Signals and Outcomes: What's the Predictive Value of Online Chatter?

Electronic medical records appear ripe for abuse. All of that patient data just waiting to be aggregated and mined for insights on who might need, or better yet, who might be susceptible to the pitch that they might need some sort of drug, treatment or intervention.

But there are rules and everyone has access to the same information, more or less, which can be normalized statistically for regional, cultural and socio-economic factors. Or any other factor of potential interest.

So where is the advantage in that? Hard to come by, though surely some algorithms are better than others. There are, however, as yet untapped resources on the margins and we, as a society, have gotten there more quickly than was, perhaps, expected. We have blown past the easily accessible stuff and are digging deeper into the weak signals to see whether they can provide some sort of edge, whether it be financial, operational, organizational and, well, let's be honest: did we already mention financial?

That enterprises are attempting to make sense of whatever data is available should not be a surprise. Relatively inexpensive computing power and the relatively inexpensive human resources required to interpret it make this opportunity harvesting a given. What is not yet clear is whether the information gleaned is of sufficient moment to be worth the effort. But it almost doesnt matter: because if that knowledge isn't worth much, there's plenty more where it came from. JL

Caroline Chen reports in BusinessWeek:

Data hasn’t proven yet that signals lead to distinctive change in outcomes.
On the Internet, people talk about everything, even their illnesses. Treato, an Israeli data-mining company, monitors conversations on Facebook (FB), Twitter (TWTR), and patient forums for information on drug side effects and prescription patterns. It then sends weekly or monthly analyses of the chatter to hedge funds and money managers that invest in pharma stocks.
Ofir Levi, head of life-science research at Israel’s Adamas Healthcare Fund, turned to Treato with a hunch. The prostate cancer drug Xtandi, co-developed by Astellas Pharma (4503:JP) and Medivation (MDVN), is approved in the U.S. only for patients who’ve already had chemotherapy. Levi suspected physicians might be prescribing it off-label for use before chemo. “By looking at patient discussions, we figured that pre-chemo patients were also getting this drug,” he says. That appeared to confirm Levi’s guess that the market for Xtandi was probably larger than the approved patient population. His hedge fund bet on Medivation and was rewarded when the company reported better-than-expected first-quarter sales of the drug on May 8 and its shares gained as much as 8 percent in trading that day. “For the first time, investors don’t have to listen to the CFO” to find out how a drug is doing, says Ido Hadari, chief executive officer of six-year-old Treato.
So-called expert network companies, such as Gerson Lehrman Group and MedaCorp, have thrived by putting investors in touch with doctors, researchers, and other health professionals. Treato’s selling point is big data. Its software reviews “tens of thousands” of patient forums and informal discussions daily, Hadari says. Individuals aren’t identified; instead, the company seeks to find trends in how drugs are used and what problems consumers are experiencing. Companies such as Saama Technologies and Signals Intelligence Group pitch what they call social listening services to drugmakers and health insurers. But closely held Treato is one of the few also putting its algorithms to work for the investment community, according to Brigham Hyde, vice president for data science at Decision Resources Group, a health-care consulting firm. “If it’s proven you can actually make better investment decisions by using this, investors will flock to it,” says Les Funtledyer, portfolio manager for Esquared Asset Management, which isn’t a Treato client.

The patient conversations monitored by Treato range from broad to granular. “I recently stopped using singular as I had a severe episode of depression,” a patient wrote on Medications.com, in a thread the company’s software dredged up even though the name of Merck (MRK) asthma drug Singulair had been misspelled. Treato’s archive on Pfizer (PFE) cholesterol drug Lipitor includes more than 40,000 discussions, including one on Bodybuilding.com about cholesterol content in protein shakes and whether they’re suitable for someone on Lipitor. Representatives for Pfizer and Merck declined to comment on the conversations Treato had gathered on their drugs.

“If it’s proven you can actually make better investment decisions by using this, investors will flock to it.”—Les Funtledyer, portfolio manager

While hedge funds now subscribe to its reports, Treato’s main customers are pharma companies. Hadari says he counts nine major drugmakers as clients—he won’t name them, saying he’s bound by confidentiality contracts—and a “handful” of finance companies. He also declined to discuss pricing. According to the director for analytics at one pharma company who asked not to be identified for competitive reasons, the appeal of Treato’s service is that it can deliver answers to such questions as why patients switch from one drug to another.
Treato could face greater competition from patient forums. When Ed Sikov, a writer in New York, was diagnosed with Parkinson’s disease in 2008, he joined PatientsLikeMe, a website that hosts forums on 2,000 conditions and has more than 250,000 members. “I’m not comfortable with investors making bets on what people are experiencing without the people themselves knowing they’re being monitored,” Sikov says. But he’s all right with PatientsLikeMe, which clearly states that it sells the data it gathers to drugmakers listed on its site, including Abbott Laboratories (ABT) and Bristol-Myers Squibb (BMY). Co-founder Ben Heywood hasn’t ruled out selling information to health investors. The site’s privacy policy says names, e-mails, and dates of birth aren’t distributed.

Hyde, of Decision Resources, who has no firsthand experience with Treato, has yet to be convinced about the value of using social listening tools for investing. Patients may discuss a drug’s side effects online, he says, but the issues they raise would affect a drugmaker’s stock only if they attract the attention of the U.S. Food and Drug Administration and lead to restrictions on a drug’s use. Says Hyde: “Data hasn’t proven yet that signals lead to distinctive change in outcomes.”