November 8th, 2011
Wrestling with Uncertainty in Clinical Practice
About a week ago I was asked to consult on a patient I’ll call Betty. This delightful 92-year-old woman, who lived alone, was admitted during the night with pain in the upper chest and shoulders. She had been feeling this discomfort off and on for 2 days, and when it woke her from sleep, she came to the ER. She also reported a history of hypertension, hyperlipidemia, and mild diabetes treated with oral medication. She was pain-free when I met her, and her exam was unremarkable. Both her troponin and her CPK levels were borderline elevated, and her EKG showed extensive T-wave inversion in the precordial leads.
So what was I to do with these worrisome clinical findings? Evidence-based medicine provides me with very little, well, “evidence” to apply to this patient. At 92, she was much older than the mean age of patients in trials such as CURE (64 years), FRISC (69 years), and the original trial of heparin in unstable angina (58 years). We know that treatments for acute coronary syndrome are all double-edged swords, with both benefits and risks. What about the dictum “first do no harm?”
I must admit that I didn’t pay much attention to the kind of medical uncertainty that Betty’s case illustrates until I made a transition from academic medicine to private practice years ago. In academia, we focused a lot on what was known. Betty’s case would have presented a bundle of teaching opportunities. If Betty had been discussed on our rounds, we would have made a group decision and moved on to the next patient. But my busy new practice had no teaching rounds, no team. In practice, such decisions made me feel utterly alone, and the uncertainty of medicine hit me like a ton of bricks.
I am pretty sure that no clinical trials apply to Betty. But am I even certain of that? This is the uncertainty that drives us to specialize in narrow fields, hoping to develop some sense of security in our decisions.
I discussed the options with Betty. She was firm that she wanted to pursue medications and avoid any cardiac procedures. Given the extensive EKG changes, I was a little uneasy about her decision, but I was pleased that she resolved my own decisional conflict. Her pain resolved, too, and she was discharged back to her home on a variety of medications, including a beta-blocker, nitrates, a statin, and dual antiplatelet therapy.
Betty returned to the ER yesterday, this time complaining of mild shortness of breath that occurred in the middle of previous night. The ER physician noted that she seemed quite comfortable without any particular treatment. Her exam was normal, labs were unremarkable, chest X-ray was clear, and EKG was unchanged. An echocardiogram showed an ejection fraction of 70% with LV hypertrophy, aortic sclerosis (no stenosis), and moderate mitral annular calcification — all expected findings for a 92-year-old woman.
On her first visit Betty had presented a therapeutic dilemma; on her second the uncertainty I experienced was of a different sort. What was troubling her? Was she simply anxious about being at home alone with a new cardiac diagnosis? What was the true origin of her symptoms?
Often a patient’s true condition is obscure. The human body is infinitely complex, and patients are infinitely variable. We use tests that are imprecise because of resolution constraints and artifacts. Many of those tests are indirect, creating problems with reproducibility, sensitivity, and specificity. Their predictive value is affected by how they are used. All of these factors contribute to diagnostic uncertainty.
Predicting an outcome for an individual patient is even more imprecise. A second order of uncertainty called ambiguity arises. It’s like an error bar around a point estimate. A patient may ask you, “Doc, how long do I have?” We can give a number, but the range of uncertainty around that number can make your answer almost meaningless.
Uncertainty is why we use clinical reasoning, intuition, induction, and inference. Probability is uncertainty quantified, and most probability calculations in clinical medicine are subjective and intuitive. It’s imperative that we apply basic scientific principles, evidence from clinical trials, and hard clinical data as we make decisions, but doing that for an individual patient requires judgment.
Sometimes we wear a mask of certainty, often an arrogant one: “Pay no attention to the man behind the curtain. I am the all powerful Wizard of Oz!” We all know a colleague who is “sometimes wrong but always certain.”
A better approach is to recognize uncertainty and to become familiar with the ways that human decision makers adapt to it, as I’ve discussed in previous posts. In doing so, we replace arrogance with confidence and develop a greater sense of comfort as we struggle to provide the best care for our patients.
Is the optimal treatment for Betty uncertain?
When you are uncertain about something in your care of a patient, do you discuss it with the patient?