We all know that Google collects a phenomenal amount of personal data – a perennial question is whether it’s healthy for one company to know so much about us all.
But what if all that collective data could be used to spot disease epidemics before they take hold? Good thing or bad thing? I’d say good.
If Google’s Flu Trends is anything to go by, the team at Google.org have pretty much cracked it for influenza, and can predict a flu outbreak with a delay of one day, as opposed to the two weeks taken by the official body CDC. It’s a big claim, but so far, the evidence presented is impressive.
Google has been very open with their research, publishing their calculations, references, the historical data they have used and providing a draft manuscript that has been accepted by the journal Nature.
The most persuasive picture is this one, which shows how Google’s search data (in blue) tracks the incidence of influenza as measured by the CDC over the last five years.
This is serious stuff, and the implications are pretty big. If Google can predict flu outbreaks with what would be considered scientific certainty, then you could argue that we are entering a new era of epidemiology. Taking two weeks off recognising disease outbreaks could potentially save thousands of lives, giving the health authorities a far higher chance of inoculations or other preventative measures.
Google do strike a suitable note of caution, saying that “past performance is no guarantee of future results. Our system is still very experimental, so anything is possible, but we’re hoping to see similar correlations in the coming year.”
I would add to that: just because we search for something doesn’t mean it will happen. It just means we are interested. Google searches for Palin outnumbered Obama for most of September and some of October, but we all know what happened come election day.
Equally, the Google trend on, for example, HIV doesn’t reflect infection rates, but rather the news agenda. Which is fair enough, but we need to be careful. There will be more people searching for flu just as a result of this research, and capturing that isn’t helpful to predicting disease outbreaks. Google would also seem unlikely to be able to predict rarer diseases with non-standard symptoms, as the search query range would be too large and random.
But for now, this looks like a good start. If flu, which kills up to 500,000 people per year, could be spotted earlier, it would be nothing short of a major breakthrough.
