Could social data be used as a diagnostic tool? Mounting evidence suggests that the prospect of using Tweets and posts as an early warning system has legs, and not just at a population health level.
Earlier this summer, UCLA behavioral psychologist Sean Young wrote about social media as a “new vital sign” to be monitored along with your blood pressure. Patients, he said, tell their social networks things they don’t tell healthcare professionals.
“People don’t just use social media to talk about celebrity sightings or joke about why TSA agents take so long to search them,” wrote Young.” They’re also using it to publicly tell the world about health issues that they often don’t even tell their doctors. They’ll share openly on Facebook and elsewhere about their feelings, their plans to do healthy things like exercise, and their intentions to do unhealthy things like use drugs.”
Indeed, researchers on a Microsoft study aimed at identifying Major Depressive Disorder through Twitter were able to construct a model that identified depression with 70% accuracy. They went on to identify several linguistic ‘tells’ in Tweets that correlated with postpartum depression. And a team of researchers at the University of Utah is working on laying the groundwork for “a natural language processing algorithms capable of automatically identifying depression-related symptoms from Twitter feeds.” In a recently published study, they determined that “Automatically identifying depression-tweets with a moderate degree of accuracy is feasible.”
It’s reminiscent of the recent news that Microsoft researchers (yep, them again!) were able to pinpoint cases of pancreatic cancer based on search terms. Researchers have long salivated over the potential epidemiological uses of such real world data sets, and Google attempted to use search data to track the progress of influenza as early as 2009, with their ambitious Google Flu Trends project (though, alas, that particular effort did not prove the best proof-of-concept case).
What’s striking about these inquiries into the use of search and social media as an early warning system for health is their potential to pinpoint disease in individual patients – and facilitate potentially life-saving interventions. It’s all a bit Big Brother-ish, and the privacy implications will take some working out, but you can see insurers and health systems incenting future patients to opt into a monitoring program – and reaching out to ask if everything’s OK, offering support and broaching therapeutic options.
Social data holds similar promise for brands, lending insights that go far deeper than traditional data sets. Prescription data can tell you how many times prescriptions were switched last month, but social data can tell you why. As my colleague Steve Reeves said in a recent post, social data offers “the largest, most unbiased source of patient opinion that exists,” and “with the right methodology in place, life sciences companies can pivot across use cases and business questions in a repeatable and scalable way.”
We’re just getting started with social data, whether in patient care or business analytics, but a future in which social analytics allows providers and manufacturers alike to respond in real time to patient needs.