EBM and the rule of thumb

aka American ER Doc Gone Walkabout… 026

Wise use of EBM makes us much more reasonable Doctors.

I meet a young woman, previously healthy… She’s had a little fever, some congestion, has been coughing and sneezing for a few days, and now her chest hurts. She’s been a little wheezy and minimally dyspneic with exertion. Her exam is normal except for some costochondral tenderness.

The thought of pulmonary embolism drifts by my mind.

Being an EBM Doctor, I quickly review mentally the costs and benefits of advanced imaging – radiation, dollars, future cancer risks, risk of missing an important diagnosis, as well as my own (and my patient’s – gotta be collaborative in our decision making) tolerance for ambiguity, and conclude that I’d be willing to do – let’s say 30 to 50 CT PE studies to find one more PE.

I then apply some clinical gestalt, perhaps reality check with a Wells score or Geneva score, and follow up with a PERC rule. And, conclude that the estimated pre-test probability that this presentation could be due to a PE is low enough that it doesn’t warrant a CT PE.

The devil whispers in my ear: “But, what if you get a d-dimer and it’s positive?”

So, I quickly do a back-of-the-envelope meta-analysis and take a bunch of studies reporting sensitivity and specificity numbers for d-dimers, and come up with a positive LR of 1.3 – but that isn’t significantly different from 1.0 – so I conclude that a positive d-dimer would be insufficient to change my pre-test probability, and the post-test probability would remain the same and is still below my threshold for ordering a CT PE. Therefore, there is no point in ordering a d-dimer.

(This gets really interesting when somebody else at triage/intake has ordered the d-dimer and it returns at 550. I now have to have a junior resident restrained, sedated, and perhaps paralyzed to keep her from ordering a CT PE.)

That’s a lot of calculations to do on every low risk patient that has a vague complaint that just “might” be from a PE. It would probably take you through lunch and afternoon tea, and maybe the opposition will have brought on their nightwatchman by then (for you Yanks, it’s a cricket thing).

So I can develop a shortcut:

“If your clinical or calculated estimate of the likelihood of PE is so low that, if the lab d-dimer machine was broken and no d-dimers would be available today, you would not order a CT, do not order a d-dimer.”

One can think of other similar shortcuts:

“D-dimer is a one sided test: make your clinical decision as to imaging. A positive d-dimer does not change that decision, whether it is yes or no to begin with. A negative d-dimer may change your CT decision from yes to no.”

So, just what is a “Rule of Thumb”?

A rule of thumb is not a guideline as to what to do with your first digit while doing a bimanual pelvic exam.

Here’s Wikipedia:

“A rule of thumb is a principle with broad application that is not intended to be strictly accurate or reliable for every situation. It is an easily learned and easily applied procedure for approximately calculating or recalling some value, or for making some determination.”

OK, so that helps understand what we’re talking about, but how does that help me if I try to be evidence based?

Why would I, an experienced clinician, use something that simplistic? Well, if I actually was very conversant with the literature, and understood statistics, a rule of thumb might simply save time. I’d know why the rule worked, could derive the rule step by step to wow the students, but when I actually was making real-time decisions could apply the rule in a matter of seconds (often without even explicitly thinking of the rule) rather than working out the numbers at all.

I suspect that we all unconsciously use implicit, personally developed, rules of thumb more often than we realize. (Implicit RoT: If you are doing so much imaging that your patient, by age 30, glows in the dark, you should rethink your diagnostic strategies.)

Which brings up a second place that a RoT is useful: for the less knowledgeable practitioner, the RoT doesn’t summarize information he knows, it summarizes information that an expert has condensed into a useable format for the less knowledgeable. If my interns remember the RoT, they don’t themselves have to know the literature based sensitivity and specificity, or how to derive the LR’s.

Maybe some examples would help:

If you go to Amazon.com (boy, did I miss having Amazon available to deliver stuff to me in Tasmania!), and search for Rules of Thumb, you get 17,483 hits! I only looked at about the first 100 – all books full of RoT’s. Wow, who would have thunk?

Some are books of pretty “hard” rules that take complex formulae and simplify them: RoT’s for pilots “Barometric pressure varies 1 millibar for each 29 feet of altitude change” – obviously meant for American pilots.

Some are books with “soft” rules: “When the going gets tough, the tough relax.” (From a book of RoT’s for businessmen.) Yeah, let’s try that in my ER.

