Patient social networks have evolved from simple message boards, where users share moral support and practical advice on their diseases, into sophisticated data-entry points producing new findings that are reshaping biomedical research.
On sites like MedHelp, Inspire and PatientsLikeMe, users still message each other about their diseases, but they also enter structured data on their health conditions, with the aim of advancing medical knowledge about their diseases.
“It seems like a win-win situation,” said Dr. William Hersh, professor and chair of Medical Informatics & Clinical Epidemiology at Oregon Health & Science University in Portland. Analyzing this data “can help patients understand their disease better and help scientists carry out better research,” he said.
Hundreds of Thousands Offer Data
This is a form of crowdsourcing like Wikipedia. Hundreds of thousands of patients are offering their experiences with a variety diseases and conditions, including cancer, diabetes, multiple sclerosis and depression. The sites combine social networking tools like on Facebook with the health-management tools that are used for personal health record systems. Divided into specific disease groups, users communicate with each other, compile their own personal health data and have access to reports about their conditions — which are based on a rich stream of personal health data from everyone on the site.
According to site operators, privacy rules under the Health Insurance Portability and Accountability Act, or HIPAA, do not apply to them because they are not providing medical care. Nonetheless, they say they de-identify data that is shared with industry and researchers and allow users to add some privacy controls. The sites also monitor activity on message boards to keep out web-crawlers looking for patient information.
These sites are free to users, so operators cover their costs with payments from the pharmaceutical industry and researchers, in return for allowing them to connect with patients and use their data.
Operators say the information is a boon for biomedical researchers who usually have to spend a great deal of time and effort tracking down research subjects. Even after going through the normal selection process for research, the core sample of patients the sites can offer is usually much larger than what researchers normally have.
The sites say their patient-based data provide many insights. For example, researchers gathering information on lens implant failures from MedHelp’s eye-care forums found that multi-focal implants had a much higher failure rate than other types, giving manufacturers tips to improve the product.
Patient social networks are particularly useful in research on rare diseases, where finding enough subjects is very difficult. Inspire patients with a rare disease called spontaneous coronary artery dissection (SCAD) worked with a researcher at Mayo Clinic to produce original findings on the largest SCAD population ever been studied.
Adding Up to Big Data
In the past few years, patient social networks have become expert at slicing and dicing users’ data into particular diseases, disease subgroups, combinations of diseases, particular drugs used and specific symptoms. These can immediately be used to test hypotheses and come to some conclusions, like a formal research study.
“Our platform itself is a formal research study,” Ben Heywood, co-founder and president of PatientsLikeMe, told HealthBiz Decoded. The group is now applying for a research grant as a Patient-Powered Research Network from the Patient Centered Outcomes Research Institute, an independent group authorized by Congress to fund biomedical research.
While PatientsLikeMe primarily uses structured reports from patients, MedHelp has also started mining unstructured data, according to President and Chief Executive John De Souza.
Renting time on powerful computers owned by Amazon, MedHelp crawls through large amounts of unstructured data on the site, applying search algorithms such as latent semantic analysis, says De Souza. For example, when MedHelp users complained about skin allergies, an analysis of unstructured data traced the cause back to a specific type of bra made in a Chinese factory for Victoria’s Secret.
Brian Loew, founder chief executive of Inspire, says his site also looks at unstructured data, but follows up by querying specific users. Using this approach, the site tracked the experiences of patients with bladder cancer and found they were having pains caused by the administration of a particular drug, which was valuable information for the drug’s manufacturer.
While software to analyze unstructured data is getting more sophisticated, “I’m skeptical about just using computers to analyze masses of data,” Loew said.
De Souza says data from patient social networks is a much more promising source of information than electronic medical records (EMRs). “It’s hard to pull large amounts of data out of EMRs, and EMRs are disconnected from each other,” he said.
Heywood agrees, adding: “Other than lab values, there isn’t that much meaningful data in EMRs.”
He believes that gathering healthcare data through patient social networks will revolutionize biomedical research. More attention will be given to subtypes or subclasses of diseases, leading to a better understanding of what drugs work best for a small number of patients.
“Our current understanding of diseases is going to break down,” said Heywood.