As a cardiologist, I often find myself talking to patients about invasive tests and procedures. On busy clinic days, it is not unusual to refer a few patients for a heart catheterization or, perhaps, a transesophageal echocardiogram or device implant. During such conversations, patients often ask, “Doc, if it were you—what would you do?”
I admit I found those questions irritating at first. We’re not talking about me, I thought, impatiently, we’re talking about you. But after practicing medicine for a few years, I’ve begun to see such questions (“If it were your mother, what would you tell her?” is another favorite) as an understandable—if ultimately unanswerable—way for patients to make sense of medicine’s tangle of individual choices.
A patient I saw recently, F., asked me about treatment of her atrial fibrillation, a common irregular heart rhythm. She had, for several years, undergone various attempts at cardioversion (an electrical shock delivered to the heart to “reset” it) as well as antiarrhythmic drug therapy. Now, she was told by her electrophysiologist, it was time for an ablation—a procedure in which catheters are inserted into the heart to “burn” the offending excitable cells. Mostly, she said, she felt fine. Or, at least, she thought she did. She was still able to pick up her grandkids from school and work in her garden, and if she was tired from time to time, well, who wasn’t? But, still, she was worried about the arrhythmia that was recorded on her Apple Watch. What should she do, she asked?
These days, as a first step, the answer might involve what’s called “shared decision-making,” a process in which patients, often through the use of validated online tools and backed by the work of behavioral scientists and decision theorists, make informed choices that are in line with their goals. Good health, the thinking goes, is the result of good decisions; good decisions, in turn, ought to be driven by patients, who are “experts in what matters most to them.” In a structured way, shared decision-making encourages patients to ask: Just how limiting are my symptoms? How much do I mind taking medications? In theory, therefore, shared decision-making transcends the old paternalistic model of medicine by aligning treatment with patients’ values and preferences.
On its face, this process is eminently reasonable. It is a way of making rational decisions in the face of uncertainty. However, despite its aim of weighing possibilities and assigning utility values to each before arriving at an answer, the process remains incomplete. I have concluded that shared decision-making attempts to untangle what is, in fact, a complex set of metaphysical problems at the heart of health care.
The philosopher L.A. Paul has written about “transformative experiences.” These are experiences that change us both personally (they change who we are and what we care about) and epistemically (they change what we know—and how we know it). They are so life-altering, in fact, that they remain unknowable until we actually experience them. Paul uses the tongue-in-cheek analogy of becoming a vampire. One cannot know what it is like, she says, to “gain immortal strength, speed, and power” while “cut[ting] a fabulous figure in black clothing” until one is bitten by Bram Stoker’s nocturnal creature—the rub being, of course, that, once one is bitten, there is no going back.
Vampires are the stuff of fiction. But I’ve started to wonder if many of the procedures we perform in medicine, even ones we consider quite minor (certainly ones less extreme than becoming a vampire), ought to be considered transformative in a similar way. Because even after certain routine procedures, patients can find themselves being changed in some qualitative and irreversible way. And often not for the better.
This is hard to square with the cavalier way in which we recommend many procedures in the first place. Consider percutaneous coronary intervention (PCI), a treatment that involves placing stents to relieve narrowing in the arteries of the heart. Well over half a million of these little devices were implanted in the United States last year. Many were for non-acute indications such as an abnormal stress test. The widespread use of stents and their relative ease of deployment tempt us to forget that they also represent permanent alterations to a patient’s body, the delicate milieu intérieur. After receiving them, patients are committed to long-term medications. They are sure to undergo more tests. Complications arise. Practice cardiology long enough, and you will encounter patients for whom PCI not only failed to relieve symptoms (if there were any to begin with) but also accelerated their overall decline.
The transformative and irreversible nature of even our most commonly performed procedures reveals that our standard model of shared decision-making—in which we ask patients to plumb their own depths and assess their values and preferences—is not up to the task. In an important sense, it cannot be.
