Cadmore media player playing audio F. Douglas Srygley, MD, and Dr David L. Simel, MD, MHS, discuss the clinical examination for upper gastrointestinal bleed.
Audio Player
Accessibility
Using forms mode, you can utilize keyboard shortcuts to make usability more efficient.
Space bar will play and pause the audio. Pressing escape will close all displays and pressing it again will jump to the end of the player.
You can press A to open the Keyboard Shortcut Dialog which also provides an audio description of the player and all of its features.
Transcript
Result Count
of
Search
Segments
Result Count
of
Search
Name:
F. Douglas Srygley, MD, and Dr David L. Simel, MD, MHS, discuss the clinical examination for upper gastrointestinal bleed.
Description:
F. Douglas Srygley, MD, and Dr David L. Simel, MD, MHS, discuss the clinical examination for upper gastrointestinal bleed.
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/94d3412e-1277-418e-b10d-c98659faa693/thumbnails/94d3412e-1277-418e-b10d-c98659faa693.jpg?sv=2019-02-02&sr=c&sig=rSubMIqtKoLqJDTFJAgjdWjYbk7YWBbleMnHamTDpaQ%3D&st=2025-01-02T19%3A02%3A27Z&se=2025-01-02T23%3A07%3A27Z&sp=r
Duration:
T00H32M39S
Embed URL:
https://stream.cadmore.media/player/94d3412e-1277-418e-b10d-c98659faa693
Content URL:
https://cadmoreoriginalmedia.blob.core.windows.net/94d3412e-1277-418e-b10d-c98659faa693/11698114.mp3?sv=2019-02-02&sr=c&sig=504R87wo0S6Br3mRsfLpqgYIp%2F17aviNYnR8xv92qlM%3D&st=2025-01-02T19%3A02%3A27Z&se=2025-01-02T21%3A07%3A27Z&sp=r
Upload Date:
2022-02-28T00:00:00.0000000
Transcript:
Language: EN.
Segment:0 .
>> Hello and welcome to JAMAevidence, our monthly podcast focused on core issues and evidence-based medicine. This is Ed Livingston, Deputy Editor of Clinical Reviews and Education here at JAMA. Just to let you know, CME is available for listening to this podcast, if you refer back to the original paper published in JAMA several years ago. We'll tell you more about that at the end of the podcast. Let's start with why the Rational Clinical Examination Series of articles are important. While preparing for this podcast, I happened to be looking at the medical record of a patient who had two hospitalizations.
The problem that I was looking at was that the patient was in the hospital after having emergency surgery and a bowel resection. He had GI bleeding, received a couple of units of blood transfusions, and did pretty much okay. He then was readmitted to the hospital for an upper GI bleed from which he ultimately died. The question was whether in this case the patient had evidence of an upper GI bleed the first time he was in the hospital. Today we take on this topic of upper GI bleeding, which causes nearly 400,000 hospitalizations a year.
There's another 600,000 hospitalizations a year where GI bleeding occurs secondary to some other problem. Every physician needs to know something about GI bleeding. Here's the bottom line for all comers with GI bleeds, about two thirds of them have upper GI bleeds. Of those who have an upper GI bleed, about one third are severe enough to need some sort of an intervention. Here an intervention means needing an endoscopy, radiological procedure, or blood transfusion.
Six features are associated with a need for an intervention. They are the patient reporting a history of having melenic stools, finding melenic stools on physical examination, finding fresh blood in an NG aspirate, having a BUN-to-creatinine ratio of greater than 30, or the presence of a malignancy or cirrhosis. Three features predict the likelihood of needing an intervention. These are having a nasogastric lavage with red blood, a patient having tachycardia exceeding 100 beats per minute, or having a hemoglobin below eight.
There are scores to assess the need for interventions. One of the most well defined is the Blatchford score. This Rational Clinical Examination systematic review found that the Blatchford score is really only useful when it's zero. When the Blatchford score is zero it tells you there is very little likelihood of needing an intervention. When it's not zero it's not quite clear how helpful the Blatchford score really is. That's the end of the bottom line. In this JAMAevidence podcast we discuss the chapter from the Rational Clinical Examination entitled "Making the Diagnosis Upper GI Bleeding." Along the way we'll put a nail in the coffin of gastric lavage as an evaluative method for assessing patients with GI bleeding.
Joining us is Dr. Douglas Srygley, a gastroenterologist at the Austin Gastroenterology Clinic in Austin, Texas, and one of the authors of this chapter. Later in the program we'll talk to Dr. David Simel, editor of the JAMA Rational Clinical Examination series and Chief of the Medical Service at the Durham VA Medical Center, and Professor of Medicine at the Duke University School of Medicine. When first presented with a patient who is GI bleeding, the first question is do they require an urgent intervention?
If you suspect even remotely that a patient might require an intervention, they need to be resuscitated as soon as possible. Once the patient is resuscitated, then you need to know if they need an intervention at all. >> We know from studies that about a third of patients will require some sort of urgent intervention, that being endoscopy, interventional radiology, or blood transfusion. And the best thing is to try to triage these patients out so that not all of them have to be admitted to the hospital and undergo an emergent endoscopy, which is both costly to the system as well as costly to the patient.
And we know that by using some of these predictors, especially the clinical prediction rule, we can rule out at least 15 to 20% of the patients from having to undergo those interventions, at least in an emergent or urgent basis. And while that might not sound like much, it's actually pretty nice for the patient and for the healthcare system if we can avoid those admissions. >> What do you find are the most useful findings to suggest that there is a clinically significant GI bleed? Or clinically important that might require intervention?
>> The most important findings from a historical basis are a history of malignancy or cirrhosis. That carries a likelihood ratio of about 3.7, and that makes sense. We all know that patients who have, particularly cirrhosis, are more prone to things like variceal bleeding or even portal gastropathy, either of which could require endoscopic therapy or potentially even interventional radiology and certainly blood transfusion. So we know that those two, from a historical basis, are very helpful to know. The other one is pulse rate. So a patient who presents with tachycardia has a likelihood ratio of almost five.
And obviously that suggests volume loss, which suggests a more rapid source of bleeding, which is likely going to require some sort of intervention, be it blood transfusion, endoscopic intervention, surgery, or interventional radiology. Lastly, a hemoglobin that's less than eight is obviously helpful. That one is a little more clear. If someone has a low hemoglobin they're likely going to need a blood transfusion or intervention. We define an urgent intervention in this study as anyone who needed blood transfusion, had stigmata or recent hemorrhage on endoscopy, required surgery or an interventional radiological procedure.
