Human Doctors Are Still Better Than Online Symptom Checkers

New research has looked into how the performance of fleshy human physicians compared to digital diagnosticians. It might not come as a surprise, but trained human doctors left online diagnosis checkers in the dust.

The study, published in JAMA Internal Medicine by Harvard Medical School (HMS), tested 234 medicine physicians against 23 digital symptom checker apps, including Web MD, the Mayo Clinic, and the Isabel Symptom Checker. They were asked to evaluate 45 clinical cases, all of varying rarity and different degrees of severity. Along with establishing one diagnosis, they also had to provide two additional possible diagnoses.

Overall, doctors were able to reach an accurate diagnosis first 72 percent of the time, while the digital platforms were correct just 34 percent of the time. Nearly 85 percent of doctors were accurate within their top three possibilities, while 51 percent of digitals platforms were.

The level of human doctor accuracy was even greater, in comparison to the digital platform accuracy, when it came to rare or more severe conditions.

It’s fairly worrying stuff, considering HMS say that “hundreds of millions of people” use online programs or smartphone apps to self-diagnose each year. But that’s not to say that "robot doctors" should be banished to the garage.

The UK National Health Service (NHS) is collaborating with Google’s Deep Mind in the hopes of applying machine learning to diagnose different forms of cancers. IBM's Watson is also working within the health and medicine sector to assist doctors.

“Clinical diagnosis is currently as much art as it is science, but there is great promise for technology to help augment clinical diagnoses. That is the true value proposition of these tools,” Ateev Mehrotra, study author and associate professor of health care policy at HMS, said in a statement.