Is a symptom checker app ready to replace a real-life doctor? (Image via iStock)
A recent audit by Harvard Medical School researchers tested 23 online symptom checkers and found them to be badly inaccurate.
The study, which included industry leaders such as Isabel, Mayo Clinic and WebMD, concluded that “although the programs varied widely in accuracy of diagnoses and triage advice, as a whole they were astonishingly inaccurate”.
The online systems ranged in accuracy from 71 percent to 29 percent. Similar systems exist in MENA (check TebChecker and AlTibbi), but no similar study has been conducted to measure their diagnostic accuracy.
While such results sound alarming and might pose a risk to the industry when digital health is considered to be in a golden age it’s worth taking a step back and looking at the findings from various angles.
Misdiagnosis is not a pitfall of online systems only, even the best doctors make mistakes. Would symptom checkers be as useful to doctors as they’re supposed to be for patients? If so, how will the business model be impacted?
Real doctors make wrong diagnoses too: in the medical world, misdiagnoses are believed to affect 10 to 20 percent of all cases. According to this article, 12 million Americans are misdiagnosed every year and in about half of these cases, the misdiagnosis has the potential to result in severe harm.
This doesn’t mean that because real doctors don’t make error-free diagnoses, we can accept faulty online diagnosis. Rather, it suggests that absolute knowledge is hard to attain by one system alone.
Online symptom checkers are not there to replace doctors: instead of looking at those systems as potential replacements for doctors and cry foul at their low performance, why not consider them as a bridge between (otherwise totally) uninformed patients and doctors. A patient with an online diagnosis might be better informed and in a position to have a more constructive discussion with his doctor.
Taking this one notch further, what if doctors use the symptom checker as a diagnostic aid? This could then be complemented by the human elements that systems are yet to be able to replicate such as gut feeling - the instinct that something is right or wrong.
New business model, new design: the first objection to the previous suggestion would be that doctors are already overwhelmed by systems. With less time to talk to patients, they don’t need an additional distraction, which is why systems should undergo a major design overhaul.
Such systems are now designed with only the patient in mind. Maybe the next generation of online symptom checkers should be designed for providers and patients alike.
Technology will evolve: as artificial intelligence and machine learning advance, chances are that symptom checkers will get smarter and more accurate.
Furthermore, convergence of multiple technologies that might feed into a system checker and provide it with more insights is an interesting area to watch.
Economics: don’t forget the billion dollar question in healthcare: who will pay?
Designing a great system that reduces misdiagnosis rates and is accepted by doctors is not enough. Someone has to carry the cost of such system unless you’re giving it away for free. The current payment system doesn’t allow hospitals to recover expenditures related to symptom checkers. Unless this changes, adoption of such systems at scale can’t be expected anytime soon.
There are definitely different ways to look at this. What are your thoughts?