Falling into the Diagnostic Trap

 

by Danielle Ofri
New York Times

He was sprawled on the floor of his apartment surrounded by empty beer and wine bottles when E.M.S. broke down the door. He was 63, but his elderly mother still kept tabs on him. When he hadn’t answered the phone for three days, she’d called 911.

He was an alcoholic, she told E.M.S., with many admissions to the hospital for “alcohol withdrawal,” but was otherwise healthy.

Working in an inner-city hospital, our medical teams admit so many alcohol withdrawal cases that the treatment is nearly second nature. Basically, all you have to do is treat the shakes with sedatives like Valium until the withdrawal symptoms abate, or the patient ducks out of the hospital for another drink.

Our patient exhibited every sign of alcohol withdrawal – a rapid heart rate (tachycardia), hypertension and tremors. He was awake but lethargic.

He was given intravenous Valium every few hours through the night, but the next day, he still had the shakes, as well as the elevated heart rate and blood pressure. Hard-core alcoholics are known to be “tolerant” of sedatives, so we cranked up the dose.

We continued to increase his Valium dose, and by the end of the day, his shakes were starting to improve, though his heart rate hadn’t yet come down. But he was a textbook case of alcohol withdrawal, we told ourselves, he just needed more sedative.

We were about to leave for the day and about to fall into a trap.

The trap that we nearly fell into is called anchoring bias. The patient was admitted to our team with the diagnosis of alcohol withdrawal. Once we had that label in our minds, we fit everything into that diagnostic box, anchoring all of his symptoms to that diagnosis, even ones that didn’t quite fit.

Just before we left, one of the medical residents reviewed the patient’s vital signs for the previous 24 hours and noted that the heart rate did not drop – even transiently – after each Valium injection. That gave us pause. Was it still alcohol withdrawal, with an “atypical” presentation, or was there something else going on?

Occam’s razor is the principle that one should hunt for the simplest single diagnosis to explain a patient’s condition. Certainly this patient, who reeked of alcohol when he first arrived in the E.R., had every symptom of withdrawal.

But Hickam’s dictum is the opposite philosophy, that patients can have as many illnesses as they damn well please. Maybe there was something else in addition to alcohol withdrawal.

Lots of things cause a rapid heart rate in addition to alcohol intoxication: dehydration, electrolyte abnormalities, hypoglycemia, fever, anxiety, drugs, bleeding, heart attacks, blood clots, infections, hyperthyroidism, even too much coffee.

We’d already hydrated him, corrected his electrolyte problems, checked for infections, heart attacks and thyroid disease. His oxygen saturation was 100 percent, without any supplemental oxygen needed. A CT scan of his head and chest X-ray were normal. Other than the rapid heart rate, so was his EKG.

But then the d-dimer, a nonspecific blood test for clots, was elevated. A stat CT scan of the chest demonstrated clots in both lungs — a life-threatening condition called pulmonary embolus that required immediate treatment with blood thinners.

In the textbooks, patients with pulmonary embolus are short of breath and complain of sharp pain when breathing in deeply. Their oxygen saturation is low, and their EKGs show a classic “right-heart strain pattern.”

Our patient had none of those symptoms, and he had another condition, alcohol withdrawal, that could account for his rapid heart rate.

This was a frightening near miss, a potentially fatal condition almost undiagnosed because of our anchoring bias. Only the astute observation – that the patient’s heart rate was not dipping after the Valium injections – led us to the additional diagnosis.

Anchoring bias is often considered the Achilles’ heel of diagnostic reasoning. It’s as though our brains close ranks around our first impression, then refuse to consider anything else. Once a patient is “billed” as a heart attack, or gastroenteritis, or anxiety, we view every data point through that particular lens.

If the data don’t fit, we tend to assume that it’s merely because the illness is presenting atypically rather than that our diagnosis might be wrong or incomplete. Anchoring bias casts an even longer shadow in today’s shift-oriented medical world, in which patients are serially handed off from one team to another. The label that is attached to them takes on a life of its own.

Once our patient was treated with blood thinners, the clots could no longer grow bigger. As they began to heal, his heart rate normalized. He still needed heavy doses of Valium to ease his alcohol withdrawal symptoms, but those, too, eventually subsided.

For the doctors, this was a harrowing lesson in the trap of anchoring bias. It is so easy to slip into it without even knowing. But this case reminded us to keep reciting the mantra: if something doesn’t fit, don’t try to make it fit. Ask what else might be going on. Don’t fall into the trap.

The patient, though, ended up in another clinical dilemma. He needed to stay on blood thinners to prevent a future clots, but his history of alcohol use made this a risky proposition. A drunken fall while on blood thinners is a recipe for a brain hemorrhage. Without blood thinners, however, his next blood clot could be fatal.

But that’s a treatment quandary, rather than a diagnostic one. Fodder for another discussion.  (From the New York Times)