by
Lauren Dubinsky, Senior Reporter | October 12, 2016
Disparity increases with rarity of diagnoses
Can computers do a better job than physicians at diagnosing patients? The answer is no, according to a new Harvard Medical School study published in
JAMA Internal Medicine.
Computer-based checklists and other digital apps have become popular over the last two decades to help reduce medication errors and streamline infection-prevention protocols. Millions of people use Internet programs and apps every year to diagnose themselves.
Ateev Mehrotra, an associate professor of health care policy at Harvard Medical School,
told CBS News that he and his colleagues decided to conduct this study because some recent findings suggested that physicians make a lot of mistakes. They wanted to explore whether computers could make fewer mistakes.
“About 10 to 15 percent of the time a patient has a misdiagnosis. Doctors make diagnostic errors, and there’s been this idea in the computer science world, in Silicon Valley and such, that 'can’t computers do it better?'” Mehrotra said.
The researchers had 234 internal medicine physicians evaluate 45 clinical cases that involved both common and uncommon conditions with different degrees of severity. For each case, they had to choose the most likely diagnosis as well as two additional potential diagnoses.
The physicians made correct diagnoses more than twice as often as 23 popular symptom-checker apps. They listed the correct diagnosis first 72 percent of the time, compared with 34 percent of the time for the apps.
In addition, 84 percent of the physicians listed the correct diagnosis in the top three possibilities, compared with 51 percent for the apps.
The difference was the most significant in the more severe and less common conditions. Conversely, it was smaller for less acute and more prevalent illnesses.
Even though they outperformed, the physicians still made errors in about 15 percent of the cases. The researchers think that computer-based algorithms that can be used along with human decision-making will help to reduce those errors.
They believe that it's important to study the future generations of computer programs because they may be more accurate.
"Clinical diagnosis is currently as much art as it is science, but there is great promise for technology to help augment clinical diagnoses," Mehrotra said in a statement. "That is the true value proposition of these tools."
