September 19th, 2011
Decision-Making Shortcuts: The Good and the Bad
A patient comes to the ER complaining of 2 hours of substernal chest pain. An electrocardiogram reveals ST-segment elevation in 3 leads. A critical, ad hoc decision is made to call a “STEMI alert,” thereby activating the cath lab team and an interventional cardiologist. As the late Alvan Feinstein, the Yale educator and father of clinical epidemiology, once noted, “Every observant clinician has discovered that certain ‘short-cuts’ or other maneuvers, either of intellect or of action, can increase the efficiency of his work in clinical practice.”
These cognitive shortcuts are also known as heuristics. Understanding how we use them in medicine can help us improve practice. Because heuristics simplify difficult decisions, they help us avoid “analysis paralysis” under conditions of uncertainty that demand speed. In that way, they can improve decision-making effectiveness. But they can also lead to mistakes. Let’s start by exploring the good side.
The Benefits of Heuristics
Psychologist Gerd Gigerenzer uses an analogy, called a “gaze heuristic,” of a baseball player catching a fly ball. To do it successfully, a player simply fixes his gaze on the ball and starts running. If he maintains a constant angle of gaze by adjusting the direction and speed of his running, he will arrive at just the right spot to make the catch. By concentrating only on the angle of gaze, he can ignore the speed, trajectory, and spin of the ball, as well as the wind and many other factors. In effect, less is better. Gigerenzer has identified an “adaptive toolbox” of heuristics that we commonly use to address various types of problems. Here are a few:
The recognition heuristic enables us to use a single cue or a recognizable pattern of cues to quickly form a conclusion or size up a situation. Rapidly analyzing an ECG to diagnose a STEMI is one example. Seeing a pattern emerge from a patient’s historical narrative, leading to a diagnosis of chronic stable angina, is another.
The one-good-reason heuristic involves analyzing a short series of cues, then stopping when we perceive a strong or compelling cue. An initial ECG showing ST-segment elevation is, for example, a strong enough cue to prompt the immediate action of activating the cardiac cath lab. The trick is to start by first analyzing the high-impact cues.
The tallying heuristic allows us to organize cues in deciding among competing options. In the ER, I recently saw a patient with chest pain and a history of gastroesophageal reflux, which she had hoped was the cause of her pain. But she also had a history of bypass surgery and multiple cardiovascular risk factors. After weighing all the factors, we proceeded to the cath lab. She had two critical lesions and received two stents, and her pain resolved. Research shows that simply tallying up unweighted cues is quite effective. You just need to know which ones to consider.
Anchoring and adjusting, a heuristic I discussed in my previous blog post, describes how we assess subjective probabilities starting with an initial (anchor) impression and then adjust the probability estimate by incorporating new information such as a test result. Used properly, this heuristic can turn you into an intuitive Bayesian thinker.
Expert clinicians know how to filter out weak cues and focus on strong cues, as if separating signal from noise. Strong cues may be a key detail from a patient’s medical history, a bead of sweat on the brow of a patient complaining of chest pain, or certain ECG findings. Weak cues may be unreliable markers such as a soft carotid bruit or the lack of an S3 gallop.
The Risks of Heuristics
Like a medical procedure, heuristics can have both risks and benefits. Psychologists Daniel Kahneman and Amos Tversky studied many of the pitfalls of heuristics, such as these:
The base-rate neglect fallacy, explored in my previous post, surfaces when we misuse the anchoring and adjusting heuristic.
Representativeness involves jumping to an erroneous conclusion that is unlikely to be accurate, on the basis of an initial impression. ECG findings of ST-segment elevation due to early repolarization could lead to the erroneous diagnosis of acute MI in a young patient for whom that diagnosis is very unlikely. The medical adage “when you hear hoof beats, consider that it is a horse not a zebra” helps us avoid this trap.
Availability is a pitfall in which judgment is clouded by salient or recent events that happen to be more available and accessible to our working memory and intuition. Missing an uncommon diagnosis such as aortic dissection can be very troubling and memorable, but we should not then give this possible diagnosis undue weight in assessing subsequent patients.
By guarding against these tendencies, we can improve the chances that our heuristics — which, after all, are often useful — will yield good judgments.
How to Increase Awareness of Heuristics
Most physicians, whether trainees or seasoned clinicians, do not think consciously about heuristics. Becoming more aware of them and developing a common vocabulary will help us use them more effectively. There are two key domains where this kind of change could have a big impact.
Clinicians can be made more conscious of heuristics starting in medical school and continuing during fellowship training. Trainees may subconsciously learn about heuristics through experience, but that method is slow and unreliable. We should be able to teach these simple thinking processes overtly, just as we explicitly teach a one-hand tie to a surgical trainee. On my teaching rounds, I often include a brief discussion of how we use heuristics in medical practice. For example, I talk about anchoring and adjusting to teach the proper use of stress testing. I also discuss the recognition heuristic to illustrate the value of taking a detailed narrative history from a patient — patient-reported cues emerge as a recognizable pattern, like stars in a constellation. Including more explicit training on the use of heuristics would undoubtedly improve the consistency and quality of medical decision making.
Research into Medical Decision Making
Cognitive psychologists may discover other heuristics, but medical research is unlikely to invent new ones. After all, humans evolved to use heuristics long before modern medicine existed. Nonetheless, the cues that heuristics employ are domain-specific, with particular ones in each medical specialty and subspecialty. Analyzing the validity of those commonly used cues may be one way to advance research about decision making in the field of medicine. Addressing the basic science of medical decision making will require new ideas and true creativity.
What are your ideas for how to improve the use of heuristics in the practice of medicine?