Higher Stakes
Years later and hundreds of miles away a nocturnist gets a call from the emergency department (ED) on the seventh new admission of the night.
“I’ve got another rule-out myocardial infarction (MI) for you” said the ED physician, who briefly provided the assessment that the patient was low risk, with negative enzymes, chest X-ray, and electrocardiogram.
The nocturnist noted the atypical severity of the pain, systolic blood pressure more than 200, and positive cocaine history. But this did not alter the plan as the patient was passed from the ED physician to the nocturnist and then to the hospitalist who assumed care the next morning. Unfortunately, it took the patient experiencing a severe increase in tearing pain radiating to his back during the exercise stress test to prompt the discovery of his ascending aortic dissection. The patient died on the operating room table, leaving all three physicians wondering how they could have missed the diagnosis when in retrospect it seemed so obvious.
Present the same clinical scenario at grand rounds and the third-year medical students could tell you dissection should have been considered. How did three smart experienced people all make the same fatal mistake?
This case demonstrates a number of heuristic failures. Availability bias is a form of pattern recognition and arises from our habit of perceiving the things we see often as more likely than those which we have not seen or thought about recently. Hoof beats in Kentucky, as they say, are usually not a herd of zebra. ED physicians see what at times seems like hordes of patients with low-risk chest pain, the vast majority of which lack a life-threatening etiology. Thus, we can become complacent in assuming that the next admission for chest pain reflects the same cause as the seven before.
Pattern recognition serves a vital role. Most expert physicians rely on this more than classic deductive reasoning and, much less, Bayesian analysis. Casino operators exploit this tendency to see false patterns to their profit by installing displays that show the last 10 to 20 results over the roulette table. However, just as each turn of the roulette wheel is not influenced by prior spins, each patient is unique. One must beware of the misleading power of the availability bias.
Once the initial misdiagnosis had been made, the anchoring bias and confirmation bias continued the cascade of events—turning a mistake from a temporary error to a disaster. The phrase “chest pain rule out MI” not only encourages the physician to minimize the potential severity of the symptom via the framing effect but also telegraphs the anchoring phenomenon by fixing on a single disease concern for a symptom whose etiologies are legion.
However, even accepting that the initial diagnosis by the ED doctor was influenced by the availability bias, why was this not corrected by the nocturnist or by the hospitalist on the next day? The answer lies in diagnosis momentum.
Each physician does not evaluate the patient in isolation but rather has a tendency to include the assessment of the prior clinician as part of their own decision-making process. The more people who have seen the patient and agreed with the diagnosis, the higher the mental hurdle becomes to disagree and take the work-up in a different direction.
click for large version
click for large version
What You Can Do
Does the mere existence of these many heuristics condemn the physician to a career of repeating these potentially fatal errors? The obvious answer is no, but the solution requires a concerted effort on the part of the physician to avoid these mistakes.