Podcast – Diagnosis in Emergency Medicine Part 1 – SpIN and SnOUT

Hello and welcome to this St Emlyn’s podcast. I’m Iain Beardsell and I’m Simon Carley. What we’re going to try and talk a little bit about today is the core of emergency medicine work. Patients come to us and ask, “Doctor, what’s wrong with me?” We endeavour, through a series of history, examination, and tests, to find the answer for them, to tell them what’s wrong. We just want to have a little bit of a chat about what that really means, what it means to make a diagnosis.

Simon, to you, what does it mean when you’re actually diagnosing a patient and you say to them, “You have this condition?” Well, in day-to-day practice, what I’m doing is I’m applying a label to them. So I’m saying either you have this condition and that I’m going to do something on the basis of that or, more commonly in emergency medicine, we’re saying that you don’t have something because we’re rule-out physicians and it allows somebody to go home. So I’m saying things like, “You don’t have a myocardial infarction, I’m fairly confident this isn’t chest pain due to cardiac disease, you can go home.” So we’re using a label to make decisions. That’s how we do it in day-to-day practice.

It’s probably worth just thinking a little bit more about that. So we tend, I think, to work backwards. We don’t start off with trying to say what’s wrong with a patient, we tend to start with saying, “We’re going to take the serious stuff and rule that out.” And work backwards if you like. Is that a reasonable way for us to work? Yeah, I think so because we’re most interested in things that are going to kill you and in particular things that are going to kill you quite quickly and in the UK certainly things that are going to kill you within four hours. So we’re really interested in spotting those at an early stage so we can do something about it.

We are a bit backwards. It means that we’re very interested in diagnostic tests which tell us what people don’t have. So we’re looking at tests which we would describe as very sensitive that we’ll pick up anybody who has the condition. And those tests, I guess, don’t just need to be blood tests, but anytime we ask a question or perform part of a clinical examination, these are all parts of our testing strategy. So asking a patient about their clinical presentation, we’re always doing tests, aren’t we? It’s just not necessarily that they’re what we might regard as tests in the blood test kind of idea.

Absolutely, and that’s really important because people think about sensitivity and specificity of this blood test or this x-ray, but everything we do, absolutely right, is so important. The most powerful tests that we have are clinical history and examination. A sensitive test for have you had a cardiac event is, do you have chest pain? A really specific test would be something like temporal arteritis: do you have jaw claudication? So there are questions that we can simply ask patients or examine them about, far more powerful than a serum rhubarb that you might get back from the lab in two hours.

In all of this discussion, we’ve got to put inside of that that our key work is done in the history and then in the examination to a certain extent, but it’s all about the history. Maybe if we start thinking of our questions within the history as diagnostic tests, that’s going to help us get to where we need to be. So we’re going to think a little bit about how we tell a patient they don’t have a certain illness, how they don’t have that diagnosis. We start with the life-threatening stuff, I guess. So how certain do you think we have to be to say that a patient doesn’t have one of these life-threatening illnesses? The ones that we all worry about, the SHO is in a dead panic about in their first few weeks, the MI’s, the missed PE, the aortic dissection, the subarachnoid haemorrhages, everything that keeps us awake at night. How sure can we be? Are we ever 100% sure that patients don’t have those?

It’s really interesting, isn’t it? If you go out and ask people, they think they are, but they’re not. So let’s explore that. How certain do you need to be? Well, if you are thinking about sending a patient home or about not doing something, so you can declare them as not having the condition. What does that mean? To me, it means that you’re identifying a group of patients who are highly unlikely to actually have the condition. Secondly, that if they do have the condition, it’s either a very low level of it, so it’s a very minor type of a very low-grade event. Thirdly, that if they do have a complication, it’s unlikely to be very sudden, very precipitous, and very serious. So at the extreme end, you mentioned subarachnoid hemorrhage. If you miss a subarachnoid haemorrhage, it could be a sudden dramatic deterioration very quickly, and you might not be able to get somebody back from that. So that’s a very, very important thing to spot. There are other things which wouldn’t be so important to spot first time around. Something like a small flake avulsion fracture of a lateral malleolus, does it really matter that much if we miss it? So we can tolerate different levels of certainty depending on the clinical condition we’re looking for.

So if we take some of those life-threatening conditions we’ve talked about, the ones that we really worry about, let’s use pulmonary embolus because that’s always talked about with diagnostic testing and other things. How sure do you need to be that a patient who’s presented with shortness of breath and/or chest pain hasn’t got a pulmonary embolus before you can say, “I can stop testing you now.” Or “I can stop asking your questions or I can stop what I’m doing.” and you can move on to ruling out the next diagnosis on the list of problematic things. Where do we have to be with that? Because from what I think we’re saying is you’re never 100% sure.

