Wednesday, April 22, 2020

The Facebook Covid-19 Symptoms Map

Facebook and Carnegie Mellon University have released an interactive map of where people in the United States are self-reporting having Covid-19 symptoms. The map is based on a Carnegie Mellon University survey of Facebook users in the United States.

The Facebook Covid-19 Symptom Map shows the percentage of people in each county who report having symptoms of Covid-19. If you hover over a county on the map you can view the actual percentage of people reporting having symptoms. It is also possible to switch the map to view the percentage of people reporting symptoms in each hospital referral region.

One problem with this map could be the accuracy of the data. Facebook argues that it is "uniquely suited" to carry out health surveys because it has billions of users and it can therefore carry out statistically accurate sampling. The survey, carried out by Carnegie Mellon University Delphi Research Center, had over a million responses within the first two weeks. This is a very large sample size. However I do wonder if Facebook users (or Facebook users who trust Facebook enough to give them data about their health) is in itself a non-representative demographic.

The image above shows the Facebook Covid-19 Symptom Map (top) compared to the fatality rate from Covid-19 on the Johns Hopkins Coronavirus Dashboard (bottom). Where people are dying from Covid-19 is probably the most accurate way we have measuring the prevalence of coronavirus. I think there is a weak correlation between the two maps, which makes me doubt the accuracy of Facebook's map. However it is possible that the Facebook map is ahead of the curve and could be highlighting  areas (not yet showing up on the Johns Hopkins map) which may be about to see a sharp rise in fatality rates.

No comments: