If you live in cooler climes (or in my case, near the sunny beaches of Canada), this time of year marks some fun rituals. Flu shots, chicken soup, and Buckley’s Cough Syrup are just as festive as the snow that blankets the city. While not fun for the flu sufferers, city public health officials see this time of year in a different light. Spikes in disease activity often come as a surprise, inundating emergency rooms and walk-in clinics with the relief-seeking masses. Being able to predict when those masses will turn up at your emergency ward has some distinct benefits. You can schedule the appropriate staff to ensure that people get the right attention promptly, and you can get a jump on pinpointing the causes of these outbreaks (food, water, the person beside you while you were trapped on the tarmac). So how do these medical oracles get their information? And what does Google have to do with it?

The Centers for Disease Control and Prevention call this type of disease monitoring Syndromic Surveillance, where they use information about things that often happen before a diagnosis to gauge the likelihood of a public health concern. I had done some work at school in 2005 looking at a project from the University of Pittsburgh called Real-time Outbreak and Disease Surveillance (RODS), which used over-the-counter medication sales as a tool for predicting community health issues. The sales data is collected through the National Retail Data Monitor Program, which uses the UPC data from each sale, and aggregates them. Any spikes determined statistically significant seem to correlate well with emergency room visits, and appear to do it faster. A study of the system’s effectiveness suggested that electrolyte sales and medications like Gravol/Dramamine spiked before children started presenting in emergency rooms for care.

Google has now weighed in with their Google Flu Trends which, instead of using OTC sales data, uses search keywords. The idea is that if you’re searching for terms like “flu” and “cough medicine” there’s a good chance you need them. Pretty neat.

I think this has the potential to be even more sensitive than the RODS program, as people only buy medicines when they’re sick enough to need them. They might Google about their symptoms before symptoms necessitate the meds, adding to the early-warning lead time.

To think even bigger, could this type of predictive use of Google search words bear fruit in other disciplines? Like, has anyone taken a look at a stock price’s volatility as a function of its search traffic (or keywords paired with a company name)? Perhaps some negative web chatter could presage rough times for the mentioned company, ideally beyond the typical rumour-mongering.

I know. The stock idea seems like a stretch. It’s just that the markets have me feeling unwell. Or is it the flu? Maybe I’ll ask Google.