How to figure out what's ailing you without becoming a cyberchondriac.

How to figure out what's ailing you without becoming a cyberchondriac.

How to figure out what's ailing you without becoming a cyberchondriac.

Innovation, the Internet, gadgets, and more.
Dec. 12 2008 5:08 PM

The Web May Be Hazardous to Your Health

How to figure out what's ailing you without becoming a cyberchondriac.

Illustration by Rob Donnelly. Click  image to expand.

Sometimes, muscles twitch. Doctors don't quite know why: The muscles in the human forearms, thighs, calves, feet, eyelids, and even the tongue sometimes begin to spontaneously, involuntarily contract, a condition that can go on intermittently for years. The medical name for this collection of symptoms, benign fasciculation syndrome, suggests that it is nothing to worry about—a cruel irony, as constant muscle twitching tends to instill unshakable dread in those who experience it. The twitching, they fear, must mean something—in particular, as a bit of Web searching will tell you, it could be a sign of ALS, or Lou Gehrig's disease, the fatal neurodegenerative condition whose early symptoms include muscle twitching.

"Please Help! So afraid," writes one of the thousands who've posted about the sudden onset of muscle twitching on, the main online forum for sufferers of benign fasciculation syndrome. "I am PETRIFIED. I am a grown man and find myself crying with fear a lot. Please try and give me some answers." Everyone here is sick with worry; even after undergoing batteries of tests and being assured by doctors that they don't have ALS, many are convinced that they're gravely ill because nearly everything that they read online seems to confirm their worst fears. "The funny thing is we Google symptoms for reassurance and a lot of the time we get that," writes another AboutBFS regular. "But like a bad gambler we're never satisfied. We spin the wheel again, and then again, until we have finally found all the exceptions and what-ifs and worst cases. And there we stand, completely stripped of self confidence and peace of mind."


Some of the people on AboutBFS may be hypochondriacs—that's not a dismissive term but an actual, DSM-recognized disorder that affects about 3 percent of the population. But many of them say they rarely feared terrible diseases in the past; it was only when they searched for "muscle twitching" and found a flood of information about ALS that they became alarmed. They're not hypochondriacs but cyberchondriacs. Like e-tailing or webinar, cyberchondria is an ugly, faddy neologism, but according to Ryen White and Eric Horvitz, scientists at Microsoft's research division, it's also a real phenomenon. Through a survey of employees of Microsoft and an in-depth analysis of anonymous traffic logs of thousands of people who use the Windows Live Toolbar, the researchers determined that many of us exhibit the same tendencies that bring folks to AboutBFS. We search the Web for common symptoms—muscle twitching, headaches, chest pain, cough, fever—and freak out when presented with a few extremely serious conditions. On the Web, White and Horvitz found, we Google ourselves into paroxysms of fear as we often fail to appreciate how rare the most terrible diseases really are.

The Microsoft study grabbed many headlines, and it seems destined to bolster what lots of doctors have long advised about the Web—stay away. But that's an unreasonable prescription. We're used to going online to solve every mystery in our lives; who wouldn't run to the Web if his tongue began to twitch? Horvitz, an artificial intelligence researcher who also has a medical degree, says that despite its flaws, the Web can be more helpful than harmful to your health. It would be even better, he says, if search engines would change the way they deal with medical queries. Horvitz argues that Web companies should build medical search algorithms to reduce the chance that people will leap to grave concerns when they look for diseases online. When you type in any term, today's search engines essentially show you what's the most popular. They might reduce collective anxiety, he says, if they instead tried to show you what you're most likely to have.

In lieu of a new kind of search engine, the most important thing you can do when searching the Web for health information is to pay attention to the "quality" of the stuff you find online. Is that post written by a doctor? Is she an authority in the field she's talking about? Is the information up to date? Checking these things may sound logical; a lack of discernment about health information is a classic sign of hypochondriasis, and you'd expect nonhypochondriacs to be more selective about what they choose to believe on the Web. But 75 percent of people surveyed by the Pew Internet and American Life Project in 2006 said they rarely or never investigate the source or the date of the health-related articles they find online.

