Rt: Effective Reproduction Number of Covid-19 | Hacker News Reader

Rt: Effective Reproduction Number of Covid-19

rt.live

158 pointsradkapital88 comments

IMO this looks to by a symptom of lack of testing. Overall the ratio of tests coming back positive is unchanged:

https://raw.githubusercontent.com/lettergram/covid19-analysi...

Even though we are doing more tests. Meaning the reduction in speed-of-spread likely has more to do with lack of testing.

That being said, I’m sure spread is being reduced, we just don’t know to what extent because we don’t have enough effective testing.

It’s also possible (probably likely) our tests are not exceptionally accurate. It’s the best we have, but given all of this is <6 months old the false positive/negative rate can be high.

Generally, Testing isn’t an effective way to measure spread. Probably we should use “hospitalizations with flu symptoms“ as the best indicator.

You can compare testing per state and deaths here:

Tests: https://raw.githubusercontent.com/lettergram/covid19-analysi...

Deaths: https://raw.githubusercontent.com/lettergram/covid19-analysi...

Overall repo: https://github.com/lettergram/covid19-analysis

Correct. Testing is only a good measurement with massive testing, even randomized testing.

Iceland's data is probably the closest to actual infections and many states probably look similar. (https://www.covid.is/data)

The reported specificity and sensitivity of RT-PCR testing is quite good, but obviously not 100%. RT-PCR is considered the "gold standard" in viral testing.

If you want to contact trace, we would need many orders of magnitude more testing.

If you want to know prevalence over time, and therefore Rt, we could ascertain it very accurately with much less testing, but it would need to be done against statistically valid and unbiased random samples of the population, not at all what is going on today.

Since the testing we are doing today doesn't really satisfy either condition, and there's also no front-line approved treatments for those that are definitively diagnosed (rather, we treat the symptoms, which are self-evident), frankly I'm not sure I understand the extreme focus on testing.

These tests have a false negative rate of 30%

My understanding is that it’s under 10% false negative and false positive typically for this type of testing.

Other tests like the rapid saliva test are totally different.

False negative is higher.

I'm not sure I understand the extreme focus on testing.

3blue1brown on YouTube created a simple model that suggests that testing and quarantine of positive cases is the most effective way to slow the spread (within the simplified model). Testing should allow us to isolate just the infected, and ease off the restrictions on others.

With statistically valid random testing, even detecting clusters / communities (as opposed to individuals) and clearing same would be conducive to reopening the economy.

Unfortunately, the path to doing so "I'm going to force you to get tested, despite you having no symptoms" is political anathema.

> "I'm going to force you to get tested, despite you having no symptoms" is political anathema

This seems less extreme and more palatable than "I'm going to force you to stay home and lose your job, despite you having no symptoms" which is very nonetheless popular right now.

No, no, there’s never going to be enough testing to test asymptomatic cases — Thats everybody.

And a test doesn’t tell you you aren’t going to be positive tomorrow. It just tells you you weren’t positive 48 hours ago!

Fauci explained all this live at the press conference. I wish people would listen. With AIDS it’s different. You can get tested, be negative, take no at-risk behaviors, and a year later that negative test is still relevant. With COVID by the time you have a negative test result back it’s no longer even relevant. Only positive tests are relevant.

The fact is that there’s no point in testing people to demonstrate that they are negative. And there will never be enough testing to test every asymptomatic person, and a significant proportion of spread is from asymptomatic carriers. Those three incontrovertible facts lead to some very basic and sobering conclusions.

So this fear of people being forced to show a negative test assumes we’ve somehow found our way into an anti-science twilight zone kabuki theater to begin with. Which frankly wouldn’t be far off from where we are already, but can we please at least acknowledge it’s based on fear and hysteria and not any kind of science?

> statistically valid random testing

Believe you missed a word in there?

Random testing for surveillance purposes would absolutely require testing asymptomatic people.

Not because you care if they're sick (though that's useful information), but because you care if the community has a change in the Sars-CoV-2 infection rate.

> With COVID by the time you have a negative test result back it’s no longer even relevant.

