Tracking Flu Outbreaks with Wikipedia

additiveanalytics.com

15 points·lauradhamilton··4 comments

I wrote a script to import Wikipedia traffic data for the Influenza page and graph it. (Updates daily)

Graph: http://www.additiveanalytics.com/solutions/flu_tracker

Inspired by this study published last week: http://www.ploscompbiol.org/article/info:doi/10.1371/journal...

The researchers found that Wikipedia page view data provided better real-time reporting on influenza outbreaks when compared to the CDC's data (which has a typical lag of 1-2 weeks) and Google Flu Trends (which had trouble with the 2012-2013 flu season and 2009 H1N1 panic.

Graph was done with d3.js.

I was very recently at a conference where some folks presented their work on using Wikipedia for disease tracking - it's considerably more nuanced than just "better than CDC". Large variation by area, disease, etc.

Disease surveillance hasn't been well served in the past decade by breathless attempts to replace existing systems with a "Big Data" solution - syndromic surveillance, then Google Flu Trends, then Twitter, now this.

Combinations people! Different data streams tell you different things.

Yep, great point. The wikipedia data does seem to be more real-time than the CDC data, which has a typical lag of 7-14 days. Obviously it has its downsides, including lack of localization, basically zero clinical information, no demographic info, etc.

Perhaps one day we will see a real-time API for anonymized influenza reports from electronic medical record data? Then that could be combined with wikipedia data and perhaps other social media data.

Reply on news.ycombinator.com