Research: Symptom checkers’ diagnostic, triage accuracy is low


Share post:

The accuracy of digital symptom checkers for each triage and prognosis is mostly low, which may pose a danger to affected person security, in keeping with a assessment of research printed in npj Digital Medication. 

Researchers compiled 10 research that evaluated a complete of 48 on-line symptom checkers. Half of the research used actual sufferers, whereas the opposite half studied simulated instances. The assessment discovered primacy diagnostic accuracy, or itemizing the proper prognosis first, ranged from 19% to 38%. Prime three diagnostic accuracy, which was evaluated in seven research, ranged from 33% to 58%.

Triage accuracy, or sending customers to the proper stage of care, was increased, starting from 49% to 90%. Three research examined instances by emergency standing as nicely and two discovered symptom checkers had been extra correct at triaging emergent instances. Nevertheless, one research discovered triage accuracy for ophthalmic emergencies was decrease than non-urgent instances, at 39% in contrast with 88%. 

“Our assessment highlighted that each diagnostic and triage accuracies had been usually low. Furthermore, there’s appreciable variation throughout symptom checkers regardless of being introduced with uniform symptom parameters. We additionally word that the diagnostic and triage accuracies of symptom checkers, in addition to the variation in efficiency, had been significantly depending on the acuity of the situation assessed,” the assessment’s authors wrote. 

“As an entire, these points elevate a number of issues about using symptom checkers as patient-facing instruments, particularly given their more and more endorsed function inside well being techniques as triage companies that direct sufferers in the direction of acceptable remedy pathways.”


Researchers famous some limitations to the assessment, together with using simulated affected person instances and the research’ concentrate on high-income nations just like the U.S. and the UK. In addition they mentioned bias might be launched in retrospective research the place data is collected after the occasion, variation in efficiency might be attributed to the distinction between research, and AI-enabled symptom checkers could enhance over time. 

Nevertheless, the outcomes elevate issues about utilizing symptom checkers, significantly once they’re promoted by well being techniques to ship sufferers to the proper care website. The research’s authors mentioned there are additionally points with transparency surrounding how symptom checkers are constructed and what context, affected person demographics and medical data are used. 

“Given the rising push in the direction of adopting this class of applied sciences throughout quite a few well being techniques, this research demonstrates that reliance upon symptom checkers may pose important affected person security hazards,” they wrote. “Giant-scale major research, based mostly upon real-world information, are warranted to reveal the sufficient efficiency of those applied sciences in a fashion that’s non-inferior to present greatest practices. Furthermore, an pressing evaluation of how these techniques are regulated and carried out is required.”


There are a variety of corporations that supply digital symptom checkers. Digital well being agency Babylon, which till not too long ago was embedded with a number of NHS Trusts, provides an AI-backed triage chatbot. The corporate has been dialing again these NHS contracts because it grows its value-based care phase within the U.S. 

Japanese startup Ubie launched its AI-backed symptom checker within the U.S. earlier this yr. It not too long ago raised $26.2 million in Sequence C funding. Different corporations within the area embrace Ada Well being, Ok Well being and Infermedica. 

Supply hyperlink


Please enter your comment!
Please enter your name here


Related articles