It happens all the time: we build impressive, exhaustive differential diagnoses, and then…we leap to ordering tests- a sloppy practice that exposes a chasm where clinical reasoning belongs.
Here’s an example: “Possible causes of this patient’s chest pain include MI, PE, aortic dissection, pneumothorax, pneumonia, pericarditis, pleuritis, costochondritis, and shingles. We’ll get an EKG, troponins, an echo, a CTA, sputum cultures, an ANA, and zoster serologies.”
On the floors and in teaching conferences, we skip over clinical reasoning and, as a result, we order useless tests for unlikely diagnoses. Who knows why we do this, but I suspect at some point, we came to equate good medicine with extensive testing, ignoring cost and consequence. We should instead focus on likely diagnoses and those too dangerous to miss, letting epidemiology and patient information guide us.
A recent Report was the exception that proves the rule. We heard about a young woman with complete heart block. We’re in peak tick season, so Lyme Disease shot to the top of the list even before we saw pictures of her rash. Her positive Lyme titers confirmed the diagnosis, but no one was surprised. Other possible explanations for heart block—sarcoidosis, ischemia, drugs, amyloidosis, etc.—were appropriately discounted. We should do this more often.
Many patients tell you the diagnosis. “Ripping chest pain” suggests aortic dissection. “Gradually increasing substernal chest pressure” suggests MI. “Sudden pleuritic chest pain and dyspnea” suggests PE. Exam findings like pulse deficits (aortic dissection), new MR murmurs (MI), and unilateral leg swelling (PE) drive you towards pre-test probabilities, which can help you decide which tests to order (if any).
Some validated tools can help you estimate pre-test probability. The Wells Score, Revised Geneva Score, and PERC can help you decide whether to test for PE. More commonly, you’ll need to calculate your own quasi-scores. If a patient has fever, productive cough, hypoxemia, a lung infiltrate, and leukocytosis, you don’t need a calculator to tell you they have pneumonia. Draw blood cultures, check a lactate, and start antibiotics. But don’t get a CTPA; PEs don’t present this way.
I’d love to eliminate unnecessary testing, like ANAs for patients without evidence of lupus and serial troponins for patients with nonischemic chest pain. Unnecessary testing is expensive and distracting, and more likely to yield confusing false positives than true diagnoses. We need to elevate clinical reasoning. So, the next time you’re done building a broad differential diagnosis, stop, think, and prioritize.
Enjoy your Sunday, everyone. I’m driving down to New Jersey to see my mom.
For further reading:
- A classic of clinical reasoning: Cope's Early Diagnosis of the Acute Abdomen
- Some of the best clinical reasoning anywhere: the NEJM’s Interactive Medical Cases and Case Records of the Massachusetts General Hospital
- A great review: Teaching Clinical Reasoning and Critical Thinking: From Cognitive Theory to Practical Application
- An old PD Note: Sets and Venn Diagrams (see attached).