Data Analysis Shows That US Is Significantly Outpacing Italy | Hacker News Reader

Data Analysis Shows That US Is Significantly Outpacing Italy

covid.bio

27 pointsgsibble47 comments

Please if you're going to create stuff like this provide some context and answer questions like "What does this actually mean?". Without those things, this is just another scary looking title and graph that will make people more panicky than they already are. Given the ramp up in testing seen in my state, the large jump up to the cases in the graph isn't surprising. We know that the US has a far higher population than Italy and that testing rates in the US have been very low. Large spikes in confirmed cases are obviously going to happen as the testing ramps up. That's when we can start drawing conclusions and make better decisions. When we have a more clear picture of how widespread this is and what that actually means. What's the R0 and what's the mortality rate? On the other hand, if we accept that there are far more infected people right now than we know, the logical conclusion is that the mortality rate is low, otherwise there'd be dead people everywhere.

The huge issue with the numbers is that they always lag behind the actual situation right now. There is a delay of around 5 days until symptoms show, and something on top of that until you get tested and until the test is processed. And the deaths lag even further behind, it takes very roughly something like 3-4 week to die as far as I've read.

The deaths of today are the people that were infected 4 weeks ago, 4 week is an eternity on an exponential curve like this.

I think the differences in testing policies make it difficult to compare the absolute numbers. I think in a week or two it will be more meaningful to compare deaths rather than cases, because deaths seem to be more uniformly reported than cases.

Yes, I gave up on case numbers. Seeming CFR differences between countries are mostly artifacts of testing policies.

I created this both for now and the future.

Should we be comparing this in a per capital manner? USA population is five times Italy's.

At the beginning of the epidemic and the exponential growth the size of the population doesn't really matter, the number of available people isn't the limiting factor. The structure of the population probably matters quite a bit, but the total number of people in a somewhat arbitrary region probably doesn't. Unless containment measures prevent it, at the beginning you'll always have exponential growth no matter the size of the country.

I'm certainly not an expert, but I would assume that the individual, isolated outbreaks would be the most interesting numbers. Below a certain level they can be potentially contained by targeted measures, above a certain number of infected people you can't do any contact tracing or similar stuff anymore. And hospital resources are location-bound, there is only a limited amount of resource sharing you can do across the country.

I agree with your points. I am still a little surprised that the US outpaces Italy, because the social distance there is traditionally much shorter than in the US, e.g., people kiss each other on the cheeks, and give each other hugs more frequently than what I observed in the US.

The bay area shelter in place area had 467 cases yesterday and 107 on March 11. On that day New York had about 200 cases and today, 9 days later, it has close to 4000. Does it really matter if we have 5x more people if we grow that fast? We can catch up on that 5x in a pretty short time.

It is very dangerous to draw any conclusions from the number of cases in the US. They follow the number of tests that are done and those are limited by the test kits available.

I spoke with one doctor in upstate New York. She said that the local hospital, Albany Medical Center, was stopping community testing and going back to the old procedure of only testing people who met a narrow range of parameters. (Exposure, travel, etc.) Then she said, "So it looks like we'll start plateauing."

The rate of case growth will vary with the rate of testing.

I've heard multiple cases of people in NY turned away from testing as well even though they have been in close contact with others who tested positive or live with them.

It's very bad there I'm hearing, I think the best we can draw is that NY is probably very much underreporting the true number of cases due to test kit limitations.

Provided the testing strategy doesn't change (as the parent poster has stated may be the case), then it is a reasonable assumption that the testing is sampling the population in the same way, day by day.

So the important thing to look at is the growth rate. On that measure, it looks like the growth rate of cases in the US is outpacing that in Italy. That's worrying.

Then she said, "So it looks like we'll start plateauing."

Could you qualify how that statement is to be read? Is that an observation about what the results are going to look like as a result of limited testing, or is it a statement about the intention beind behind decreased testing (to make the number appear to be plateauing artificially).

Aside: 'Plateauing' is a funny word to type with all those vowels together.

She did not suggest anything about the intention. Only that wide-open testing would yield many more positives than tightly controlled testing. And if they're going to tighten testing that will reduce growth in positive cases.

But since testing in the US has been so extremely limited... should we be worried that the actual conclusion here is that "US is far outpacing Italy"?

Of course I'm not sure how available or limited testing is in Italy either.

If you're willing to accept a lot of hidden cases you will have to accept a much lower mortality rate. Just saying.

For now.

pity we dont have the number of tests conducted by each country also

Doesn't this data need to be normalized per-capita? It looks like it is just number of cases, and that doesn't get you a comparable. US population is about 5 times that of Italy.

I'm not trying to show percentage of infected, but the spread over time. And it's spreading a lot faster in the US and/or we are testing significantly.