A family friend died on Friday. His heart went out on him. Big guy on his 60's. He got tested for COVID-19 before he died at the hospital. He and his now widow had some sinus congestion. She couldn't enter to he hospital to be with him when he died. A few days later, his widow got the negative result for her husband. She is alone, scared, dealing with this loss. can't see her kids or grandkids. She can sleep a little better at night knowing that death is not knocking at her doorstep.

Relevant? From a human perspective it is most definitely relevant. Maybe not as much from a epidemiology or number cruncher's perspective but like all expert opinions they are open to interpretation and second opinions. Statistics are easy to cherry pick for whatever agenda is being pushed behind the scenes.

Just like masks were not relevent at first. They were scarce. Suddenly they are relevant for the public to use. I believe this testing scarcity and related lack of any cohesive national mobilization is the biggest failure of our US government. From a public health perspective this dismissal of negative testing is just a way to manage the demand side for testing.

Public health and epidemiology experts have very different motivations than individuals dealing with loss and fear during this crazy time.

> I'm not sure I understand the extreme focus on testing.

Group think.

"hospitalizations with flu symptoms" is a good indicator, but you have to account for the fact that people who catch it in different areas (or with different incomes) will have different levels of access to a hospital. many people are dying (or recovering) at home.

Not just that.

NYC has seen the number of people dying at home increase from around 20 per day to 200 per day. The difference is likely a mix of COVID-19 cases, people not seeking treatment out of fear, and people just spending significantly more time at home. But, that kind of change makes accurate statistics even more difficult.

Just read an interesting article using ICU admissions as a proxy. Of course it does lag, but it may be more consistent than hospital admissions.

http://medrxiv.org/cgi/content/short/2020.04.13.20063388

It does saturate at some point, though. If you're turning patients away who are in need of ICU treatment (as some Italian hospitals were), your ICU admissions will be constant and equal to your rate of release from ICU, so it will appear that you've "flattened the curve" even if the number of sick people continues to grow exponentially.

True, the paper talks about patients needing ICU, if you can get that statistic instead.

Here's an alternative set of per-US-state Rt estimates from the London School of Hygiene & Tropical Medicine, with somewhat different results: https://epiforecasts.io/covid/posts/national/united-states/

They believe there's insufficient data to estimate Rt today, though, due to lags in symptoms/diagnosis/testing/reporting. So their estimates lag about 10 days, with the x-axis for the Rt-over-time graphs currently running February 17 through April 7.

They also have per-country estimates: https://epiforecasts.io/covid/posts/global/

I've been following their maps over the last week and I have noticed a lot of variation day-to-day. I have seen three different statuses for California so far, for example.

Great looking site. For those who don't know/realize - this is the work of Kevin Systrom, Mike Krieger (and team?) - co-founders of Instagram.

Statistically speaking - love that it has error bars and timeseries. Makes it infinitely more informative to track relative to news, events, and mitigation strategies and have a sense of the confidence of the estimate. The provided Jupyter notebook is a great resource.

Alphabetically speaking - I enjoyed the puzzle of figuring out how the states were sorted in the timeseries list (e.g. why was Alaska before Alabama, Iowa before Idaho, etc.). Turns out it's alphabetical by the two letter state abbreviation which is not shown - just a fun observation :)

Ah, thank you for pointing out the author. Helps with the inferiority complex for my past attempts at data visualization dashboards.

It does say in the footer: "Data analysis by Kevin Systrom and site built by Mike Krieger, with thanks to Ryan O’Rourke and Robby Stein."

I particularly liked browsing the code for the hero chart, which you can find under static/js/StackViz.js. 630 lines of JS and you can see the level of attention to detail.

I couldn't find a copyright or licensing statement, but there are a few pieces of StackOverflow code copied and pasted into the bundle, so in theory the whole code base is therefore open source.

Where did you find static/js/StackViz.js?

It's in the bundle. Using dev tools in your browser you can get an expanded view of the file/folder structure.

This really illustrates that the parameter in question is not just a characteristic of the virus, but also of the environment.

I think a lot of people forget that if try to return to normal activity this is just going to shoot back up. I have yet to see an exit strategy other than "modulate social distancing such that the health care system is near capacity, we can gradually open up as more people are immune due to prior infection."

> more people are immune due to prior infection

This is not a given. Other coronaviruses tend to confer only temporary immunity, and this one might be similar.

This thing could become seasonal unless/until we make a very effective vaccine against it. <shudder/>

> an exit strategy

Widespread vaccinations. Yes, I understand how far away that is likely to be. Yes, the economic ramifications are relatively terrifying.

It's not obvious that it will shoot back up, depending on how careful we all about the return. People talk a lot about the multiple waves of the Spanish Flu, but the logical consequence of that is that transmission of a respiratory disease can be sustainably reduced; otherwise there would have only been the one wave.

The flu is seasonal, the first wave died down during the summer and the second wave started in the fall.

"Seasonal" just means that we see this pattern regularly every year. It's not that there's some specific property of the flu which stops it from spreading during the summer - there could be, but we don't know.

It's wild that California has been locked down for a full month and the reproduction number is still hovering around 1 and even went up a couple of days ago.

If that's actually true, we shouldn't expect to see a quick decrease in cases and we probably aren't on track to end the lockdown any time before at least June. I can't help but wondering if the unlimited outside time under the guise of "exercise", but which people are clearly taking advantage of by going to hang out with friends in the park, has anything to do with it. Or perhaps it's the lack of masks in indoor areas like grocery stores that has prevented a bigger drop (in SF, only starting today has that finally changed).

For all that California has been praised for acting quickly, it still feels like we've really been slow playing this lockdown by being so lenient, and without really having gained anything. And it doesn't really seem like there's any urgency anymore around continuing to increase testing, even though we know that's needed to open back up. Now we're going to be stuck inside for weeks or months longer than we would have needed to be if we had just copied what successful countries (Taiwan, South Korea) did from the beginning.

Or may be the infection is already so widespread as some studies suggest that any observed changes is just a noise and all the lockdowns/etc. don't affect it much. Sweden for example isn't that different despite different approach.

Without wide testing we don't really know and are just blindly drifting along the 5 stage model - denial, anger (that is lockdown), and now we're moving onto the negotiating with all that conditional reopening. It is pretty telling that anger to negotiating stages change is happening without improvement of the real situation.

Sweden has one of the highest per capita death rates from COVID-19 in the world, and one of the lowest testing rates in Europe.

nice play.

Those are the numbers https://www.statista.com/statistics/1104709/coronavirus-deat... . The whole Europe and US have the highest deaths per capita in the world. Comparing to the peers - Western Europe - Sweden is midway in the list (and just above US) pretty much evidencing the point that whatever actions have been taken in all those countries it just doesn't seem to matter much.

> Sweden is midway in the list (and just above US) pretty much evidencing the point

Sweden is surely not "midway in the list" unless in an imaginary world where one thinks that all European countries fit on the accidental 13-item (!) part of the whole list of the Statista site.

Try it again but with the list of all European countries and considering the actual values and let me know if that "point" still exists. Especially when adding the countries which were able to do more testing than Sweden and introduced at least a bit stronger measures. E.g. Austria is comparable to Sweden in size, was potentially significantly more exposed in the begin, bordering Italy, and having three times less deaths per capita at the moment.

Moreover, people also mention Sweden as an example of "different" but in Sweden the Universities are also closed and everybody who can works from home too. The difference in "appearance" exists, but also among how much testing is being done and what is reported.

Also:

Sweden: 550 recovered (?!), Austria: 10,501 recovered

Sweden: 1,511 deaths, Austria: 443 deaths.

>Sweden is surely not "midway in the list" unless in an imaginary world where one thinks that all European countries fit on the accidental 13-item (!) part of the whole list of the Statista site.

Sweden peers are the old EU-15 plus Switzerland and Norway (i.e. Europe excluding the former Eastern Bloc countries). That makes the Sweden's 8th position in the list exactly the middle.

>Try it again but with the list of all European countries

you sure wouldn't think that Poland, Albania, Russia or Ukraine are the peer countries to compare Sweden with?

> makes the Sweden's 8th position in the list exactly the middle

From all countries and entities in the world, around 200 here:

https://www.worldometers.info/coronavirus/#countries

Sweden is currently 11th in deaths per capita, as everybody can see.

The only ones worse in the world currently are: San Marino, Belgium, Andorra, Spain, Italy, France, UK, Netherlands, Sint Maarten and Switzerland. From these, San Marino has 30K citizens, Andorra 70K, Sint Marteen 40K, so these three are practically irrelevant and misleading to compare.

