Incorporating Diagnostic Testing Into Your Clinical Decisions

Author: Jesus Trevino, MD (EM Resident Physician, PGY-1, NUEM) // Edited by: Andrew Ketterer, (EM Resident Physician, PGY-3, NUEM) // Expert Commentary: Mark Courtney, MD

Citation: [Peer-Reviewed, Web Publication] Trevino J, Ketterer A (2016, June 7). Incorporating Diagnostic Testing Into Your Clinical Decisions [NUEM Blog. Expert Commentary by Courtney M]. Retrieved from

Clinical context

Meet Mr. PE, a 45 year old male with history of renal cell carcinoma, now POD#7 status post partial nephrectomy, who presents with shortness of breath. Among other things, you worry about a pulmonary embolus and are one click away from ordering a CT angiogram when you see that his creatinine is elevated with a GFR < 30. Foiled! As it dawns on you that a V/Q scan is your next diagnostic tool, your head starts to spin as you try to anticipate how a low vs. intermediate vs. high probability result will affect your management of Mr. PE.

How should diagnostic testing influence clinical management?

At the conclusion of the initial clinical assessment, one produces a suspicion for a particular disease process (i.e. pre-test probability). In cases where this suspicion is equivocal, diagnostic studies are ordered to provide additional information (i.e. increase or decrease the likelihood of disease) and ultimately adjust your prior belief with an updated suspicion for disease (i.e. post-test probability).

This process of honing your clinical suspicion can be represented mathematically as:

Pre-test odds * Likelihood ratio = post-test odds

What the what?!? No need to fret, this blog entry will explore the impact of diagnostic testing on disease likelihood by 1) highlighting the concepts behind each term of the above equation, and 2) applying these principles to Mr. PE and his V/Q scan results.

Pre-test probability

Pre-test probability (Ppre) represents the chance that a particular disease process exists upon your initial assessment based on readily available information. How do you estimate this probability? This depends on the amount of information you have about your patient. In the case of aortic dissection, any comer with atraumatic chest pain presenting to the ED would have a pre-test probability of 0.102% (point A on the probability spectrum below) based on a retrospective study of the prevalence of aortic dissection in New York and New Jersey ED encounters [1].


However, if any comer were to present with chest pain and pulse deficits, this probability would intuitively be higher; as a matter of fact, this scenario would have a pre-test probability of 27% (point B) as suggested by the presenting features documented in the International Registry of Acute Aortic Dissection [2]. 

The take away is that Ppre is your suspicion for a disease based on information that you glean from the first-pass assessment. As a side note, you probably noticed the “test” and “treat” points on the line as well. While the locations of these points will vary from one situation to another, having a Ppre higher than the “treat” mark means that you should treat the patient for the suspected disease without further testing; having a Ppre below the “test” mark means that your suspicion is low enough that further testing is unnecessary. The money from a diagnostic perspective is the area between these two points, namely where point B lies above. Realistically, you’re not going to draw this out for every possible test for each diagnosis in your differential, but the point is that if your suspicion is sufficiently low for a disease, you probably don’t need to test for it; if your suspicion is sufficiently high you should probably just treat the patient.

Likelihood ratio

Equipped with Ppre, it’s time to consider the utility of certain diagnostic tests to achieve your goal of diagnosing or ruling out a disease. The “utility” of tests can be viewed from several perspectives; the view we will cover is the likelihood ratio (LR).

Each test is associated with a positive and negative LR that is a function of the sensitivity and specificity of that test:

Here are two examples of LRs associated with diagnostic imaging in thoracic aortic dissection:

For LR+(-), the higher (lower) the ratio, the greater the impact the test outcome will have on Ppre. In other words, tests with LR+ > 10 and/or LR- < 0.1 are considered useful tests for significantly revising Ppre. Therefore we can categorize CXR mediastinal and aortic root abnormalities as decent for ruling out thoracic aortic dissection but poor for ruling in, whereas CTa is excellent in both regards.

Post-test probability

Lastly, determining post-test probability (Ppost) from Ppre and LR+/- takes straightforward multiplication. Technically, this requires a few extra steps where the pre-test probability is converted to pre-test odds, then multiplied by LR to obtain post-test odds, and finally conversion of the post-test odds to a post-test probability.

    In the presentation of chest pain with pulse deficit and concern for aortic dissection (Ppre = 0.27), a negative chest x-ray would decrease Ppost to 0.10 whereas a negative CTa would decrease Ppost to 0.004. This demonstrates the importance of selecting the right diagnostic tool for the presentation at hand: with such a high Ppre, a negative chest x-ray yielding a Ppost of 0.10 in conjunction with the morbidity associated with aortic dissection would not be sufficient to rule out this disease, so unless you can get a portable plain film within seconds, you should forego x-ray and rush the patient over to CT. On the other hand, let’s say you had a patient with solitary chest pain (Ppre = 0.00102) and you were worried about aortic dissection, a negative chest x-ray would result in a Ppost of 0.0003 and approach a satisfactory conclusion to the work-up of aortic dissection.

    An alternative method to determining the Ppost involves the use of the Fagan nomogram.  This nomogram allows one to visually determine the Ppost by drawing a straight line from the point that represents the Ppre (left axis) through the point of the LR (central axis) and finally to the right axis which yields the Ppost. See below for an example of the Fagan nomogram in action in the evaluation of chest pain and pulse deficit with positive CTa:


Created in R with fagan.plot function of TeachingDemos package


Case resolution

Back to Mr. PE and his V/Q scan quandary. Having reviewed the caveats of diagnostic testing, we now appreciate that interpreting the V/Q scan requires knowing a) Ppre for a PE, and b) LRs for V/Q test outcomes.

