My shift in the emergency department (ED) with the new intern was winding down. He was dressed in scrubs covered by his starched white coat displaying his photo and name on an identification badge: “Jacob Tapper, MD” (names and identifying features changed, here and later). His face radiated enthusiasm, intensity, and attentiveness, and he listened carefully to every word I said when we discussed his patients. We had seen five patients together over the past four hours and were preparing to hand off our patients to the oncoming team while tying up loose ends and discussing a few pending tests. But the charts of newly arrived patients began to accumulate in our rack, so I decided to look through them and make sure there was no one who needed attention immediately. I came across one chart with a chief complaint of chest pain in a 40-year-old man. An ambulance had delivered him to the ED 20 minutes ago. “Let’s take a look at this one,” I said to Jake, partly to make sure nothing serious was going on and partly to give Jake the opportunity to learn more about the differential diagnosis of chest pain.
As we walked into the room I noticed that the patient, Mr. Sanchez, a balding man with wire-rimmed glasses, was sitting up in bed with his jaws clenched and his hands pushing down on the mattress. His wife sighed with relief as we walked into the room. Jake introduced us and began asking questions while I scanned the chart for the EKG that typically was ordered on all chest pain patients immediately upon arrival in the ED. Since there was no EKG on the chart, I ordered one and listened to Jake continue to elicit the medical history. Because I was there watching, he was probably being even more careful than he might have been normally. Mr. Sanchez described a 24-hour history of pain that was in the middle of his chest and also in his left shoulder and back. I watched as Jake touched the painful areas, and Mr. Sanchez, with tears running from his eyes, winced when Jake pressed against the muscles of the back of the shoulder and below the sternum.
“He never complains of pain,” said his wife, a fastidious woman in a pink dress carrying a large white bag in one hand and a cell phone in the other.
“We’ll give him something for pain,” I said. I watched as the EKG technician entered the room, placed the leads on the patient’s chest, and soon handed me the resulting EKG.
“It looks normal,” I announced to both the patient and his wife. “No sign of a heart attack,” I added as I handed the EKG over to Jake. He nodded and continued with his physical exam and more questions about the character of the pain, previous problems, family history, occupation, and health-related habits. I learned that Mr. Sanchez rarely visited doctors and that this was his first visit ever to an ED. We spent 20 minutes listening to the heart and lungs, feeling pulses, and examining the abdomen, investigating every possible cause of pain that might accompany a normal EKG—pulmonary embolus, aortic dissection, and acute coronary syndrome among the serious possibilities, and costochondritis, esophageal reflux, and upper respiratory infection among the less serious ones. No diagnosis was apparent, but it was clear that, based on the severity of the pain, the age of the patient, and the fact that Mr. Sanchez rarely sought medical care, we needed to pursue all the possibilities. We concluded the exam and explained to the patient and his wife that we were not sure what the cause of the pain was, but would order pain medications, some blood tests, and X-rays. Jake and I walked back to our workstation and watched as the patient, his wife, and a nurse ambled over to the bathroom. They later waved to us as they were returning to their room and I waved back, feeling satisfaction that we had established good rapport with the patient in spite of having no answer for the cause of the chest pain.
Two minutes later a nurse came running to tell me that Mr. Sanchez had gone into ventricular fibrillation. My replacement attending, who had been walking by the room, had taken him back to the critical care area and was doing CPR. When I got there, I watched as the defibrillator discharged and Mr. Sanchez jerked and opened his eyes. The monitor still showed ventricular fibrillation, and another shock followed. Jake and I watched in horror unable to comprehend how someone who had been waving to us moments ago could now be dying. What had we missed?
The second defibrillation worked, and the monitor showed a normal sinus rhythm. “He has pulses,” said the nurse who was holding her fingers on Mr. Sanchez’s neck. The rest of the management went by in a blur. The cardiology team arrived; we described the history and showed the normal EKG and the monitor strip with the ventricular fibrillation; before long, Mr. Sanchez was being wheeled to the catheterization lab. I thanked the other attending for being in the right place at the right time and taking charge of the situation, and Jake and I went off to speak with the family. By this point Mr. Sanchez’s daughter and son had joined his wife. They had heard about the cardiac arrest, and I could tell they expected bad news. The guarded report I gave was actually better than they had expected. Finally, Jake and I talked. “You didn’t miss anything,” I said, answering the question I imagined he was thinking before he could ask it. “No one could have anticipated the fibrillation.” But I asked myself whether I had given undue weight to the normal EKG and had relaxed prematurely. I realized we were lucky. The nurse had noticed the fibrillation on the monitor immediately; if she had been busy with other responsibilities, there could have been a delay, and my colleague might not have been walking by the room at just the right moment and jumped into action.
