How to assimilate information accurately and why aortic dissections and PEs get missed- the pretest probability

Medical school and medical training teaches us that we do tests to confirm the presence or absence of disease. This is the wrong way to think about things. A better concept is to realise that we start with a certain pre-test probability of a disease, which is determined by the base rates of that disease in the population and the patient’s clinical history. Tests can only ever modify this pre-test probability into becoming more or less likely. At a certain point the disease may become so unlikely that testing for it causes more harm than good. This greater harm may come from radiation, reactions to things such as contrast dyes, harmful therapy that might be initiated as a result of a false positive result e.g. antibiotics for a blood culture result that is a contaminant, or simply the fact that time is wasted not pursuing the most likely diagnosis. Other times the disease remains so likely that you may have to pursue repeat testing (take for example the high false-negative rate of COVID swabs).

Consider this scenario. You are the on call house officer. You get paged to the ward to review a 35 year old patient who is having abdominal pain. He was admitted 6 hours ago with severe central chest pain that came on over a matter of seconds and lasted 2 hours. His troponins and ECG have been normal. He has now developed abdominal pain of the same severity and also reaching its peak over a matter of seconds. Concerned about the possibility of aortic dissection you look for mediastinal widening on the chest Xray, pulse defecits, or any neurological symptoms as you know these are the things to look for in a dissection. None of these things are present. Satisfied, you order further ECGs and troponins. The next day you find out he died overnight of an aortic dissection. The next day your consultant tells you “it just shows you how useless clinical exam findings are for aortic dissection- you can’t rely on them. Most dissections have a normal Xray!”

Is this correct? Are these clinical exam findings useless? Is the chest Xray normal in most dissections, as commonly quoted? Well, not quite. They are actually reasonably good tests, including the chest Xray (1,2). The problem is not taking into account the pre-test probability of an aortic dissection, which in this case is high based on the clinical history.

Aortic dissection is a rare disease. About 0.3% of all chest pain presentations to ED will have one. Therefore if you assume that a patient in front of you has the characteristics of an average patient presenting to the ED with chest pain, their starting probability is 0.3%. If they have no pulse or neurological defecits and a normal Xray their probability of dissection falls to 0.08% based on the sensitivity and specificity of these findings (you can use an online Bayesian calculator to calculate this). Most people would consider this below the threshold where testing does more harm than good. Over time without progression of symptoms and with the elucidation of an alternative diagnosis this will fall even further.

Our man is not your typical ED patient with chest pain. His history reveals pain that has a lot of the features of aortic dissection- severe, reaching maximal intensity within seconds, and migratory. He also has no clear alternate diagnosis at present. Therefore his starting probability of dissection is much higher than 0.3%. No amount of negative exam or chest Xray findings will reduce his probability to the point where we don’t need to do a CT of his aorta.

So it’s not really that the test is useless. It’s that we haven’t interpreted the test in the context of the clinical history or pre-test probability.

Consider the day 3 post-operative patient who has an isolated tachycardia. Does he need a CTPA to look for a PE? And does the fact he is on DVT prophylaxis change this assessment? It seems that people on prophylaxis develop PE at a rate of about 2-5% and those not on it have PE at about 10-20%- we can take these as the starting probabilities. Someone on DVT prophylaxis without any other features of PE has a low chance (it will drop below 2-5%) and will probably do better with a watch and wait approach (the discovery of an alternate diagnosis with the benefit of time drops our probability of PE dramatically). In contrast, someone not on prophylaxis should probably just get the scan- even if their probability went from 20% to 10% (extreme and unlikely), a chance of PE of 10% is still too large to sit on. Of course every situation is unique and this is why all these cases should be discussed with your seniors!

In contrast, if your patient has tachycardia, chest pain and hemoptysis, it really doesn’t matter whether they received prophylaxis or not- you need to look for PE either way.

Finally, the most common cause of serious pathology in your chest pain patients will be cardiac ischemia (because ischemic heart disease is common). So even a patient with negative troponins and ECG should be considered to have ischemic heart disease if their chest pain sounds ischemic. Ischemic chest pain can present without a troponin rise if it lasts for an hour or less (a rough estimation- generally cardiac ischemia lasting for hours on end should produce a troponin rise). In contrast you can be more reassured in the patient who has atypical chest pain, minimal risk factors, and a normal troponin and ECG. Once again, it’s all about the pre-test probability.

This is a guide to basic diagnostic reasoning- again remember that this stuff is complex and you should always ask for help when you need it!


  1. Meta-Analysis JAMA. 2002 May 1;287(17):2262-72. doi: 10.1001/jama.287.17.2262. Does this patient have an acute thoracic aortic dissection? Michael Klompas 1
  2. Meta-Analysis Acad Emerg Med. 2018 Apr;25(4):397-412. doi: 10.1111/acem.13360. Epub 2018 Jan 24. Clinical Examination for Acute Aortic Dissection: A Systematic Review and Meta-analysis. Robert Ohle 1, Hashim Khaliq Kareemi 2, George Wells 3 4, Jeffrey J Perry 4

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s