Sickweather, a disease surveillance company based in Baltimore, has made its illness data available to developers so they can create apps that present disease forecasts and outbreak maps. The company's Sickweather mobile app is already available for iTunes and Android users, touting itself as a Doppler radar for sickness.
In a phone interview, CEO Graham Dodge suggested that disease forecasts could become common conversational fodder alongside weather forecasts, thanks to social media, the source of the company's illness data. Already, AccuWeather has incorporated disease forecasts into its StoryTeller content platform. Meanwhile, Johnson & Johnson and thermometer-maker Swaive are using the company's data in their respective mobile apps.
Through Sickweather's API, developers can fetch JSON-formatted data about illness reports at specific map coordinates, disease forecasts for a given area, and contagion threat level scores for leading sources of illness. The API can also receive illness reports from developers' apps.
[Never underestimate the human capacity for foolishness. Read 10 Stupid Moves That Threaten Your Company's Security.]
Sickweather's primary source of disease information, however, comes from machine analysis of social media posts to detect personal declarations of illness.
Crowdsourced weather reports work well in conjunction with meteorological science. But crowdsourced epidemiology has proven to be problematic for medical science.
In 2008, Google launched Flu Tracker based on the notion that search queries about the flu could be correlated with actual flu infections. There's some truth to that, but not enough to reflect outbreak data with the accuracy demanded by medical professionals.
During the 2012-2013 flu season, Google's models overestimated the number of flu cases in the US by more than six percentage points. Google subsequently retired Flu Tracker, though researchers from Harvard and Boston Children's Hospital have since improved on Google's model with a system called ARGO (AutoRegression with GOogle search data).
Dodge said Sickweather's approach differs from Google Flu Tracker. "The main difference between what we're doing and what Google was doing is the difference between implicit data and explicit data."
In other words, where Google's model inferred flu infection from search activity, Sickweather's model focuses on detecting personal declarations of illness from social media posts, and from illness self-reported through apps.
Asked about the merits of crowdsourced disease data, the Centers for Diseases Control did not immediately respond to a request for comment.
Dodge, however, insisted that Sickweather's data has real epidemiological value. "In our own flu forecasting testing, we were able to achieve 91% accuracy in forecasting flu activity," he said. He said that Sickweather has been invited to participate in the CDC's Epidemic Prediction Initiative, and plans to do so.
Marketers appear to be another target for Sickweather's disease forecasts. As Dodge explained in a blog post, being able to forecast illness at least eight weeks in advance allows advertisers to "know when and where to advertise products and services related to the illnesses that we are tracking (e.g. a cough product in a city where people are coughing)."
Sickweather is also offering its data to HR groups in enterprises. The company's Sickweather Pro analytics dashboard provides HR directors with data about the risk of absenteeism and "presenteeism." The latter describes a phenomenon in which "people show up to work sick but aren't productive," Dodge explained.
"Illness forecasts will be a lot more ubiquitous in coming years," said Dodge.
Rising stars wanted. Are you an IT professional under age 30 who's making a major contribution to the field? Do you know someone who fits that description? Submit your entry now for InformationWeek's Pearl Award. Full details and a submission form can be found here.