Columns, Patient Care

Sensitivity and Specificity as Guides to Clinical Decision Making

Issue 12 and Volume 42.

From headache to helping hand

Many of us get headaches and go cross-eyed when we see statistical concepts like sensitivity and specificity in print. Usually, these appear in articles about a diagnostic test’s ability to tell you when a condition or disease state is present (i.e., sensitivity) or, conversely, to tell you when that condition or disease state isn’t there (i.e., specificity).

An initial 12-lead ECG, for example, has a sensitivity of about 68% for diagnosing an acute ST-elevation myocardial infarction (STEMI).

That means that a little more than two-thirds of patients who eventually rule-in for MIs are going to have ST elevations on their initial ECGs.

What this translates to on the street is that a nondiagnostic ECG (no ST elevations) probably shouldn’t be used to withhold aspirin in a suspected MI, or to support a patient’s decision to refuse transport; otherwise, there’s a reasonable chance that you and your patient are going to get into trouble about one-third of the time (32%).

On the flip side, an ECG’s ability to detect a STEMI has a specificity of about 97%. In other words, if the patient isn’t having a STEMI, then there’s only a very small chance that they’re going to have ST elevations that look like they’re having one (3%).

What this all means is that a 12-lead ECG is a good way to rule-in an MI when ST-elevations are present (68%), but it’s a bad way to rule one out when those changes aren’t (32%). There’s also a small chance that you might see ST-elevations when the patient isn’t actually having a STEMI (3%).

Fortunately, we’re all trained to rely on more than just the 12-lead ECG. We also base our clinical decision-making on patient history and risk factors, review of systems and physical exam.

Taking It to the Street

Translating sensitivity and specificity from the more abstract realm of statistics or research to the reality of the street can provide some unexpected benefits. These concepts not only help us to more appropriately interpret ECGs, but they may also give us a more accurate perspective on our histories and physical exams.

When we arrive on scene, our patients aren’t usually standing there with signs saying, “Hey, I’m having a pulmonary embolism,” or, “I think this pleuritic chest pain and shortness of breath must be a spontaneous pneumothorax.”

Instead, all we really have to rely on is our history, our stethoscope and a few diagnostic tests to narrow down the differential diagnosis and figure out what’s going on. To make things even more complicated, all this must happen while we’re managing a complex scene and trying to keep our “task times” under the radar.

Hanging this all together is a little like being a judge in court. We have evidence and it’s our job to weigh it-either to make a treatment decision or appropriately advise a patient of the risks and consequences of refusing transport.

Putting It to the Test

By keeping a file on the sensitivity and specificity of history and physical exam findings in the backs of our heads, on-scene clinical decision-making becomes more evidence-based, efficient and accurate.

For example, tachypnea is typically thought of as one of the hallmark presenting symptoms of pulmonary embolism (PE). Unfortunately, the sensitivity and specificity of a respiratory rate greater than 20 are 53% and 57%, respectively.

What that means is that a little more than half of patients with a PE will present with elevated respiratory rates (53%) and, conversely, a little more than half who don’t have PEs will also be tachypneic.

We could probably flip a coin instead of counting the respiratory rates in these patients: “Heads they have a PE, tails they don’t.”

The point, though, is not to get cynical, but instead to use this information wisely. If we use the absence of tachypnea to rule out a PE, then we’re going to miss the boat in half our cases.

Similarly, most cases of PE are thought to arise from deep venous thromboses (DVT) of the lower extremity. The sensitivity and specificity of the physical exam for detecting calf or thigh swelling, erythema, edema, tenderness or a palpable cord are 47% and 77%, respectively-again, a virtual coin toss.

If you use the absence of these signs to rule out a DVT, you’re going to miss one in about half of cases. On the other hand, a little more than a quarter of your patients who do have these findings won’t actually have DVTs or PEs.

Coin Tosses are OK

There’s nothing wrong with a little uncertainty, as long as you know that’s what you’re dealing with.

No individual symptom, sign or test should guide our clinical decision-making, but an understanding of the sensitivity and specificity of these diagnostic elements can turn what would otherwise be a coin toss into a more sophisticated and effective approach to critical thinking.

References

1. Ioannidis JP, Salem D, Chew PW et al. Accuracy and clinical effect of out-of-hospital electrocardiography in the diagnosis of acute cardiac ischemia: a meta-analysis. Ann Emerg Med. 2001;37:461-470.

2. Stein PD, Beemath A, Matta F et al. Clinical characteristics of patients with acute pulmonary embolism: data from PIOPED II. Am J Med. 2007;120(10):871. 

Learn more from Neal Richmond at the EMS Today Conference in Charlotte, N.C., on Feb. 21-23, 2018. EMSToday.com