Puzzling out the PERC rule

aka  016

The intern from Hematology Hoodwinker 001 is stuck with another dilemma. Needless to say, he has another patient to discuss with you. The patient is a 26 year-old female with mild shortness of breath and intermittent right-sided chest discomfort that is vaguely pleuritic in nature. Her chest x-ray is normal, as are her vital signs.

Because his head is still spinning from yesterday’s ‘D-dimer talk’, the intern decides to talk you about how to rule out a pulmonary embolism in this patient. You take a deep breath, crack your knuckles, and prepare yourself for the ‘PERC rule’ talk….


Q1. What is the PERC rule?

PERC = Pulmonary Embolism Rule-out Criteria

Pulmonary embolism can be ruled out clinically if none of the 8 PERC criteria are present in a patient with a low pretest probability of PE (e.g. Wells PE CPG score of <3) that is consistent with the ‘gestalt‘ of an experienced physician:

  • age < 50 years
  • pulse < 100 beats min
  • SaO2 >or= 95%
  • no hemoptysis
  • no estrogen use
  • no surgery/trauma requiring hospitalization within 4 weeks
  • no prior venous thromboembolism (VTE)
  • no unilateral leg swelling

Can’t remember all of this?

Try this mnemonic suggested by the guys at Keeping Up in Emergency Medicine to remember which patients the PERC rule can’t be used on:


Hormone, Age >50, DVT/PE history, Coughing blood, Leg swelling, O2 >95%, Tachycardia 100+, Surgery/trauma <28 d

Alternatively why waste valuable brain space? Use @grahamwalker‘s MDCalc.com online and/or download Tschopp and Pfiffner’s MedCalc for free onto your smart phone.

Q2. In which patients may the PERC rule be unreliable?

The original ‘PERC rule’ study by Kine et al (2004) excluded patients with these characteristics:

  • patients in whom shortness of breath is not the most important, or equally most important, presenting complaint
  • cancer
  • thrombophilia
  • strong family history of thrombophilia
  • beta blockers that may mask tachycardia
  • patients with transient tachycardia
  • patients with amputations
  • patients who are massively obese and in whom leg swelling cannot be reliably ascertained
  • with baseline hypoxemia in whom a pulse oximetry reading <95% is long-standing

However, since the derivation study there has been two validation studies. One of the validation studies (Kline et al, 2008) was a multi-center study that included patients with a primary complaint of shortness of breath or chest pain. Thus, I think that the PERC rule can be safely applied to patients with chest pain as the primary complaint.

Kline et al (2004), the creators of the PERC rule, chose 1.8% as the point of equipoise between (1) the benefits and risks of further investigations for PE, and (2) the benefits and risks of not investigating further.

Q3. What does this mean?

Equipoise is reached when the risk-benefit ratio of one course of action is balanced by the risk-benefit ratio of an alternative course of action. When deciding to investigate further for PE, this point is reached when the pre-test probability of PE is 1.8%.

Kline et al (2004) calculated this number based on a formula that incorporated the risks from CT angiography:

  • cancer from radiation exposure
  • anaphylaxis or severe pulmonary edema requiring intubation
  • requiring hemodialysis from the dye

as well as the risk of death from missing a PE (defined as a 20% risk reduction for detected vs. missed).

So, to use the PERC rule you have to choose the right patient to apply it too, and then accept a miss rate of 1.8%.

According to Carpenter et al (2008), the negative likelihood ratio for the PERC rule is 0.17 (95%CI 0.11—0.25).

Q4. What is the maximum pre-test probability of VTE such that the post-test probability (after correctly applying the PERC rule) is 1.8% or less?

There are a few way of working this out.

1. Use the Fagan nomogram (the link is to an interactive version — click here for an old-school non-interactive version)

2. Do the math…

Step a. convert post-test probability to post-test odds

post-test odds = probability / (1 – probability) = 0.018 / 0.982 = 0.018

Step b. Divide post-test odds by the negative likelihood ratio to get pre-test odds

pre-test odds = post-test odds / LR = 0.018 / 0.17 = 0.106

Step c. Convert pre-test odds to pre-test probability

pre-test probability = odds / (1 + odds) = 0.106 / 1.106 = 0.096

Thus the PERC rule can be applied to patients with a pre-test probability of about 10%
(this decreases to about 7% if a negative likelihood ratio of 0.25 is used, which is the upper limit of the 95% confidence interval)

3. Back calculate an estimate using MedCalc on your smart phone.

Note that when using the Well’s criteria for PE, low risk (score 0 or 1) is a probability of 1.3% and moderate risk (score 2 to 6) is a probability of 16.2%. Thus the PERC can be safely applied to low risk patients, but not moderate risk patients when using the Well’s criteria for PE.

Q5. Is inappropriate use of the D-dimer assay a problem in medicine?


As an example consider the paper by Durieux et al (2001). This describes a single centre prospective study of 168 patients to analyze D-dimer usage after an educational intervention to implement guidelines for the diagnosis of PE. Despite the intervention 20% of the D-dimer requests were inappropriate and 45% of those with a positive D-dimer test had no imaging procedures performed.

Whether it is being used appropriately or not, over half of D-dimer positive tests are false positives. Hence there is a definite need for something like the PERC rule.

Q6. What problems can arise from inappropriate use of the D-dimer test?

