Hacker News Reader

Comment by abduhl | Hacker News Reader

I’m sorry, but is anyone else getting a real “this is good for bitcoin” vibe from the comments here? It seems we are all attributing this to a successful lockdown rather than considering other scenarios. It feels very “rock that wards off tiger”-y to me.

View on HN

A curve was trending exponential, and flattened with lockdown. What's the alternate scenario for what caused the curve to flatten?

Wrong assumptions/data/conclusions from the epidemic model

Misattribution of cause

Seems like it should be fairly trivial to test.

Different countries went into lockdown at different times. When did each one's curve start to flatten in relation to their lockdowns?

'the curve' is based on a model. All models are wrong. No matter how much it is broadcasted.

I'm not referring to the future models, I'm referring to the "how many cases have we confirmed per day" charts. (You could also do it with the excess death charts, given enough time for that data to settle out.)

Stuff like https://91-divoc.com/pages/covid-visualization/ and https://www.nytimes.com/interactive/2020/04/21/world/coronav....

Data from Italy makes for a good head-to-head comparison. Five regions locked down much sooner than their neighbors, and fared, at least initially, much better.

Preprint: https://t.co/JYf1F5GnYu?amp=1 Figure S3 is probably the most relevant.

We don't know when each country really got their first COVID-19 infections. That will always complicate analysis, and it will never be trivial. Never mind that we don't fully understand the virus itself, it's new.

Use the spikes in excess death rates, then.

https://www.nytimes.com/interactive/2020/04/21/world/coronav...

The relationship between the spike and the beginning of spread is changed by the effectiveness of lockdown/other measures. Not to mention other variables like population density, population activity, etc.

That SARS-CoV-2 was in our environment for months longer than thought, that 20% or more of Americans may already have antibodies, that tons of people suffered from not being able to get essential heart, dental and other surgery in preparation for some mythical surge that wouldn't have even happened anyway. That there's no evidence social distancing did anything at all?

There are tons of other possibilities here. There simply isn't enough data, and won't be until Feb 2021 where we can retroactively look back at data that is cleaned, pruned and properly averaged.

What data do you think will exist by next year that we don't have now? Doctors have recorded those who died with respiratory distress and other Covid19 symptoms, and that tested positive, as being victims -- that data is here now and shows increased deaths haven't been just to co-morbidities not being treated.

Test results are available to those in power.

The data is around, it's just not public because - I warrant - politicians can't manipulate the story if the public has the full data.

We have never had time series data for any other illness, to this granularity. The data is also terribly dirty.

The CDC doesn't track flu. It tracks pneumonia deaths, and then retroactively uses surveillance data to estimate the number of deaths from different infections.

U Washington also seems to be using entirely different data sets than Johns Hopkins and the metrics for international data is not the same at all. I did a post on some of the issues I found:

https://battlepenguin.com/tech/fighting-with-the-data/

> That SARS-CoV-2 was in our environment for months longer than thought, that 20% or more of Americans may already have antibodies

It's generally a good habit to do a little bit of simple arithmetic to see if the claims you are making pass a basic sanity check. If 20% of Americans already have antibodies, then how come NYC has 10x as many deaths per capita as the rest of the country? Did 200% of New Yorkers get infected?

Dr. John Ioannidis released the result of his Serology Study in California:

https://www.youtube.com/watch?v=jGUgrEfSgaU

We've found earlier cases in the US:

https://www.cnn.com/2020/04/22/us/california-deaths-earliest...

and it was spreading in France earlier than we thought:

https://doi.org/10.1016/j.ijantimicag.2020.106006 / https://news.ycombinator.com/item?id=23072956

We don't have the testing yet to really get good numbers/samples though. We need more random sampling antibody tests. A number of people claimed they got sick this year after leaving conferences in Las Vegas, but it's all anecdotal without wide scale efforts to just test random people for antibodies in cities in all US states.

NYC's numbers could be due to population density (amount of exposure can increase severity in some viruses, including SARS-CoV-2) and some of the increase could just be medical malpractice:

https://www.youtube.com/watch?v=CvhTQV5FNUE

> Dr. John Ioannidis released the result of his Serology Study in California:

That study showed that the seroprevalance in Santa Clara was less than 2%. That directly contradicts your earlier suggestion that 20% of the country was infected.

https://statmodeling.stat.columbia.edu/2020/04/19/fatal-flaw...

The seroprevalence number was also approximately the same size as the confidence interval on the false positive rate of the antibody test they were using, which means the confidence interval for the denominator of the fatality rate goes all the way down to zero, i.e. the fatality rate estimates are complete garbage.

> NYC's numbers could be due to population density (amount of exposure can increase severity in some viruses, including SARS-CoV-2)

Other death hotspots include New Orleans, Detroit, rural Georgia, etc. none of which have high population density.

> That SARS-CoV-2 was in our environment for months longer than thought, that 20% or more of Americans may already have antibodies

That would be a reasonable conclusion, perhaps, if the curve was flattening somewhat informally, and not specifically in places that have stronger lockdowns.

> That there's no evidence social distancing did anything at all

There’s reams of evidence in data across the world, and comparing locations in the US, that distancing has a substantial effect. Even most of the nutballs clamoring for radical reopening, like the President and his Republican allies, acknowledge that there is a trade-off in reopening, but argue that the lives to be lost from COVID-19 are a price worth paying for the near-term economic gains.

We raised testing capacity exponentially, but weren't accurately tracking the spread until recently.

A possible alternative - and I'm categorically not saying this is true, not even likely - is that Covid19 is more infectious, and more widely infected people. But, that fewer people were susceptible, so there would be more asymptomatic carriers but fewer people hospitalised. If a virus runs out of people to infect then it will peak.

You can prove that scenario wrong by testing (if number of 'already infected' is relatively low then this hypothesis would be wrong and the risk of further peaks is higher; I don't have enough data to attempt to disprove such a hypothesis.

This is what (reliable, and I repeat reliable) serological testing might tell us.

The curve will always flatten lockdown or no. It can't be exponential forever.

> A curve was trending exponential, and flattened with lockdown. What's the alternate scenario for what caused the curve to flatten?

You cannot honestly believe this. 'Trending exponential' is such a useless buzzword. Exponential growth is only scary depending on the exponent itself. Our models on what that exponent were were absolutely, completely off by orders of magnitude. Just look at Sweden, they were estimating 40k deaths by now and only have three thousand something.

Certainly, early action helped flatten the curve, but the original estimates for what the curve would be were incredibly off. There was no overwhelming of hospitals in the US at all, despite warnings that they were imminent, even with behavioral changes. Our country's scientific advisers need to look inward at this point and figure out what went wrong with the models.

> It seems we are all attributing this to a successful lockdown rather than considering other scenarios

The curve in parts of the US without a strong lockdown isn't flattening. NYC just dominated the counts so much that it makes the whole country look like it is flattening if you don't break things out. There's similar indications in data from international experience

So, there's a lot more basis for conclusions than “anti-tiger rock” situation.

The curve has flattened i Sweden too, despite the lack of strong lock down. Countries are different and so on, still I think it is way to early to conclude that there is a causative effect between lock downs and reduced virus spread. My own theory is that improved hand hygiene is what stopped the spread but we'll see what the future data says.

Reply on news.ycombinator.com