Most high-severity misdiagnosis cases in the ED fall into two groups, vascular events and infections, according to the authors of an analysis of malpractice claims. Some interventions:

Simulation training to improve diagnostic reasoning skills;

Training modules based on real-world cases;

Portable eye movement recording gear;

Data sets to look for patterns of unexpected post-ED discharge hospital admissions.

Most high-severity misdiagnosis cases in the ED fall into two groups: vascular events and infections, according to the authors of a recent analysis of 11,592 closed U.S. malpractice claims (1,323 in the ED) involving diagnostic error from 2006 to 2015.1

The top misdiagnosed conditions:

  • Within the vascular group: stroke, heart attack, pulmonary embolus, aortic dissection, and arterial embolism;
  • Within the infection group: sepsis, meningitis/encephalitis, spinal abscess, pneumonia, and endocarditis.

To learn more about these findings, ED Legal Letter spoke with the study’s lead author, David Newman-Toker, MD, PhD, director of the division of neuro-visual and vestibular disorders in the department of neurology and the Johns Hopkins Armstrong Institute Center for Diagnostic Excellence. Newman-Toker expounds on what the findings mean for EDs in terms of patient safety and malpractice risk. (Editor’s Note: Newman-Toker’s comments have been lightly edited for style and clarity.)

  • On what EDs should be doing right now in light of the study’s findings:

One immediate implication for the ED is that error reduction initiatives that are trying to improve diagnosis should definitely target major vascular events, No. 1, and infections, No. 2.

  • On the unique nature of the ED setting in terms of misdiagnosis:

The ED probably is the toughest place to practice medicine, in terms of diagnosis. There is a high level of variability of illness severity and high risk for dangerous diseases, relative to a lot of other practice settings. It is the perfect storm of challenging conditions to practice in.

We should not have the expectation that no one will ever be misdiagnosed. The goal is to improve diagnosis where we can. There certainly are areas where there are known evidence practice gaps.

One of the things that we showed in this study is that the vast majority [86%] of cases involved bedside clinical reasoning. That was way higher than the next-highest contributing factor [communication, involved in 35% of cases]. If more than 80% of cases involve clinical judgment, we are not going to eliminate diagnostic errors unless we tackle that problem head-on.

In the long term, those bedside problems are presumably going to be solved by things like artificial intelligence, computer-based decision support, and simulation training to radically transform the way we educate physicians to improve their diagnostic reasoning skills.

People are using machine learning to comb through the electronic health record to figure out if there are subtle clues, subtle changes in vital signs, that people are not picking up on. That is a little bit easier in the hospital setting when you have a longer set of data to work with. In the ED, when you’ve only got a few hours of data, rather than a few days, it is a tougher challenge.

Those kinds of methods hold some long-term promise. But those solutions are mostly decades away. In the short term, we are going to have to identify situations in which there is a known problem in bedside diagnosis, a known evidence-practice gap, and a known solution to close it.

  • On the most promising approaches to reduce misdiagnosis:

One thing we know is that general solutions, like, “Take a diagnostic time out,” or “Pause to consider what you might not have thought of,” are highly unlikely to work. They are weak interventions. Instead, we are going to have to take a more problem-specific approach, then use multifaceted interventions to target them one by one until we can move the needle. The No. 1 missed vascular event is stroke. We have been working on this issue for some time. Strokes do not get missed when patients are weak on one side and they cannot talk. They get missed when the stroke looks like something else — isolated headache that looks like a migraine, or isolated dizziness that looks like an inner ear problem. We really have to home in, in a symptom-oriented way, on how we are going to fix this missed stroke problem.

For too long, [diagnosis] has really been the exclusive purview of the one emergency physician [EP] taking care of the one patient. We are turning diagnostic errors from a singles tennis kind of sport to more of a team game. Everybody — nurses, allied health professionals — is contributing information that keeps the patient on the right diagnostic track. For example, if we have a protocol that says if you need an image, it needs to be an MRI and not a CT scan — because we know that the CT scans miss 90% of these strokes — the nurse can look at the order and say, “The protocol says you should get an MRI.” The nurse also can say, “By the way, it also says before you do that, to check the patient’s eye movements to see whether the pattern suggests it’s an inner ear problem or brain disease. Do you want me to call somebody from neurology to take a look at that?”

We also are creating training modules using virtual patients. We are taking real-world cases — dizzy patients’ digital histories and physical exams from an ongoing clinical trial — and turning them into simulation material. We have recently trained medical interns who are basically just out of medical school to be twice as accurate as senior internal medicine residents nearing graduation. Essentially, at least for this problem, nine hours of simulation training worked better than three years of residency.