>> When you say that tachycardia is associated with a likelihood ratio of five, suggesting the need for an intervention, what does that mean? What does the likelihood ratio of five mean to you? >> So if we take all comers who come in with a GI bleed, we take our pretest probability. If we use that with the likelihood ratio graph, then we can get our post-test probability to be much higher. Typically we want a likelihood ratio greater than two to provide any clinical benefit, and the higher the likelihood ratio the more benefit it provides.
So for instance, if a patient has a pretest probability of 35%, which is in general what a patient presenting to the emergency room would have, then a likelihood ratio of five would certainly suggest to us that we should proceed with an evaluation and admit the patient to the hospital. >> In the article there's a table that has clinical factors of severity of bleeding from the history and physical used to determine the need for urgent evaluation of upper GI bleeds. So coffee-ground emesis has a likelihood ratio of 0.41. So why is it that someone who has coffee-ground emesis has, if I'm interpreting this correctly, a low likelihood of needing an intervention?
>> What one person's coffee-ground emesis is it could be another person's normal bowel contents, I think is part of the issue. I personally think that coffee-ground emesis is one of the more difficult historical features to tease out. Additionally, if a patient is bleeding more rapidly, then the blood will not have time to go into a more coffee-ground material. Remember, blood is changed into a darker material via an enzymatic process, and the longer it stays in the bowel the more dark it becomes, so one would surmise that if the emesis is darker it's less likely to have a more active rapid bleed than one who is throwing up bright red blood.
That's less likely to need intervention endoscopically, radiologically, or with blood transfusion. >> On a similar note, the likelihood ratio for hematochezia was only 0.224 for requiring an urgent evaluation. How does that stack up? >> So I think that really breaks down into we're getting two populations of patients. One population is probably patients who truly have a lower GI bleed. And as we know, oftentimes patients will present with a lower GI bleed which looks like clinically a significant amount of blood loss, but when we follow the patient they do not have nearly as much blood loss as an upper GI bleed.
For instance, we've all taken care of patients with diverticular bleeding where it looks like they're almost bleeding to death and the hemoglobin hardly drops at all over two to three days, versus a patient with upper GI bleeding that clinically might not look like they're bleeding as much but really has a significant bleed. So I think one population that we're getting there is probably patients who actually have a lower GI bleed. The other thing that is concerning, and it makes me a little bit hesitant for that hematochezia, although that's what the data shows us, is that patients can have hematochezia with a rapid upper bleed.
And so that's where, I think, the tachycardia also comes in. If you have a patient with hematochezia, you would say oh, well, this likelihood ratio would suggest it's less likely to need an intervention. But if that patient has a pulse of 120, or shock, then we know we've got to take that a little more serious, and that would push us to act in a more aggressive manor. >> One of the big surprises in this Rational Clinical Examination article is the relative ineffectiveness of NG tubes in assessing upper GI bleeds. When I was a resident, one of the most unpleasant things we had to do was place NG tubes in all the GI bleeding patients and lavage them.
I have many unhappy memories of doing that in the middle of the night in the ICU. >> Where I trained, as a rule, just about anyone who came in with coffee-ground emesis or hematemesis or melena received a nasogastric lavage to try to determine both the severity of the bleed and the location of the bleed. The data for that is actually not as robust as one might wish, as much as anything because most patients who you would go ahead and put a nasogastric tube in you're going to proceed to endoscopy anyways, so it doesn't necessarily add anything.
It's not a great predictor of whether or not a patient is going to require an urgent intervention. But the most striking thing to me was reading articles about how much patients dislike nasogastric tubes. I think as a provider we oftentimes think of this as a benign intervention roughly akin to putting in an IV. Patients do not feel that way. In at least one study, patients rated having a nasogastric tube as being worse than fracture, if they have to get a fracture reduced, or if they have to have an abscess drained, a lumbar puncture, or an arterial stick.
All of those things I would consider things that I personally would want to avoid, and I think we have to be a little bit hesitant to think about nasogastric lavage in these patients if it's not going to add much and put them in significant discomfort. >> Is there a role at all for nasogastric tubes in evaluating a GI bleed? >> They've been evaluated both to try to predict how severe the bleed is, and we know, based on several studies, that the likelihood ratio for a lavage with red blood in it is 3.1 for someone who would need an urgent intervention. If you go down to just red blood or coffee grounds, it goes down to 2.0. So that can persuade someone to go ahead and proceed with an urgent intervention.
But the vast majority of patients are going to have something else in their history, such as a low hemoglobin, tachycardia, melena, that would persuade you to go ahead with a intervention in an urgent basis. So that leaves whether or not placing an NG tube makes endoscopy easier or better. And there is some limited data that doing a nasogastric lavage might improve visualization at the time of endoscopy. It should be noted, though, that that data is quite limited, and there's other much less invasive agents that we could use such as erythromycin or Reglan, which has also been studied and are affective.
I will say in my practice, as someone who primarily takes care of hospitalized GI patients, as a rule I don't put a nasogastric tube in, and I recommend it in general when the emergency department physicians call me. The other thing that I think needs to be brought up is the possibility of harm with a nasogastric tube. There's a teaching out there that patients with varices should not receive nasogastric tubes. There is not really any evidence that that is true, with the one exception of someone who has just recently undergone banding. There's no data to suggest that placing a nasogastric tube in that situation is concerning, but I think anyone who has placed bands on someone who has recently had variceal bleeding would recommend against that.
>> What about rules and things like the Blatchford score to assess bleeding severity and predict the need for an intervention? >> I think the clinical prediction rules have a limited use. The big use of the clinical prediction rule is the patient who appears healthy and appears to be having a minor bleed that we can determine, that we can safely send them out. In the studies with the Blatchford score, depending on which study you look at, somewhere between 5 and 20% of the patients will have a Blatchford score of zero, which means that we can basically safely send them out. >> So a Blatchford score of zero pretty securely tells you that there's no clinically significant bleed and no intervention will be required.
What do you do with a Blatchford score of one or greater? >> In those patients, I think the clinical prediction rules have very limited utility, and I also think that the clinical prediction rule's positive likelihood ratio is so low as to, quite frankly, be totally unhelpful. It technically doesn't cross one, but a likelihood ratio of 1.4 is not going to push you one way or the other. >> So they're really good for ruling out the diagnosis not ruling it in. Is that right? >> That's correct.
