Welcome, Clinical Problem Solvers, to our first post in the “Clinical Reasoning Corner,” where we will take a deeper dive into key concepts that shape how we think through cases.
The “Clinical Reasoning Corner” will provide an introduction to the topics that keep our reasoning rooted in the core principles of clinical problem solving. We hope this can serve as a practical and accessible resource for when you need a quick refresher or want to prepare your own materials for teaching and discussing clinical reasoning!
For our first post, we are going to talk about two concepts that help us decide whether we treat, test for, or toss specific diagnoses (i.e., moving them much lower on our differential): pre and posttest probabilities.
Let’s dive in!
Understanding how specific tests alter the relative probabilities of our diagnostic considerations is an important aspect of clinical reasoning. In order to do that, we need a mental map of how likely different diseases are and how those probabilities shift as we gather and synthesize data.
Pre and posttest probabilities provide a framework that helps us decide which diagnostic tests to send, interpret their results, and prioritize our differential.
Let’s practice with a case:
You are called to admit a 72 year-old woman with hypertension, diabetes, and knee replacement seven days prior who presents with acute, pleuritic chest pain and dyspnea.
Each of the diagnoses you’re considering fall somewhere on a spectrum of probability, from unlikely to
likely, based on her:
(1) Clinical presentation
(2) Risk factors, and
(3) Base rate of the disease
Your pretest probability is the probability you give to each of these potential diagnoses being the correct diagnosis before you get more data.
On exam, she is afebrile, tachycardic, and normotensive. Her lungs are clear, you notice swelling of her left lower extremity. Initial workup is notable for a WBC count of 8.2, an EKG with sinus tachycardia without PR depression or ST segment changes, and a normal CXR.
How, if at all, has this data changed the way you’re thinking about the case?
This new framing of diagnostic likelihood is your posttest probability, or the probability of each diagnosis after initial diagnostic tests have returned.
Diagnostic reasoning is an iterative process over multiple rounds of testing. This posttest probability from your first aliquot of data will become your pretest probability for the next round of testing.
When considering the probability that a given disease is causing our patient’s symptoms, there are essentially three potential outcomes for that possible diagnosis.
It should be noted, however, that the exact thresholds for “toss, test, and treat” (the “3 T’s”) change based on factors beyond probability alone, such as the morbidity of the disease (e.g., we’re more likely to test for a “can’t miss” diagnosis even if our pretest probability is very low), the morbidity of the treatment (e.g. we have a higher threshold of diagnostic certainty to start toxic chemotherapy for cancer than we do antibiotics for possible infection), and patient preferences. There are certainly other factors we haven’t captured that influence these decisions. Let us know what else alters your diagnostic thresholds on Twitter or via email at firstname.lastname@example.org. We love to hear from you!
This framework of the “3 T’s” correlates with increasing probabilities of the disease, as you can see in the graphic below, inspired by Dr. Catherine Lucey’s Clinical Reasoning Series.
Here is a sample representation of how our probabilities for some diseases have changed as we gathered data.
Now, we’re left to grapple with the question of whether or not she has a pulmonary embolism.
When PE is on the differential, we can use a CT pulmonary angiogram (CTPA) or a D-dimer to adjust its probability. The utility of each test is a product of the test characteristics and the clinical scenario.
When positive, a CTPA allows us to make a diagnosis of pulmonary embolism with a relatively high degree of confidence, and, when negative, allows us to rule it out with nearly equal certainty.
A D-dimer can alter the probability of a PE and the need for further testing depending on your pretest probability and the test results.
In a patient with a low pretest probability, a negative D-dimer significantly decreases the probability of a PE, and effectively rules it out. A positive D-dimer in a low risk patient may not sway our pretest probability enough to warrant a CTPA. Treating the alternative diagnosis and studying the response to therapy is often used as a diagnostic tool. If our patient gets better, it further reassures us they likely do not have a PE. If they do not improve, we have to revisit the possibility of a PE and reconsider obtaining a CTPA.
The utility of a D-dimer in patients with intermediate risk is controversial. A negative D-dimer generally excludes the diagnosis of a PE in intermediate risk patients. However, some patients may still warrant a CTPA, including those with limited cardiopulmonary reserve (i.e., those in whom a PE would have high morbidity) or those in the upper ranges of intermediate risk based on risk stratifying systems, such as the Geneva or Wells Score. A positive D-dimer in an intermediate risk patient generally warrants a CTPA. However, as above, if there is a single, convincing, alternative diagnosis of similar probability, we can use the treatment of that alternative diagnosis as a diagnostic tool.
When our pretest probability is high, a negative D-dimer is not enough to rule out a PE, and we need a CTPA to make or exclude the diagnosis. As with the “3T’s” above, there is so much ambiguity in this topic, so we would love your input on how these diagnostic tests influence your clinical reasoning. Let us know on Twitter or via email at email@example.com!
This leads us to a key concept of how pretest probabilities guide diagnostic testing:
To recap, we have an elderly woman with recent orthopedic surgery presenting with acute onset pleuritic chest pain and shortness of breath and found to have asymmetric lower extremity swelling, sinus tachycardia, and a clear CXR.
Whether we use clinical decision tools or our own judgment, our pretest probability for PE is quite high. So, we bypass a D-dimer and go straight to the CTPA, which shows a segmental pulmonary embolism. You initiate anticoagulation and she recovers well.
That covers it, Clinical Problem Solvers! To recap, today we reviewed that:
What tools can we use to determine how much or how little positive and negative test results influence our pretest probability?
Stay tuned for the next blog post to get the answer!
Check out this tweetorial