Skip to Main Content

Reasoning Rules

February 08, 2020
by Mark David Siegel

Hi everyone,

Medical deities dwell among us. You know them. They’ve been around for decades. They haven’t just read the books; they’ve written them. They’ve seen it all before, even the most arcane disorders. They make obscure diagnoses instantly. They know the answers before most of us start thinking.

Watching master diagnosticians work is both inspiring and frustrating: inspiring because we hope to think like them some day, frustrating because those days are a long way off. When we begin our careers, we know the first diagnoses that enter our heads are often wrong, and we want to avoid premature closure. But even masters can fall into cognitive traps, so we should always check ourselves, using deliberate reasoning rules. Consider these five:

Rule #1: Summarize the salient data: Do you ever feel like you’re drowning in a flood of details? If so, a succinct summary can be your lifeline, pulling you past whirlpools of meaningless details, towards the safe shore of useful, manageable information.

The most helpful findings are specific, severe, and reliable. Patchy erythema is probably non-specific, while palpable purpura points to a diagnosis. A temperature of 99.9oF doesn’t tell you much, but a temperature of 103.5oF can’t be ignored. An elevated JVD is a crucial insight, but only if the examiner knows how to find it.

After deciding which data need your attention, create a pithy summary like this: “In summary, this is a middle aged man with advanced HIV, presenting with one week of progressive dyspnea on exertion, non-productive cough, fever, tachypnea, hypoxemia, diffuse pulmonary crackles, a high LDH, and bilateral infiltrates on the chest x-ray. Now you can get to work.

Rule #2: Decide how many problems to address: Classically, we seek to fit Occam’s Razor, identifying one diagnosis that explains everything. An all-encompassing diagnosis is elegant and satisfying when it works, which it often does for otherwise healthy patients who present acutely. Consider a young woman who suddenly develops right leg swelling, chest pain, and shortness of breath after a 12-hour plane flight. One diagnosis, pulmonary embolism, probably explains everything.

In many of our patients, following Hickam’s Dictum (“a man can have as many diseases as he damn well pleases”) is more reliable. Patients with chronic disorders and those who haven’t seen a physician for a long time often present with multiple problems simultaneously, sometimes in parallel, sometimes in sequence. The patient’s polyuria and polydipsia may be due to diabetes, the peripheral edema due to chronic kidney disease or heart failure, and the throbbing big toe due to gout. These problems are related but need to be addressed separately.

Be sure to distinguish chronic and acute problems. Sometimes chronic diseases predispose patients to acute complications. For example, consider a patient with chronic alcohol use disorder who presents with weight loss, jaundice, fever and abdominal pain. Your exam reveals gynecomastia, spider angiomas, and a nodular liver, which point to a chronic problem- cirrhosis. You also discover ascites and bacteremia, which point to an acute problem- SBP. In these cases, one differential diagnosis isn’t enough; you need to generate two. Other examples include chronic HIV leading to opportunistic infections, COPD leading to exacerbations, and cardiomyopathies leading to heart failure.

Finally, avoid false dichotomies. Patients don’t have septic shock or pneumonia. Rather, the septic shock is caused by the pneumonia.

Rule #3: Simplify: We make our work too hard by generating differentials for multi-component syndromes like fever, headache, rash, and thrombocytopenia. It’s much easier to create differential diagnoses for each element, following illness scripts we can remember, and then see how they come together. In this example, it’s easy to generate an illness script for each finding, which brings us to a unifying diagnosis- the fever signifies disseminated infection, the headache is due to meningitis, the rash is purpura fulminans, and the thrombocytopenia is caused by DIC. The patient has meningococcal meningitis.

Rule #4: Get quantitative: Before ordering tests, ask yourself how concerned you are about each diagnosis on your list. You don’t want to subject the patient to the expense and discomfort of unnecessary testing, and you should only order tests after considering how you will use the results.

Restrict yourself to tests that will address the most likely diagnoses and the ones you “can’t miss.” For a patient with fever, cough, rusty sputum, and crackles in the left lower lobe, you should get a chest-ray to diagnose pneumonia; you don’t need a CT to rule out PE, because PEs don’t present this way. However, if the patient also has metastatic cancer and his right leg is swollen, you should get the CT because PE is a real possibility. You don’t want to miss it.

Thinking quantitatively will also help you interpret results correctly. How will a test move you towards or away from a diagnosis? Consider a 35-year-old woman with atypical chest pain. She probably isn’t having an MI, so if her EKG is normal and her troponin is negative, you can probably stop worrying about coronary artery disease. On the other hand, if your patient is 70 with classic substernal chest pain, a normal EKG and negative troponin don’t come close to ruling out heart disease. You can only interpret results correctly, and act accordingly, after considering pre-test probabilities.

Rule #5: Explain everything: You need to explain all your patient’s findings. We can commit huge errors by ignoring, rationalizing, or overlooking salient data (see Rule #1). Does your diagnosis of constipation explain the high anion gap? Does your diagnosis of cellulitis explain why the patient needs oxygen?

Avoid writing off chronic, unexplained abnormalities. Maybe you’ll be the one to finally crack the case. Did you notice that your heart failure patient also has a chronic sensory neuropathy, anemia, and an elevated creatinine? Has anyone considered amyloidosis? Should we do a biopsy? Don’t pass off unexplained findings to future workups that may never happen. Your patient’s problems are your problems. Address everything, and if you can’t solve every problem immediately, arrange follow up. You may be the only one standing between your patient and a tragic oversight. Your work isn’t done until every finding is explained.

I’m sure we can identify more clinical reasoning rules, but I think these five are key. Even when diagnoses come to you in a flash, it pays to slow down. Take a moment to summarize the salient data, decide how many problems to address, simplify the questions, think quantitatively, and ask yourself if you’ve covered everything. After that, you can close the chart and move on to your next challenge.

And with that, I’ll close this note and move on to finishing the rank lists.


Submitted by Mark David Siegel on February 09, 2020