So the only "real" (in size) countries in the whole world currently reported worse than Sweden are:

Belgium, Spain, Italy, France, UK, Netherlands and Switzerland.

How is that some "middle"? Note that specifically UK, Netherlands and Switzerland had longer bet on "herd immunity building, avoid other measures" than the rest of Europe. And I guess Belgium was just as incompetent as years before, having for years no government.

If anything, it is a good evidence of lockdowns working.

Many people are commenting from an American perspective, where "universities closed and everybody who can works from home" is the post-lockdown state we're trying to get to. American schools and universities end their academic year between mid May and early June, so we probably won't be sending people back to school until the fall term starts.

His source data for California is almost certainly wrong. If you go look at it in the csv its got a bunch of repeats and obvious gorp.

The data assumptions in the model itself is questionable; as someone pointed out below, he's basically modeling the generating process involved in giving people tests (put a different way: the spread of test availability). Which is pretty useless. For myself I wouldn't have gone through the trouble he did to make it more sciencey looking; exponential smoothing and sqrt(n) would give you virtually the same picture.

I don't know about California, but around me, not sure the social mixing is primarily outdoors. A decent number of people are still going over to friends' houses, taking kids on play dates, etc. Granted, far fewer than a month ago, but people aren't all staying in their own homes.

Especially in places who have not had a significant impact as of yet, I expect compliance with the lockdown to wane as we go forward. To use Oregon as an example, it is definitely plateaued and looks like it's slipping a bit. Probably because out of 4 million people we're only seeing perhaps 5 deaths a day from COVID19, and it doesn't seem to be increasing.

It doesn't help that we have a weak governor and nothing even resembling a plan. So people are starting to make their own judgement call and going about their business. I see a noticeable increase in the number of people having gatherings of friends & family at their home.

At the same time other countries like Belarus have done nothing and haven’t done worse than CA. Or consider Sweden has ~81% of the rate of COVID per capita as the UK and has ~65% of the death per capita as the UK but no lockdown.

Edit. I don’t mean to go against the prevailing opinions but it seems like something that merits more research. There’s a lot we don’t know about this virus.

Here's an interesting interview with a senior epidemiologist partly responsible for Sweden's unconventional strategy:

https://youtu.be/bfN2JWifLCY

TL;DW: Preventing transmission in the long term is impossible and the actual fatality rate is likely on the order of 0.1%, so general lockdowns are pointless and being driven mostly by political considerations (something must be done and this is something). Just protect the elderly and weak.

> Just protect the elderly and weak.

Given the catastrophic impact on nursing homes in the U.S. so far, that’s clearly something we’ve completely botched.

> actual fatality rate is likely on the order of 0.1%

This is ludicrously optimistic. The virus has already killed over 0.1% of the entire population of New York City, and it's barely starting to wind down (i.e. will probably hit 0.2%). It's also approaching 0.1% of the entire populations of several of the hardest hit areas across the US (Bergen County, NJ; Chelsea, MA; Orleans Parish, NJ), with the most optimistic antibody and random sampling studies suggesting no more than one-third of the total populations infected at this point.

Some of the most reasonably optimistic numbers that are being bandied about nowadays are three to four times higher than that (0.3-0.4%), though I think the majority of the evidence points to a range more like 0.5-1.0%.

Agreed. The Swedish doctor was throwing around speculative numbers (e.g., 50% of UK and Sweden have already been infected) with absolutely no data or references to back it up. I find it hard to believe that he can be taken seriously by anyone.

> The virus has already killed over 0.1% of the entire population of New York City

Wait, what? That would be one person in every 1000? Population of 8,398,748[1] that would be... huh. Confirmed deaths, 8,448[2].

Nice, as they say on another site.

[1] https://en.wikipedia.org/wiki/New_York_City

[2] https://www1.nyc.gov/site/doh/covid/covid-19-data.page

His thinking is that a) fatalities are on the order of 0.1% (might be 2-3x higher) and that b) 50% of the population of the UK, Sweden and (presumably) NYC is already infected.

I agree that this seems overly optimistic, but the high end of that range seems plausible in light of the figures we're seeing, no? That is, if we're at 0.1% fatalities at (say) 25% infected, then that would imply 0.2% at 50%.

