Institutions that notice diagnostic errors need to have a better way to notify doctors when their diagnoses are incorrect, researchers say, because internists' overconfidence in their decisions may be hurting patient care.
Even though they were right only 6% of the time and should have had doubts, internists asked to diagnose tough patient cases express nearly the same strong confidence in their diagnoses as they express for much easier cases, when their accuracy rates were much better, at 55%.
That's the finding of researchers at Baylor College of Medicine who conclude that physicians' "overconfidence" in their diagnostic decisions may be hurting patient care, perhaps because seasoned doctors don't think they need to look harder for an explanation behind a patient's symptoms.
"The point of our study was to figure out what happens to their confidence when they go from an easy case to a harder case; we asked do they adjust their confidence or not, and whether there's overconfidence in their decisions? And that is indeed what we found," says Hardeep Singh, MD, the principal author of a paper published Monday in JAMA Internal Medicine.
Additionally, the researchers found, when dealing with these tougher problem cases, physicians were not more likely to seek opinions from their peers, ask for second opinions, or recommend referrals, although they did request more reference materials to help them determine their patient's illness.
The take-home message from the paper, says first author Ashley N.D. Meyer, M.D., is that institutions that notice these diagnostic errors need to have a better way of giving doctors feedback, to notify them their original diagnosis was incorrect.
"I think one of the issues is that doctors don't really get a lot of feedback on how accurate they really are, so they can never align their confidence with that accuracy," Meyer says. "Sometimes patients just go to a different doctor, or somehow, the doctor never hears what eventually happened."
She adds that the doctors who were overly confident were part of the problem, but the issue was compounded when they didn't ask for additional resources "when they should have, because they were wrong in their diagnoses."
Singh, who also is Health Policy, Quality and Informatics Program chief at the Houston Veterans Affairs Medical Center, adds that institutions such as hospitals could provide much of this feedback, especially as bundled or global payment models are increasingly applied, or with accountable care organizations that track patients' cost and quality of care over time.
"If I see a patient and then later, that patient shows up in the ED, the institution could tell the doctor that and provide feedback on a missed opportunity for learning," Singh says. "Maybe last month one doctor had four patients who ended up in the ED. And over time, the doctor's thinking readjusts and would get better."
Singh emphasizes that no one really knows how accurate internal medicine physicians are at diagnosing their patients during the course of their practice, although estimates tend to hover around 85% to 90%. A15% error rate, however, is large enough to be a worrisome concern.
The accuracy rates in this research project were lower because these cases were more unusual, and because for most internists, "their day-to-day practice is colds and coughs, diarrhea, bellyaches, pneumonia, and heart failure," Singh says. These tougher cases were not, however, beyond the scope of the typical internal medicine practice.
He adds that physicians may express higher levels of confidence, whether they truly are that sure of their opinions, in part because "of something that's ingrained—what we call the hidden curriculum—" learned in medical school.
"The whole medical enterprise is based on the fact that one goes to a doctor in the belief that doctors usually know what they're doing, otherwise you won't go. If a doctor said, 'you know, I'm kind of wrong half the time,' no one is going to come to them."
Singh's area of research focuses on diagnostic accuracy, an issue that's increasingly important for measuring quality of care, and which some say should be among the measures used to determine physician performance and payment. In a February report, he quantified common errors in physician practice resulting in diagnostic error, for example, a failure to take an adequate medical history or deficiencies in the physician's performance of the physical exam.
In this study, 118 physicians from around the country agreed to participate, and were asked to correctly diagnose four case "vignettes," based on real cases. They were recruited through QuantiaMD.com, an online community of physicians who review clinical evidence from experts and exchange knowledge.
Two of the cases were easy and the other two more difficult. There were four phases of opportunity to gather information about the case, as close to possible as an actual physician-patient encounter: the chief complaint and medical history, physical exam, laboratory and imaging studies, followed up with definitive or specialized lab and imaging tests.
They were then asked "Given all information provided, what one diagnosis concerning the chief complaint is most likely?"
They were then asked if they would order additional labs and imaging, seek a second opinion from a colleague within internal medicine, seek a "curbside" or more informal consult with another doctor, seek a formal referral from a specialist or use reference materials, such as internet and electronic medical documents or books.
The physicians tested had a mean experience level of 16.5 years. Six in 10 were born in the U.S. and nearly two-thirds received a U.S. medical education and 61% were affiliated with non-academic institutions.
In an invited commentary, Gurpreet Dhaliwal, MD, of the University of California San Francisco Department of Medicine, said Singh's and Meyer's report "suggest that physicians may not request the full complement of point-of-care resources—references and colleagues—to facilitate diagnosis when we need it the most, and that confidence is not a reliable indicator of when we need help."