Worse, says Horvitz, much of the medical information on the Web isn't written with an eye toward reassuring people that they're not gravely ill—even though must of us are not. Your chance of being diagnosed with a brain tumor is about one in 10,000; if you have a headache, it is extremely unlikely that it's caused by a tumor and much more likely the result of some other less serious condition, like tension or caffeine withdrawal. Similarly, ALS affects just one in 55,000 Americans, most likely not you or anyone you know. But when Horvitz and White analyzed the set of documents that a search engine would return if you typed in "headache," they found a 26 percent probability that a given article pointed to a brain tumor as a possible cause of headaches—far higher than your chance of having a brain tumor.


This makes sense: Even though they're rare, serious conditions like ALS or brain tumors merit a lot of medical attention and, therefore, many articles on the Web. But human beings are easily confused into thinking that quantity represents likelihood. Psychologists call this the availability bias—if we hear about something a lot, we tend to think it's common. For instance, studies show that because newspapers cover deaths caused by accidents and homicide more often than they do deaths caused by disease or suicide, people erroneously believe that accidents and murders are a common way to die. In the same way, when you search for "headaches," you become entranced by the number of articles on brain tumors—and in the avalanche, you quickly forget that your headache is probably caused by something more mundane.

To escape this bias, you've got to become a smarter searcher. Doctors use a process known as differential diagnosis, in which they examine a patient to eliminate unlikely causes of disease. (If you want to know how this works, watch House.) The trouble is that finding the respective probabilities of certain diseases for different patients can be difficult online. If a 25-year-old woman experiences a sudden, sharp chest pain, she is most likely not having a heart attack; a doctor, knowing these facts, will dismiss that idea quickly and would try to find some muscular, gastrointestinal, or psychological condition causing her discomfort. (In one study, 25 percent of emergency room patients who experienced noncardiac chest pain were also diagnosed with a panic disorder.)

But Horvitz says that it took him a while to find the online statistics suggesting that heart attacks are uncommon in young people—and he's a medical doctor who knows where to look. If you experience chest pain and look to the Web for help, you'll see heart attacks displayed prominently. In their survey of Microsoft employees, White and Horvitz found that half of Web searchers took search rankings as a proxy for likelihood; if something was listed up high, they were more likely to assume it was what ailed them.

This suggests a way for search companies like Microsoft and Google to improve their health offerings, the researchers say. Search engines could detect when people are looking for symptoms of diseases and then switch to diagnostic mode. In this mode, the search engine would not only ask you to fill out symptoms but also things like risk factors (your age, sex, weight, and so forth). Then the search engine would display a list of Web results in order of the likelihood that the diseases mentioned in those articles might affect you. By acting in this manner, the search engine would be closer to what doctors call "decision support systems," sophisticated programs that doctors use to help in their diagnoses. Several studies show that these systems can be accurate in spotting even complex diseases like schizophrenia.

Until we get such a search engine, White and Horvitz's research suggests that you might be better off looking for information at dedicated medical sites like WebMD, the Mayo Clinic, or the health sections on MSN or Yahoo. Compared with articles found by searching the Web, pages found on dedicated health sites were far more likely to point toward innocuous—rather than grave—causes for certain symptoms, the researchers found. WebMD even has a "symptom checker" tool similar to the diagnostic search engine I describe above. You tell it your sex, age, and a few symptoms, as well as descriptions of what you're feeling (is your chest pain sharp, dull, stabbing?), and it lists the conditions you might have.

There's only one downside to the WebMD page: Most people don't go to WebMD when they're sick. According to the Pew study, two-thirds of people looking for health information go to a search engine first, not a dedicated medical site. That's why search engines need to build in diagnostics—something that works much like WebMD's symptom checker but that includes articles from all over the Web. True, such a site probably wouldn't cure every cyberchondriac. But it would make the Web a lot less scary.