Discover and defeat your diagnostic errors

Experts explain the causes and solutions for common mistakes in diagnosis.


While statistics show diagnostic error to be common in medicine—responsible for an estimated 40,000 to 80,000 deaths a year—most physicians don't think they make many such mistakes, Mark L. Graber, MD, FACP, told attendees at Hospital Medicine 2017.

But a simple shift in perspective can clarify the problem. “Take off your doctor hat. Can you recall a diagnosis you were given that was wrong or a diagnosis that should have been made much earlier? Or is there somebody in your family who has a medical condition that still causes symptoms but hasn't been diagnosed? That's what we're dealing with,” he said.

Mark L Graber MD FACPslashPhoto courtesy of Dr Graber
Mark L. Graber, MD, FACP/Photo courtesy of Dr. Graber.

The causes of diagnostic error are both individual and systemic, according to Dr. Graber, president of the Society to Improve Diagnosis in Medicine, senior fellow at RTI International, and a leader in the effort to reduce diagnostic error.

“Diagnosis is too important a process to rely solely on our intuition,” he said. During sessions in the conference's diagnostic reasoning track, he and other experts offered advice on how hospitalists can reduce their and their hospitals' diagnostic errors.

Check your work

One reason hospitalists might not think they make errors in diagnosis is because they never see their patients again. “I bet that most of the outcomes of the patients that you care for in the health care system are unknown,” said Andrew Olson, MD, FACP, assistant professor of medicine and pediatrics at the University of Minnesota in Minneapolis.

Hospital medicine is very focused on conveying important information to other clinicians, he noted. “I try really hard to call primary care providers after we have a patient discharged. I try really hard to sign out in a systematic way to my colleague coming on. But there is no information that goes the other way in any feedback loops,” he said. “When the patient is readmitted, the discharging physician doesn't always find out. When the discharge diagnosis is different from the admission diagnosis, we don't know that either.”

To improve their diagnostic skills and avoid errors, hospitalists need to know whether they've gotten a diagnosis right or wrong. Although that doesn't happen often now, it's not that hard to make it a regular practice, according to Dr. Olson.

“The cool thing about hospital medicine is that pretty much every patient will be seen by more than one person,” he said.

Feedback can be collected individually, for example, by asking a colleague on handoff to reconsider a patient's diagnosis. “When I work nights, it's really kind of fun if somebody says, ‘Could you take a look at this case?’” said Dr. Olson.

Andrew Olson MD FACPslashPhoto courtesy of Dr Olson
Andrew Olson, MD, FACP/Photo courtesy of Dr. Olson.

This practice can identify individual issues as well as more general problems. “If I find that I'm actually really suboptimal at diagnosing bronchiolitis, my partners can help me with that and give me feedback about that,” said Dr. Olson.

This review process doesn't have to be lengthy to be effective. “That doesn't mean we redo the whole H&P,” he said. “I'm not encouraging you to take an hour for each holdover patient.”

Feedback can also be low tech, Dr. Olson said. “Leave a list for your colleague of your census on Monday afternoon. They can fill it out on Wednesday: What's their name? Did their diagnosis change: yes or no, and why?”

Such systems could also be worked into electronic health records. “We all think we look in the charts, but there's really compelling data saying that we don't. The patients we look in the chart for are the ones we're worried about,” said Dr. Olson. “It needs to be automated. When a patient gets discharged that I admitted, I should get a discharge summary.”

Fear of failure

All this potentially negative feedback might be an intimidating prospect. “All of us are worried that … the reason the diagnosis changed or evolved is because I'm stupid. That's not it,” Dr. Olson said, noting that most diagnoses will change because of new information. “The daughter flew in from California and gave more information. The test came back positive. The patient spiked a fever the next day.”

He advocated accepting more uncertainty about diagnosis throughout a hospitalization. “On day 1, you write this whole big long thing about what the patient might have or might not have, but on day 2, we stop doing that. Now it's just a diagnosis,” he said. “Keep the idea of a working diagnosis.”

In cases where the diagnosis is particularly uncertain, that information should be shared when the patient is handed off. “We are terrible at talking about uncertainty,” Dr. Olson said. “We assign diagnoses and whether we're sure or not, the note looks the same. Why don't major transitions of care have a quantification of ambiguity?”

A transition presents an opportunity for a diagnostic timeout, either in your own head or in conversation with another clinician. “What else could it be? You don't have to do another test. Just think about it at the time of transition of care,” he said. Also consider the worst-case scenario, Dr. Olson advised. “What's the worst thing this could be? Come up with that, because once in a while, you need to test for it. The great example of that is pulmonary embolism.”

While they're accepting uncertainty, physicians should also acknowledge their own trouble spots, he said. “How many of you have found that you've had delayed diagnosis of a spinal epidural abscess in a patient on chronic opioids?” asked Dr. Olson. “The patient comes in with back pain, and we have certain assumptions.”

Whether it's your patient or your day causing you diagnostic problems, consulting with a team can help avoid an error. “Has anybody ever had a day where you're just off or you feel like you can't do it? Give yourself permission to talk about that,” said Dr. Olson. “Say, ‘You know, I'm just not hitting on all cylinders today, I need help.’”

Teams and tools

Working with a team was one of the recommendations offered by the National Academy of Medicine's 2015 report “Improving Diagnosis in Health Care,” Dr. Graber noted. Team efforts to combat diagnostic error should include other physician specialties, of course, but also nurses. “Nurses spend a lot of time with our patients. They are in a great position to know whether our communication was effective, whether things are playing out the way we think they should,” he said.

Patients should also be considered part of the diagnostic team. “If the patient is your partner, you have a new safety net. Tell him, ‘I think you have gastroenteritis, but I'm not sure. If you get worse by tomorrow, you have to let me know,’” said Dr. Graber.

He reiterated Dr. Olson's advice to consider, and ask others, what else could be causing the patient's symptoms. “Always make a differential diagnosis. Always. Not just on the cases you're puzzled about. I'm not worried about those. You'll take plenty of time on those cases. Do it on the cases you're really sure about,” Dr. Graber said.

Checklists can help, and a number of diagnostic ones are available on the website of the Society to Improve Diagnosis in Medicine , an organization Dr. Graber founded. “You can download the smartphone version, so you can just tap on the app,” he noted.

Other online tools can also help, but not the one you're probably thinking of. “Friends do not let friends use Google for diagnosis,” said Dr. Graber. “The sensitivity of Google is about 60%, so 40% of the time what you're looking for is not going to be on that list, as opposed to these engines that have been specifically designed for diagnosis, which have a sensitivity of 95%, 98%. And the specificity of Google is zero. It's an unordered list.”

Tools that he does recommend include Isabel , Dxplain , and VisualDx.

Dr. Graber also offered some advice to hospitalists on fixing the systemic causes of diagnostic error. “None of these are going to be a surprise to you,” he said, listing off causes identified by past research, including failures of communication, coordination of care, subspecialist availability, trainee supervision, and test and lab reliability.

The problem with everyone's familiarity with these shortcomings of hospital care is that they become normalized, he explained.

“We get used to the way things are; we don't have time to go report every little problem to our boss, and if we do, we won't be well thought of anymore. And yet that's what we need to do to improve medicine,” he said. “Every once in a while—I don't know what the right interval is, once a month, once a year—bring something to attention. Say, ‘Hey, this is really bugging me. Can we please fix this in our system?’”

Each small improvement helps physicians come closer to acing the herculean task of accurate diagnosis. “I think it's the most difficult cognitive challenge that humans face,” said Dr. Graber.