When Heuristics Cause Harm
Dumbledore, finance guys, and two patient stories
Recently, on a drive home from the lab, I listened to an episode of Code Switch that discusses the racial biases inherent to the tools of diagnosis and treatment that we use every day in medicine. Dr. Andrea Deyrup, a pathologist who has spent her career devoted to uncovering race-based medicine, speaks about the harmful impact of “facts” we are taught in medical school, ranging from keloids, to spirometers, to kidney disease. Some of these things, I had heard of; others were new to me.
The episode is well-produced and engaging, while also providing a nuanced and neutral discussion of facts, with the inclusion of verified, peer-reviewed research. I highly recommend a listen when you have time.
At one point in the episode, Dr. Deyrup talks about heuristics — heuristics are simple cognitive shortcuts that can potentially help diagnose patients, recognize life-threatening conditions, and help us avoid ‘analysis paralysis.’ We use them in everyday life, too: when I ask a stranger to take a picture of me in a tourist location, and I gravitate towards a young woman because I assume that young women are likely to be savvy with technology (and particularly, social media), that’s using a heuristic.
Or, if you’ve seen those videos floating around that show a picture of a person, and others are asked to guess what their job is — the guesses are all based on heuristics. What we know about people, what we’ve observed in the past, and the connections our brain makes all play into these kinds of assumptions (so when I see a 6’5” man with blue eyes wearing a Patagonia vest, walking around FiDi, I will assume he works in finance)1.
In medicine, heuristics can be incredibly useful in providing efficient and timely care. But heuristics, just like any shortcut in life, can be dangerous — they can cause patient harm if not applied correctly, or misused.
In the Harry Potter series (yes, I am referencing my favorite childhood books, and it probably won’t be the last time) Dumbledore makes a distinction in the choice between what is right, and what is easy, implying that the right choice might be the more challenging of the two. But with heuristics, this distinction breaks down. The easy choice might be the right choice.
So… how do we know when it’s right, and when it’s wrong?
One night in the hospital on consults, I got a call from the emergency room for a patient with suspicion for compartment syndrome. Compartment syndrome is a limb-threatening (and sometimes life-threatening) condition that occurs when pressure builds up in anatomic spaces to the degree that it can compromise blood flow and interfere with normal physiologic functions. It is one of the very few cases that requires immediate surgical attention – not tomorrow, not in an hour; now.
When I got downstairs to examine the patient, knowing nothing other than what I had been told over the phone,2 I started with an inspection of the leg — it looked normal to me, but the patient had dark skin, which I knew would impact my assessment of pallor. I barely brushed against his shin with my gloved hand and he yelped in pain. I couldn’t palpate a pulse in his foot, and I couldn’t exactly locate a signal when using the Doppler ultrasound, either — so far, not very reassuring.
I asked about his pain, when it started, what brought him to the hospital. He started to tell me, in a roundabout, convoluted way, about his medical history, his symptoms, and his pain. I touched his leg again while he was talking, trying to see if he’d still show discomfort when distracted – he winced slightly. Neither a positive nor a negative result.
There were some suggestions of compartment syndrome there, and yet — something wasn’t adding up. The patient’s appearance (disheveled, poor grooming), his halting speech, his previous visits to multiple hospitals, and his questionable description of his injury all pointed to one thing: malingering.
Malingering is a word we use to describe a faked injury, an exaggerated symptom — any time a patient misreports or lies about their subjective illness experience in order to gain something (to avoid work or responsibility, seek attention, or gain treatment).
I stayed with him for longer than I normally would in an emergency situation, asking him more questions, trying to get him to talk more — the more he said, the more clues I would have at my disposal when trying to form an assessment. I was reading closely, something I have done hundreds of times before — but this time I was reading a person, and a situation — not a book.
But the longer he talked, the less sure I became. Maybe this man’s pain was real (in over two millenia of developing and advancing medical care, we still have no good way to measure someone else’s pain, which surely says something deep and philosophical about the human experience and the limits of science), but I just didn’t think it was life threatening. My professional assessment was that this patient was drug-seeking.
I called my senior resident so she could assess the patient too (compartment syndrome is not a diagnosis you can afford to miss). She was concerned. She thought he had compartment syndrome. We called the fellow, proceeded with the diagnostic workup, and expedited whatever was in our power to expedite.
