Healthcare Listens To Social Media 2

Patients often reveal more details about their medical conditions in discussions on social networks, blogs, and other online forums than they share with their doctors. They talk about their experience with illness and the medications they're taking in more detail. They bring up issues they don't realize are relevant to their treatment or that they're too embarrassed to share face to face.

And now companies are starting to mine online chatter for valuable information that can identify trends in patient symptoms and outcomes, track the effectiveness of treatments, spot complications with drugs and drug interactions, identify patients for clinical trials, and identify market opportunities.

One of the early market leaders is PatientsLikeMe, which offers an online data-sharing platform for patients with "life-changing" illnesses. About 80,000 people participate in 11 disease communities, including ALS, chronic fatigue syndrome, depression and other mood disorders, epilepsy, fibromyalgia, HIV/AIDS, multiple sclerosis, organ transplants, and Parkinson's.

PatientsLikeMe works with pharmaceutical makers UCB and Novartis; various research centers, including Johns Hopkins, University of Wisconsin, Penn State, and the Milton S. Hershey Medical Center; and several non-profit organizations. It provides them with aggregated and anonymized data from its patient profiles and discussion forums that they can analyze themselves. PatientsLikeMe also offers its own analytics services.

Novartis has used information gathered from PatientsLikeMe discussion forums and patient profiles to determine that multiple sclerosis patients have difficulty managing the injectable drugs that are a standard part of their treatment. It has also used PatientsLikeMe services to recruit patients for clinical trials of a pill alternative, called Gelenya. Novartis has since gotten Food and Drug Administration approval for Gelenya.

PatientsLikeMe is among a small number of companies starting to track and analyze patient-generated online data. Another is NielsenHealth, which provides BuzzMetrics services that include data collection and analysis, focusing on what people are saying about healthcare and pharmaceutical companies' products. Some pharmaceutical companies are doing this sort of research and analysis on their own, as well.

Intimate Details Revealed

"It's still the very early phase in how to learn from this unstructured data." --Dr. Roni Zeiger, chief health strategist, Google

Health-related online discussion forums aren't novel either. There are thousands of them. What's new is the ability to track and analyze the information generated from these sites to spot trends that give healthcare providers, researchers, pharmaceutical companies, and public health agencies insight into patient experiences with various diseases and medications.

Patients using Twitter, Facebook, blogs, discussion forums, and other social media often volunteer more intimate details about their illnesses than they do in person, including with their doctors, says Lisa Gualtieri, an adjunct professor at Tufts Medical School who studies the use of health-related social media. They offer up details on their illness's impact on family relationships, sex, work, as well as their symptoms, side effects, depression, and ways of coping, she says.

Discussing their illnesses online provides patients with an emotional outlet, Gualtieri says. Patients often feel that doctors don't have the time or interest in hearing about the details of their medical problems, and that other patients might benefit from their experiences.

"Illness is very disempowering," Gualtieri says. Social networking sites provide connections.

Unstructured patient-generated data on social media sites will become an important source of medical information within the next decade, says Google chief health strategist Dr. Roni Zeiger. "It's still the very early phase in how to learn from this unstructured data," he said in a presentation at Partners Healthcare's Connected Health Symposium last month. And the extraneous "noise" is a big problem, he said--people looking up symptoms of illnesses that they don't have or checking out side effects they think might be caused by a drug.

Google has been using Web search data to spot trends in flu activity in a project it started about two years ago with the Centers for Disease Control and Prevention. When the flu breaks out in a region, Web search soars for topics like "influenza symptoms." By analyzing aggregated search data, Google has been able to estimate in near-real time flu activity around the world.

Flu patterns that Google has identified track closely with more conventional--and typically much slower--outbreak surveillance and reporting methods by the CDC and other public health agencies. Those organizations use paper and Web-based forms to get healthcare providers and individuals to report illness symptoms and suspected medication side effects.

Eventually, Web and EHR data could be combined to find "pointers" to trends. --Jay Pieper, VP of Partners Healthcare

Eventually, data from the Web could be combined with data in electronic health records to find "pointers" to trends worth further investigation, Jay Pieper, VP of Partners Healthcare and president and CEO of Partners' International Medical Services, said at the symposium. By searching patient-generated data on the Web and in electronic medical records, healthcare organizations may be able to identify problems with drugs sooner than they can now, Pieper says.

Eventually, EMRs may have a standardized "slot" where the patient or doctor could enter side effects a patient is experiencing. EMRs across a patient group could then be analyzed for unusual or unexpected side effects. "With that sort of standardization, you'd be able to do rigorous analysis," he said.

Overheard

For its part, PatientsLikeMe has collected profile information from tens of thousands of patient volunteers. The volunteers are told up front that their information will be viewed by other site participants and that anonymized data will be shared with PatientsLikeMe clients. Participants self-report their medical histories, symptoms, test results, treatments, and outcomes in their profiles, in addition to participating in discussion forums and surveys.

Some of the 11 disease communities on the site have enough participants to allow for statistically relevant trending and modeling. For example, about 22,000 multiple sclerosis patients--5% of all diagnosed MS patients in the U.S.--participate in PatientsLikeMe, generating about a half-million postings over the last three years, says company co-founder and president Ben Heywood.

Tools on the site provide participants with charts and graphs that track their conditions, symptoms, and treatments. Using the data collected, PatientsLikeMe compares a patient's history, symptoms, and treatment outcomes with the larger population of patients having the same illness. It then models predicted outcomes based on various treatment options and other factors.

"We're looking at the data to see how the real-life experience of patients is different from clinical trials." --Sue Curro, VP at drugmaker UCB

PatientsLikeMe collects both structured and unstructured data from patients. That combination "is what's so powerful," Heywood says. The company's data set will get even more significant as more patient information is available, including genomic mapping and testing, he says.

Since PatientsLikeMe started its epilepsy community in January, about 3,000 patients have joined, including about 1,000 who have completed as many as three outcome reports for pharmaceutical company UCB. UCB uses data from those reports to track how patients are doing on its and other companies' drugs. It also tracks disease symptoms, adherence to treatments, side effects, and other issues. It's using the data to tweak and develop new clinical trials. "We're looking at the data to get insight into how the real-life experience of patients is different from what you might see in a clinical trial," says Sue Curro, a UCB VP and senior global project leader of patient solutions.

UCB is planning to extend its relationship with the site to study patients with other diseases, including rheumatoid arthritis and Crohn's disease, Curro says.

Among the services PatientsLikeMe offers is "passive listening," where the company's analysts listen in, in order to monitor for keywords, tag discussions, and note the context in which patients are discussing topics. The site is working with natural language processing, text analytics, and other analytics technologies to further automate the monitoring and extend its reach even further into patient-generated data.