The following originally appeared on The Upshot (copyright 2016, The New York Times Company). In response, one reader wrote me a poem (captured as a pic here).
Before they see a doctor, most patients turn to websites and smartphone apps.
Caution is advised. Research shows they aren’t very good.
A few years ago, doctors from the Mayo Clinic tested the wisdom of online health advice. Their conclusion: It’s risky. According to their study, going online for health advice is more likely to result in getting no advice or incomplete advice than the right advice.
The doctors assessed the quality of advice on the top sites returned from Google, Yahoo and Bing for searches on common health complaints — like “chest pain” or “headache.”
No site they examined listed all the necessary symptoms so that a user could obtain an accurate triage — whether to rush to the emergency room, call the doctor or treat the condition at home. A third of the sites did not list any of the key symptoms. Among sites that checked any critical symptoms, four in 10 provided no triage advice.
Even when online symptom checkers suggest diagnoses, they can suggest so many of them that patients are unlikely to be able to deduce which diagnosis is most likely. One study found that older adults could find the correct diagnosis of an illness only half the time using Google or WebMD.
Rather than searching the internet, what about using a symptom-checking app you may have loaded on your smartphone or tablet? Such apps — some from trusted institutions like Harvard Medical School or the Mayo Clinic, among many others — are specifically designed to provide fingertip access to diagnostic and triage advice. Yet they are not regulated for safety or accuracy.
Last year, researchers from Harvard Medical School and the RAND Corporation assessed the diagnostic and triage accuracy of 23 of the most popular symptom-checkers, some with tens of millions of users. They found that only one-third listed the correct diagnosis first, half got it among the top three suggestions, and 58 percent included the correct diagnosis among the top 20 suggestions.
Diagnostic abilities were better for those problems reasonably treated at home, for which the top diagnosis was correct 40 percent of the time, or for those that are more common, for which the top diagnosis was correct 38 percent of the time.
Tested apps gave correct triage advice 58 percent of the time, and at a higher rate for more serious problems requiring urgent care. One reason is that symptom checkers tend to be risk-averse. They’re biased toward advising patients to seek professional care even when self-care is appropriate. Some of the apps tested always advise professional care, even for conditions that don’t warrant it.
Though it’s far less convenient, talking to a medical professional may be a more likely route to accurate medical advice than using the internet or apps. But how much more likely depends whether you talk to a nurse by phone or see a doctor in person.
Calling a nurse-staffed, telephone triage line — offered by some health insurance plans — may be comparable to or slightly better than apps in diagnostic accuracy and appropriate triage.
A 2012 systematic review found that most telephone triage services provided accurate advice over two-thirds of the time. A study of pediatric abdominal pain found phone triage accuracy of 61 percent. Another put it at 69 percent. However, the services have also been faulted for increasing visits and driving too many people to emergency departments with overly cautious advice.
Though apps may be in roughly the same range as phone consultations for triage accuracy, they are poor substitutes for consulting a doctor in person. Studies find physician diagnostic error rates to be much lower than that of apps, though still in the 10 to 15 percent range by some estimates (others put it as low as 5 percent).
Seeing a doctor in person takes a lot of time and, for some, costs a lot of money. So it’s understandable that people will turn to more efficient ways, like using websites, apps and the phone. But, as they do so, they should not (yet) presume greater accuracy.
