SoylentNews is people
SoylentNews
from the and-your-dr.-is-always-right-first-time dept.
For your typical hypochondriac, online symptom checkers are a rabbit hole of medical information and the anxiety that comes with it. But according to a new study led [PDF] by researchers at the Harvard Medical School, most of these sites are so inconsistent and inaccurate that patients shouldn't rely upon them for correct diagnoses.
In the study, which was published in the British Medical Journal, the researchers looked at 23 web sites from around the world that claim to offer information for diagnosis and triage (assessing how urgently a condition needs to be treated). They used 45 patient vignettes, about half of which were common conditions, to assess the sites' accuracy.
They found that the correct diagnosis came up first only 34 percent of the time. Half the sites had the right answer in their top three results, and almost 60 percent had it in the top 20. Triage advice fared a bit better, with accurate suggestions coming up first 57 percent of the time. The sites in which the right answer came up as the first result most often were: DocResponse (50 percent), Family Doctor (47 percent), and Isabel (44 percent).
What has been the experience of SN Members? Do they rely on online medical diagnosis?
[Also Covered By]: http://commonhealth.wbur.org/2015/07/self-diagnosing-health-websites-study
Original Submission
(Score: 2) by MrGuy on Tuesday July 14 2015, @11:26AM
There's an unstated assumption here when you throw around the term "accuracy." Specifically, that when a certain symptom is described, a physician will look primarily at the most common causes of that condition, and determine what's most likely.
I'm not certain (IANAD) that that's necessarily the way things work, especially early in the diagnostic process (where we have little information other than a few symptoms that COULD mean a lot of things). Rather than start with the most likely, we might want to ask questions to eliminate the most problematic or dangerous possible cause, rather than to try to confirm a somewhat benign or more common one. For example, if the three most likely causes of a given set of symptoms are (in order) irritation from airborne pollutants, a common cold, and lung cancer, I'd be more worried about making sure it's not lung cancer.
Given that online Q&A can only ask so many pre-determined questions, can't follow-up off-script, can't order tests, and can't consider nuances of the patient history, maybe the goal isn't strict likelihood.
Because the second question is what the patient will do with the follow-up information. If I present three possible diagnoses, with accompanying detailed description, do I expect the patient to read all of them and see if they apply? Or do I only expect they'll read the first one? Maybe the one I want them most to read is the lung cancer one, because if they feel like that actually fits, they'll seek a follow-up from an actual doctor who can run some tests. That might be a better result for the patient.