Some take some geometry and engineering and apply the numbers to everyday life:

“You can’t always back out of what you drive forward into, but you can always drive forward out of what you backed into.”

Some contain myriads of empirically derived rules: heights of toilet seats, countertops, etc.

So, in medicine rules of thumb may be explicit, or more commonly are used as implicit common sense applications. They take formal evidence, or consensus knowledge bases, and apply them to clinical scenarios. They make the knowledgeable more efficient: and they provide a way for the less well informed to function safely. Used wisely, they allow safe function as a physician while learning the material in greater depth, and give a knowledge base to build upon.

Some RoT’s (even if not called that) are part of the accepted knowledge base passed from Doctor to Doctor (If he has been in the tropics, has a fever, has such bad retro-orbital pain that it hurts to move his eyeballs, think dengue), some are self-derived rules that constitute what we call “clinical experience” – the next time I see somebody short of breath the day after flying home from Sochi, I’ll test him for a PE, rather than giving him a salbutamol inhaler.

Hmmm! I just had a thought: What if we were to string together a whole bunch of rules of thumb? Would we have a pathway? Would that be a good thing?

I’ll have to think about that. Stay tuned.

More later, mates.

Print Friendly


  1. arbitraryTag says

    Good thoughts, and one that has definitely occupied much of my time (PEs and d-Dimers and CTPAs).

    I rule of thumb (similar to a heuristic), is one of those things that work great when they work. The change in pressure with altitude is a simplification of a well known theory and those who use it would know the limitations well.

    I don’t think I am an EBM ‘zealot’, but I am often very wary of the use of RoTs in medicine, primarily because history has shown how often we get it wrong. It irritates my sensibilities that the sense of judgement of doctors is mainly being replaced by lots of testing, but the scientist in me has to recognise that ‘clinical gestalt’ is a very unreliable mechanism for assessing things that are uncommon.

    I think there is a lesson from the field of obstetrics. If ever there was an EBM utopia, it is obstetrics. The sheer numbers and relative ease of measuring outcomes means that very rare risks can be quantified. And I note that things that used to be ‘see how it goes’ or ‘give it a chance’ are now ‘intervene’. The reason being that our society will not tolerate a single perinatal or maternal death unless every possibility has been explored. We know (from large statistical studies) that the risk of, say, a vaginal breech delivery is only slightly higher than cephalic, but this small chance has drastically changed practice -- no one wants their patient/child to be the one in a thousand (or one in ten thousand) that would have benefited from earlier intervention.

    And so we scan all the ‘chest pain with positive d-Dimer’, mainly because we don’t want our patient to be the one ‘healthy 42 year old who died suddenly from a large PE’ -- because we know it happens (even though it may never happen during my career or your career, it will happen to someone). The cost of testing and the possible negative effects of radiation etc. are judged to be more socially acceptable than sudden death from large PE (likely because it is almost impossible to attribute a person’s cancer to medical radiation imaging, even though we believe some cancers are caused by imaging).

    Perhaps I have ended up back where I started. You made the comment ‘if the d-dimer in the lab was broken, would you image’. Unfortunately, the d-dimer is not broken, and there is a chance that your gestalt + PERC rule will miss a PE that the d-dimer and CT PA would have found. And (though it may never happen to you and it may never reach a Coronor) an investigating Coronor would ask “Knowing that this simple test (d-dimer) may have directed you to image this patient, why did you not perform it?” -- I think an uncomfortable conversation would ensue. As much as I hate the attitude of ‘image them all’, I simply cannot deny that there is a group of people that have been found to have PEs by d-dimer and CT PA that would not have been found without the d-dimer, and their lives may have been saved by anticoagulation.

    (As to what you would do with the CT PA, while future research may invalidate current practice, commencement of anticoagulation is the treatment for all proven PEs. Whether this has any effect is perhaps debatable, and the treatment carries its own risk, but there is a clear management change when a PE is found on CT PA).

  2. Dave mountain says

    I am not sure that EBM would agree with you re image them all take. PERC rule in a low risk pt is perfectly acceptable and a valid defence.In the end if the risk of PE is < 2% then you are not doing your patient a favour by submitting them to unnecessary testing and radiation, you are certainly removing resources from some of your other patients in the ED/ system (both in cost and timeliness)and the risks from overtreatment of false postive PE is very signficant.