Facebook for Disease Tracking?

[Pangloss]

I keep seeing Facebook friends from different parts of the world reporting that they have the flu. Not that I'm suggesting anything apocalyptic -- it just got me wondering if scientists have ever tried using Facebook data to track the spread of a flu outbreak.

 

The appeal here is pretty obvious -- hundreds of millions of users communicating over a vast network in a centralized manner. But the problems seem pretty robust too -- people's public wall posts don't really constitute good medical data (the user could have anything, or nothing at all). You'd have to sit down and figure out what kind of wall post text constituted a "hit", but here I think information science could come in useful, providing some parsing algorithms.

 

I wonder if you could use it for a "rough read". And of course you could track it against the general population to see how accurate it turned out to be. What do you all think?

[Mr Skeptic]

I think it wouldn't work. As you said, they could have anything, and also there would have to be an automatic way to understand which people were saying they have flu. And furthermore, you can't assume that people who don't mention it don't have the flu either.

[Pangloss]

No, I think you're right, there's no way you can come up with anything conclusive.

 

But what about general trends? What if someone just whipped up a little filter that scanned for instance of the word "sick" in Facebook posts, and then sorted them by time and location, and then compared that data with statistics over the time domain from the CEC, etc?

 

If there was a correlation, that might suggest that the tool could be useful for seeing general spread patterns. I mean, we are talking about half a billion people. Seems like that potential data pool ought to be useful for something. (grin)

 

It also occurs to me that one could offer a Facebook application that allowed participants to opt in to an epidemiological study. The app could have them report their health status on a weekly basis, and report immediately if they contract something. Of course that would be a much smaller number of people, and with still no way to confirm the reporting, but that would address the automation problem.

[Dan6541]

No, I think you're right, there's no way you can come up with anything conclusive.

 

But what about general trends? What if someone just whipped up a little filter that scanned for instance of the word "sick" in Facebook posts, and then sorted them by time and location, and then compared that data with statistics over the time domain from the CEC, etc?

 

If there was a correlation, that might suggest that the tool could be useful for seeing general spread patterns. I mean, we are talking about half a billion people. Seems like that potential data pool ought to be useful for something. (grin)

 

It also occurs to me that one could offer a Facebook application that allowed participants to opt in to an epidemiological study. The app could have them report their health status on a weekly basis, and report immediately if they contract something. Of course that would be a much smaller number of people, and with still no way to confirm the reporting, but that would address the automation problem.

 

If you searched for the word sick, you'd get all kinds of posts like: "I'm sick of Chelsea beating Man Utd," or "I'm sick of school."

[Rickdog]

You`d also get lots of participation by small inmature kids who would be constantly playing with the tool and report false data to the system. lots of them from diferent computers and with diferent user names. As an example, my 11 year old child has more than 40 user names in facebook, which he uses randomly and his schoolmates the same, where they usually play among each other trying to fool themselves of who is actually comunicating with them. Imagine all those users reporting a highly contagious fatal disease. It would outbrake total chaos.....

 

Edited September 11, 2010 by Rickdog