(Reuters Health) – When multiple doctors are on the job, there’s a higher likelihood that a correct diagnosis will be made, a new study suggests.
FILE PHOTO: A photo illustration shows a stethoscope and blood-pressure machine of a French general practitioner displayed in a doctor’s office in Bordeaux January 7, 2015. REUTERS/Regis Duvignau
Researchers found that diagnoses were far more accurate if they were the result of online physician crowdsourcing, according to a report in JAMA Network Open.
“We are saying, what if you crowdsourced a case?” said lead author Dr. Michael Barnett, an assistant professor of health policy and management at the Harvard T.H. Chan School of Public Health. “You get a bunch of doctors together solving the case and put the answers together as if they were in the same room and come up with a ranked list. If you pick the diagnosis at the top of the list, you do way better than individual doctors do.”
For the new study, Barnett and his colleagues turned to The Human Diagnosis Project (Human Dx), an international online project in which physicians and fellows, residents and medical students practice authoring and diagnosing teaching cases.
Users create teaching cases from their own clinical practice with key elements of the history, physical and diagnostic tests (for example, laboratory and imaging studies). Once other doctors put their answers in – they are allowed five guesses – they are told what the real diagnosis is.
Ultimately the researchers concentrated on 1,572 cases from the dataset that 2,069 users had offered diagnoses for. Among the users were 1,228 residents or fellows, 431 attending physicians and 410 medical students.
On their own, doctors and doctors-in-training didn’t do so well, arriving at the correct diagnosis just 62.5 percent of the time. That would be a “D” grade, Barnett said. But when the researchers put groups of nine physicians per case together and looked at the diagnosis chosen most often by the nine, they found the odds of getting the correct answer were vastly improved at 85.6 percent.
“When you add people in groups you get up to a C and a B+,” Barnett said. “And you can get all the way up to an A.”
Barnett isn’t sure how a similar process could be implemented in real world situations. That’s a subject for further research, he said.
The new paper is looking at ways to improve diagnosis, said Dr. Allen Kachalia, senior vice president for patient safety and quality and director of the Armstrong Institute for Patient Safety and Quality at Johns Hopkins Medicine.
“There’s no question that diagnostic error is a big problem in medicine,” Kachalia said. “It’s something we’re still trying to address. In the past, that’s been done by giving clinicians guidelines and pathways to follow to help them in decision making. This study helps us think about different solutions to that problem. It puts in our minds the question: how do we get two or three minds together to help solve diagnostic issues.”
There are two messages you can take from this study, said Dr. Ben Miller, clinical services director at the Paul C. Gaffney Division of Pediatric Hospital Medicine at the Children’s Hospital of Pittsburgh.
First, “even though this is simulated, it shows that the collaborative process of multiple physicians thinking about a patient probably leads to better diagnostic accuracy,” said Miller, who was not involved in the new research. “The other takeaway is the inherent uncertainty of medicine.”
Many patients assume that if they are seen by a doctor, diagnosis is simply a black or white issue, and “if you’re a doctor you should know the right answer, “Miller said. “But even with the collective wisdom of multiple doctors, one in 10 times they will miss the mark.”
SOURCE: bit.ly/2Uby11m JAMA Network Open, online March 1, 2019.