Can a 'Health Weather Map' detect COVID-19 trends?

A company that maps digital smart-thermometers may be detecting predictive connections between fevers and COVID-19 trends.

Recent data collected and analyzed by smart digital thermometer-maker Kinsa Inc. suggest the overall level of fevers (NOT COVID-19 cases) in the U.S. may be falling rapidly.

As many as 150,000 of those individual fever readings are crowdsourced through the internet and that data produces a “health weather map” tracking illness trends across the United States.

The company is clear to say that their map and data are not direct evidence of COVID-19 cases. But they do say unusual regional fever levels, and what they detect as “atypical illness” may be an early indicator of the spread of COVID-19.

Here’s is how Kinsa describes their data in relation to COVID-19.

Please note: We are not stating that this data represents COVID-19 activity. However, we would expect to pick up higher-than-anticipated levels of flu-like symptoms in our data in areas where the pandemic is affecting large numbers of people. Taken together with other data points, we believe this data may be a helpful early indicator of where and how quickly the virus is spreading.

With that in mind, I decided to take a look at what Kinsa Smart Thermometer data is showing in various cities. Some interesting trends may be emerging. Maybe.

New York

New York City is clearly a COVID-19 hot spot. But the Kinsa data shows an interesting recent trend in overall fever reductions.

Fevers in New York County (Manhattan) were near expected seasonal levels on March 3. Then fever levels surged in the following two weeks. The level of what Kinsa calls “atypical illness” peaked 6.53 percent on March 17. That’s nearly 3 percent above expected levels.

Kinsa data also shows atypical fevers in NYC have fallen since March 17. They now show a 4.2 percent rate, just 0.7 percent above the expected range. The number of fevers in NYC is dropping and nearly back into the upper part of the “expected” range. The data shows a decrease in “illness trend” of 5.93 percent in New York County.

Seattle

Seattle was likely the first COVID-19 cluster in the U.S. Kinsa data shows fevers were near expected levels on March 1. Then the number of fevers surged well above expected seasonal levels and peaked March 9. They peaked at 5.48 percent, nearly 2 percent above expected seasonal levels.

Since March 9, the level of fever has fallen into the expected range on March 22. The “illness trend” has recently decreased by 5.62 percent according to Kinsa data.

Twin Cities

Minnesota has been fortunate to see far lower levels of COVID-19 so far. We also had more time to adjust and increase social distancing. The Kinsa data shows Hennepin County has seen below-average seasonal illness levels overall.

Overall cases in Minnesota continue to increase. Thursday’s update from the Minnesota Department of Health shows 346 confirmed cases in Minnesota. While Minnesota’s numbers are lower than many states, the curve is still rising.

Data analysis

Here are my observations of what the Kinsa data may be telling us — again if accurate. Keep in mind this is only one data set. I am skeptical if the sample size is large enough to capture trends accurately.

• The best information suggests that fever trends may be predictive of COVID-19 trends. If Kinsa fever data trends in places like New York and Seattle are accurate, that could suggest a reduction in the rate of new COVID 19 cases in the coming weeks.

• Keep in mind it appears that seriously ill hospital and ICU admissions lag fever data by several days to a week or more. So you would expect ICU peaks to be reached as much as 1 to 2 weeks after fever numbers peak in the hardest-hit areas.

• The recent trends in most areas show (rapidly) falling levels of fever. If this is accurate, that may be showing how well effective social distancing and stay at home policies are working to reduce disease transmission.

Again, this data is relatively new and great caution needs to be exercised before reading too much into the trends. There’s still so much we don’t know about how COVID-19 works. And temperature trends could also be reflecting a reduction in seasonal flu. New regional hot zones may emerge, especially in areas that are not practicing social distancing.

But if this data accurately reflects disease spread trends with COVID-19, it suggests many places may be on the way to reduced new COVID-19 cases in the next few weeks. Given the lag time between fever and confirmed tests, we may not know for a couple of weeks if the trends manifest.

Crowdsourced Kinsa smart thermometer data seems to have accurately predicted the rise in COVID-19 cases. Could current trends accurately predict a slowdown in news cases in the weeks ahead?

Time will tell.