It can be a very effective tool to rule out a bleed requiring urgent intervention, but it is not an effective tool to rule in. >> To recap, about two thirds of all GI bleeds that come into the hospital are upper GI bleeds. Of these, about a third are going to need some sort of urgent intervention. That means that they'll require a blood transfusion, GI endoscopy, or some radiological procedure. The most reliable clinical feature suggesting the need for an intervention is the presence of tachycardia with a heart rate of more than 100 beats per minute or a hemoglobin of less than eight.
An NG lavage showing red blood is also a good indicator of needing an intervention. The problem with this is that an NG lavage not showing any blood doesn't necessarily rule out the need for an intervention. This, coupled with the fact that NG tubes are so painful, has led most gastroenterologists to no longer recommend NG tubes when evaluating GI bleeding. Keep in mind this is different than needing them or resuscitating a patient in shock who needs the NG tube to prevent aspiration. Lastly, clinical prediction rules such as the Blatchford score are helpful when they're zero, indicating no need for an intervention, but higher scores are not helpful for predicting those who do need an intervention.
Let's return to the case that I was looking at of the patient who had two hospitalizations and GI bleeding. The big issue was whether or not the patient had an upper or lower GI bleed. There was GI bleeding, but assessment of its clinical features could distinguish between an upper and lower source, a critical distinction in this case. >> In distinguishing between a lower and an upper GI bleed, the most important thing is not be color blind. A patient who presents with a history of melena or has melena on examination is, quite frankly, the best way to determine whether or not a patient has an upper or a lower GI bleed.
Remember, we define an upper GI bleed as one that's proximal to the ligament of Treitz. Now, certainly there can be some bleeding in the small valve, but that's less common. Most lower GI bleeds are from the colon or anorectal region. In a patient who presents with a self-described melena, they have a likelihood ratio of five to six for having an upper bleed. But if a physician actually views the melena, then that likelihood ratio jumps to 25. So we know that if we actually see the stool, that is a very important factor.
In fact, when I examine patients who present with bleeding, if they are not able to re-stool I do a rectal examination because I know that if I get that information it's almost certainly going to be correct. Alternatively, blood clots in the stool are a high likelihood that it will be a lower GI bleed. In that situation, the positive likelihood ratio for an upper GI bleed is 0.05, which means it's much less likely. Now, one has to take a little bit of caution with that because occasionally a rapid upper GI bleed can present in that manner, but that's uncommon.
>> How does melena factor into this? >> Melena is characterized by the passage of black tarry stools. Most patients will notice that they have black tarry stools, but many patients will not know the clinical significance of that. So I always say that we need to specifically ask patients about their stool because they may not volunteer that information themselves. The other thing to remember is melena is made from the enzymatic breakdown of blood as it transitions through the GI tract, but the biggest determinant of the color is how long the blood has been in the GI tract.
So blood that comes from the proximal GI tract has typically been in the GI tract for several hours to days as it traverses through the tract. However, in a rapid bleed there's cases where patients will pass bright red blood within four hours in a rapid bleed, and that's because the blood has traversed the GI tract very quickly. Alternatively, in a slow bleed from the right colon, one might see melena as it takes quite a bit of time for blood that's slowly been bleeding to traverse the lower GI tract. >> What specific question do you ask the patient?
What words do you use to ask a patient about melena or to describe melena? >> When I ask a patient about melena I always start by saying, "do you have any stool that looks like tar?" Oftentimes providers will ask about dark stool, and many things can cause dark stool. Some patients will call dark green stool dark, or they won't call melena dark. So I particularly like the question about if it looks tarry. Many patients will describe it as particularly more foul smelling than typical stool.
It's a little bit funny to ask a patient if their stool smells, but they will, in general, tell you it smells different than their normal stool. And lastly, there's a stickiness to melena, and I always ask patients about that. If the stool sticks to the bottom of the bowl, or even if they have more difficulty wiping then removing stool at the end of their bowel movement. >> So an upper GI bleeding source is likely if there's a history of melena, but if melena is seen on exam the likelihood for an upper GI bleed goes way up. Upper GI bleeding sources are also likely if the BUN-to-creatinine ratio is greater than 30, and if there's red blood or coffee grounds seen on an NG lavage.
But as we said before, we're not going to do NG lavages anymore. If blood clots are seen in the stool, the likelihood for an upper GI bleeding source goes way down. So we've just covered the factors associated with upper GI bleeds and the characteristics of the upper GI bleeds. Many of the clinical features associated with the presence of an upper or lower GI bleed, or the severity of the bleed requiring an intervention, are discussed in the context of the sensitivity, specificity, and likelihood ratios.
Let's shift gears. Let's just talk about the basic epidemiological concepts of sensitivity, specificity, and likelihood ratios. When talking about diagnostic testing, the tests are always discussed in the context of sensitivity, specificity, predictive values, or likelihood ratios. All of these are presented in the Rational Clinical Examination articles. However, they're confusing, so it's useful to review the concepts. Clinicians are probably most familiar with the sensitivity and specificity of tests.
The sensitivity is the proportion of patients who have the disease that will have a positive test. Similarly, the specificity is the proportion of patients without the disease who have a negative test. All of this is oriented to populations of patients who have or do not have a disease. None of this tells you very much about what to do with a patient in whom you don't know if they have the disease or not, and you're going to get a test to figure it out. For that, you need a measure of how a test result will help you figure out if a patient has some disease when you don't know they have that disease.
We all learn about sensitivity and specificity in medical school. What do those numbers help you with? To better understand all this I spoke to Dr. David Simel. >> So one of the better mnemonics was described by Dave Sackett's group early on as they were popularizing the use of sensitivity, specificity, and predictive values and likelihood ratios. And the mnemonic he developed was for sensitivity, the mnemonic is SnNOut which is S-N-N-O-U-T.
I've modified it a little bit, but the way I interpret it is high sensitivity, negative test result tends to rule out. And the mnemonic for specificity is SpPIn, S-P-P-I-N, which is high specificity test, when you get a positive result tends to rule in. Now, the reason that I say tends, which is the way I've modified it, is that tends simply points in the direction. So a high sensitivity test will tend to rule out when you get a negative result.
A high specificity test will tend to rule in when you get a positive result. But it doesn't absolutely rule in or rule out, and that depends on where you're starting. >> So sensitivity and specificity are of limited value in helping you establish a diagnosis based on some finding or test result. To get a sense for how well a test performed from the perspective of getting the test in a patient in whom you do not know has the disease, you need to think in terms of predictive values. A positive predictive value is a proportion of patients who have a positive test and have the disease.
