Facebook meets WebMD: How Treato mines social data to uncover health insights

On thousands of message boards, community forums and health sites across the web, patients and caregivers leave detailed descriptions of their battles with disease, their experiences with drugs and their paths to recovery. But those online notes don’t just provide catharsis and comfort. For big data startup Treato, they’re a rich source of data for uncovering health insights that could help patients, health care providers, pharmaceutical companies and other health organizations.

Plenty of other social media monitoring tools “listen” to digital chatter by keyword to help companies track the sentiment of online conversations about their brand.  But Treato CEO Ido Hadari said his company goes further. Over four years and with $7.5 million in venture funding, Treato developed technology that collects and processes billions of natural language patient conversation in real-time, as well as qualifies the sources of those conversations.

What Treato is doing is the next-generation of patient intelligence,” he said. “It’s about an in-depth understanding of patients and caregivers and their real-life personal experiences.”

To start, it analyzes health-oriented web sites to make sure that they include quality health content and personal experiences. To date, Treato has qualified 2,000 sites, including Topix.com, Drugs.com and smaller condition-specific health sites. Interestingly, while it includes public posts from Facebook, Hadari said, Treato does not include comments on Twitter because they found that 90 percent of its data is spam.

Once the company aggregates the commentary, it runs an analysis with its own medical ontology that organizes the unstructured social data into knowledge about patients’ personal experiences. To date, Hadari said, it’s analyzed 1.3 billion conversations from millions of patients about 11,000 drugs and 13,000 conditions.

Since launching in 2011, the company has focused on patient comments related to drugs — on Treato.com, anyone can look up thousands of different medications to learn about side effects, effectiveness, whether patients switched to alternatives and more. And a service launched in October enables pharma companies to monitor patient responses to their products.

Treato declined to share the number of clients that had signed on to the service but said, anecdotally, that its service has helped customers uncover the reasons behind a product’s poor adoption, uncover new uses for a medication and reposition a drug’s marketing plan. Treato could also serve as an early warning system for a drug with harmful side effects to a large population.

With a new $14.5 million round of financing from OrbiMed Israel Partners, New Leaf Ventures and Reed Elsevier Ventures, announced this week  (see disclosure), Hadari said the company plans to build up a presence in the U.S. and expand its “Facebook meets WebMD” concept to unearth patient insights about broader medical conditions or clinical trials.

It also plans to roll out a set of features that would let patients view a more personalized set of data. For example, instead of just seeing how thousands of patients have responded to a drug, a user could create a profile of their medical history and then filter data to see the results most relevant to his condition.

As patients go online to seek out others experiencing similar conditions and symptoms, they’re leaving a valuable trail of data that could lead to advancements in care. Lucine Biotechnology, a Las Vegas-based startup, is attempting to improve women’s healthcare by drawing connections between social comments online and information gathered through salivary hormone monitoring. Alliance Health Networks uses algorithms to analyze conversations in its online communities to better understand how patients influence each other and enable pharma companies and clinicians understand how patients make medical decision.

Disclosure: Reed Elsevier, an investor in Treato, is also an investor in GigaOM through its Reed Ventures arm.