 is a concept I made up on the spot. I think it is clear what I mean but experts in this may have better terms.
- tripletao 6y ago
  A first-order SIR model with unmodeled pre-existing immunity will be accurate initially, but get gradually more wrong with time. For a simplified, non-representative example with fake numbers, let's say half the population already had perfect sterilizing (they don't get sick, and the virus doesn't replicate at all so they can't spread it to anyone else) immunity. Then an apparent R0 of 2.5 would imply an actual R0 of 5. But R0 = 2.5 would imply you need 1 - 1/2.5 = 60% of the population infected for herd immunity, while in that hypothetical case you'd need only 30% infected (so the 50% initially immune plus that 30% make 1 - 1/5 = 80%). The above still assumes a well-mixed population, but with heterogeneous susceptibility (using the terminology from [1]). There's also heterogeneous connectivity (i.e., the fact that some people have more social contacts than others), which further reduces the herd immunity threshold. To be clear, the above is a fake example. We know that almost no one has perfect pre-existing sterilizing immunity to the novel coronavirus, because we've seen highly-connected populations (homeless shelters, slums, boats, etc.) where almost everyone tested positive. But that doesn't mean everyone is equally susceptible! Immunity is a continuum, not a binary thing, and the 20-50%[2] of people with reactive T cells very likely get less sick given the same initial viral dose (or, stated in the other direction, would need a greater viral dose to get equally sick). They also might spread the virus less, though that's more speculative. Finally, I'll note that the herd immunity threshold is the point where R becomes < 1, and the number of people currently infected starts to shrink instead of growing. People still get infected on the downslope, so the fraction of people infected can overshoot higher (and potentially much higher, especially if no precautions are taken to slow the spread). 1. https://www.medrxiv.org/content/10.1101/2020.07.23.20160762v2 https://www.medrxiv.org/content/10.1101/2020.07.23.20160762v... 2. https://science.sciencemag.org/content/early/2020/08/04/science.abd3871?ijkey=a3504aae1c4fe757a654a7b4c742d55a3ebc5828&keytype2=tf_ipsecsha https://science.sciencemag.org/content/early/2020/08/04/scie...
spfzero 6y ago
Well, its a bit unfair to say that since Sweden didn't reap all of the benefits of protecting their economy, because of neighboring countries' following a different path, that something was wrong with Sweden's approach. If I had the right idea, but am surrounded by people with the opposite idea, that doesn't mean my idea became wrong. It only means I maybe wasn't the best at convincing my neighbors.
- Orou 6y ago
  Where did they say something was wrong with Sweden's approach?
- heylook 6y ago
  Because you exist within a system whether you like it or not, so the optimal/correct solution must account for those variables, including your ability to entice cooperation. So yes, you would still be wrong, just not for the narrow way you’re choosing to redefine “correct.”
9HZZRfNlpR 6y ago
Any thoughts on Sweden (and many other EU countries) saying masks have very little scientific evidence for the little benefit they supposedly offer. How can scientists in different countries have so different views. Some countries preach masks like holy water and they both say God (science) is on their side.
TheSpiceIsLife 6y ago
Where would I look of I wanted to follow geographicly diverse epidemiology news / publications / groups / individuals, not just covid related. Thanks in advance.
kybernetikos 6y ago
Comparing Sweden with its neighbours really highlights how poorly it has done: https://ig.ft.com/coronavirus-chart/?areas=fin&areas=swe&areas=nor&areas=isl&areasRegional=usny&areasRegional=usca&areasRegional=usfl&areasRegional=ustx&byDate=0&cumulative=0&logScale=1&perMillion=0&values=deaths https://ig.ft.com/coronavirus-chart/?areas=fin&areas=swe&are... Pretty much any way you slice the data, a lot more people died in Sweden than in its neighbours. You can see the same pattern in excess all cause mortality too: Denmark 200, Finland 600, Norway 0, Iceland 0, Sweden 5500. Now Sweden has a population slightly less than Finland and Norway combined, and figures can be difficult to compare, but still it's hard to avoid the conclusion that so far, around 4000 people in Sweden have died because of differences in policy compared to its neighbours.
- kzrdude 6y ago
  The least you can say is that Sweden didn't manage to test as many as its neighbors did. It is claimed the influx of infected was greater etc, but lack of tests meant that it was hard to get it under control too.
- api_or_ipa 6y ago
  The better question that Sweden may help answer is whether shutting down the economy was worth it. Was the deaths of 5,500 people worth keeping the economy open, in other words, could you have saved more lives by pursuing other health priorities and let covid run it's courses as Sweden elected to do?
  - EricE 6y ago
    This. The deaths directly attributed to the virus alone aren't the only relevant deaths. Suicides are up. Domestic violence is up. Child abuse may be up but schools are all "virtual" so good luck detecting it (as if it wasn't hard enough when kids were mostly in school). Talk about throwing out the baby with the bath water...
    - barumi 6y ago
      Could you show any data that supports your assertion about rates of suicides, domestic violence, and child abuse?

