There could be some selection bias here. It's based on thermometer readings from Kinsa. And the articles states:
Kinsa is receiving two to three times more thermometer readings per day than in previous flu seasons
Many more people may be taking thermometer reading that don't really need to, but are being cautious. That would cause the appearance of a proportional drop in high temp readings compared to prior years when really it's just that lots of not-so-sick people are taking their temp lots more.
> Due to the emergence of the coronavirus, Kinsa is receiving two to three times more thermometer readings per day than in previous flu seasons. Historically, Kinsa’s methods have been able to account for a sharp rise in the number of people taking tests, said Benjamin Dalziel, a professor at Oregon State University who studies infectious diseases. For his research, Dalziel has ensured that a spike in testing didn’t lead to inaccurate predictions, and so he believes the current decline in illness is a result of social distancing, and not a statistical anomaly. (Dalziel has had research funded by Kinsa in the past).
That'll teach me to rush to comment when I see something "off" instead of reading the whole thing. I'm tempted to delete a post that isn't useful now, but I should leave it as a minor embarrassing lesson to myself. I wish people would stop upvoting me though... Each one makes me cringe: another person who didn't read very carefully.
> I wish people would stop upvoting me though... Each one makes me cringe: another person who didn't read very carefully.
I don't think upvotes mean something's correct, I think they mean that something (and the ensuing discussion) is worth reading.
Even though you were a little off base, it was a useful chain of discussion because presumably a large nonzero number of people would think the same thing. :)
I agree with your original point, even if you no longer stand by it.
>Historically, Kinsa’s methods have been able to account for a sharp rise in the number of people taking tests
What historical precedent is Kinsa referring to? There has never before been a pandemic of this magnitude, in a time and place where many people owned internet-connected thermometers.
Perhaps I am being uncharitable, but it seems that they are leaping to conclusions here. I'm sure that they have done their best to account for this, but the claim that "Dalziel has ensured that a spike in testing didn’t lead to inaccurate predictions" is one I find difficult to take at face value. Ensured?
You said that "there could be some selection bias here." Could. Sorry to cause you to cringe again, but +1.
I don't think this is uncharitable at all. I'm sure Kinsa has made a good effort at controlling for testing frequency, and I'm sure it's helped. But there's no reason to think the dynamics of COVID-motivated testing are the same as for flu-season, or new-buyer novelty, or anything else.
And more importantly, how could we know if it is? That's not just a Kinsa problem; we see this over and over again with peer-reviewed studies that "control for" certain factors like socioeconimics or health history. They're inherently limited to controlling for what they know about, and it's never perfect. Often, the entire effect is from an undiscovered variable. Take, say, the widely-promoted study finding that visiting a museum, opera, or concert just once a year is tied to a 14% decline in early death risk. The researchers tried to control for health and economic status, then concluded "over half the association is independent of all the factors we identified that could explain the link." [1]
Now, what seems more likely: that the unexplained half is from the profound, persistent social impact of dropping by a museum or concert once a year? Or that some of the explained factors like "civic engagement" can't be defined clearly, others are undercounted (e.g. mental health issues), and some were missed entirely?
I suspect Kinsa did much better than that, because they're not trying to control for such vague terms. But I think "even after controlling for" should basically never rule out asking "what if it's a confounder"?
Yeah, I wish that science reporting would either a) mention specifics on methodology or b) link directly to papers.
Without this plus a stronger push for open-access publishing, readers often have absolutely no way to verify claims like this - and, moreover, have no way to learn how to apply similar methodologies to their own work. There's a lot of people in data science / analytics positions right now who could benefit from sharing knowledge around statistical tools to correct for highly unusual situations.
It’s a reasonable question, and the reply to your OP suggests it was accounted for but I don’t know enough about stats to say that it fully accounts for it. I would say it’s still very reasonable to be skeptical and ask if it is still biased, or still partially biased.
Eh, the response isn't exactly a slam dunk. There's a difference between saying "I thought of that, and it's not a problem" and whatever it is actually not being a problem.
The reply doesn't necessarily have to be a slam dunk or complete rebuke. Quoting the single sentence, without the next 5, impartially paints the story. It's imbalanced. It denies the reader the level of nuance necessary to pass fair judgment.
Even if you jumped to an incorrect conclusion in this case, it's still a useful one in other cases. And anyone who maybe would not have thought of that might learn something from your comment, even if it wasn't applicable in this specific instance.
I think your parent, parent post should stay up. This discussion is valuable. So many show up in the comments when the first thing seems amiss before having a full understanding of what is linked. It's a cycle that repeatedly plays out. It's something I've done before, too. We could all collectively do better.
It is just a linear interpolation and backward looking. I’m definitely not convinced how this would control for how more healthy people take temperatures more frequent after govt announced social distancing to stop COVID-19.
The folks at TWIV (This Week In Virology) noted that regular influenza numbers were sharply down in Japan as a result of people's changing hygiene behaviors.
Exactly what we'd expect. It's a silver lining on this whole mess.
>Recent data clearly show the spread of Covid-19. On March 19, the share of Americans with temperatures indicating they had flu-like symptoms was about 4.9% when it typically would be expected to be about 4.0%. This was likely a result of the spread of Covid-19, according to Kinsa’s researchers.
... means that 1 week ago 2,964,898 Americans (0.9%) had Covid-19. Today there are 65,950 recorded cases in the United States: https://www.worldometers.info/coronavirus/
It's roughly consistent with the data from South Korea which was quite liberal with testing potential cases based on contact tracing (so they have records of also the infected but asymptomatic or very mild symptomatic carriers) and they got 0.6%-0.7% case fatality rate, and the median time from infection to death is something like 3 weeks.
You get a much higher fatality rate if you're testing only people with severe symptoms or those who get hospitalized, which is what many countries are doing. For example, if you have mild symptoms, UK health service will tell you to isolate at home and call them for hospitalizatin if you get worse, and you'll get tested only if that happens. You also get a much higher fatality rate if hospitals are overloaded and people can't get proper lifesaving treatment, which wasn't the case yet in USA, so an estimate based on these numbers seems reasonable now.
Pick whatever numbers you think are reasonable based on the data you have read - I think you will end up with a worrying result.
I use this analysis because the number of deaths is hard to fudge, and the fatality rate after x days, and the infection rate are somewhat measurable (I believe there are statistically valid measurements of those numbers).
Any analysis based upon detected cases is deeply flawed IMHO, because detected cases can be anything (look at different countries with wildly different values with no basis in reality).
You can assume that as long as you're far below saturation. Yes, the logistic curve comes to play at some point, slowing down the spread, stopping at some large proportion of the population - it's plausible that two thirds of Americans will be infected, but that will take longer.
However, until the total number of infected is less than, say, 10% of total population, it certainly makes sense to just use the average doubling rate.
It is not an assumption just arithmetic. Median time to first symptom is 6 days. Median time from first symptom to hospitalization is 5-7 days. Median time to death is about 7 days. Add them up and the median time from contact to death is about 3 weeks.
I dont understand what millions of millions of test kits will do. I’m not going to get tested because I have no symptoms. If I had symptoms I would stay in lockdown until it got worse. I don’t see rationale for anyone going to get tested just because (and based on some of the photos from drive through testing, a very easy way to contract COVID yourself)
It's not just about people with symptoms to get tested. It's about people who are known to have had contact with infected to get tested. Also importantly you want to regularly (at least once a week) test health workers and others who are in contact with vulnerable people. And finally, by doing lots of so called sentinel tests (random testing around the country) you can get a much better picture of how the virus spreads.
In short it is not about your interest if you get tested it is about societies interest.
Hey I agree with you, but I am not a health worker. I think we need to remember what spawned this thread and not move the goalposts for talking points.
Even if you are not showing symptoms stay inside so you are not spreading the disease. Getting tested won’t prevent the spread, staying isolated will.
Why do you keep saying China instead of countries that have contained the virus like South Korea and Taiwan? Aggressive testing are key aspects of their strategy for containing the virus.
Because South Korea govt was allowed to look at citizens credit card transaction data and cell phone location data to enforce quarantine, something US cannot do without creating very unpopular laws.