The current estimate is that 2/3rds of deaths "due to" C19 are just people who would have died anyway dying a little early.

The relevant number is excess deaths due to C19, which are hard to estimate during a pandemic. That will be much lower than these numbers being reported.

UK was slow to lockdown and currently has the fastest-rising death counts in Europe, so they're probably not the best comparison. Sweden has several multiples of the per capita death rates of its Scandivanian neighbors, and it's growing faster than many other European countries (passed Switzerland's total death count yesterday and will probably pass its per capita count tomorrow).

It's still possible it will look wiser in the long-term, but it certainly doesn't so far.

Don't forget the effect of voluntary lockdown. Looking at the Google mobility data for Belarus and Sweden suggests there is a lot of self-isolation going on.

It’s crazy that things aren’t getting more out of control in Sweden. I guess we’ll see if it’s just early or if they continue to be fine over the next month and beyond.

Sweden has a higher per capita death rate than the US, for a different comparison. And the president of Belarus said on Monday that "No one will die of coronavirus in our country. I publicly declare this". He is the first and only president of Belarus and has been in power for 26 years.

The USSR only died for a moment in Belarus.

https://en.wikipedia.org/wiki/Union_State

What's your metric for "on track to end the lockdown"?

> I can't help but wondering if the unlimited outside time under the guise of "exercise", but which people are clearly taking advantage of by going to hang out with friends in the park, has anything to do with it. Or perhaps it's the lack of masks in indoor areas like grocery stores that has prevented a bigger drop (in SF, only starting today has that finally changed).

Unlikely to both. Outdoor transmission is rare and I'd question if grocery store transmission (with heavy amounts of social distancing) is common given other countries historically being able to trace the majority of covid infection sources.

First off, as you allude to lockdowns are less effective than you think - we managed in norcal to drop from 1.43 to 0.98 (https://www.medrxiv.org/content/10.1101/2020.04.12.20062943v...). Which is really good (uncontained to contained) in one sense, but shows you how slow containment is.

The containment difficulties lie in:

1. Household transmission still takes place after lockdown.

2. Huge variances within population sub-groups. R within folks with little human contact (WFH, etc.) was probably barely above 1 before the lockdown -- essential workers with heavy human contact already had the highest R and because they are still working (i.e. not really locked down), still have an R > 1 for quite some time.

3. Essential workers tend to live with other essential workers -- creating a multiple generation transmission chain (#1 and #2)

4. Group living outbreaks (symptoms of #3). e.g. in Santa Clara county, nursing homes are becoming a larger percentage of new cases over time. Something like 20% of new cases in the last 2 weeks. A single homeless shelter in SF is responsible for 20% of cases in the last 2 weeks.

5. At some point, you just are going to reach some level of herd immunity within the most susceptible groups -- in a sense the lockdown is more of a quarantining of a large percent of the population and letting everyone else get herd immunity.

CA definitely has high urgency in increasing testing - honestly, if you just hired a bunch of people to contact trace, you could probably go back to the March 15 pre-SIP world now.

All good points, I’ve wondered about the differences between different groups as well as far as herd immunity. Haven’t heard it discussed much, maybe because it’s super problematic to explicitly call that a plan - let’s let the working class all get it, and then everyone else can go back to normal.

How was the reproductive rate 1.5 before the lockdown though? The latest estimates I’ve seen put R0 at 5.7.

I'm not the first person to have thought about group variations. :) Here's an old paper (which also highlights how this important fact is omitted in a lot of analysis): http://web.eve.ucdavis.edu/sschreiber/reprints/Moving_beyond...

R0 is before interventions AND is environmentally related. Wuhan & NYC covid has way higher R0 than California Covid because of higher density/public transit usage. Additionally, effective R was already well below R0 by the first week of march (pre-lockdown) because of heavy (mostly voluntary) measures already in place. (WFH, more washing of hands, avoidance of mass gatherings, etc.)