asah 6y ago

q for you: early on, there was a big variance (region vs region, one region over time, etc) in how mortality was reported, e.g. heart attack vs COVID, single cause vs multi-cause, etc. Is mortality more standardized now?

craftinator 6y ago
I've been wondering this myself; last I heard (2 months ago, fairly credible investigative journalism), there was a massive disconnect, state to state, how COVID-19 deaths were assessed and recorded. IIRC, Florida was on one end of the spectrum, tagging deaths as from any comorbidity present (only tagging as COVID if they couldn't find anything else), to California, which was tagging as COVID without test confirmation, if enough of the symptoms were present.
- username90 6y ago
  California and Florida have about the same gap between excess mortality and registered covid deaths, with Florida being slightly better. You can see it here: https://www.nytimes.com/interactive/2020/05/05/us/coronavirus-death-toll-us.html https://www.nytimes.com/interactive/2020/05/05/us/coronaviru...
- sunsetMurk 6y ago
  Do you have any links/sources for this?
- spfzero 6y ago
  I think the CDC numbers say that, at least nationwide, 6% of deaths list only Cov 19 as the cause of death. 94% list other causes. This data is so messy, I don't think there is much of a take away. Of the 6%, could there have been other reasons which were missed? Certainly some. Of the 94%, would they have died anyway sans Cov 19? Certainly not all, and not then. As others have pointed out, we should pay attention to the delta in deaths over some other year with an average flu season, and a year with a bad flu season. Someone will run these numbers at the end of 2020, I hope, and then we'll have a better idea of the effect of this particular pandemic.
  - chrisco255 6y ago
    They've already been running those numbers and they're available on CDC's website: https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm (Scroll way down for the chart) We had a bad 4-6 weeks a couple months ago but we're already down in the range of a normal flu season as far as deaths go.
beagle3 6y ago
Not GP. In my opinion, the only reliable measure is excess death (if you believe death count, which I trust for most countries, but not all countries), because it is invariant to these (often political) classifications. See e.g.[0] (it is up-to-date for 11-sep, despite the link date saying April 21). In fact, these graphs would look the same even if you didn't have access to any testing/diagnostics/causes-of-death. It is easily visible which countries have gone back to baseline, which countries haven't, and which didn't even have a "first wave", which likely means that they are in great danger of one happening. [0] https://www.nytimes.com/interactive/2020/04/21/world/coronavirus-missing-deaths.html https://www.nytimes.com/interactive/2020/04/21/world/coronav...

xorfish 6y ago

Would't any preexisting immunity be accounted for by any decent estimate for R0 and the IFR?

bluGill 6y ago
Not really. If there is existing immunity then the measured R0 will be lower than the actual R0, but we don't know how to measure the true R0 in absence of the current population with their immunity so we can never know what the true R0 might actually be. Note that true R0 is a concept I made up on the spot. I think it is clear what I mean but experts in this may have better terms.
tripletao 6y ago
A first-order SIR model with unmodeled pre-existing immunity will be accurate initially, but get gradually more wrong with time. For a simplified, non-representative example with fake numbers, let's say half the population already had perfect sterilizing (they don't get sick, and the virus doesn't replicate at all so they can't spread it to anyone else) immunity. Then an apparent R0 of 2.5 would imply an actual R0 of 5. But R0 = 2.5 would imply you need 1 - 1/2.5 = 60% of the population infected for herd immunity, while in that hypothetical case you'd need only 30% infected (so the 50% initially immune plus that 30% make 1 - 1/5 = 80%). The above still assumes a well-mixed population, but with heterogeneous susceptibility (using the terminology from [1]). There's also heterogeneous connectivity (i.e., the fact that some people have more social contacts than others), which further reduces the herd immunity threshold. To be clear, the above is a fake example. We know that almost no one has perfect pre-existing sterilizing immunity to the novel coronavirus, because we've seen highly-connected populations (homeless shelters, slums, boats, etc.) where almost everyone tested positive. But that doesn't mean everyone is equally susceptible! Immunity is a continuum, not a binary thing, and the 20-50%[2] of people with reactive T cells very likely get less sick given the same initial viral dose (or, stated in the other direction, would need a greater viral dose to get equally sick). They also might spread the virus less, though that's more speculative. Finally, I'll note that the herd immunity threshold is the point where R becomes < 1, and the number of people currently infected starts to shrink instead of growing. People still get infected on the downslope, so the fraction of people infected can overshoot higher (and potentially much higher, especially if no precautions are taken to slow the spread). 1. https://www.medrxiv.org/content/10.1101/2020.07.23.20160762v2 https://www.medrxiv.org/content/10.1101/2020.07.23.20160762v... 2. https://science.sciencemag.org/content/early/2020/08/04/science.abd3871?ijkey=a3504aae1c4fe757a654a7b4c742d55a3ebc5828&keytype2=tf_ipsecsha https://science.sciencemag.org/content/early/2020/08/04/scie...

spfzero 6y ago

Well, its a bit unfair to say that since Sweden didn't reap all of the benefits of protecting their economy, because of neighboring countries' following a different path, that something was wrong with Sweden's approach. If I had the right idea, but am surrounded by people with the opposite idea, that doesn't mean my idea became wrong. It only means I maybe wasn't the best at convincing my neighbors.

Orou 6y ago
Where did they say something was wrong with Sweden's approach?
heylook 6y ago
Because you exist within a system whether you like it or not, so the optimal/correct solution must account for those variables, including your ability to entice cooperation. So yes, you would still be wrong, just not for the narrow way you’re choosing to redefine “correct.”

9HZZRfNlpR 6y ago

Any thoughts on Sweden (and many other EU countries) saying masks have very little scientific evidence for the little benefit they supposedly offer. How can scientists in different countries have so different views. Some countries preach masks like holy water and they both say God (science) is on their side.

TheSpiceIsLife 6y ago

Where would I look of I wanted to follow geographicly diverse epidemiology news / publications / groups / individuals, not just covid related. Thanks in advance.

kybernetikos 6y ago

Comparing Sweden with its neighbours really highlights how poorly it has done: https://ig.ft.com/coronavirus-chart/?areas=fin&areas=swe&areas=nor&areas=isl&areasRegional=usny&areasRegional=usca&areasRegional=usfl&areasRegional=ustx&byDate=0&cumulative=0&logScale=1&perMillion=0&values=deaths https://ig.ft.com/coronavirus-chart/?areas=fin&areas=swe&are... Pretty much any way you slice the data, a lot more people died in Sweden than in its neighbours. You can see the same pattern in excess all cause mortality too: Denmark 200, Finland 600, Norway 0, Iceland 0, Sweden 5500. Now Sweden has a population slightly less than Finland and Norway combined, and figures can be difficult to compare, but still it's hard to avoid the conclusion that so far, around 4000 people in Sweden have died because of differences in policy compared to its neighbours.

kzrdude 6y ago
The least you can say is that Sweden didn't manage to test as many as its neighbors did. It is claimed the influx of infected was greater etc, but lack of tests meant that it was hard to get it under control too.
api_or_ipa 6y ago
The better question that Sweden may help answer is whether shutting down the economy was worth it. Was the deaths of 5,500 people worth keeping the economy open, in other words, could you have saved more lives by pursuing other health priorities and let covid run it's courses as Sweden elected to do?
- EricE 6y ago
  This. The deaths directly attributed to the virus alone aren't the only relevant deaths. Suicides are up. Domestic violence is up. Child abuse may be up but schools are all "virtual" so good luck detecting it (as if it wasn't hard enough when kids were mostly in school). Talk about throwing out the baby with the bath water...
  - barumi 6y ago
    Could you show any data that supports your assertion about rates of suicides, domestic violence, and child abuse?