Base Rates: The Most Ignored Number in Emergency Medicine

If you are like the average emergency physician, you’ll see about 3,000 patients a year. In that time, you are unlikely to see one aortic dissection, maybe one or two subarachnoid hemorrhages, and doubtfully a necrotizing soft tissue infection. But you will think about these conditions every time you evaluate the hundreds of patients with chest pain, headache, and cellulitis that you will see.

That mismatch is the environment we practice in. We are trained—and rightly so—to not miss the dangerous diagnoses hiding in that noise. But in doing so, we often overlook a key number in clinical decision-making: the base rate.

The base rate is how common a disease is before you learn anything new about the patient. The base rate for dissection in patients with chest pain - about 1 in 400, spinal epidural abscess in back pain - about 1 in 2,000, mesenteric ischemia in abdominal pain - about 1 in 1,000. The base rate is your starting point. All of the data you collect - each question you ask, your exam findings, every test you order is an attempt to refine that initial probability.

The problem is that we are very good at assigning value to this data, but we don’t always start from a probability that accurately reflects the true likelihood of a condition.

This is not a knowledge gap. It is how we are wired.

We remember the misses. The dissection we almost sent home, the bounceback with a problem that looks so clear in hindsight. We do not remember the thousands of patients who didn’t have the diagnosis - the unremarkable CTs, the normal CRPs, the negative LPs. Because of how we think about them, rare diseases start to feel common.

Moreover, doing something feels safer than doing nothing. Ordering the CT feels like good care. Not ordering it feels like a risk. Base rates often push toward restraint, and restraint is uncomfortable because of the uncertainty that remains.

We are trained to think in worst-case scenarios. Don’t miss this. Rule this out. So we tend to focus on the test itself—its sensitivity, its specificity, whether it will “rule out” disease. But the performance of any test depends on where you start. A good test applied to a very low-risk patient mostly generates noise. And a test avoided in a patient whose risk is underestimated misses the diagnosis entirely

Where it matters most.

When the base rate is low, most testing happens in patients who do not have the disease. This leads to more imaging, more labs, and more downstream testing.

That harm is real. For example, a CT angiogram delivers a meaningful radiation dose, with a small but measurable increase in lifetime cancer risk. Multiply that across thousands of low-risk patients and it is no longer theoretical, someone is getting harmed. It is not just the radiation to account for. Testing also creates answers to questions that were not being asked. Incidental findings show up in imaging and labs that trigger follow-up testing, procedures, and anxiety. Once that cascade starts, it is hard to stop.

And more testing does not just add information. It adds noise. Borderline abnormalities, incidentalomas, conflicting data. All of this increases cognitive load and makes decision-making harder, not easier. More data does not equal more clarity.

Where clinical judgment matters most.

We are responsible for not missing the 0.05% diagnosis. That responsibility is real. But so is this: the other 99.95% of patients are exposed to how we look for it. Good emergency medicine lives in that space. Not in ignoring rare disease. Not in chasing it indiscriminately. But in navigating it calibrated to reality.

Instead of asking “could this be a dissection,” start with “out of 100 patients like this, how many actually have one?” Then ask whether the test you are considering will meaningfully change that number—or if there is a safer, more appropriate step that will get you there.

The point is not to do less. It is about doing the right amount for the right patient for the right reason. You will see thousands of patients this year. And you will see high acuity/low occurrence conditions. You are expected to recognize them. But you are also responsible for how we care for the patients who do not have them.

The highest risk number in emergency medicine is the one that was not properly calibrated.

Of course, none of this is easy in practice. It is much simpler to think about these decisions in hindsight than it is in a busy department, with constant interruptions and a heavy cognitive load. But the more we work to better understand and calibrate how we make decisions, the better we care for our patients.

• Hagan PG et al. The International Registry of Acute Aortic Dissection (IRAD). JAMA.
• Davis DP et al. The clinical presentation and impact of diagnostic delays in spinal epidural abscess. J Emerg Med.
• Crum-Cianflone NF. Spinal epidural abscess: diagnosis and management. Am J Med.
• Kassahun WT et al. Unchanged high mortality rates from acute mesenteric ischemia. J Gastrointest Surg.
• Croskerry P. The importance of cognitive errors in diagnosis and strategies to minimize them. Acad Med.
• Brenner DJ, Hall EJ. Computed tomography — an increasing source of radiation exposure. NEJM.