I really wonder what the difference between these countries is:

- Germany: 15,320 cases / 44 deaths

- France: 10,947 cases / 243 deaths

- Spain: 17,963 cases / 830 deaths

There are various discussions and theories especially as to the low death rates in Germany. Germany has had distributed testing capabilities very early (end of January) and has performed a large amount of testing compared to other European countries, which means cases are cought early and not when they hit the ICU. Other theories are the large amount of ICU beds, local medical production facilities, still ongoing contact tracing and some other things. The expectation however is that the death rate will go up a fair bit.

Also - different reporting strategies. Germany is not always reporting someone as a CV death if the dominating factor was something else, whereas Italy reports them all as CV deaths. Or at least that's a widespread theory I've seen. It's hard to imagine such a huge difference can be down purely to testing.

I think the next days will see people dig into these stats and start a more complex discussion about what exactly it means to die of this virus, what exactly capacity means in a world where doctors can go to enormous lengths to save every life in quiet times but can't in crisis times. Just comparing raw numbers doesn't tell us the whole story.

You even see it among states in the US:

- New York: 4152 cases / 29 deaths

- Washington: 1228 cases / 75 deaths

In this case at least it appears that in Washington there was a connected set of old age homes that got hit pretty early before more preventative steps were being taken -- https://www.latimes.com/world-nation/story/2020-03-18/corona...

Age of the infected population?

The basis for “significant outpacing” is only one day / data point. Previous days follow Italy closely.

Besides that, we have a much larger total population. We’ve asserted counter-measures earlier. We are able to change our course based on our observed response of Italy.

This seems irresponsible to reactively share a “trend” based on a single data point of notable deviation.

I feel like comparing large countries should require some normalization.

How would the results change if we split the US into two countries - US East and US West? The current number of cases would be cut in half(?) but the "start" date would be pushed forward.

The total population of these two countries is pretty different though.

r0 doesn’t take per capita into account

clusters originally were isolated

Eventually this will matter though for absolute numbers, but our healthcare capacity is the bigger issue its not really a contest

Why is no one comparing the North of Italy with the South?

Why is no one comparing Madrid to the rest of Spain?

Germany and France have a very similar curve to Spain, Italy and even China, with a much smaller population as US.

It would be useful to know what "shows" means in this context. The normal meaning doesn't work.

Now, imagine a world where a large percentage of reality is presented to people in this manner, what might be the result? Does it look anything like what we've been living in for quite some time now, where it seems nobody can agree on what is actually true?

I may be way off base, but I think this is a much bigger issue than people realize, but it requires some effort and the will to see it.

This seems like a really nice chart. I haven't seen the data "normalized" to start date like this before.

I made one in logarithmic units. But it's on a teeny digital ocean instance and Streamlit, so please don't trample on it.

http://covid.superlabs.club:8501

Wouldn't data like this make more sense in a per capita comparison? I get that we're outpacing Italy, but are we outpacing.

US population is about Italy, Spain, Germany, France, UK, and Netherlands put together.

For that matter does China even register compared to the impact this is having on Italy on a per capita basis?

Also I think you have to look at the clusters (in a state) as independent threads growing in parallel. It makes more sense to compare individual state to European countries of similar population.

Very true. The most basic modeling of by country is not helpful, it'd be far better to show it by city, metro area, etc. And as you said each area has their own start date.

Shoot in the US most metro areas have their own quarantine date too.

If you look at China and South Korea, they both took about 11 days between crossing 100 cases to having a strong effect on their case# curve. (18 if you ignore the roughly 7 day incubation + getting tested period)

With incubation time removed:

Italy: 19 days from crossing 100, and counting

Iran: 15 days and counting

Spain: 11 days and counting

France: 12 days and counting

US: 9 days and counting

- Why would you report absolute numbers?

- You know that private labs are widely not reporting negative results back to CDC, right? (See Alexis Madrigal)

- We just activated testing in general availability yet virus has been here for months

This is bad statistics, and should be removed.

Absolute numbers seems like a bad comparison on a country with 6 times as many people.

Adding death's would be interesting. Especially Italy vs. Germany.

How is day zero chosen? Unclear if this is a reasonable comparison.

Yes, a very fair point. At the left when the tail is sooo thin, it's easy to move it left and right by quite a few days. When there aren't many cases, it's hard to choose a "day zero."

It doesn't seem to say here, but in other analysis I've seen it as the day the 100 case threshold was crossed.

Testing in Italy is also limited.

Italy has a decent testing coverage. US has almost no coverage.

https://www.nytimes.com/interactive/2020/03/17/us/coronaviru...

The US is terrible. But Italy is still limited, the only country that I'm aware of that is able to test everybody they need to test is South Korea. Everywhere else there are still pre-qualifications.

Italy did test a lot during the first days over 100 cases; after, since the number were so high, it limited testing to people with symptoms. I think most countries are straight out testing a small fraction of people with symptoms.

Reply on news.ycombinator.com