Checking the symptom checkers

I have a small problem
I have a small problem

Online “symptom checkers have deficits in both diagnosis and triage, and their triage advice is generally risk averse,” according to a new study in BMJ by Harvard researchers. Some of the press coverage of the study has been pretty critical of the symptom checkers, but the study itself is quite balanced.

Symptom checkers are a lot better than “Google Diagnosing” (typing a list of symptoms into Google and seeing what comes up). They’re very similar to nurse triage lines used by health plans, which is no great surprise considering that many of the triage lines use the same logic that drives the symptom checkers. They’re inferior to a primary care physician, but of course physicians aren’t infallible either.

People use symptom checkers to self-diagnose and to figure out if they need to go the emergency department, doctor’s office or can treat at home. Not surprisingly, the symptom checkers err on the side of suggesting patients seek care. No producer of these tools wants to get sued for recommending self-care to someone who should have called an ambulance. But in this regard symptom checkers are similar to nurse triage lines and also –at least in my experience– to physicians who are covering call. Too often they suggest a trip to the ED.

There’s an opportunity to reframe the next generation of symptom checkers as tools for navigating the medical system. That means not just suggesting a diagnosis and level of care, but pointing to an appropriate specialty, facility, or individual clinician to follow up with. This could be especially useful for patients in areas with fewer specialists and sub-specialists, and those with rare or hard to diagnose conditions. Primary care physicians could use such a tool as well to help direct referrals.

Image courtesy of David Castillo Dominici at

By healthcare business consultant David E. Williams, president of Health Business Group.

One thought on “Checking the symptom checkers”

  1. The issue with using such symptom checkers for diagnosing rare diseases is that many of the findings needed to diagnose rare diseases are not symptoms, but rather signs observed on examination, or testing results. These can be difficult for members of the general public to get right.

    As I indicated in my “rapid response” included with the BMJ article, I’d prefer that people deal with difficult to diagnose diseases by getting input from a medical professional, which includes many people with advanced training other than physicians. Genetics has a good model for such interactions, involving genetic counselors. This model is worth expanding to other areas of medicine, as is already done using the triage systems from which many of these symptom checkers evolved.

    The idea of patients using such symptom checkers for choosing specialty of such counselors makes a lot of sense.

Leave a Reply

Your email address will not be published. Required fields are marked *