Googling can mislead people seeking health information


(Juliette Borda)

Network News

X Profile
View More Activity
Carolyn Butler
Washington Post Staff Writer
Tuesday, November 10, 2009

It always starts out innocently enough -- for example, with an eye twitch. It's just a little tic, but it keeps coming and going over the course of a few weeks, and so I decide to do a little medical investigation online. I plug "recurrent eye twitch" into my friendly search engine and, after several hours poring over a range of health-related Web sites -- skimming over likely explanations such as fatigue, stress and too much caffeine in favor of dozens of worst-case scenarios, and growing increasingly panicky all the while -- I am utterly convinced that I have multiple sclerosis, at the very least, and quite possibly Lou Gehrig's disease.

But what really ails me? Cyberchondria, loosely defined as the baseless fueling of fears and anxiety about common health symptoms due to Internet research, or, as I like to think of it, Googling oneself into a state of absolute, clinical hysteria over every last pain, itch and strange freckle on your body.

Apparently, I'm not alone. Last year, Microsoft researchers Eric Horvitz and Ryen White documented the growing trend in "Cyberchondria: Studies of the Escalation of Medical Concerns in Web Search," which included a survey of 515 Microsoft employees and Web-search tracking of hundreds of thousands of consenting Windows Live toolbar users.

The report showed that about 2 percent of all the Windows Live searches were health-related. Of the 250,000 or so users who engaged in at least one such query during the study, roughly one-third "escalated" their subsequent Web surfing to focus on far more serious -- and much less common -- conditions. In addition, the employee survey showed that this type of escalation interrupted the everyday life of more than half the respondents at least once.

Of course, it's important to acknowledge that there is a lot of high-quality health content on the Internet that has helped a lot of people, both on respected, vetted Web sites such as WebMD and Medstory, and also within the myriad online support groups for particular illnesses, where people can seek information, encouragement or a shoulder to cry on. In addition, Horvitz and White's follow-up study found that while two in five people report that surfing the Web for health-related information has made them feel more nervous about a perceived medical condition, just over half of folks say that it reduces anxiety.

The problems arise when people turn to a broad Web search to diagnose their ills, says Horvitz, whose professional credentials include an MD degree.

"People have come to look at search engines as question-answering systems," he explains. "We now see [the Internet] as a general oracle, in our pockets and desktops, that we can just ask questions to, and people think it's going to answer all questions in a quality manner; therefore, people turn to the system and say, 'Diagnose me; here are the symptoms.' "

Horvitz notes that medical diagnostics requires taking in sets of symptoms, reflecting, having an interactive dialogue with a patient and then converging on a list of likely conditions. "It's a relatively sophisticated task that's quite different than information retrieval, which is what search engines are good at. They do not have a good sense for how to reason under uncertainty, or for probabilities. . . . The Web is really great at finding out who played the role of Gilligan on 'Gilligan's Island,' but not so good at weighing the evidence to give you good information about concerning and unconcerning health situations."

Instead, Web search rankings are often based on such things as relevance and click-through rates, which skew the results you see. For example, Horvitz and White use the example of headaches, which are just as likely to be associated with "brain tumor" as "caffeine withdrawal" in search, although the annual U.S. incidence rate of brain tumors is about 1 in 10,000, and missing your daily cup of java is one of the most likely explanations for a common headache. Yet the research also shows that the vast majority of people have interpreted the ranking of search results as a list of likely ailments, in order of probable diagnoses.

Clearly, psychology is just as much at play as technology. Stephen Josephson, a clinical associate professor of psychology at Weill Cornell Medical College in New York, says that a lot of the health concerns people have fall broadly under the category of anxiety, which can prompt compulsive behaviors such as constantly cruising the Web for information. He explains that it's well-proven that people are prone to selectively attend to negative information -- like the fact that a mole might be melanoma -- and to ignore the actual low prevalence of dire diseases,

"It's a paradox: The more you read in an attempt to reduce your fear, the more you try to figure things out, the more anxiety peaks. Very few people know how to navigate the Internet and evaluate information when they're anxious, and yet that's when they tend to go online."

Microsoft's Horvitz says the challenge for the Web is to improve health content so it more accurately reflects probability and likelihoods, and to adapt search engines to factor in many more complexities, such as family history, to be able to properly diagnose an ailment and then intervene when people are escalating inappropriately. Take, for example, someone younger than 35 with no family history of cardiac trouble who plugs "chest pain" into a search. Right now "heart attack" will pop up the most frequently, about 37 percent of the time, says Horvitz. "But if the system just knew their age and family history, it would say, 'Just take it easy; don't run to the hospital,' " because it would be extremely unusual for a young person with no history to be having a heart attack.

In the meantime, fellow cyberchondriacs, try to keep everything in perspective and seek out credible information. The Medical Library Association has some great tips for evaluating health research online.

Oh, and about that twitch? Eventually I asked my doctor about it, and he helped me figure out that I'm actually allergic to a new eye cream I bought to stave off the effects of aging -- not in need of pricey and invasive tests for a rare tropical disease.


© 2009 The Washington Post Company

Network News

X My Profile
View More Activity