As an EM Doc you do a lot of tests.  You know about them, and you know that negative tests don’t always rule out a disease, you know that positive tests can be false positives (patients don’t have the disease you are looking for, but the test suggests they do).  You have seen this normogram about likelihood ratios, and maybe even used it once at a teaching day.

That Normogram (Faagan’s normogram) is your friend, because it defines the relationship between the possibility of the disease before and after you test for it, based on what you think.  This depends on three things, the quality of the test, how likely you think the disease is, and whether you are trying to diagnose it (rule in) or exclude it (rule out).

If you have a really low pre-test probability (<1%), and your test has a decent LR of say 10, you still only manage to get a 5% post-test probability that you’ve found pathology.  Most tests don’t have a +veLR of 10, most of them run from 0-10, so if after looking at the patient, talking to them, and examining them you really don’t think it’s X, and maybe you think there is 1 % chance its X to rule the diagnosis in with a test you need an amazing test, one with a LR of over 200 to get you to a coin-toss likelihood (seriously print out a normogram and draw the lines!).

Now this is hard to get our heads around, mostly because we don’t trust ourselves, we have the entire medical establishment telling us not to miss things, but paradoxically only ever refer things we’ve found, and never admit someone who turns out not to have high consequence disease X. The other thing that people don’t realise is that big ticket, high consequence genuine medical emergencies are rare events.  They are rare events in the ED population, and they are even rarer in the general population.

This means that if you test lots and lots for these rare events you are going to find more false positives than true positives.

So if it doesn’t sound like X, and you don’t think it’s X, and the patient doesn’t think it’s X, why do the test? (Please don’t do the test!).

We also need to talk about ruling-out here.  Unless you’ve got a really good rule out test like (for example CTA for AAD) if you don’t think its AAD, doing the scan isn’t going to change the post-test probability in the slightest.  This is counter-intuitive, so I’ll go step by step.

  1. You’ve seen someone.  You’ve “Thought Aorta”.  They look ok to you.  You estimate a <1 % chance of AAD.
  2. You do your CTA.  Lets say CTA has a -LR of 0.01
  3. Well done your post-test odds are now 0.2% (but you thought that was the case already…), and now your patient has a 1 in 300 lifetime risk of cancer (according to our radiology requesting system)– you’re a monster…

Essentially if you’ve seen a patient and you don’t think they’ve got the disease they probably haven’t got it.  If you’ve seen a patient and think there’s a good chance they might, then you do the test and see, but for most tests your pre-test probability has to be above 1% at least, and probably higher.  If you’ve seen the patient and it clearly is X, then why do the test?  This is the problem with super rare, high consequence medical conditions.

Also:  your pre-test probability will almost certainly be an over-estimate.


Should it be low? Should it be high? It is a term that peppers lectures and learning modules about specific conditions all the time. Have you noticed? It should always be high or low, it can never be right. It’s probably worth thinking about, and what I think people really mean when they talk about it, is your gain.

I mean gain in terms of amplification  – “the factor by which power or voltage is increased in an amplifier or other electronic device, usually expressed as a logarithm”

Now if we think of ourselves as detectors of pathology (which is incredibly pretentious but bear with me) we can increase or decrease our sensitivity to signals (in this case pathology) depending on a variety of factors.  These could be patient factors, or factors associated with the clinical question that has been asked.  For example we all know that neonates can have very subtle signs when they have a serious underlying infection.  So, we are careful, and exceptionally risk averse when assessing them.  Similarly, we should be careful in elderly patients with frailty for the same reasons.  Presentations are non-specific and can mask pathology that can be of high consequence (I’m planning on writing about non-specific presentations later).  We all do this to a certain degree, this can be seen with referral and admission patterns in the elderly and in the very young (higher conversion rates to admission).  However we probably don’t need to be similarly alert to subtle signs in a 24 year old man.

I think people find it difficult to switch modes here.  The likelihood of finding significant subtle pathology in a normally healthy person with nothing overt in the history to point you in any particular direction is so low that you probably don’t need to spend as much time with them as you would with the other categories of people I’ve described.  Now a number of doctors do this instinctively, but the majority I see use the same mental model, and the same ‘gain’ for every patient interaction.  This is inefficient, it will be too low for the cases that require subtlety and far too high for our ambulatory majors area. 

It is theoretically possible that with your gain turned down, you might be more likely to miss something; however, the sub-populations (Farmers anyone?) you are working within here are different as well.  The risk is probably smaller than you think, because the prevalence of subtle stuff in our young healthy population is less than in the elderly and frail.

So this is what comes with experience, you can see people incredibly quickly if you understand the population they are coming from, and the relative likliehoods of the pathology they or their GP are worrying about.  Now most people who work in EM have a good but tacit understanding of this.  They know for example that ACS is incredibly unlikely in a 24-year-old who attends with chest pain and are more wary of a 54 year old woman with ‘heartburn’ who is feeling tired all the time.  So this is probably a combination of gestalt and that term (that I hate) ‘index of suspicion’ or, as I’ve described, your gain.

So if after assessing someone you feel they are low risk, you aren’t really obliged to do any confirmatory or rule out testing.  You can send them home.  Equally if you have assessed someone and the diagnosis is clear, you don’t have to go through a risk stratification or Boolean algorithm to get to the diagnosis.  That wastes time.  Just skip to the end.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.