No, and we can put some numbers on that later if you want, but the important point is you’re never completely sure because most of the tests that we would use, we talk about some certain tests being very sensitive. There’s a ballpark figure which is knocked around and it’s kind of culture and practice as much as anything else. But if you’ve got a test that’s more than 95% sensitive or certainly 98% sensitive for serious conditions, that’s a pretty good test. Mathematically it is. If I told you that something picks up 98% of something, it sounds good, doesn’t it? Yeah, that’s a 98%. If I’d got that in my exams, my mum would have been well-chuffed. I would have got a cookie and a glass of milk. So 98%, I’m fine, thank you. Yeah, there’s probably been an investigation about how you got such a high mark. But 98% means you miss one in 50. Hold your horses, I miss one in 50, now I’m not happy. Which is why just expressing something as a natural frequency rather than as a percentage makes people feel differently. There’s loads of evidence out there, I’m not going to talk about it today, but loads of evidence that how you present data makes people feel very differently about it. But a 98% sensitivity means that you miss two out of a hundred people with a condition. So that’s one in 50. That’s a fairly accepted level of high sensitivity, but we’ve got to accept that still means you miss a significant number of patients.

I don’t think people understand that. I think people are deluded a lot of the time, I think they believe that high sensitivity tests rule out everything completely. That’s how people behave anyway. Well, I think there’s a couple of things there, isn’t there? The first expectation of our patients upon us and of ourselves is that we’re never wrong. Certain publications in the UK would be absolutely delighted if we went ahead and said, “You know what? We sometimes get it wrong.” But perhaps it’s the language we use that makes that difficult. Are we missing cases or is this just part of being good doctors? Or do we have to keep investigating until the point you never miss something? Even though those investigations, I think as we’ll talk about another time, could end up doing harm. So we’re a bit between a rock and a hard place really, aren’t we? We need to have a nice balance of where the risk is. One of the reasons we do use diagnostic tests in the ED, which don’t have 100% sensitivity, an imperfect rate of picking up everybody, is that by pursuing a diagnosis, you can actually do harm.

Let’s take a DVT for example, we use testing processes which have got 98% sensitivity to avoid having to do things like venography, which is an invasive test, unpleasant, expensive, and uncomfortable for patients. We use diagnostic testing processes for PE to avoid having to do, in the past certainly, CTPAs and even pulmonary angiograms which have got a significant risk associated with them. So you reach a balance point where pursuing the diagnosis further is probably not wise. There’s another point as well, conditions that you miss, that 2% that you miss are probably, and in most cases, the patients who have the smallest burden of disease. A massive PE is unlikely to be missed compared to a teeny tiny one and therefore the consequences of the miss aren’t as great.

We deal in absolute, don’t we? You have a PE, you don’t have a PE. We rarely put a quantity on that, I guess, because we’re not really sure what that means. But I guess massive PE, we will say this is a massive PE because it changes what you do next. But we just give them a label of disease rather than how bad that disease is. You’ve had a non-STEMI, we don’t necessarily say how bad that non-STEMI has been. So when you make the diagnosis, so you say you’ve got the label, are we saying you’ve actually got the disease or are you saying you’re now in a population of patients and if we treat that population of patients which is a mix of people who actually have the disease and who don’t have the disease, overall you’ll benefit? That’s how medicine actually works. When you go through a diagnostic process and you come out the other end, if your tag positive diagnosis is positive, you’re a mix of true positives and false positives. We then take that group of patients, we do trials on them, RCTs and decide whether that group of patients, which is always a mix, benefit or don’t benefit from treatment. Similarly, we look at patients who don’t have the label applied and they’re a mixture of people who generally don’t have disease and some misses, so some false negatives. We look at that group of patients and see if they benefit from not having had the diagnosis made. Well, that’s a less common study.

More and more we’re dealing in uncertainty and we’re removing some of those absolutes. It’s probably worth just taking a couple of minutes out if you can cover it in a couple of minutes and thinking a little bit about what we mean by sensitivity and specificity just because we’re going to be talking about it a lot. If I’ve understood it correctly, sensitivity, snout and spin, so snout being sensitivity, that’s a test that helps you rule out disease. But it’s actually measuring the number of people who test positive over the number of people who actually have the disease. Is that right? Sensitivity is basically a measure of whether or not you can pick up the disease. So what you want from a sensitive test is that it will be positive for everybody who potentially has it. Now, that will be a mixture of people who actually have the disease and also some people who don’t. A test like D-Dimer for thrombotic disease is very sensitive. It picks up lots of people who’ve got the disease, most people who’ve got the disease in fact, but also picks up a bunch of other people who haven’t as well. But it’s still very sensitive.

The key for us is that there’s very few false negatives with a sensitive test. So because there’s a few false negatives, if you get a negative test, you’re reassured that that negative test is correct. Am I getting down the right line here? Yeah, absolutely. So a negative test means that you’re unlikely to have the disease. Just because that false negative rate forms part of the sensitivity equation. Yeah, sort of, but that one in 50 that we talked about before, that will be the one in 50. You will be falling into that group who actually had a negative test, but do have the disease. They’re false negative. Exactly. Absolutely.

Then if we just flip that round quickly hoping that we can take specificity in amongst all this, we’ve said that specificity is a rule-in test. So if you’ve got a highly specific test, it’s a rule-in. And again, in my head, that means that we have few false positives. So if you get a positive test result, it’s likely the patient has disease. Yeah, if the test is positive, so if you have that characteristic or you have that feature, they are very likely to have the disease. The one I always remember, and I use it in a lot of examples, is jaw claudication and temporal arteritis. Only 4% of people with temporal arteritis have jaw claudication. But everybody with jaw claudication has temporal arteritis. So it’s a very specific test, but you couldn’t use it as a screening test because only 4% of people have got it.

I get you. So we’re going to be using these terms quite a lot, sensitivity and specificity. I think we’ll probably be doing a lot of that sensitivity because that’s our rule-out thing that we’re thinking about more in the emergency department. So hopefully we’ll come back and I think repetition is learning. So what we’ll do is we’ll go over those terms a bit more because if we can get those implanted in our brains and our listeners’ brains, then we might be getting somewhere. So where we’ve got to so far is that the patient comes to us and they want to know what’s wrong with them. But actually we’re going to work in reverse and say well let’s work out first what isn’t wrong with you and work out that you haven’t got these life-threatening conditions. So we’re working backwards. We thought a little bit about how certain we need to be and we reckon one in 52% or 98% certainty is about right? Yeah it’s culture and practice but yeah that’s what most people are working to. So that may be a bit of a bombshell for some people who are listening. We have to become more accepting that we’re not always right and that we may be getting what we’re going to call misdiagnosis even though perhaps the terminology isn’t as helpful. But even if that happens the burden of disease may be so little although they have it that they come to know harm. Is that what about we are?

When we’re thinking about misdiagnoses and how important they are then you can investigate that in several different ways. The first thing is that they’re probably got a slightly less severe degree of burden of disease. So for instance big PEs are more likely to be picked up. If you’ve got a massive pancreatitis you’re more likely to spot it. If you’ve had a massive MI you’re more likely to spot it than a small MI that kind of thing. So burden of disease is important so it means the consequences of your missus isn’t as great. The other thing about missing diseases is again how quickly it’s going to come back and get you. So it’s acceptable to miss certain conditions if you know that they will gradually get worse and therefore the patient is likely to come back to the hospital means you can do something about it when they come back. Missing things isn’t a disaster. It’s a natural consequence of being a good diagnostician and using good high quality evidence in your diagnostic protocols. But it does mean we’ve got to tell the patients that. If you tell the patient you definitely don’t have the condition don’t come and darken my door again. You’ve definitely got an MI why you bothering me. Then that’s a really bad practice because a number of them will come back. And if you told them that when they went out of the door they’re not going to be very happy.

I’m pretty comfortable with the idea that we’re working through a series of tests. They might be part of the history, the examination. There might even be blood tests, radiological tests to say a patient hasn’t got an illness. And when they haven’t got an illness we just don’t treat them obviously and we reassure them and we move on and we try and find out what is causing the illness. If we flip that the other way around and actually perhaps work in the way that most people think we work which is to say when a patient has got a disease how sure do we need to be that a patient has an illness in order to then firstly dismiss all the other things that are on the differential but also go ahead and start giving them treatment for whatever that illness may be. Are we always sure that people have illness or disease and a positive diagnosis or is that again uncertain? If you want to get really pedantic about it and those are different terms of course illness and disease are different things. Illness is what the patient experiences, disease is the process. What the important thing is you’re asking about I think is what happens, the consequence of making a diagnosis in a positive way. If you apply the tag to somebody and say you have this what are the consequences of treatment? If we take the biggies again something like myocardial infarction there are significant consequences to making that diagnosis. That patient is going to get thrombolysis or go to PCI and be exposed to a whole bunch of different drugs which are going to potentially expose them to bleeding events and all sorts of things. So there are risks in making a positive diagnosis and we don’t always know whether they’re going to benefit, whether they’re going to suffer harm, whether it’s going to make no difference at all.

I don’t want to go there but we could think about strokes or thrombolysis here but we haven’t got six hours have we? The internet and phone is so awash with strokes or thrombolysis I’m not sure that we need to add anything to debate I guess. When we’re making a diagnosis, we have to know not just add up what it is we’re saying but what the consequences of treatment for that diagnosis might be, and those consequences can be very different for us in the emergency department, say, from in family practice or general practice. The idea that you have an MI is very different from the idea have you got a streptococcal bacterial throat infection or a viral throat infection the consequences are very different. How sure do we need to be if we were to put a percentage term on it like we did before? How sure do I need to be that a patient has something in order to treat them for something? I think it entirely depends on what the consequences of your actions are. Let’s take a simple example, I’m very big into chest ultrasound at the moment so I’m chest ultrasounding all our major trauma patients. If they’ve got a pneumothorax on the ultrasound when they come through I’m putting a chest drain in without a chest x-ray. So that’s a significant consequence to that patient. Would you agree? Yeah especially if they don’t have a pneumothorax and your test wasn’t. If I’m rubbish then there’s a bit of a problem that’s a significant consequence to it. Let’s just put out there I’m not at all suggesting that professionally I’m sure you’re brilliant at it but we can even take that diagnosis before the arrival at hospital when we’re seeing major trauma patients who’ve been seen by pre-hospital providers who may or may not have access to ultrasound are trying to do a clinical examination in different difficult circumstances. So we could describe their sensitivity and the specificity of their tests be that history or examination it’s all much more limited than perhaps we get in hospital. These patients are then getting thoracostomies on the best available evidence in the pre-hospital environment and then in our clean brightly lit huge scanner of a hospital we find out they didn’t have a pneumothorax but there has been a consequence to that decision being made and that’s based on the testing that’s been gone before it.

Absolutely and you can apply that kind of logic to many of the things that we do so how sure you’ve got to be definitely depends on what you’re going to do about it if you make a diagnosis but actually there are no consequences from it for instance you make a diagnosis of a viral sore throat and I’m not doing anything about it there’s no consequences no harm you can do it as often or as little as you like really but when you’re talking about a therapeutic event which takes place as a result of that diagnosis being made then you’ve got to be pretty sure and particularly if there are harms associated with it. For the biggies for us things like thrombolysis things like PCI things like going to an operation if you dissect free fluid on your ultrasound about imaging the chest about intervening in the chest about making airway interventions etc etc etc all of these big things that we do in the ED I think you’ve got to be fairly sure how sure is fairly sure it’s a balance and that’s why we are clinicians we’re not automotons we don’t follow strict protocols and we have to have some judgment but that judgment is much better informed if you understand the probabilistic nature of the diagnosis that we make.

With that I think you’ve just dropped the p-bomb there at the end the word probabilistic, which I think is something we’re going to come back to in episode two to discover a little bit more about discuss a bit more about we’ve just started to scratch the surface of what it is we really do in the emergency department we aim to rule out disease those life-threatening diseases but when we can’t rule them out we need to have a degree of certainty about ruling them in and we need to use good diagnostic testing and also much more importantly good clinical acumen to make the decision to do the treatment and these are all topics that we’re going to discover a little bit more about in further podcasts.

Just want to leave you to have a think about everything we’ve talked about reflect a little bit think about how that affects your practice and we’ll be back very soon with part two discussing a bit more of how we make decisions and diagnoses in the ED. Take care.

Before we go, what’s your favourite diagnostic test? My favourite diagnostic test? I would say that my favourite diagnostic test is taking a history. That’s kind of the standard answer. No, just at the end we could do what’s your favourite diagnostic test. Okay let me think. My favorite diagnostic test that I like the most? Tricky. Go on let’s just do something on the fly. Ian what is your favorite diagnostic test? Anyway, Simon, I have to say I’m renowned in my department for being the man who’s anti-testing. I’m always trying to reinforce to our doctors and nurses that history and examination is the key but I suppose there are times when we need tests to make decisions for us and we need to enhance our decision-making capability by using technology be that blood or radiological testing. I guess the piece of kit that I enjoy using the most because I can use it is the ultrasound machine and I actually like that most in the resus room. I enjoy, I think the pneumothorax example is a good one, it’s the thing that you can do immediately, you can make an intervention and arguably that test has better characteristics than any of the other tests we can do. So I’m going to plump for ultrasound in query pneumothorax in trauma patients. How about you? It’s interesting you’ve gone cool, you’ve gone for the cool test it’s got nice shiny kit and electronics and I’m with you on that. I’m going to go for D-Dimer. Really but everyone hates D-Dimer and that’s why I love it because D-Dimer is not a bad test it’s just badly used. What I love about D-Dimer is it can be incredibly valuable and incredibly helpful if you understand how diagnostic tests work. So I love D-Dimer because when anybody comes up to me and says you know what I hate D-Dimer is the rubbish it gives me the opportunity to talk to them for between 45 maybe 50 minutes on diagnostic testing and you know what I love that. I really reckon that once this goes public no one will ask you about D-Dimer’s again envisaging 45 minutes spent with you talking about that.

Any obviously 45 minutes with you would be delightful otherwise and on that final note we look forward to speaking to you again soon on this Emlyn’s podcast. Enjoy your emergency medicine and take care.