Accuracy of Health Information Online

Your physician is trained to deal with patients in ways that won't provoke unnecessary anxiety. For example, if you go to the doctor's office complaining of constant headaches, the chances are little that your doctor would suggest, "Oh, you may have a brain tumor!" Rather, the doctor would check for other symptoms and investigate more likely routes. After all, the chance per year of having this serious condition is VERY low - 0.000116 (around 1:10,000). Thus, the physician avoids mentioning this condition if clear signs are not present, as any anxiety provoked otherwise would more than likely be unnecessary.

Dr. Google doesn't operate in the same manner. In a comprehensive study, Ryen White and Eric Horvitz of Microsoft found that using co-occurence statistics, the internet's "probability estimates" of serious conditions to be much higher than these conditions' actual incidence rates. For example, using the top-10 Web search results, they found the probability estimate of a brain tumor based of the search term "headache" to be 0.26; as mentioned above, the actual annual chance is just 0.000116!

What does this mean? The presentation of online health information is based off of algorithms, and does NOT consider the potential anxiety of a searcher in the same way that a physician might with a patient. Therefore, the internet will often provoke the searcher with suggestions of serious illnesses right away! As unlikely as some of these serious conditions may be, internet searches will more than likely bring them up since they are matched with the symptoms searched.

Sources used for the page:

1. White, R. W.; Horvitz, E. Studies of the Escalations of Medical Concerns in Web Search. ACM Transactions on Information Systems 2009, 27 (4), 1–37.