We also are bringing technology to bear. We have shown that experts are able to differentiate inner ear disease from strokes even more accurately than imaging by looking carefully at people’s eye movements. We have put portable eye movement recording gear in the field that allows us to look at those eye movements remotely by telemedicine. Eventually, we hope it will be as commonplace as something like the ECG.

Finally, we are trying to give people the feedback they need on performance. If we do not start measuring this, no one will make any headway at all. Right now, it is out of sight, out of mind.

We do not have that many operationally viable ways to way to measure diagnostic errors and the harms that result from them. We have been developing new methods. We have shown that these work for stroke, and probably will work for other vascular events and infections in the ED. These use relatively simple administrative data sets to look for patterns of unexpected post-ED discharge hospital admissions that likely have been missed and suffered harm afterward. This feedback mechanism will allow us to give at least whole EDs, if not individual clinicians, feedback on their overall diagnostic performance.

  • On misdiagnosis of dizziness in the ED:

Of the top complaints misdiagnosed in the ED, dizziness is by far the worst in terms of how often it is misdiagnosed. For the half or so of dizzy patients with brain or ear problems, we essentially operate at chance in terms of diagnostic accuracy. We are misdiagnosing about 80% of the ear cases, and missing about 35% of the strokes. It is not a happy current state of affairs. We know we could be doing better.

While complete solutions will take a little while to disseminate, there are things we can do right now. We need to make calling for help easier, more systematic, with a lower barrier.

Some of that can be done with the protocolization of care. That is one of the things we did in our ED. The Tele-Dizzy Consultation Program is a subspecialty service that is provided to the ED. (Editor’s Note: Learn more about the Tele-Dizzy Consultation Program at: http://bit.ly/2py1dp9.) We created a pathway where if you [are unsure if it is] an ear problem or brain problem, and you are thinking about sending the patient home, you just call the Tele-Dizzy service, and we’ll take care of it for you.

Reaching out to neuro-otologists is not a perfect solution, because not every ED has that kind of access to subspecialty expertise. There are tons of patients, and not enough subspecialists to go around. That said, many places have other providers that have similar diagnostic skills — neurologists or vestibular physical therapists.

  • On the variability of diagnostic error rates in the top misdiagnosed conditions identified in the study:

For heart attack, average misdiagnosis rates are down around the 1% to 2% mark. That is because we have spent decades protocolizing care and developing new diagnostic tests, turning them into decision support tools and making sure the whole team is properly trained ... For all the other nine vascular events and infections, the diagnostic rates start at about 9% for stroke and sepsis and work their way up from there, to as high as over 50% for spinal abscess.

Sepsis is almost certainly the biggest bucket of harms, just in terms of the overall frequency of the disease. It also is likely that in the ED, the ones we are missing are not people who are already severely septic; it is people who are going to be septic in a day or two. That means you are looking for different things. You are looking for people sent home with a fever or viral syndrome diagnosis, when in fact there is a bacterial infection you have not recognized.

Spinal abscess is a particularly bad one. It is a relatively rare disease, but it is literally misdiagnosed more than half the time. We have now estimated that the rates of harm associated with misdiagnosis are sufficiently high that roughly one in three of every spinal abscess patients suffers permanent disability or, occasionally, death as a consequence of misdiagnosis.

That is an astonishing rate. If you have this disease, there is a one in three chance that you will be diagnosed, and you will be harmed in a way that is permanent and irrevocable ... It compares, for instance, to heart attack, where that number is probably one in 100 or less. That is a staggering difference.

That is an example where there is really just a huge education and training gap. The studies that have looked at spinal abscess have found over and over again that in almost all missed cases, the bedside red flags in history or neurological examination are missed in the assessment of these patients.

  • On what really needs to happen to make headway in preventing misdiagnosis in the ED:

The highest-ranking person — the department chair or safety leader — has to say, “This is a problem, and we’re going to address this.”

Then, you need to have someone, or some team of people, whose responsibility it is to be experts in this area.

Then, they need to organize groups of people on the front lines to identify one problem they can measure and tackle at their home institution, then do it. If you can mobilize the whole team to do that, it is a powerful combination. That might actually move the needle in reducing patient harms.


  1. Newman-Toker DE, Schaffer AC, Yu-Moe CW, et al. Serious misdiagnosis-related harms in malpractice claims: The “Big Three” – vascular events, infections, and cancers. Diagnosis (Berl) 2019;6:227-240.