  • Ppre: Using clinical gestalt, we surmised his Ppre was around 0.40 as he was post-op with a history of malignancy but without chest pain and with a reasonable alternative diagnosis of pulmonary edema.
  • LR: The 1990 PIOPED trial determined sensitivities and specificities of high vs. intermediate vs. low vs. normal probability V/Q scan results using angiography as the gold standard [3]. The table below illustrates the range of potential Ppost pertaining to each V/Q scan result:

Mr. PE’s scan returned an intermediate probability, which was not very helpful in ruling in or ruling out PE (Ppost.= 0.38, essentially the same as our Ppre). In the end, we started therapeutic anticoagulation with heparin as well as gentle diuresis and observed a reduction in his dyspnea over the next two days. Ultimately, he was discharged to home on hospital day 3 with a prescription for Coumadin and recommendations to repeat V/Q scan in 2 weeks to ascertain whether the V/Q scan abnormality resolved after anticoagulation, which would support the diagnosis of a PE.

Expert Commentary

This is an in-depth, mathematical description of pretest and post test probability for two commonly considered, high-stakes, emergency medicine diagnoses: PE and Aortic Dissection. There are 3 major points this well written analysis brings to mind:  

First, what is fascinating about pretest probability is it is something we use all the time in emergency or acute care medicine but often fail to consciously recognize.  These cases above are well thought out, involve conditions for which there are clear tests with published diagnostic accuracy that allow one to calculate LR+ or LR- values, but this is often not the case in many day-to-day situations in care – yet we still employ the same approach above.  When the tech or nurse interrupts me on what is now a q 5 minute basis to show me an ECG I am often at a loss.  Sure I can look for a STEMI…and I guess that is the point….but how “abnormal” the cardiogram is vastly depends on the patient.  My first question is: “What’s going on with this patient in the waiting room?”  I am seeking some input into the pre-test calculator that is my brain.  I may have a much different response if they are 65 years old with prior CABG and indigestion than a 32 year old with tingling and dizziness.  So the point is to recognize the importance of taking EVERY test into context with some component of pretest probability.  This will help when you get that equivocal CT brain read as “can not rule out age indeterminate ischemic changes vs. artifact” in the 38 year old vague dizziness patient referred from urgent care.   Their post-test probability given the exact same CT read is less than for a 70 year old diabetic hypertensive patient with similar symptoms.

Second, sometimes your gestalt pretest probability is just as good as pretest probability scoring systems.  Limited research has been done in this area for a variety of reasons, but it is well known that physicians even when they are aware of standardized pretest probability scoring systems often fail to use them, claim to use them but often don’t document them, can not remember them, or believe their gestalt is just as good or better.  There is some literature to support these views for PE [4,5]. 

Third, integration of pre-test, post-post test approaches to decision making still often must involve the patient.  The increasingly strong move toward involving patients in their decision making with respect to testing choices and treatment choices has been methodically applied to chest pain with leading work being done by emergency medicine investigators [6,7] But it is still critical to take the time to talk with patients about their preferences, listen, provide education as to test/no-test implications and arrive at a plan.  This is challenging.  It is well documented that many patients lack health care literacy.  Many also lack understanding of numeracy.  Having a discussion about probability even with a well educated patient can be difficult yet is at the heart of true shared decision making based on evidence, probability and choice.  More research is ongoing in this area, and hopefully this coupled with pre-test probability tools embedded in electronic medical records that function more as a reflection of care and decision making rather than what is essentially a glorified billing platform, will transform the future of acute care – at least that is my dream.

D. Mark Courtney MD MSCI
Director of Research; Associate Professor; Department of Emergency Medicine Northwestern University, Feinberg School of Medicine [Pubmed]

Related Posts


  1. Alter SM1, Eskin B2, Allegra JR2. Diagnosis of Aortic Dissection in Emergency Department Patients is Rare. West J Emerg Med. 2015 Sep;16(5):629-31.
  2.  Golledge J1, Eagle KA. Acute aortic dissection. Lancet. 2008 Jul 5;372(9632):55-66.

  3.  PIOPED Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism. Results of the prospective investigation of pulmonary embolism diagnosis (PIOPED). JAMA. 1990 May 23-30;263(20):2753-9.

  4. Runyon MS, Webb WB, Jones AE, Kline JA. Comparison of the unstructured clinician estimate of pretest probability for pulmonary embolism to the Canadian score and the Charlotte rule: a prospective observational study. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2005;12(7):587-593. doi:10.1197/j.aem.2005.02.010.

  5. Runyon MS, Richman PB, Kline JA, Unknown. Emergency medicine practitioner knowledge and use of decision rules for the evaluation of patients with suspected pulmonary embolism: variations by practice setting and training level. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2007;14(1):53-57. doi:10.1197/j.aem.2006.07.032.

  6. Hess EP, Grudzen CR, Thomson R, Raja AS, Carpenter CR. Shared Decision-making in the Emergency Department: Respecting Patient Autonomy When Seconds Count. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. June 2015. doi:10.1111/acem.12703.

  7. Hess EP, Brison RJ, Perry JJ, et al. Development of a clinical prediction rule for 30-day cardiac events in emergency department patients with chest pain and possible acute coronary syndrome. Annals of Emergency Medicine. 2012;59(2):115–25.e1. doi:10.1016/j.annemergmed.2011.07.026.