Finally, I'm speaking a bit tongue-in-cheek calling it the "essential worker herd immunity plan". I doubt public health officials, rushing to do something, explicitly realized this would happen -- but with a bit of thought [1], it was a pretty expected outcome. [sort of how nursing home outbreaks occurring when we didn't - and still don't -- quarantine employee or give them heavy PPE was expected)

[1] It's pretty unfortunate there isn't a public or academic comment period to evaluate the SIP orders (after issuance) and almost no useful data being given about cases. There's a lot that doesn't make a lot of sense -- important restrictions missing and unnecessary restrictions present. My biggest pet peeve is allowing babysitters to come to any workers' houses, but it being disallowed to send your kid to a babbysitter's house, when adults are more likely to transmit than children.

1. The symmetrical curve in some of the influential models is so wildly unhelpful from a policy+PR perspective. It's so far from what is likely to happen.

2. There's some thought that one of China's most impactful moves was moving confirmed cases to dedicated quarantine centers, so they don't infect household members. Household transmission is huge - so reducing that number will reduce the scope and length of the epidemic.

The need for social distancing isn't going to diminish until there's herd immunity, whether through exposure or vaccination. That's at least one and possibly several years out. We're going to need to settle on distancing measures we can actually sustain for those years. Masks in grocery stores are part of that. Wiping the live entertainment industry off the face of the earth is part of that (sadly). Cold turkey on every form of IRL socialization isn't.

> Wiping the live entertainment industry off the face of the earth is part of that (sadly)

The industry, yes, which is different from live entertainment in general.

If my local coffee house is allowed to reopen in any capacity, I don't expect the marginal risk of their open mic night happening to be significant.

>> even went up a couple of days ago

I speculate that everywhere in the US will see slight increases currently due to Easter weekend, regardless of shuts downs and mandated distancing. It didn't stop everyone from church going or visiting family.

Anyone else think this was Russia Today from the URL on first glance?

I was pretty worried it was.

These are pretty, but given the inconsistent testing regimes, difficult to interpret.

I believe that inconsistency is factored into, and indicated by the error bars and the width of the range indicators for each graph.

I'm not sure that "we might not be measuring the right thing in the first place" is the kind of error that can be sufficiently accounted for by error bars.

Since it's not from subject matter experts, expect things like sampling bias and administrative reporting issues to be completely neglected. That is to say, the uncertainty estimate is simply based on observed variance.

Great visualizations - I'd love to see this done with hospitalizations though as case count overestimates r given increasing testing. (Granted hospitalizations are also breaking down recently due to recent disproportionate nursing home infections)

E.g. CA is unlikely to have had an r above 2 even at beginning of March, but this is calculating close to 3. (Source: https://www.medrxiv.org/content/10.1101/2020.04.12.20062943v...).

They list the following known issues:

1) Changes in testing will affect numbers.

2) The delay between infection and testing is ignored so actual Rt values are delayed by some amount.

Both are huge actual issues.

The ramp-up in testing has an effect on these observations (thinking specifically about the early spikes in many places):

> Absolute testing levels should not affect this algorithm much, but a fast ramp or decline in testing will affect numbers.

Does anybody know if the testing numbers are per patient or per test? I was trying to run some bayes simulations yesterday and while debating it we couldn't figure out anything useful, because the statistics we were able to find are already rolled up and the methodology is hidden.

Great job! I live in HI (top 4 states according to the website), it looks like they are gonna start loosening up restrictions in a couple of weeks. It will be interesting to see how Rt evolves after that. There's a couple of states where Rt started increasing again after some time (e.g. WA).

Most of those increases are artifacts of testing parameters changing or reporting delays, not actual infection rates. WA data here is quite misleading.

This could also be an interesting data point for anyone who needs to pick a location for establishing a new business.

Lesser Rt number might mean safer operating environment and healthier employees.

This is awesome! Sure the data is not perfect but it's what we've for and the r naught is exactly what we need to be monitoring.

This monitors Rt (r-effective) not R0(r-naught)

These are just toys while testing remains as sparse as it is.

Don't look at these and think it's over.

Beautiful! What gets measured gets managed.

Another COVID tracker that leaves Puerto Rico out of the United States. I don't even know why I'm still surprised.

You should write to them: hello at rt.live

I did, before I commented here.

Looks like you can create a PR about on the notebook and/or run it now if you're curious: https://github.com/k-sys/covid-19/blob/master/Realtime%20R0.... as the data source has territories and the capital but they filtered out anything that's not a state.

The District of Columbia is not a state, and yet is included.

Reply on news.ycombinator.com