But… after a few hours of working on this patient, we came to the realization that he did not have compartment syndrome. In fact, careful review of his outside medical records showed that this patient had already had a fasciotomy (a procedure to relieve the pressure within the compartments), and it is nearly impossible to develop compartment syndrome once your compartments have been permanently opened up.
So it turned out that I was right. He was just here to obtain drugs. A patient’s appearance, mannerisms, attitudes — all of these things are important in painting a picture of who they are, and what they might be suffering from. And I painted the right picture.
But there was another time, the year before, when relying on my gut didn’t help things; in fact, it may have done some harm.
A young girl had undergone abdominal surgery and was recovering on the floor — I was covering nights, which meant there wasn’t much help around if I needed it. Each night, I got a call about pain. And each night, I went to her room to assess and examine her.
“Where does it hurt?” I asked her, and she just grimaced, whining slightly. She was 17.
Her mother, sitting next to her, was quiet but looked up at me with large eyes, both sympathetic and desperate. She didn’t say anything, like others might have; she didn’t demand that I get the medicine right now, she didn’t insist to talk to the attending. She simply thanked me, silently, for coming in to see her daughter.
I palpated across the girl’s abdomen, but the pain didn’t seem to be elicited by my touch. I didn’t know what to do with this information. If the pain wasn’t worse when I palpated, then surely it wasn’t anything terrible? Maybe she just had a low pain threshold.
She was already on Tylenol and opioids at regularly scheduled intervals. What else could I do? I went on UpToDate (the doctor Google), looked through old textbooks, even tried Google itself, to see what I could do. I settled on Toradol, and hoped for the best.3
Two nights later, I learned from the daytime team what the problem was — suspicious of the pain, they had ordered a CT, and it revealed that she had a splenic infarct. An organ inside her was dying, slowly cut off from its blood supply, shrinking and crumbling inwards.
Her pain hadn’t just been a temporary problem to take care of. It was a sign that something was wrong. I chastised myself for not realizing this, for not remembering what was probably the first thing we learn as medical students. Pain is not a diagnosis. It is a symptom. And it was my job to figure out what that symptom was.
Did I assume that she was just complaining because she was young? Had I internalized the medical establishment’s tendency to ignore women’s pain? I wondered how I could have missed such an exceptionally bright clue.
In relating these stories to you, it is important for me to share that neither of these patients were white. Both patients were people of color. Both had probably experienced dismissal and racism at the hands of healthcare professionals.
In my time taking care of both of them, I tried to make sure that race didn’t get in between their stories and my assessments. I took care of them as human beings. I constantly think about race, sex, gender, and identity when it comes to my patients (and, for that matter, my colleagues and community) — but. We all have internalized biases. We are all products of the world we live in.
So can we trust heuristics? In one of these cases, I was right. In the other, I was probably wrong.
In the first case, when heuristics helped me deduce that a patient was malingering, I had been a doctor for nearly two years, having four years of clinical experience under my belt by that point. In the second case, when I failed to understand the cause for a girl’s pain, I had barely been a doctor for two weeks. I try to give myself grace in these situations.
But the question of age and experience doesn’t mean nothing; it is probably why people react with such alarm when I walk into the room and announce that I’m their doctor: “You look so young!” Part of it, surely, is the societal expectation that a doctor must be an old, white, man (I could not pass for any of these). But the other part is that people expect someone with experience; someone who has seen thousands of cases before and therefore will just know what to do.
The implication there is that heuristics are more helpful the longer you’ve been a doctor. But even there, I have my doubts.
What I will say is that a heuristic is not a direct path from a constellation of symptoms to a diagnosis. It is one tool in our armamentarium, and it should be used carefully and wisely. Just as we would never chalk up chest pain to gastric reflux without ruling out a cardiac source with a careful history, examination, EKG, and blood tests, we should never rely on a heuristic to point us directly to a diagnosis. It is just one tool. It should be used wisely.
Sorry, I couldn’t resist referencing his hilarious meme.
Typically, when called with a consult, I take the time to look over the chart, to ascertain their medical history, most recent labs, and any imaging — I want to know as much as I can about the patient before I go talk to them. But in an emergency, that process is flipped: the first, and most important thing to do, is lay eyes — and hands — on the patient.
Though the common teaching is to avoid NSAIDs after surgery due to risk of bleeding, most studies and randomized trials are inconclusive, and many sources suggest that a short course is safe.
Enjoyed this immensely.