I was just looking at the Google Flu Trends map. Google Insight could be used to gain similar information for other trends in disease keyword searches.

I started wondering how valid are keyword searches to represent the spread of diseases. In some ways I think it could be a good representative, but in others I think it would be susceptible to influence by mass media.

Do maps of clinical data exist that could be compared to this Google data? If so, where could I find some. =)


2 Answers 2


You can use official sources such as hospital admissions, prescriptions for drugs fighting the disease you are tracking, sales of over-the-counter medicines. CDC (cdc.gov/flu/weekly/, cdc.gov/outbreaknet/outbreaks.html, etc), WHO, EuroFlu Weekly Electronic Bulletin map official clinical data, Aurametrix uses these sources.

Several scientific studies have compared GFT, twitter, even facebook with official sources. Johns Hopkins study (pubmed/22230244) used data from local hospitals in addition to CDC. My sampling studies too showed that social trends are surprisingly good, despite all the noise, but provide good estimates only for highly populated areas.

You might also want to check Sickweather's algorithm, healthmap, usgs disease maps and industry databases providing coverage of pharmaceutical companies and product sales.


For myself, I've not been terribly impressed with the use of keyword searches and the like as methods of predicting influenza or other diseases. They are, when it comes down to it, somewhat glorified extensions of a syndromic surveillance system, just for "I feel sick" instead of "Honey, pop down to CVS and pick up some NyQuil?"

It's also apt to get you some odd results. I tried to use a "Google-style" analysis to match some rotavirus data to Google searches - I believe I was 7 entries down before I ever got to anything rotavirus related, and even that was "rotavirus", which doesn't suggest it would be useful as a symptom detector, but rather people Googling a disease their doctor said they had. The top results, some of which fit really well? A series of prom dress-related searches. In order for it to be useful, you'd have to extend Google's system.

But, as a thought experiment, if you'd like to match it to data, there's a number of potential data sources publicly available from the CDC. Morbidity and Mortality Weekly Report for example publishes weekly counts for a number of notifiable diseases.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .