A Debate on PE Decision Rules

This may be of interest to you EBM Wonks out there.Ā  At one of the resident conferences, I debated David Newman of SMARTEM fame on whether a low risk Wells’ score can be used as a PE rule out criteria. Video and audio quality are only so-so.

You can view it on Vimeo as well


Thanks to Dena Asaad for filming and Reuben Strayer for hosting the video.

You finished the 'cast,
Now get CME credit

Not a subcriber yet? Why the heck not?
By subscribing, you can...

  • Get CME hours
  • Support the show
  • Write it off on your taxes or get reimbursed by your department

Sign Up Today!


Subscribe Now

If you enjoyed this post, you will almost certainly enjoy our others. Subscribe to our email list to keep informed on all of the ED Critical Care goodness.

This Post was by , MD, published 4 years ago. We never spam; we hate spammers! Spammers probably work for the Joint Commission.


  1. says

    Great discussion of not just the investigation of PE, but the philosophy behind when and why we do diagnostic tests in general.
    Had to wait 1h 25min til the elephant in the room was explicitly named: gestalt!
    This is probably the most important part of the PERC rule, and is also what determines whether you score +3 on the Well’s score (i.e. PE Is #1 Diagnosis, or Equally Likely) – think of the unsubtle example of the SOB pt with HR 105 – and who has a known thrombophilia, on OCP, who was a passenger in an 8 hour bus journey before the onset of symptoms, and whose father died of a massive PE… – Wells 1 or 4?
    How this +3 is applied surely drastically effects how different people investigate PE.
    Also, great comments on whether we should even investigate (PE: so what? Is it a ‘bad’ PE or a ‘good’ PE?).
    Well worth the sleep deprivation!

  2. says

    Gosh, I wish you were there Nickson, you are the master of putting this stuff into the proper context. And your accent sounds so much cooler than mine.


  3. Matt Brown says

    I enjoyed the debate. Like everybody apparently, I use the Wells and PERC in concert, but I felt uncomfortable doing so until you pointed out how similar the two were.
    If we accept 2% as an acceptable miss-rate in the ED for PE, then combining two rules (for the sake of simplicity lets say they are both 98% sensitive) that aren’t the same will push your miss-rate over the threshold. The more different they are, the higher it will be pushed, because we’re not applying the second rule to the population it was intended for at all. For example, we’re applying Wells to a population who failed PERC – a much higher risk population. Another way of thinking about it is to imagine that in 2021, we’ve validated 1000 different rules that meet the 2% threshold. If we keep trying, there’s a good chance we’ll find one that excuses any patient from a workup.
    I was making a pavlova so I missed a little of the debate, maybe you did mention this somewhere.
    Fascinating disease, everybody has it to some degree, we have trouble figuring out who to test for it in, our tests for it suck, and out treatment is unproven.
    I say, if you use both, keep it on the DL. And if you really want to scan them or do a D-dimer then do it. If clinical decision-making could truly be boiled down to a 5-step algorithm then we’d be out of a job.

    • says

      Matt – hope the pavlova was good – great Kiwi invention that! šŸ˜‰

      I have a further comment on the application of the PERC – as always an opportunity to use the word gestalt as much as possible… the only thing better is a comment saturated with repetitions of the word ‘equipoise’…

      In essence, I think there is no point combining the PERC rule with another clinical decision rule (Wells, revised Geneva, etc) unless the clinician’s gestalt is that the patient has a very low probability of PE (<7% if you're conservative – see http://lifeinthefastlane.com/2010/05/pulmonary-puzzler-010/ – note this post probably needs to be updated in light of the Hugli study mentioned below).

      In Kline's 2008 validation [PMID: 20836787], the PERC rule was validated on patients for whom the clinician's gestalt was <15% (pretest) probability before the application of the PERC rule. My understanding is that the doctors ticked a box to indicate their 'gut feeling' of this low probability – it wasn't explicitly based on the Well's criteria (though much of the current generation's 'gestalt' must be heavily influenced by having grown up with the Well's criteria). Nevertheless, in my ED the official guideline is that the PERC rule is only allowed to be applied to patients with a Wells score of 0 or 1, and a senior clinician's gestalt is that the patient has very low risk of PE.

      Interestingly, the recent Hugli 2011 study [PMID: 21091866] failed to validate the PERC rule in low risk patients identified using the revised Geneva score. Was this because clinician gestalt was not a criterion for the use of the PERC rule? Among other issues, I think it was probably a contributing factor.

      Perhaps the PE mantra should be: think, then trust your gut!


  4. says

    Loved this debate – great stuff. The problem, or perhaps the beauty, of it all is that not all gestalt is created equal!
    On a tangent, I have always found it slightly odd when a clinical decision rule has gestalt as one of the criteria (as in both Well’s and PERC), and yet many of the other variables which make up the score also contribute to the development in the clinician of THAT VERY GESTALT!!! Just my tuppence.
    Would also like to say – if I pass my fellowship (Australian FACEM) over the next couple of months, it is in no small part thanks to emcrit.org!!!

    • pik says

      Nice comment, that.

      I look at it as: so these data points get this score, huh? Well, my INTERPRETATION of those data points gets this (slightly higher or lower) score. So there.

  5. Casey Parker says

    Hi Scott, Chris et al

    Loved the debate – “when huge-brained nerds clash” – twas like watching the nature channel!

    Here is the Broome Docs perpsective. About once a day I get a call from somewhere a long way away to say: “we got this man/woman with vague chest symptom x. We have done D-dimer and it is x.x (usually agonisingly close to the reference range)… so we want to send them in for a CTPA”

    Sound familiar? Well the you gotta decide – do I spend a ridiculous amount of money to mobilise the flying docs, etc and fly them 1000 km for this test.

    Amazing how the decision -making changes when you put a big cost / logistical hurdle in the middle. So we go back to the criteria and gestalt, it always comes down to the gestalt.

    So Scott, you should ask your residents – would they still want the imaging if it meant they had to fly the patient to Texas to get it! My guess – you would do less CTPAs if you really have to think hard about the “gestalt” before ordering the test, not just whelling them up the corridor to Xray.

    My 5 cents

  6. SnoopyDoc says

    There’s a definite risk I’m missing the point (maybe I could get funding to develop a criteria to determine if I’m likely to be below threshhold for doing so?) but along with Domhnall and Pik, I’m wondering about the actual utility of scoring systems (and debating the same) based on excruciatingly careful consideration of the statistics involved, when those very same scoring systems involve “using the force” gestalt clinical impressions which render the scoring criteria a wee bit random/subjective, and at least partially self-referential.

    The mooted non-subjective score, the Geneva (or the revised or simplified revised variants thereof), obviates these shortcomings but alas, doesn’t seem to offer much help in actually deciding who needs further testing; the predictive power of the score is only valid in conjunction with a D-dimer/XDP result. While it could be used to determine who should go straight to CTPA, do not pass pathology, do not collect a D-dimer result… it cannot help in reducing unnecessary CT’s, as since the score requires you to do a D-dimer on everyone not going straight to CT, you still have the same number of patients who have a false-positive D-dimer (positive for reasons other than PE) who then cop a CTPA anyway.

    Thus my take on the situation is:

    – PERC: Avoids unneccsary imaging but relies on zen ninja gestalt impressions.
    – Geneva: Objectively identifies “straight to CTPA” patients but doesn’t reduce imaging.


    • says

      PERC, the way I see it, only asks if PE is the most likely dx you are working up. A little gestalty, but not zen-ninja level.

      Spec of the newer d-dimers is in the 40-50% range so geneva or wells will cut your imaging in half.

  7. SnoopyDoc says

    Fair comment on PERC… though I still question the idea of the standardisation achieved when the leading question is as fuzzily categorised as “Is PE likely or unlikely?”. Ditto for the Wellsian “Is another Dx more likely than PE?”… when the clinical criteria by which we make that determination overlap considerably with the criteria of the scoring system itself.

    I guess I just long for a utopia in which there was a more objective way to accurately & reproducibly qualitatively assess the pre-test probability. :-)

    How can applying Geneva cut imaging in half? All of the high risk Geneva punters are getting a CTPA anyway… The low (or low & medium) risk patients require a D-dimer to make the Geneva score interpretable… so of the population you identify as not-high-risk (the vast majority) you still end up with essentially the same number of patients getting a D-dimer… with the same risk of a false positive and a subsequent unnecessary CTPA.

    More than willing to concede I’m missing the obvious point here… I’m at the wrong end of a night shift and using far too many clock cycles just to avoid faceplanting into my breakfast cereal…

    Loving the podcasts, btw. :)

    • says

      In that utopia, there probably would be no doctors. we do not earn our bread by following guidelines (though of course we should when they are available) we are paid to deal with uncertainty nd to wind up wrong sometimes. we earn our living by the stress of not being sure.

      Geneva (or wells) allows you to characterize the low/moderate patients, who like you mention are the vast majority. You can then do a d-dimer in this group.

      A vast majority of these patents will be true negatives. A d-dimer with a 50% specificity will correctly characterize 50% of these. You have therefore cut your imaging in half in this group.

  8. Snoopydoc says

    The Geneva / Wells allows identification of a sub-group in the suspected PE population that has a low enough pre-test probability such that the application of a D-dimer with, for example, a ~95% sensitivity, will drive the post-test probability below the threshold (call it ~2%) considered reasonable, derived from the risk-benefit analysis of further investigation versus missing the diagnosis.

    Was rather stuck on the “but everyone you characterise as lower risk is getting a D-dimer anyway” train of thought whilst happily ignoring the rather salient point that the removal of the high-risk sub-group is what you want/need, in order to drop the pre-test probability in the others to bring them “in range” of a useful assault by the dreaded D-dimer.

    Gotta love the blonde days… really… *sigh*

    (crawls quietly into the corner… rocking slowly…)

  9. Reedsposer22 says

    Hi, im a new 4th yr med student. Im doing a rotation in the ED right now (wanting to go into EM). ipI had a resident tell me yesterday that the PERC criteria are good to know for PE rule out. I didnt get to ask him my question because he was slammed, but I learned Wells criteria on my IM rotation, Do you not use Wells in the ED? Is PERC easier to use? Ive been researching online and cant find anything that explains this. Thank you for any help. Confused M4.


  1. […] First he featured a fascinating intellectual slugfest between himself and the cerebrally top-heavy David Newman who you will know as the brilliant ER doc (with help from his friends) behind SMARTEM, TheNNT and even the Annals of Emergency Medicine podcast (be sure to catch his great interview ‘What would Hippocrates Do?‘ on Rob O’s ERCast). Check out all 15 rounds of this ‘clash of the titans’ at EMCrit: A Debate on PE Rules. […]

Speak Your Mind (Along with your name, job, and affiliation)