Non-scientific Internet trend analysis can mislead


There is a difference between scientific surveys or analysis and the assumptions made about contextual trends found on social media, and elsewhere, on the Internet. Sometimes, it's a big difference.

The New York Times highlighted such a case in which Google (NASDAQ: GOOG) took a stab at disease control and missed the mark significantly. Tracking an influenza outbreak using algorithms that tracked, among other things, mentions of sneezing, coughing and other maladies online, Google also tracked people's location, social media and flu-related search queries on Google.

Google Flu Trends showed that nearly 11 percent of the population of the United States had influenza. The Centers for Disease Control and Prevention says the actual numbers were about half that at around six percent. And an article in Nature said traditional flu monitoring relies on national networks of physicians who report cases of patients with influenza-like illness and refines that estimate by testing a subset of people with these symptoms to determine how many have the flu and not some other infection.

Some countries, including France, have automated the tracking and surveillance process for influenza, as has the CDC in Atlanta. Those hoping that the data analytics of social media and the Internet could supplement data will have to wait for more refinement of the algorithms.

For more:
- see the Nature article

Related Articles:
Pathologists: Big data's next 'rock stars'?
Helping newborns, infants fight their fight
Big data helps find possible Alzheimer's-immune system connection

Filed Under