US can do what China did, heavy handed lock people inside with supplies and test anyone who moves outside of the house. You test positive you either go into mandatory quarantine (hospital) or you get an ankle monitor.
China method works better than SK method when you think about how we got the TSA and how it helps our daily lives
By the time you feeling symptomatic, our current understanding[0] is that you've been shedding the virus all over the damn place for the past 2-14 days[1], spreading germs everywhere you've been. By the time you feel sick, it's too late. It's spring, are you really going to self-isolate for 14 days just because you sneezed once? What sounds better, going to get tested, so you can know if it's a sneeze and you can go about your life, or self-isolate because of you sneezed once. During spring.
We failed at preparing for this. Let's not double down on failure.
Even if they go and get tested, and it comes back negative, that doesn't prevent them from being infected (possibly even at the testing site, as they mentioned) and then being an asymptomatic spreader.
Where am I spreading germs all over the place if I’m isolating myself in my home? Please read my text.. I agree Let’s not double down on failure. Look at what China did, it worked let’s follow them. Practice social distancing until July. There is no other option until our hospitals are under control...
I have been locked in for two weeks, if I am spreading it everywhere then my house is a biological war zone. But that’s the point of self isolation, it’s to make sure you are not spreading it to others. Why would I go to get a test if I’m isolating? If I’m negative that doesn’t mean I will never get it. So why not just stay inside until hospitals are less overwhelmed (and THEN get tested on a regular basis to avoid contamination — you know what we should have been doing since day 1)
Everyone should be doing this so you wouldn’t have to assume I am spreading it everywhere.
The tests can tell you if you have covid-19. What they can't do, because they have such a high false negative rate, is tell you that you don't have it.
But if you catch co19 t-0 after leaving the testing site then you can spread it, right?
So what is testing the world going to solve if no one is preventing the spread of contamination? Because the answer is that you will only overwhelm hospital resources
I don’t think anyone is proposing that testing is a magical cure for the overall problem; obviously you’d have to use testing in conjunction with treatment and isolation and presumably other tactics.
However, this idea that testing is worthless is frankly bizarre. You can’t control what you can’t measure, and “number of people occupying hospital beds” is a very poor measure for a disease that spreads silently for weeks before symptoms emerge.
the human element will always be a failure mode. Every photo I have seen in my area has shown the testers either reaching into the car or having contact with the patient. So you may be asymptomatic or you may not, but then you possibly end up catching it at the test site but your test was negative so you continue about your routine and start spreading it for real.
Not saying the above scenario is a reason not to get tested, I’m saying it’s an argument against getting tested for the sake of getting tested. I am safer and I am making the world safer by locking myself down until this passes. See China as example. I urge you to do the same..
I don’t believe that when we have a shortage. The guidelines even state don’t come to a testing site if you are not showing symptoms as you may catch the covid. I mean why are you not even considering my viewpoint?
Look at China, is it testing that’s working there or the lockdown?
Another example is to look at South Korea, which got a handle on things with lots of testing (and tracing) . I think this is actually a better example than china, considering that it is a democracy. Lots of testing is btw also the WHO recommendation.
The error bars on that 4% number are huge, I imagine. Like 4% - 2% or +15%. It's entirely possible you could get 4.9% with zero Covid-19 infections, just from normal variation. Not only possible but I'm sure there's a month in the past decade which had more.
Why would the error range be that high? We'd know about huge spikes like that through hospital data, insurance claims, and the sales of drugs, and there aren't spikes like that. The number of people who are ill is relatively stable year-on-year.
I'm not sure what you're basing your assertion on, but the estimated number of symptomatic flu illnesses ranges from 9.3 million to 45 million annually over the past decade, which as I said is roughly from between 3 and 15% of the US population.
There is probably some selection bias - it’s 4.9% of people using the thermometers, which would skew towards people that are thinking they have a fever.
Kinsa is receiving two to three times more thermometer readings per day than in previous flu seasons
Many more people may be taking thermometer reading that don't really need to, but are being cautious. That would cause the appearance of a proportional drop in high temp readings compared to prior years when really it's just that lots of not-so-sick people are taking their temp lots more.