A negative predictive value is a proportion of patients who don't have the disease who have a negative test. Note the difference in perspective. Sensitivity is calculated from the population of patients who have the disease. Positive predictive value is calculated from the perspective of the population of patients with a positive test. From a clinical perspective, positive predictive value is more useful than sensitivity since it answers the question "if I don't know what's going on and I get a positive test, what's the probability that the patient has the disease?" Sounds good, but there's a caveat with this.
The positive predictive value is calculated from the number of patients with a positive test divided by the sum of the number of patients with a positive test who have the disease, added to the number of patients who have a positive test who don't have the disease. Patients who have a positive test but don't have the disease are considered to have false positive test results. Let's work through an example using information from this Rational Clinical Exam article. I'm going to use the numbers for NG lavage even though I recommend that you not do NG lavages.
The only reason I'm doing this is because of the math. The math for NG lavage is really easy, and it makes it easier to understand how everything works. NG lavage has a sensitivity of 77% for the need for an intervention for patients with GI bleeds. That means that 77% of all patients who need an intervention have a positive NG lavage. For argument's sake, let's call this 75% or three quarters. NG lavage has a specificity of 76%.
Again, let's simplify and just call it three quarters. That means that of all the patients who don't need an intervention, three quarters of them had a negative NG lavage. If you have a patient in front of you, these numbers are of limited value since you're asking if the test is positive or negative in this patient and does he or she need an urgent evaluation. We also note from the Rational Clinical Exam article that about one third of patients with an upper GI bleed need an intervention. That's the prevalence of the disease, or in this case the need for an intervention.
Say you have 1,000 upper GI bleed patients. One third of them need an intervention. The NG lavage has a sensitivity of 75%, or three quarters of all patients who need an intervention will have a positive NG lavage. So if you multiply 333 times 3/4, you get 250 patients. These 250 patients are the true positives. They're the patients who need an intervention and have a positive lavage. They're not the only ones with positive tests, though.
There's going to be false positive tests. Patients with false positives, they're the patient who don't need an intervention but have a positive NG lavage. This is calculated from the specificity. We know the NG lavage's specificity is 75%, or three quarters of patients who don't need an intervention will have a negative lavage. These are the true negatives. It also means that the remainder, or one quarter, are patients who do not need an intervention but are going to have a positive NG lavage.
These are the false positives. Running through the numbers of our two thirds of patients who didn't need a lavage, one quarter of them are going to have a positive test. That means that for 666 patients in our initial 1,000 who don't need a lavage, one quarter of them, or 167 patients, will have a false positive test. The positive predictive value for a positive NG lavage is 250, or the true positives, divided by all the patients with positive tests, again the true positives, added to the false positives, which was 167.
When you do the math, the positive predictive value for an NG lavage is 0.6. Keep that 0.6 number in mind. What would happen if the prevalence of upper GI bleeds needing an intervention was two thirds instead of one third? Then there'd be 666 patients needing an urgent evaluation of that original cohort of 1,000. Given the sensitivity of NG lavage being 75%, 500 patients who'd need an urgent evaluation would have a positive lavage. Of the 333 patients who didn't need an intervention, one quarter, or one minus the specificity, have a false positive test.
This calculates out to 83 false positive tests. With a prevalence of needing an intervention of two thirds instead of one third, there's now 583 total positive NG lavages. When you do the math, the positive predictive value in this case is 0.9, much higher than the 0.6 positive predictive value when the prevalence was one third. This illustrates how the positive predictive value is influenced by the disease prevalence. This means that any scientific report providing a positive predictive value can only give you that number in the context of the disease prevalence for the population in which the test was studied.
For example, the positive predictive value of a malaria test will be much higher in Africa than if the test is used in the US where the incidence of malaria is vanishingly small. Given the limitations of sensitivity, specificity, and the predictive values, the Rational Clinical Examination authors advocate for the use of likelihood ratios. Without going through the math, likelihood ratios are not reliant on the disease prevalence, and also provide a sense for the likelihood of a disease's presence or absence in a single number.
Let's walk through the same NG lavage scenario using likelihood ratios. We know that about a third of all upper GI bleeds need an intervention. That's also called our pretest probability. When we get a test, an NG lavage, and we know we're not really going to do this, but for argument's sake we're going to get an NG lavage to help improve our guess that patient has an upper GI bleed and needs an intervention. Based on the article, the positive likelihood ratio for an NG lavage showing red blood is 3.1. Using the calculator available on JAMAevidence.com, we find that the positive test improves the probability that the patient needs an intervention from 33% to 60%.
In our hypothetical scenario where two thirds of all patients with upper GI bleeds need an intervention, the positive test, i.e., an NG lavage showing blood, increases the probability of needing an intervention from 66% to 86%. Thus, a positive NG lavage is useful. What about a NG lavage's ability to rule out the need for an intervention? The negative likelihood ratio for this scenario is 0.32. If the pretest probability is one third, then a negative NG lavage reduces the probability of needing an intervention to 14% from 33%.
If the pretest probability was two thirds, or 66%, the post-test probability of a negative test is reduced only to 38%. What if the pretest probability was one third and the negative likelihood ratio was 0.03 instead of 0.3? Then the post-test probability would be 2% instead of 14%. This shows how a test with a very small negative likelihood ratio is very effective in ruling out a disease when it's negative. In general, tests that have a likelihood ratio of greater than 10 are very effective for ruling in a disease, and those with a negative likelihood ratio of less than 0.1 are very good at ruling out a disease when negative.
Again, this is Dr. Simel. >> The problem is that in real life you don't know which test to apply to your patient. So remember, sensitivity only applies for the people who need a rapid evaluation. Specificity only applies to the people who don't need a rapid evaluation. And you've got a patient there with a GI bleed and you can't figure out whether or not they need a rapid evaluation, so you don't really know which of those to use. And that's the other reason that we use likelihood ratios because the likelihood ratios combine the information from sensitivity with specificity and allow us to determine the change in probability of disease based on whether or not they have the disease.
The individual values of sensitivity and specificity don't help you with an individual patient because you don't know which of those results to apply. >> We encourage listeners to become more familiar with likelihood ratios. They're really useful in clinical medicine and far more useful than sensitivity or specificity. So going back to the beginning of the podcast, what happened with the case of the patient with two hospitalizations and GI bleeding that I was looking at? During the patient's first hospitalization he had red blood and clots in his stool, and there was no melena.
Not having melena had a negative likelihood ratio of 0.06. If this patient was randomly selected and had a two-thirds probability of having an upper GI bleed, the lack of melena reduced the probability that the patient had an upper GI bleed from somewhere around 10 to 34%. But the patient had clots in his stool. That finding has a positive likelihood ratio of 0.05, meaning that the probability that the patient had an upper GI bleed fell from 66% to 9%.
These findings strongly suggest that the GI bleeding was from a lower and not upper source. When the patient re-presented for his second hospitalization, he did have melena on examination. That has a likelihood ratio of 25. The probability that he had an upper GI bleed when he was admitted the second time rose from 66% to 98%. Indeed, an endoscopy was performed, showing gastric blood and clots as well as a vessel that had clot on it.
This story is important. We need to use the best available evidence in all aspects of clinical practice. The Rational Clinical Examination series helps separate fact from fiction when considering the various signs and symptoms and how they relate to disease. A note about CME you can get CME for having listened to this podcast by going to JAMA.com, looking up the original article by Dr. Srygley published in the March 14, 2012, issue of JAMA, reviewing the charts, tables, figures, and other materials.
Click the CME tab, take the test, and you get CME. More information about GI bleeding is available in the Rational Clinical Examination Series, and on our website JAMAevidence.com where you can listen to our entire roster of podcasts. I'm Ed Livingston and I'll be back with you soon for another addition of JAMAevidence.
Segment:0 .
>> Hello and welcome to JAMAevidence, our monthly podcast focused on core issues and evidence-based medicine. This is Ed Livingston, Deputy Editor of Clinical Reviews and Education here at JAMA. Just to let you know, CME is available for listening to this podcast, if you refer back to the original paper published in JAMA several years ago. We'll tell you more about that at the end of the podcast. Let's start with why the Rational Clinical Examination Series of articles are important. While preparing for this podcast, I happened to be looking at the medical record of a patient who had two hospitalizations.
The problem that I was looking at was that the patient was in the hospital after having emergency surgery and a bowel resection. He had GI bleeding, received a couple of units of blood transfusions, and did pretty much okay. He then was readmitted to the hospital for an upper GI bleed from which he ultimately died. The question was whether in this case the patient had evidence of an upper GI bleed the first time he was in the hospital. Today we take on this topic of upper GI bleeding, which causes nearly 400,000 hospitalizations a year.
There's another 600,000 hospitalizations a year where GI bleeding occurs secondary to some other problem. Every physician needs to know something about GI bleeding. Here's the bottom line for all comers with GI bleeds, about two thirds of them have upper GI bleeds. Of those who have an upper GI bleed, about one third are severe enough to need some sort of an intervention. Here an intervention means needing an endoscopy, radiological procedure, or blood transfusion.
Six features are associated with a need for an intervention. They are the patient reporting a history of having melenic stools, finding melenic stools on physical examination, finding fresh blood in an NG aspirate, having a BUN-to-creatinine ratio of greater than 30, or the presence of a malignancy or cirrhosis. Three features predict the likelihood of needing an intervention. These are having a nasogastric lavage with red blood, a patient having tachycardia exceeding 100 beats per minute, or having a hemoglobin below eight.
There are scores to assess the need for interventions. One of the most well defined is the Blatchford score. This Rational Clinical Examination systematic review found that the Blatchford score is really only useful when it's zero. When the Blatchford score is zero it tells you there is very little likelihood of needing an intervention. When it's not zero it's not quite clear how helpful the Blatchford score really is. That's the end of the bottom line. In this JAMAevidence podcast we discuss the chapter from the Rational Clinical Examination entitled "Making the Diagnosis Upper GI Bleeding." Along the way we'll put a nail in the coffin of gastric lavage as an evaluative method for assessing patients with GI bleeding.
Joining us is Dr. Douglas Srygley, a gastroenterologist at the Austin Gastroenterology Clinic in Austin, Texas, and one of the authors of this chapter. Later in the program we'll talk to Dr. David Simel, editor of the JAMA Rational Clinical Examination series and Chief of the Medical Service at the Durham VA Medical Center, and Professor of Medicine at the Duke University School of Medicine. When first presented with a patient who is GI bleeding, the first question is do they require an urgent intervention?
If you suspect even remotely that a patient might require an intervention, they need to be resuscitated as soon as possible. Once the patient is resuscitated, then you need to know if they need an intervention at all. >> We know from studies that about a third of patients will require some sort of urgent intervention, that being endoscopy, interventional radiology, or blood transfusion. And the best thing is to try to triage these patients out so that not all of them have to be admitted to the hospital and undergo an emergent endoscopy, which is both costly to the system as well as costly to the patient.
And we know that by using some of these predictors, especially the clinical prediction rule, we can rule out at least 15 to 20% of the patients from having to undergo those interventions, at least in an emergent or urgent basis. And while that might not sound like much, it's actually pretty nice for the patient and for the healthcare system if we can avoid those admissions. >> What do you find are the most useful findings to suggest that there is a clinically significant GI bleed? Or clinically important that might require intervention?
>> The most important findings from a historical basis are a history of malignancy or cirrhosis. That carries a likelihood ratio of about 3.7, and that makes sense. We all know that patients who have, particularly cirrhosis, are more prone to things like variceal bleeding or even portal gastropathy, either of which could require endoscopic therapy or potentially even interventional radiology and certainly blood transfusion. So we know that those two, from a historical basis, are very helpful to know. The other one is pulse rate. So a patient who presents with tachycardia has a likelihood ratio of almost five.
And obviously that suggests volume loss, which suggests a more rapid source of bleeding, which is likely going to require some sort of intervention, be it blood transfusion, endoscopic intervention, surgery, or interventional radiology. Lastly, a hemoglobin that's less than eight is obviously helpful. That one is a little more clear. If someone has a low hemoglobin they're likely going to need a blood transfusion or intervention. We define an urgent intervention in this study as anyone who needed blood transfusion, had stigmata or recent hemorrhage on endoscopy, required surgery or an interventional radiological procedure.
>> When you say that tachycardia is associated with a likelihood ratio of five, suggesting the need for an intervention, what does that mean? What does the likelihood ratio of five mean to you? >> So if we take all comers who come in with a GI bleed, we take our pretest probability. If we use that with the likelihood ratio graph, then we can get our post-test probability to be much higher. Typically we want a likelihood ratio greater than two to provide any clinical benefit, and the higher the likelihood ratio the more benefit it provides.
So for instance, if a patient has a pretest probability of 35%, which is in general what a patient presenting to the emergency room would have, then a likelihood ratio of five would certainly suggest to us that we should proceed with an evaluation and admit the patient to the hospital. >> In the article there's a table that has clinical factors of severity of bleeding from the history and physical used to determine the need for urgent evaluation of upper GI bleeds. So coffee-ground emesis has a likelihood ratio of 0.41. So why is it that someone who has coffee-ground emesis has, if I'm interpreting this correctly, a low likelihood of needing an intervention?
>> What one person's coffee-ground emesis is it could be another person's normal bowel contents, I think is part of the issue. I personally think that coffee-ground emesis is one of the more difficult historical features to tease out. Additionally, if a patient is bleeding more rapidly, then the blood will not have time to go into a more coffee-ground material. Remember, blood is changed into a darker material via an enzymatic process, and the longer it stays in the bowel the more dark it becomes, so one would surmise that if the emesis is darker it's less likely to have a more active rapid bleed than one who is throwing up bright red blood.
That's less likely to need intervention endoscopically, radiologically, or with blood transfusion. >> On a similar note, the likelihood ratio for hematochezia was only 0.224 for requiring an urgent evaluation. How does that stack up? >> So I think that really breaks down into we're getting two populations of patients. One population is probably patients who truly have a lower GI bleed. And as we know, oftentimes patients will present with a lower GI bleed which looks like clinically a significant amount of blood loss, but when we follow the patient they do not have nearly as much blood loss as an upper GI bleed.
For instance, we've all taken care of patients with diverticular bleeding where it looks like they're almost bleeding to death and the hemoglobin hardly drops at all over two to three days, versus a patient with upper GI bleeding that clinically might not look like they're bleeding as much but really has a significant bleed. So I think one population that we're getting there is probably patients who actually have a lower GI bleed. The other thing that is concerning, and it makes me a little bit hesitant for that hematochezia, although that's what the data shows us, is that patients can have hematochezia with a rapid upper bleed.
And so that's where, I think, the tachycardia also comes in. If you have a patient with hematochezia, you would say oh, well, this likelihood ratio would suggest it's less likely to need an intervention. But if that patient has a pulse of 120, or shock, then we know we've got to take that a little more serious, and that would push us to act in a more aggressive manor. >> One of the big surprises in this Rational Clinical Examination article is the relative ineffectiveness of NG tubes in assessing upper GI bleeds. When I was a resident, one of the most unpleasant things we had to do was place NG tubes in all the GI bleeding patients and lavage them.
I have many unhappy memories of doing that in the middle of the night in the ICU. >> Where I trained, as a rule, just about anyone who came in with coffee-ground emesis or hematemesis or melena received a nasogastric lavage to try to determine both the severity of the bleed and the location of the bleed. The data for that is actually not as robust as one might wish, as much as anything because most patients who you would go ahead and put a nasogastric tube in you're going to proceed to endoscopy anyways, so it doesn't necessarily add anything.
It's not a great predictor of whether or not a patient is going to require an urgent intervention. But the most striking thing to me was reading articles about how much patients dislike nasogastric tubes. I think as a provider we oftentimes think of this as a benign intervention roughly akin to putting in an IV. Patients do not feel that way. In at least one study, patients rated having a nasogastric tube as being worse than fracture, if they have to get a fracture reduced, or if they have to have an abscess drained, a lumbar puncture, or an arterial stick.
All of those things I would consider things that I personally would want to avoid, and I think we have to be a little bit hesitant to think about nasogastric lavage in these patients if it's not going to add much and put them in significant discomfort. >> Is there a role at all for nasogastric tubes in evaluating a GI bleed? >> They've been evaluated both to try to predict how severe the bleed is, and we know, based on several studies, that the likelihood ratio for a lavage with red blood in it is 3.1 for someone who would need an urgent intervention. If you go down to just red blood or coffee grounds, it goes down to 2.0. So that can persuade someone to go ahead and proceed with an urgent intervention.
But the vast majority of patients are going to have something else in their history, such as a low hemoglobin, tachycardia, melena, that would persuade you to go ahead with a intervention in an urgent basis. So that leaves whether or not placing an NG tube makes endoscopy easier or better. And there is some limited data that doing a nasogastric lavage might improve visualization at the time of endoscopy. It should be noted, though, that that data is quite limited, and there's other much less invasive agents that we could use such as erythromycin or Reglan, which has also been studied and are affective.
I will say in my practice, as someone who primarily takes care of hospitalized GI patients, as a rule I don't put a nasogastric tube in, and I recommend it in general when the emergency department physicians call me. The other thing that I think needs to be brought up is the possibility of harm with a nasogastric tube. There's a teaching out there that patients with varices should not receive nasogastric tubes. There is not really any evidence that that is true, with the one exception of someone who has just recently undergone banding. There's no data to suggest that placing a nasogastric tube in that situation is concerning, but I think anyone who has placed bands on someone who has recently had variceal bleeding would recommend against that.
>> What about rules and things like the Blatchford score to assess bleeding severity and predict the need for an intervention? >> I think the clinical prediction rules have a limited use. The big use of the clinical prediction rule is the patient who appears healthy and appears to be having a minor bleed that we can determine, that we can safely send them out. In the studies with the Blatchford score, depending on which study you look at, somewhere between 5 and 20% of the patients will have a Blatchford score of zero, which means that we can basically safely send them out. >> So a Blatchford score of zero pretty securely tells you that there's no clinically significant bleed and no intervention will be required.
What do you do with a Blatchford score of one or greater? >> In those patients, I think the clinical prediction rules have very limited utility, and I also think that the clinical prediction rule's positive likelihood ratio is so low as to, quite frankly, be totally unhelpful. It technically doesn't cross one, but a likelihood ratio of 1.4 is not going to push you one way or the other. >> So they're really good for ruling out the diagnosis not ruling it in. Is that right? >> That's correct.
It can be a very effective tool to rule out a bleed requiring urgent intervention, but it is not an effective tool to rule in. >> To recap, about two thirds of all GI bleeds that come into the hospital are upper GI bleeds. Of these, about a third are going to need some sort of urgent intervention. That means that they'll require a blood transfusion, GI endoscopy, or some radiological procedure. The most reliable clinical feature suggesting the need for an intervention is the presence of tachycardia with a heart rate of more than 100 beats per minute or a hemoglobin of less than eight.
An NG lavage showing red blood is also a good indicator of needing an intervention. The problem with this is that an NG lavage not showing any blood doesn't necessarily rule out the need for an intervention. This, coupled with the fact that NG tubes are so painful, has led most gastroenterologists to no longer recommend NG tubes when evaluating GI bleeding. Keep in mind this is different than needing them or resuscitating a patient in shock who needs the NG tube to prevent aspiration. Lastly, clinical prediction rules such as the Blatchford score are helpful when they're zero, indicating no need for an intervention, but higher scores are not helpful for predicting those who do need an intervention.
Let's return to the case that I was looking at of the patient who had two hospitalizations and GI bleeding. The big issue was whether or not the patient had an upper or lower GI bleed. There was GI bleeding, but assessment of its clinical features could distinguish between an upper and lower source, a critical distinction in this case. >> In distinguishing between a lower and an upper GI bleed, the most important thing is not be color blind. A patient who presents with a history of melena or has melena on examination is, quite frankly, the best way to determine whether or not a patient has an upper or a lower GI bleed.
Remember, we define an upper GI bleed as one that's proximal to the ligament of Treitz. Now, certainly there can be some bleeding in the small valve, but that's less common. Most lower GI bleeds are from the colon or anorectal region. In a patient who presents with a self-described melena, they have a likelihood ratio of five to six for having an upper bleed. But if a physician actually views the melena, then that likelihood ratio jumps to 25. So we know that if we actually see the stool, that is a very important factor.
In fact, when I examine patients who present with bleeding, if they are not able to re-stool I do a rectal examination because I know that if I get that information it's almost certainly going to be correct. Alternatively, blood clots in the stool are a high likelihood that it will be a lower GI bleed. In that situation, the positive likelihood ratio for an upper GI bleed is 0.05, which means it's much less likely. Now, one has to take a little bit of caution with that because occasionally a rapid upper GI bleed can present in that manner, but that's uncommon.
>> How does melena factor into this? >> Melena is characterized by the passage of black tarry stools. Most patients will notice that they have black tarry stools, but many patients will not know the clinical significance of that. So I always say that we need to specifically ask patients about their stool because they may not volunteer that information themselves. The other thing to remember is melena is made from the enzymatic breakdown of blood as it transitions through the GI tract, but the biggest determinant of the color is how long the blood has been in the GI tract.
So blood that comes from the proximal GI tract has typically been in the GI tract for several hours to days as it traverses through the tract. However, in a rapid bleed there's cases where patients will pass bright red blood within four hours in a rapid bleed, and that's because the blood has traversed the GI tract very quickly. Alternatively, in a slow bleed from the right colon, one might see melena as it takes quite a bit of time for blood that's slowly been bleeding to traverse the lower GI tract. >> What specific question do you ask the patient?
What words do you use to ask a patient about melena or to describe melena? >> When I ask a patient about melena I always start by saying, "do you have any stool that looks like tar?" Oftentimes providers will ask about dark stool, and many things can cause dark stool. Some patients will call dark green stool dark, or they won't call melena dark. So I particularly like the question about if it looks tarry. Many patients will describe it as particularly more foul smelling than typical stool.
It's a little bit funny to ask a patient if their stool smells, but they will, in general, tell you it smells different than their normal stool. And lastly, there's a stickiness to melena, and I always ask patients about that. If the stool sticks to the bottom of the bowl, or even if they have more difficulty wiping then removing stool at the end of their bowel movement. >> So an upper GI bleeding source is likely if there's a history of melena, but if melena is seen on exam the likelihood for an upper GI bleed goes way up. Upper GI bleeding sources are also likely if the BUN-to-creatinine ratio is greater than 30, and if there's red blood or coffee grounds seen on an NG lavage.
But as we said before, we're not going to do NG lavages anymore. If blood clots are seen in the stool, the likelihood for an upper GI bleeding source goes way down. So we've just covered the factors associated with upper GI bleeds and the characteristics of the upper GI bleeds. Many of the clinical features associated with the presence of an upper or lower GI bleed, or the severity of the bleed requiring an intervention, are discussed in the context of the sensitivity, specificity, and likelihood ratios.
Let's shift gears. Let's just talk about the basic epidemiological concepts of sensitivity, specificity, and likelihood ratios. When talking about diagnostic testing, the tests are always discussed in the context of sensitivity, specificity, predictive values, or likelihood ratios. All of these are presented in the Rational Clinical Examination articles. However, they're confusing, so it's useful to review the concepts. Clinicians are probably most familiar with the sensitivity and specificity of tests.
The sensitivity is the proportion of patients who have the disease that will have a positive test. Similarly, the specificity is the proportion of patients without the disease who have a negative test. All of this is oriented to populations of patients who have or do not have a disease. None of this tells you very much about what to do with a patient in whom you don't know if they have the disease or not, and you're going to get a test to figure it out. For that, you need a measure of how a test result will help you figure out if a patient has some disease when you don't know they have that disease.
We all learn about sensitivity and specificity in medical school. What do those numbers help you with? To better understand all this I spoke to Dr. David Simel. >> So one of the better mnemonics was described by Dave Sackett's group early on as they were popularizing the use of sensitivity, specificity, and predictive values and likelihood ratios. And the mnemonic he developed was for sensitivity, the mnemonic is SnNOut which is S-N-N-O-U-T.
I've modified it a little bit, but the way I interpret it is high sensitivity, negative test result tends to rule out. And the mnemonic for specificity is SpPIn, S-P-P-I-N, which is high specificity test, when you get a positive result tends to rule in. Now, the reason that I say tends, which is the way I've modified it, is that tends simply points in the direction. So a high sensitivity test will tend to rule out when you get a negative result.
A high specificity test will tend to rule in when you get a positive result. But it doesn't absolutely rule in or rule out, and that depends on where you're starting. >> So sensitivity and specificity are of limited value in helping you establish a diagnosis based on some finding or test result. To get a sense for how well a test performed from the perspective of getting the test in a patient in whom you do not know has the disease, you need to think in terms of predictive values. A positive predictive value is a proportion of patients who have a positive test and have the disease.
A negative predictive value is a proportion of patients who don't have the disease who have a negative test. Note the difference in perspective. Sensitivity is calculated from the population of patients who have the disease. Positive predictive value is calculated from the perspective of the population of patients with a positive test. From a clinical perspective, positive predictive value is more useful than sensitivity since it answers the question "if I don't know what's going on and I get a positive test, what's the probability that the patient has the disease?" Sounds good, but there's a caveat with this.
The positive predictive value is calculated from the number of patients with a positive test divided by the sum of the number of patients with a positive test who have the disease, added to the number of patients who have a positive test who don't have the disease. Patients who have a positive test but don't have the disease are considered to have false positive test results. Let's work through an example using information from this Rational Clinical Exam article. I'm going to use the numbers for NG lavage even though I recommend that you not do NG lavages.
