Have you ever gone on the internet
to seek a diagnosis for your symptoms?
gets 6.5 million visits a month. WebMD’s “symptom tracker” gets 4 million visits a month.
Are they any good at making a diagnosis for you? Do they give good triage advice?
A recent study by Harvard Medical School faculty says “not too bad”, BUT “seeker beware”. (1)
There is significant accuracy variability between the websites. The authors identified 143 websites offering diagnoses in response to symptoms entered by the user. They submitted 45 standard sets of symptoms from “patients looking for a diagnosis” (entered by non-professionals) to each of the 43 websites directed at general medicine/primary care “symptom tracking”. The “patient’s” symptoms entered were in one of three categories (unlabeled and unknown to the website): 1) symptoms require urgent medical evaluation, 2) symptoms should be evaluated by a medical professional by appointment, and 3) self-care without medical evaluation is all that is recommended (the classic medical triage decision tree).
On average, half of the websites listed the correct diagnosis first. 58% listed the correct diagnosis in the top three. However, the range of correct number of #1 diagnoses varied from 5% to 50% between websites. Not unlike a medical professional’s diagnostic decision-making, the software on these websites correctly identified common diseases more frequently than uncommon ones. (The old medical student saw is true: “When you hear hoofbeats think of horses, not zebras.”)
A more comforting result of the study was that 80% of the symptoms needing urgent evaluation and care were identified correctly. The flip side was that two-thirds of the patients that could have been served well by self-care alone were instructed to seek a medical evaluation. As one web-designer physician remarked, “We try to neither unduly alarm nor inappropriately reassurance users.” The software behind the triage advice on these websites (“should I see a doctor or not”) is highly risk-adverse in suggesting so many medical evaluations for minor symptoms. This may not prove so useful to people using the websites to avoid a visit and the payment of a deductible.
The triage accuracy was comparable to that found in studies of nurse practitioner telephone triage which ran about 60-70% compliant with physician in-person recommendations. Of course, the symptom tracking websites do lack perceptions of patient anxiety, tone of voice, urgency of speech, and other subtle, but important clues (especially to pediatricians), of the nature of the situation. Perhaps as a practicing pediatrician I put too much stock in those kind of clues in assessing a patient by telephone, but I suspect that they do improve the person-to-person triage decision-making. At least, the symptom tracking websites were better than just entering symptoms into a Google.
Which symptom tracking website was the best? DocResponse.com got the highest score, but offers no pediatric content. Healthychildren.com by the American Academy of Pediatrics is the site to go to for that. MEDoctor had the worst results. e-Patient Dave, a respected non-medical patient care advocate, recommends trying 2 or 3 websites with the same symptoms.
The traffic to these kind of websites and their popularity with patients may spawn a new word in the lexicon of medical practice; “cyberchondriac”
1. BMJ 2015:351, July 8, 2015