The standard model falls short because our present preferences, the ones assessed by our best shared decision-making tools, are just one part of the story. We could ask patients to project themselves forward in time, encouraging them to imagine their future selves post-ablation or post-PCI. But which self should they imagine? If F. were to undergo an ablation, she may feel better, at least for a time. Or she may become subject to further medicalization and face an uncertain future of hospital visits for cardioversions and “touch-ups.” She may end up wishing she had never undergone the procedure at all.
Explaining the nature of a procedure, its average success rate, and what to expect after undergoing it are all worthwhile endeavors. So, too, is attempting to have patients clarify their own internal “utility function.” But it also seems entirely plausible that a patient like F. could know every physical fact there is to know about catheter ablation of atrial fibrillation, as well as possess intimate knowledge of her core preferences, and yet still have no awareness of what it’s like to undergo the procedure. Or how she would feel afterward.
What’s gained after passing through the catheterization suite or electrophysiology lab, in other words, is knowledge of a different sort. We might term this “phenomenal” knowledge, as opposed to scientific or empirical knowledge. And this sort of knowledge about how one feels can be acquired only ex post facto. Indeed, in medicine, the only way we know if something, A (a pesky arrhythmia, an angiographically impressive narrowing), is causing something else, B (palpitations, chest pain), is to try to get rid of A. And if it doesn’t work the first time, to try harder. Or harder still.
Such efforts, I find, are often justified by appeals to evidence. But even the best randomized trial doesn’t establish a cause, at least not in the way that we’re talking about here: Is this stenosis causing my patient’s symptoms? Will this arrhythmia eventually cause my patient to develop heart failure? In cardiology especially, which seems to privilege the epistemic techniques of “evidence-based medicine” even more than other disciplines do, we conveniently transpose the average treatment effect in a trial population, gleaned from thousands of research subjects, to the case of our individual patient. But a statistical finding at the population level need not imply a cause at the individual level.
This is related to what’s known as the “reference class problem.” The dilemma itself is hardly new. The English mathematician John Venn (yes, of “diagram” fame) observed in 1876 that because “it is obvious that every individual thing or event has an indefinite number of properties or attributes observable in it,” it therefore may belong to an “indefinite number of different classes of things.” This, in turn, “has an important bearing on the process of inference.”
Assigning a probability, then, to how a singular patient may fare following a recommended treatment is a fraught proposition. This very probability is contingent on how a patient is classified—and patients, of course, can be classified in innumerable ways. We cannot knowwhich relevant characteristics matter, though we tend to assume that certain ones, usually those collected as data in a trial population (things like age, sex, the presence or absence of comorbid conditions such as diabetes and hypertension), are the ones we ought to pay the most attention to.
But consider the current enthusiasm for “precision” or “individualized” medicine. I suspect this enthusiasm may actually reveal an uneasy tension, something that most physicians already intuitively grasp: Despite structural similarities in a data set, one patient’s illness cannot easily be compared with another’s. A particular patient’s disease combination is as unique as she is and may thus defy categorization (to say nothing of the particular life history, the accretion of events large and small, which led her to you).
F. elected not to undergo ablation, and, as I have since observed, seems to be doing just fine. But during her visits, I still find myself thinking about the counterfactual scenario. Is there an unrealized version of F., traversing an alternate timeline, who does undergo ablation? Has this spectral counterpart somehow been granted a richer life, or even perhaps a longer one? My heightened awareness of this hypothetical has no doubt crept into our clinical interactions. “But how do you really feel?” I probe. F. just rolls her eyes and laughs.
Historically, medicine’s aims have always been pretty clear: to save lives and alleviate symptoms. But even in the relatively short time I have been in practice, I have noticed a change in the doctor’s perceived role, perhaps emblematic of a broader cultural shift prioritizing “wellness” and longevity. Our mandate has grown to include the identification of the pre-diseased and, with our vast technical knowledge and array of tools, to forestall—if not prevent altogether—the journey from the kingdom of the well to the kingdom of the sick.
The question of what to do about a medical condition, or even a cluster of symptoms possibly sharing an underlying causal link, may require a response that less resembles the standard paradigm of ranking options along an axis of desirability than a Kierkegaardian leap of faith.