If 'clean,' big data can improve U.S. health care

(Correction: This story was updated to correct the name of the Reach for Peeks program in Colorado.)

SAN FRANCISCO — Less medical privacy may be good for your health.

A growing body of research has found that information Americans share on social media websites about their health and lifestyle is more up to date and accurate than what they share with doctors, employers, insurance companies and government agencies.

In other words, we're more honest with our friends than we are with those who control our access to medical care.

While that may simply reflect human nature, it has huge implications for health care as patients and providers look to the analysis of so-called big data to improve diagnosis and treatment.

The findings suggest that improvement in medical services may depend as much on widespread availability of accurate patient data as it does on advances in technologies and procedures.

"The little secret of big data is that a lot of it isn't clean," says Eva Ho, a partner with the early-stage venture capital firm Susa Ventures and a former executive at both Google and Factual, an upstart Internet-search company.

In health care, that means a patient's medical records can be filled with outdated or conflicting information that makes an accurate diagnosis more difficult.

With the federal government now requiring all patient data to be digital, there's a big opportunity for companies that can integrate health data from a variety of sources and ensure its accuracy, says Ho, a co-founder of Applied Semantics, which Google acquired in 2003 for its Web-analytics technology.

The most accurate source may be what patients themselves share on their social media accounts, research shows.

One recent study analyzed Facebook ad campaigns alongside public health records from the Census Bureau, National Vital Statistics System, Centers for Disease Control and Prevention and birth and death records from hundreds of U.S. counties from 2010 to 2013. It found something startling about the predictive ability of health-related social media data:

Knowing what Facebook users "Like" led to more accurate predictions about how long people will live, how often they exercise or smoke, and what their chances are of getting a serious illness or disease, such as diabetes, obesity or heart attacks.

By combining the Facebook data with medical-record analysis, predictions for some health outcomes — such as whether an infant would have low birth weight or whether an adult would be in general poor health — were two to four times more accurate than those based on medical or socio-economic data alone, according to the study done by New York-based MKTG.

"The power of 'Likes' is that they represent behavior," wrote Steve Gittelman, a veteran of online market research who co-authored the report "Facebook Likes: A New Source of Data for Public Health Surveillance," along with Elaine Trimarchi and Victor Lange.

The findings echo those of other studies that have hinted at the power of social media to improve medical diagnosis and treatment.

One showed that two-thirds of regular smokers keep their habit a secret from doctors and health insurers — which is why smoking-cessation plans that are shared with Facebook "friends" have proved more effective than traditional strategies for quitting smoking.

Yet another study found that tracking the Twitter accounts of new mothers could help determine how likely they were to develop postpartum depression.

While the analysis of such personal data may make privacy advocates cringe, some in the medical community predict that making it publicly available in electronic form can revolutionize health care.

Dr. Leslie Saxon, chief of cardiology at the University of Southern California's Keck School of Medicine, is among them.

One of her lectures on the subject carried the provocative title "Privacy is Bad for Your Health."

"We need indiscriminate, continuous, multisourced data streams to realize the great potential of digital health," Saxon says.

More American consumers seem to agree. Millions of them are tracking health and lifestyle activity using smartphones or wearable wireless devices — then sharing that data on Facebook and Twitter.

Employers and health insurers, meanwhile, are already using social media tools to boost worker participation in programs that promote exercise and healthier diets.

The State of Colorado, working with Kaiser Permanente and UnitedHealth Plans, used social gaming to boost such participation rates by 650% last year.

"Health managers can use incentives and rewards to encourage consumer behavior," says Jeff Margolis, CEO of Welltok, a Denver-based start-up that designed the plan, called Reach for the Peaks, used by Colorado state workers.

Welltok is using IBM's Watson supercomputer to improve the recommendations made by its social-based health platform, called CafeWell.

Analyzing health and lifestyle data from many sources — including social media accounts — can help create highly customized personal health management services for consumers, says Claudia Fan Munce, managing director of IBM's venture capital group, which invested in Welltok.

But only if the information is clean — and shared.

John Shinal has covered tech and financial markets for more than 15 years at Bloomberg, BusinessWeek,The San Francisco Chronicle, Dow Jones MarketWatch, Wall Street Journal Digital Network and others. Follow him on Twitter: @johnshinal.