Problems from over-use of the D-dimer test include:

  • Unnecessary cost of D-dimer test
  • Unnecessary delays in investigation, management or disposition
  • Unnecessary further imaging (if PERC rule would have prevented a false-positive D-dimer result from being obtained)
    • Radiation exposure from CTPA or CT venography
    • Risk of contrast nephropathy
    • Diversion of focus away from other more likely causes

Problems from under-use of the D-dimer test include:

  • Unnecessary further imaging (if negative D-dimer would have precluded further investigation)
    • Radiation exposure from CTPA or CT venography
    • Risk of contrast nephropathy
  • Missed diagnosis of venous thromboembolism resulting in morbidity and mortality

Q7. What should you do if a patient has a positive D-dimer test, but on reflection the investigation was not indicated — for example, the patient meets the PERC rule criteria, which is consistent with a senior clinician’s ‘gestalt’, but a D-dimer test has already been performed?

This is what I suggest:

  • Ignore the result with respect to assessing VTE risk (go back and use the Wells criteria for PE!).
  • Consider the other possible causes of an elevated D-dimer and manage as indicated (this may include no further action).


  • Carpenter CR, et al (2009). Differentiating low-risk and no-risk PE patients: the PERC score. J Emerg Med, 36 (3), 317-22 PMID: 19097732
  • Durieux P, et al. D-dimer testing as the initial test for suspected pulmonary embolism. Appropriateness of prescription and physician compliance to guidelines. Thromb Res 2001;101:261–6 PMID: 11248287
  • Kline JA, et al. Clinical criteria to prevent unnecessary diagnostic testing in emergency department patients with suspected pulmonary embolism. J Thromb Haemost. Aug 2004;2(8):1247-1255. PMID: 15304025
  • Kline JA, et al. Prospective multicenter evaluation of the pulmonary embolism rule-out criteria. J Thromb Haem 2008 May;6(5):772-80. Epub 2008 Mar 3. PMID: 18318689
  • Fields JM, Goyal M. Venothromboembolism. Emerg Med Clin North Am. Aug 2008;26(3):649-683, viii. PMID: 18655939
  • Wolf SJ, et al. Assessment of the pulmonary embolism rule-out criteria rule for evaluation of suspected pulmonary embolism in the emergency department. Am J Emerg Med. Feb 2008;26(2):181-185. PMID: 18272098
Print Friendly


  1. says

    Master Nickson. Fantastic Post! We need to speak about the scary implications of this study: 21091866 and whether or not it casts doubt on the use of the PERC. Dr. Newman of SMARTEM fame and I will also be debating the validity of his contention that a low risk Well’s score can be used as a Rule-Out criteria.

  2. Andrew Leonard says

    Nice post and comment. There’s such a lot of overlap between Wells and PERC (DVT, haemoptysis, surgery, signs of DVT, tachcardia are common to both scores) that you wouldn’t see many patients in a year that could tick all 8 boxes in the PERC rule and be anything other than low probability by Wells’ score.

    Which brings us to the Wells: I’m uneasy about the clinical judgement element of the Wells, and also uneasy about the low PE rate in his studies -- 9.2% (86/930) in this one Ann Intern Med. 2001;135:98-107. To my eyes this is too low to reflect real-world (at least, in the UK) practice and suggests that the studies included patients who pretty obviously didn’t have PE -- and perhaps it’s therefore not that surprising that the score performs well at identifying them. I suspect the calculated negative likelihood ratios reported by Wells would be optimistic if applied to general UK practice, and certainly other groups haven’t been able to get near them (eg Am J Med 2002;113:269-75) -- and neither have I in my own practice. The revised Geneva score derivation study was rather more realistic with 222/965 patients having PE (23%). We knowacross large populations that Wells’ (in experienced hands) and Geneva scores correctly identify similar proportions in the various risk categories, but that concordance between the scores is not perfect -- and I suspect that this, combined with a much higher (ie European) PE rate in Aujesky’s dataset is why he found in study 21091866 that PERC after low probability Geneva was an inadequate stratagem.

    I read this blog unaware of study 21091866 and thinking that the PERC rule is OK to use irrespective of clinical probability, in that for PERC to be applicable the probability would always be low. But now I’m right behind you: PERC rule should only to be used in patients with clinical probability <10%. Now I'm going to go to sleep and worry about the circular nature of the argument, given the Wells/PERC overlap problem.

  3. says

    Thanks for taking the time to make such a detailed comment Andrew.

    Considerable discussion took place on the emcrit website on the issues surrounding the diagnosis of PE (and the PERC rule in particular) after this post was originally written.

    At our institution in Australia we have found that the prevalence of PE in low risk Well’s score patients have a sufficiently low likelihood of PE that the PERC rule can be applied (based on follow up studies).
    The fairly recent Hugli 2011 study [PMID: 21091866] failed to validate the PERC rule in low risk patients identified using the revised Geneva score. I suspect a major reason for this was that clinician gestalt was not used, only the other elements of the PERC rule. Clinician gestalt is likely to be the most important part of the PERC rule! I see the rule as way of saying: “yeah, PE could be in the differential for this, but come on, let’s get real” thus allowing you to avoid going further along the PE pathway. I suspect that in Europe most patients investigated for PE have higher pretest probabilities because European clinicians are less likely to chase unlikely diagnoses in the first place (e.g. less defensive medicine) and wouldn’t think of doing a d-dimer on a very low risk patient anyway.


    Its all very interesting and perpetually puzzling…!

  4. swiftdriver says

    Still can’t figure out PERC. The patient must first be assigned a low pre-test probability, based on what? I’d base it on the kind of things found in the PERC rule, but then, you couldn’t use the PERC, unless the PERC was already negative.