If you have an opportunity, watch the Rachel Maddow show from 3/18. They talk about this in depth.
Kinsa has been selling a digital thermometer for a while. As part of the device, it uploads readings to both your phone and then sends the data back to Kinsa.
Based on their reporting (part of what you're seeing visualized on this map), they have been able to see trending signs of flu outbreaks for a while up to 2 weeks before the CDC issues its reports.
The key difference is that the CDC relies on hospitals compiling reports and sending them in, where Kinsa gets the data directly. This causes a delay of up to 2 weeks in the CDC reports.
Where this ties in with covid-19 is that in the last few months, when they layer recent data over historical data from the last few years, they see an atypical increase in fevers being reported. This can be seen in the graph where the expected trend is going down, but they're now seeing abnormal increases since February.
From the news report, Kinsa is selling thousands of thermometers per day right now and is running into manufacturing issues.
Another point that gets mentioned in the report is that Kinsa has tried offering their data to the CDC before but due to CDC policy, they (CDC) require taking ownership of the data in order to accept it. I'm honestly very uninformed about this part so expect some inaccuracy in what I'm saying. If someone who has more details could reply to this comment, I think that would be very helpful to know.
On the positive side, the Bay Area isn't looking bad at the moment. On the negative side, all of my family is in Florida and I wasn't aware until just now that it was getting that bad state wide.
Florida having outliers like that does not look good.
Florida's governor is still refusing to close beaches, which are packed with people, many of them traveling long distances to get there (spring break), with countless people being interviewed on camera saying they don't care..
Some cities are finally taking measures into their own hand and closing beaches, but at the moment it's only a few.
Given the fact that Florida's population tends to skew older, the data mixed with the public response is concerning.
Or there is sampling bias in Florida, maybe insufficient thermometers or some other error? It' seems bizarre to me that one state has way more problems, yet the numbers are not terribly high compared to Washington of CV19.
Our governors incompetence is alarming in any case, though not unexpected.
I believe that Florida is also one of the oldest states which might skew the numbers. Too many thermometers doesn’t seem to me like it would naturally skew the data in the way that it has.
This info, paired with the recent infection of a US congressman from Florida, makes me suspect they may have a lot of unrealized cases.
2. It would skew the numbers (by conjecture) if a higher proportion of people were old because then a higher proportion of people acquiring the disease would have high fever.
I think it would be good for them to indicate the relative penetration of their device into the region. For example, in Ohio the 88 confirmed cases are in counties that don't correlate well with the data presented. Most of the cases are in the northeast of the state.
The child poverty map correlates a bit better with the data from healthweather.us (esp if you factor in the population (density) of each county). The south and east counties in Ohio are pretty sparsely populated.
2) how can you tell credibly from this data it’s just more healthy ppl taking temperature now vs a true decline in fever rate? Feel this is a marketing stunt or a poorly done study.
This shows some of the potential of health-related data-collection everywhere, and cross-referencing it with time and geography (among other dimensions). It is obviously already happening, but these types of overviews are largely kept behind closed doors as of yet. Releasing the data in such an easy digestible manner would democratize the knowledge and enable crowd-sourcing of solutions.
The technical issues are solved, and there are only 4 main hurdles IMO: 1. privacy, 2. data is privately owned (it costs to have it released), 3. professional gate keeping and 4. that governments are not yet tasked with addressing the problem. Surely all of these can be overcome, with various multi-pronged approaches.
I just listened to a recent episode of 99% Invisible: The Weather Machine and I'm struck by the similarities. In the podcast they discuss the evolution of the data sharing agreements to do weather prediction and the birth of the idea that with that data weather was predictable. Your list of 4 could have been lifted right from that episode about weather. So, someday maybe!
As long as I don't have to take part, it could surely help. But I don't want to task my government to collect health data. Personally, I prefer to keep it between my doctor and myself without anyone knowing about it.
Because having the information out there are too many negative external factors. And health data is pretty hard to anonymize.
How are they inferring sickness rates from smart thermometer data? I skimmed the article and their technical paper and got nothing. Presumably sick people stay home and they're correlating stay-home rates with sickness rates?
Or people are scared about illness, so they are using their thermometers more frequently and thus finding small fevers/incidents of increased body temperature that they wouldn't have checked in a normal year.
Amazing level of detail. I wouldn’t want to be in Florida at present - implied infection rates are extremely high. Good independent data sampling whole population (well, biased, but different bias from other sources).
I would love to be able to go back in time for a few weeks (animated), and also to be able to see a comparison against last year.
Edit: also this is a fairly leading indicator of problems (and includes many that are sick but get better), compared with confirmed cases which lags and underestimates, or deaths which severely lags (but fairly reliable indicator there is a serious problem)
- There's a lot more testing in Washington state, so there aren't as many undocumented cases as, say, Florida.
- Sparse Kinsa device coverage in Washington, and since the infection rate is still low vs the population at large, the elevated rate of actual disease is within the wide variation in data due to the sparsity.
I have been taking my temperature regularly with a WiThings 'Thermo' smart thermometer and am pleased with it. I was able to purchase it using my FSA funds:
Kinsa has been selling a digital thermometer for a while. As part of the device, it uploads readings to both your phone and then sends the data back to Kinsa.
Based on their reporting (part of what you're seeing visualized on this map), they have been able to see trending signs of flu outbreaks for a while up to 2 weeks before the CDC issues its reports.
The key difference is that the CDC relies on hospitals compiling reports and sending them in, where Kinsa gets the data directly. This causes a delay of up to 2 weeks in the CDC reports.
Where this ties in with covid-19 is that in the last few months, when they layer recent data over historical data from the last few years, they see an atypical increase in fevers being reported. This can be seen in the graph where the expected trend is going down, but they're now seeing abnormal increases since February.
From the news report, Kinsa is selling thousands of thermometers per day right now and is running into manufacturing issues.
Another point that gets mentioned in the report is that Kinsa has tried offering their data to the CDC before but due to CDC policy, they (CDC) require taking ownership of the data in order to accept it. I'm honestly very uninformed about this part so expect some inaccuracy in what I'm saying. If someone who has more details could reply to this comment, I think that would be very helpful to know.
On the positive side, the Bay Area isn't looking bad at the moment. On the negative side, all of my family is in Florida and I wasn't aware until just now that it was getting that bad state wide.