The Evolving Landscape of Diagnostic Errors—Challenges and Opportunities

An 18-year-old star lacrosse player developed vertigo, vomiting, and an unsteady gait two days after his final high school game. The physicians in the emergency department (ED) believed that the patient must be taking drugs, despite the patient’s and parents’ denial. They also anchored on the idea of an inner ear problem.

In fact, the young man was in the midst of an evolving stroke syndrome from a vertebral artery dissection, presumably an injury suffered during his lacrosse game. The evolving basilar occlusion starved his brainstem of blood, resulting in a large pontine infarction. Had he been treated with clot-busting medications (and maybe even just aspirin) when he first arrived at the ED, this catastrophic outcome could have been prevented. Instead, he was left cognitively intact but quadriplegic and mute—the devastating “locked in” syndrome. Through painstaking efforts, this former top-caliber athlete, previously bound for college on a full athletic and academic scholarship, can now tip his head enough to operate a motorized wheelchair.

This patient, John Michael Night, and his family have given me permission to share his story,¹ in the hopes that preventable misdiagnosis-related harms like this will never happen to others.

The “Big Three” Diseases and Serious Harms from Diagnostic Error

John Michael’s case brings home the importance of solving the problems that contribute to diagnostic error—specifically, to preventing those that cause irreversible harms. Data from Harvard’s CRICO Strategies were used to quantify the “Big Three” buckets of disease categories that, when diagnostic errors occur, lead to serious misdiagnosis-related harms (i.e., potentially preventable permanent disability or death):²

1. Vascular events
2. Infections
3. Cancers

Three-fourths of serious misdiagnosis-related harms fall into those Big Three buckets. That fact may not come as a surprise to many clinicians. What may, however, is that the top five missed diagnoses in each bucket, taken together, account for half of all serious misdiagnosis-related harms. This means that if we improve the diagnosis of the 15 conditions below, we can take a big chunk out of serious harms and significantly improve patient outcomes.

Top five misdiagnosed vascular events: (1) stroke, (2) myocardial infarction, (3) venous thromboembolism, (4) aortic aneurysm and dissection, and (5) arterial thromboembolism.

Top five misdiagnosed infections: (1) sepsis, (2) meningitis and encephalitis, (3) spinal abscess, (4) pneumonia, and (5) endocarditis.

Top five misdiagnosed cancers: (1) lung, (2) breast, (3) colorectal, (4) prostate, and (5) melanoma.

Systems Solutions to Cognitive Problems

At this point, ample evidence exists that most diagnostic errors happen at the bedside. Failures in bedside assessment, clinical reasoning, and test-ordering decisions are found in more than 80 percent of all cases involving serious harms.² To transform diagnostic accuracy, we need systems solutions to cognitive problems, and that means we need what I call the four Ts:

Teamwork: We should engage patients, enlist nurses and allied health professionals, and adopt telemedicine as a means of working together more effectively as a team. Our team at Johns Hopkins recently created a teamwork-based ED protocol for ensuring that patients with dizziness and vertigo, like John Michael, are correctly diagnosed right away. When ED clinicians narrow the diagnostic question to whether dizziness is due to an inner ear disease or stroke, they can call a “tele-dizzy” consult and get immediate remote expertise to help in making the distinction (which relies on careful interpretation of subtle eye movements).³

Training: We must raise awareness, leverage simulations, and focus on performance in practice. Simulation training can be very powerful. For instance, we pitted internal medicine interns against senior residents on diagnosing cases involving dizziness and vertigo. At baseline, the two groups’ scores were similar, but after the interns trained with simulation for just nine hours, they roughly doubled the accuracy of their graduating senior resident colleagues, who had, on average, two years more of residency training. This finding opens new avenues for training people more efficiently to improve diagnostic performance.

Technology: We need to discover new tests, embrace mobile health, and apply decision support. Without getting swept away by the hype that “the future of medicine is robot doctors with artificial intelligence,” there are ways to design technologies that assist clinicians through diagnostic decision support by, for example, targeting the Big Three diseases described above. For stroke, we have been using facial recognition algorithms to turn commercial mobile phones into devices that can interpret eye movements the way specialists do, effectively putting “boxed expertise” into the hands of clinicians around the world at virtually zero cost.

Tuning: We have to review cases, provide feedback, and develop dashboards. A National Academy of Medicine report describes a feedback loop by which clinicians can learn from past experience⁴—but the feedback loop mostly does not exist. Our team calls this the “calibration gap.”⁵ We need to create many more opportunities for clinicians to confirm their diagnostic successes and learn from their failures. At Hopkins, we have developed a big data statistical method called SPADE to build feedback dashboards at the institutional level.⁶ Using only information about discharges and admissions, we can track missed cases of dangerous vascular events (like stroke) and infections (like sepsis). This also brings us specific insights about demographic disparities, like the increased risk of misdiagnosis faced by people of color.⁷ When tracked over time, these measures become the “needle” we are trying to move.

Our Path Forward Is Clear

Had John Michael Night been correctly diagnosed, he could have received the appropriate thrombolytic therapy and would be playing Division I lacrosse right now. We owe it to him (and to all the other John Michael Nights of the world) to make diagnostic excellence our mission.

Diagnostic error can happen anywhere, and solutions are, to an extent, specific to both disease and health system—which means that all healthcare professionals have the power, and the obligation, to make a positive impact wherever they are. I hope that everyone will join the national movement and help us to ACT for Better Diagnosis™. Find information on this initiative and other programs and resources through the Society to Improve Diagnosis in Medicine.

We are still at the beginning of a long journey toward diagnostic excellence. But I believe that if we take the four Ts to heart and execute them in a disease-specific fashion, we will make huge strides toward creating a virtuous cycle of diagnostic quality improvement that eventually gets us to where we all want to be: a world in which no patients are harmed by diagnostic error.