Predicting flu in 140 characters or less

Early on in the 2009 H1N1 epidemic, I posted about how some of our Iowa colleagues were tracking public interest in H1N1 (then still called “swine flu”) by tracking tweets in real time. Well, they’ve now published their findings, which are extremely cool, in PLoS ONE.

Following public opinion about an emerging infection is interesting enough, but I especially like their use of Twitter to predict influenza-like illness (ILI) rates. Check out figures 9 and 10 to see how closely Twitter traffic tracked reported ILI cases (once models were developed to determine the relative contributions of each influenza-related Twitter term to predicting ILI rates). As these authors point out, real-time Twitter data precedes traditional surveillance information by a couple weeks, which could be quite useful in public health planning and preparedness.

Check out other cool stuff from the computational epidemiology group at Iowa