Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Covid-19 Antibody Seroprevalence in Santa Clara County, California

medrxiv.org

Post reply on HN
Print view View original

81–90 of 209 posts

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#81

Post by usaar333 »

dsjoerg wrote: post #69
compiler-guy wrote: post #50

Earlier quoted context omitted.

These are very common statistical methods when dealing with surveys and populations. It isn't this paper's job to describe the background on something like that, just like it isn't this papers job to explain blood testing. They describe their limitations and adjustments exactly like most other studies of this sort do. For example: "Those imbalances were partly addressed by weighting our sample population by zip code,…

Adjusting for selection bias is really hard and potentially first-order.

Correct. For some things you could use baseline data, but there is zero baseline data to use here.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#82

Post by brink »

opportune wrote: post #78

I don't see anything about sampling bias here. I would assume that people who have been ill, experienced mild symptoms, or who were exposed to confirmed cases would be more likely to respond to the facebook ads recruiting testing volunteers. I'm sure we are significantly undercounting cases but I highly doubt we are off by a magnitude of 50x.

I know. 50-85x is hard to believe. How could we be that far off?

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#83

Post by greedo »

hedora wrote: post #35
greedo wrote: post #12

Earlier quoted context omitted.

You would have seen more reports in hospitals of interstitial pneumonia, as well as a higher YoY death rate during January.

The CDC is reporting an abnormally high pneumonia death rate in January, and it can’t be explained by confirmed flu cases: https://www.cdc.gov/flu/weekly/#S6 Look carefully at the black and red graph after the phrase “mortality surveillance data”.

The graph and text clearly state it refers to pneumonia and influenza, if you refer to the legend where it says Percent P&I. The P&I for the first 8 weeks of the year is clearly within normal ranges. It spikes from baseline in the first week of March continuing on to the present day.

If you look at the first set of graphs, you'll see a spike around the same time for influenza tests. I don't know how fast pneumonia occurs in people weakened by influenza, but I would expect some delay.

I think the spike in both influenza and in death due to influenza and pneumonia is probably due to testing. If you presented with influenza like symptoms in March, you were definitely given a flu test to rule out influenza.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#84

Post by gfodor »

opportune wrote: post #78

I don't see anything about sampling bias here. I would assume that people who have been ill, experienced mild symptoms, or who were exposed to confirmed cases would be more likely to respond to the facebook ads recruiting testing volunteers. I'm sure we are significantly undercounting cases but I highly doubt we are off by a magnitude of 50x.

They mention it, but state it's hard to ascertain. I agree with that assessment. Given the magnitude of the delta between confirmed cases and projected cases based on this study tho, it seems intuitively obvious there's a pretty big gap. (And not surprising, given the lack of testing, the asymptomatic nature of the disease, etc.)

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#85

Post by jplayer01 »

brink wrote: post #82
opportune wrote: post #78

I don't see anything about sampling bias here. I would assume that people who have been ill, experienced mild symptoms, or who were exposed to confirmed cases would be more likely to respond to the facebook ads recruiting testing volunteers. I'm sure we are significantly undercounting cases but I highly doubt we are off by a magnitude of 50x.

I know. 50-85x is hard to believe. How could we be that far off?

It’s not hard to believe when you consider testing criteria.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#86

Post by all_usernames »

erentz wrote: post #79

It’s hacky and I’m happy to be told how wrong it is. But based on these recent antibody studies in Germany, Finland, and now here in CA, I’ve been assuming an actual fatality rate of about 0.4, and thus an actual infected rate of 250x our known deaths, which is a much firmer number. This is only a small comfort since it means we may have had about 8.75 million infected and presumably now immune in the USA. Or about 2…

It's known that the vast majority infected do not experience symptoms, or have mild symptoms. I don't think you can draw the conclusion that they had prior immunity.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#87

Post by usaar333 »

georgewfraser wrote: post #31

There are now multiple sources of evidence pointing to a very low true infection fatality rate, as low as 0.1% in some regions. Iceland has been testing a semi-random sample using PCR since early in the epidemic, and there have been several small antibody surveys in “hot spots” showing infection rates over 15%. You can also use CDC surveillance of flu-like illnesses to estimate the total number of infected nationwide…

CDC flu surveillance data suffers from even more selection bias than this study. It's basically useless.

Iceland PCR data is reliable but it is suggestive of an IFR of more like 0.3%. Except Iceland also managed to successfully isolate their elderly so if you correct for that, we're back at 0.7% or so for the general population.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#88

Post by gfodor »

brink wrote: post #82
opportune wrote: post #78

I don't see anything about sampling bias here. I would assume that people who have been ill, experienced mild symptoms, or who were exposed to confirmed cases would be more likely to respond to the facebook ads recruiting testing volunteers. I'm sure we are significantly undercounting cases but I highly doubt we are off by a magnitude of 50x.

I know. 50-85x is hard to believe. How could we be that far off?

The # of cases reported that I believe they are comparing against is the number of confirmed positive tests, not the number of assumed cases in the county based on some other model that is trying to project # infected. I don't think anyone thought that we had 100% coverage where all cases were tested. So we're not "wrong" if such a study as this sees a gap between reported confirmed cases and expected infected, it just means we have a better understanding of the testing gap, which is self evident in existence but the magnitude of which we don't know.

Given all the other factors, an order of magnitude or more gap between tests and sick people doesn't seem completely out of the question. I would be curious if there are other models using a different methodology which could help us get a handle on the conditional probability chain leading to tests being done or not on an individual, to see if there is a similar set of conclusions.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#89

Post by all_usernames »

opportune wrote: post #78

I don't see anything about sampling bias here. I would assume that people who have been ill, experienced mild symptoms, or who were exposed to confirmed cases would be more likely to respond to the facebook ads recruiting testing volunteers. I'm sure we are significantly undercounting cases but I highly doubt we are off by a magnitude of 50x.

If the vast majority of people contracting the virus never have symptoms, or have symptoms mild enough that they don't seek a doctor, or they DO seek a doctor but can't get a test because of testing criteria restrictions, I think the 50-85x number starts to make a lot of sense.

Re: Covid-19 Antibody Seroprevalence in Santa Clara County, California

#90

Post by guscost »

_9vpl wrote: post #51
pwned1 wrote: post #5

Earlier quoted context omitted.

Based on the difference between CA and NY, I'd be willing to bet that it got here earlier, maybe in December.

I think the more likely explanation is that NY simply is a lot more dense and less reliant on cars vs the less dense and more reliant on cars California.

Especially if the infectious dose also affects severity, as is the case with other respiratory illnesses: https://www.ncbi.nlm.nih.gov/pubmed/25416753

Basically, if you're in a suburban area and pick up a few viral particles from one coffee-shop counter, it's possible that you're not in nearly as much danger as someone touching handrails and inhaling subway air all day long.

Post reply on HN