All of the emotion brought on by this experience got me thinking about the challenges of making a medical diagnosis and the dangers of diagnostic error, as well as what we can do to improve diagnostic thinking and teaching. According to a report issued by the Institute of Medicine (IOM)1 in 2015, diagnostic error is the leading cause of medical malpractice. The report has eight recommendations for reducing diagnostic error that address education, patient and team communications, information technology, surveillance and reporting, liability issues, and payment issues. In the rest of this editorial, I share ideas about how medical education could help us improve diagnostic thinking and how, in addressing diagnostic error, we might also address many related issues that could improve medical education.
Valuing the Time With Patients and Families
The first area for medical education to focus on to reduce medical error is to help students recognize the vital role that patients and their families can play in the diagnostic process. A part of the first recommendation from the IOM report states,
Health care professionals and organizations should partner with patients and their families as diagnostic team members and facilitate patient and family engagement in the diagnostic process, aligned with their needs, values, and preferences.
In the care model that I learned as a student, the patient played a passive role, providing background information. But the real focus of a diagnostic workup was on using laboratory and radiologic testing to confirm a hypothesis, and then defining and treating the pathology. In contrast, in the model described in the IOM report, the patient brings ideas, values, and priorities to the medical encounter; shares in all the information and its interpretation with the physician; and actively participates in treatment decisions when possible.
What this means to me is that as we teach our students about diagnostic problem solving, we must emphasize the importance of the time that they spend with the patient in developing rapport and in understanding the concerns that caused the patient to seek care. We must prevent the diagnostic process from being dominated by the blood tests, imaging tests, and other technologies that can provide valuable confirmatory information but can also alienate the patient from the physician. A recent report by Mamykina et al2 showed that the residents they studied spent 9.4% of their time in direct care activities with patients. Fifty percent of their time was spent on computers, either adding information about their patients into the electronic medical record or checking results of tests. For physicians to partner with patients, there must be a revaluing and reallocation of the time they spend together, and this should begin in training by having students and residents focus more on talking to and listening to patients.
We who are educators can also demonstrate that we value time with patients by encouraging faculty to personally observe our students and residents when they are with patients rather than waiting for presentations or reviewing charts. When patients like Mr. Sanchez experience a complication such as ventricular fibrillation, it is important for residents and students to know how their faculty supervisors view their findings and their ways of taking responsibility for the patient’s care. It is also important for the students and residents to observe their supervisors speaking with family and nurses to deliver bad news.
Understanding and Improving Diagnostic Thinking
In addition to engaging patients and families in the diagnostic process, it is important that we work to improve our understanding of diagnostic errors and how to make the diagnostic process safer for patients. A patient like Mr. Sanchez provides a rich opportunity to explore the complexities of the diagnostic process. While most patients—who, like Mr. Sanchez, experienced a heart attack that developed over 24 hours—would have an abnormal EKG, Mr. Sanchez had a normal EKG. This should have reduced his risk substantially according to a recent review by Hollander et al.3 However, Mr. Sanchez did have certain clues for possible risk in his history: His pain was in the middle of the chest and radiated to the shoulder, and there was nausea and vomiting. Unfortunately, the various clues and the initial test results presented a contradictory picture. This is often the case, and provides one of the most vexing challenges in medical diagnosis. The usual patterns of a disease are not always present, and patients’ interpretations of pain are variable and inconsistent.
One of the differences between expert clinicians and new students and residents is that experts are able to recognize the typical patterns of a disease quickly, but also recognize when findings do not seem to fit together. In such cases, experts take a step back and consider atypical presentations of disease. In a study by Sklar et al,4 atypical presentations were frequently associated with a misdiagnosis in patients who were discharged home after an ED visit and later died. Norman et al5 in this issue of the journal review theories of diagnostic error and conclude that the science of diagnostic error prevention is still in its infancy and that there are no easy solutions beyond improving a clinician’s overall store of knowledge through experience. They believe that many diagnostic errors are often a function of the lack of clear information (as was the case in Mr. Sanchez), and they recommend continued research in the science of diagnostic thinking. This was also a recommendation of the IOM report.
However, there may be ways to prevent diagnostic errors, and to prevent those that do occur from causing harm to patients, by turning to the growing science of quality improvement and patient safety (QI/PS). While we may not understand the intricacies of how the brain processes masses of information, we do know that medical errors, both diagnostic and procedural, can sometimes be prevented by redesigning care processes to reduce reliance on memory, which can become overtaxed, and to standardize processes so that fewer decisions need to be made. The science of QI/PS recognizes that humans are fallible and will make mistakes, and it attempts to design systems to limit the mistakes and to reduce the chances that the mistakes will harm a patient. Simple steps, such as reducing interruptions, noise, and time pressures, could help clinicians who are puzzling over a difficult case maintain focus and make the right diagnosis. Encouraging faculty to become knowledgeable about QI/PS and encouraging them to use their skills to improve care delivery systems could help reduce diagnostic error through better design of care delivery systems. In this issue, Coleman et al6 describe the imperatives and challenges for faculty development in QI/PS, and Stevens7 provides a Commentary about the tactics that might help meet those challenges. Sehgal et al8 describe the use of a quality improvement portfolio for faculty that can reward and encourage faculty involvement in QI/PS by recognizing faculty activities in QI/PS and integrating them into the promotions process.
To address the challenges of recognizing atypical patterns of disease presentation and diagnostic dilemmas, we could find more ways for less experienced clinicians to learn from senior clinicians. Iyasere et al9 have described a program of coaching for early-career hospitalists that could provide the opportunity for them to capitalize on their own knowledge bases combined with that of a senior clinician in difficult cases. Improved access of generalist clinicians to specialists could also provide added expertise and knowledge in challenging diagnostic cases, and the specialists need not even be onsite, as Keely et al10 have noted in their description of eConsults that can occur remotely through the Internet. Consultation via telemedicine can also provide support to isolated clinicians with a difficult diagnostic challenge. For cases in which diagnostic decisions cannot be made immediately, clinicians can sometimes gain time to observe the progress of signs and symptoms by placing the patient in an observation unit, monitoring and reexamining the patient there, and consulting various specialists.
While there are many ways to provide support to clinicians struggling with a difficult diagnosis, those of us who are physicians must also acknowledge the dangers from our own egos in how we deal with our uncertainty and our own possible fallibility when there is conflicting information. This can be a problem not only with diagnostic decision making but also in the performance of procedures as technology changes and experience and competence diminish over time. Hubris and arrogance can interfere with our ability to admit our own uncertainty and ask for help. We must be constantly vigilant to recognize situations in which the culture and environment, or our own personalities, encourage overconfidence and unwillingness to recognize uncertainty. In this issue, Vaisman and Cram11 raise concerns about faculty who are supervising residents and students in procedures that the faculty no longer carry out frequently, and the need for clinicians who do perform procedures frequently (proceduralists) to supplement clinical supervision of residents and students in hospitals.
Embracing the Opportunities of Patient Data
Integrated patient care information and databases offer untapped opportunities to improve diagnostic processes and reduce diagnostic error. Daschle,12 in his essay identifying five transformational forces for academic medicine, selected “big data” as the most important, noting that “big data will unleash new and innovative approaches for improved chronic illness and disease management with extraordinary opportunity for patient engagement.” I believe that the same can be said for the potential of big data to transform diagnostic processes. While information systems can sometimes overwhelm and divert the attention of clinicians, information systems also offer the potential of bringing together data that may be available in multiple venues—hospitals, imaging centers, urgent care centers—that can create a clearer picture about the cause of a problem. There are many times when I have reviewed the EKG of a patient with chest pain, like Mr. Sanchez, in which the EKG was abnormal, and have wished I had access to a previous EKG obtained at another facility so that I could compare them. But there were often too many barriers to obtaining the information, and I had to proceed without it.
As health systems begin to share data platforms, it will become increasingly possible to check records and tests that may have occurred in other hospitals that may contribute key information about the cause of a current problem. Smith et al13 have described the enormous potential of patient data currently being collected through social media and wearable devices; such data could contribute new insights into the development and causes of disease. Our students will likely be using these sources of information in their future practices for diagnosis and management, and it will be important to educate them about how to use them effectively while respecting the patient’s privacy.
A Model to Improve Medical Education
By creating partnerships with patients in the diagnostic process, helping our clinicians to become as capable as possible in analyzing data, and making accurate diagnostic decisions that are supported by safe clinical systems, integrated information systems, and expert consultation, we can use the diagnostic process as a model for how to improve medical education generally. Most aspects of medical education involve development of skills in patient engagement, growth of individual knowledge and expertise, safe and effective clinical environments, and management of information. How we give greater emphasis to these in the area of diagnostic thinking can be a model for other areas of medical education as well as a reminder of the challenges and risks involved in diagnosis. Fortunately, in the case of Mr. Sanchez that I presented at the beginning of this editorial, we had a monitoring system that quickly detected the life-threatening complications, and had a team that responded effectively. Mr. Sanchez walked out of the hospital with an implanted defibrillator several days after almost dying three times; Jake Tapper, the intern, learned valuable lessons about diagnostic testing and the need to step back and see the implications when diagnostic findings do not fit together. The rapport Jake initially established with the patient and family helped the entire care team through difficult circumstances. And the ED decided to augment the use of monitors to improve its observation of cardiac rhythms by hiring dedicated monitor technicians to reduce the risk that a patient with sudden ventricular fibrillation might experience a delay in treatment.
While this case began as a simple observation of an intern’s history and physical examination skills, its educational effects spread far beyond the initial intent. Medical education could have this same impact on many areas of our health care system by increasing students’ understanding of its many interconnections to health care providers, students, and clinical care processes. As we begin a new year, education about reducing diagnostic errors in medicine would be a great starting place to focus our efforts to improve medical education and, in so doing, improve our health care system.
David P. Sklar, MD