The only reason I'm doing this is because of the math. The math for NG lavage is really easy, and it makes it easier to understand how everything works. NG lavage has a sensitivity of 77% for the need for an intervention for patients with GI bleeds. That means that 77% of all patients who need an intervention have a positive NG lavage. For argument's sake, let's call this 75% or three quarters. NG lavage has a specificity of 76%.
Again, let's simplify and just call it three quarters. That means that of all the patients who don't need an intervention, three quarters of them had a negative NG lavage. If you have a patient in front of you, these numbers are of limited value since you're asking if the test is positive or negative in this patient and does he or she need an urgent evaluation. We also note from the Rational Clinical Exam article that about one third of patients with an upper GI bleed need an intervention. That's the prevalence of the disease, or in this case the need for an intervention.
Say you have 1,000 upper GI bleed patients. One third of them need an intervention. The NG lavage has a sensitivity of 75%, or three quarters of all patients who need an intervention will have a positive NG lavage. So if you multiply 333 times 3/4, you get 250 patients. These 250 patients are the true positives. They're the patients who need an intervention and have a positive lavage. They're not the only ones with positive tests, though.
There's going to be false positive tests. Patients with false positives, they're the patient who don't need an intervention but have a positive NG lavage. This is calculated from the specificity. We know the NG lavage's specificity is 75%, or three quarters of patients who don't need an intervention will have a negative lavage. These are the true negatives. It also means that the remainder, or one quarter, are patients who do not need an intervention but are going to have a positive NG lavage.
These are the false positives. Running through the numbers of our two thirds of patients who didn't need a lavage, one quarter of them are going to have a positive test. That means that for 666 patients in our initial 1,000 who don't need a lavage, one quarter of them, or 167 patients, will have a false positive test. The positive predictive value for a positive NG lavage is 250, or the true positives, divided by all the patients with positive tests, again the true positives, added to the false positives, which was 167.
When you do the math, the positive predictive value for an NG lavage is 0.6. Keep that 0.6 number in mind. What would happen if the prevalence of upper GI bleeds needing an intervention was two thirds instead of one third? Then there'd be 666 patients needing an urgent evaluation of that original cohort of 1,000. Given the sensitivity of NG lavage being 75%, 500 patients who'd need an urgent evaluation would have a positive lavage. Of the 333 patients who didn't need an intervention, one quarter, or one minus the specificity, have a false positive test.
This calculates out to 83 false positive tests. With a prevalence of needing an intervention of two thirds instead of one third, there's now 583 total positive NG lavages. When you do the math, the positive predictive value in this case is 0.9, much higher than the 0.6 positive predictive value when the prevalence was one third. This illustrates how the positive predictive value is influenced by the disease prevalence. This means that any scientific report providing a positive predictive value can only give you that number in the context of the disease prevalence for the population in which the test was studied.
For example, the positive predictive value of a malaria test will be much higher in Africa than if the test is used in the US where the incidence of malaria is vanishingly small. Given the limitations of sensitivity, specificity, and the predictive values, the Rational Clinical Examination authors advocate for the use of likelihood ratios. Without going through the math, likelihood ratios are not reliant on the disease prevalence, and also provide a sense for the likelihood of a disease's presence or absence in a single number.
Let's walk through the same NG lavage scenario using likelihood ratios. We know that about a third of all upper GI bleeds need an intervention. That's also called our pretest probability. When we get a test, an NG lavage, and we know we're not really going to do this, but for argument's sake we're going to get an NG lavage to help improve our guess that patient has an upper GI bleed and needs an intervention. Based on the article, the positive likelihood ratio for an NG lavage showing red blood is 3.1. Using the calculator available on JAMAevidence.com, we find that the positive test improves the probability that the patient needs an intervention from 33% to 60%.
In our hypothetical scenario where two thirds of all patients with upper GI bleeds need an intervention, the positive test, i.e., an NG lavage showing blood, increases the probability of needing an intervention from 66% to 86%. Thus, a positive NG lavage is useful. What about a NG lavage's ability to rule out the need for an intervention? The negative likelihood ratio for this scenario is 0.32. If the pretest probability is one third, then a negative NG lavage reduces the probability of needing an intervention to 14% from 33%.
If the pretest probability was two thirds, or 66%, the post-test probability of a negative test is reduced only to 38%. What if the pretest probability was one third and the negative likelihood ratio was 0.03 instead of 0.3? Then the post-test probability would be 2% instead of 14%. This shows how a test with a very small negative likelihood ratio is very effective in ruling out a disease when it's negative. In general, tests that have a likelihood ratio of greater than 10 are very effective for ruling in a disease, and those with a negative likelihood ratio of less than 0.1 are very good at ruling out a disease when negative.
Again, this is Dr. Simel. >> The problem is that in real life you don't know which test to apply to your patient. So remember, sensitivity only applies for the people who need a rapid evaluation. Specificity only applies to the people who don't need a rapid evaluation. And you've got a patient there with a GI bleed and you can't figure out whether or not they need a rapid evaluation, so you don't really know which of those to use. And that's the other reason that we use likelihood ratios because the likelihood ratios combine the information from sensitivity with specificity and allow us to determine the change in probability of disease based on whether or not they have the disease.
The individual values of sensitivity and specificity don't help you with an individual patient because you don't know which of those results to apply. >> We encourage listeners to become more familiar with likelihood ratios. They're really useful in clinical medicine and far more useful than sensitivity or specificity. So going back to the beginning of the podcast, what happened with the case of the patient with two hospitalizations and GI bleeding that I was looking at? During the patient's first hospitalization he had red blood and clots in his stool, and there was no melena.
Not having melena had a negative likelihood ratio of 0.06. If this patient was randomly selected and had a two-thirds probability of having an upper GI bleed, the lack of melena reduced the probability that the patient had an upper GI bleed from somewhere around 10 to 34%. But the patient had clots in his stool. That finding has a positive likelihood ratio of 0.05, meaning that the probability that the patient had an upper GI bleed fell from 66% to 9%.
These findings strongly suggest that the GI bleeding was from a lower and not upper source. When the patient re-presented for his second hospitalization, he did have melena on examination. That has a likelihood ratio of 25. The probability that he had an upper GI bleed when he was admitted the second time rose from 66% to 98%. Indeed, an endoscopy was performed, showing gastric blood and clots as well as a vessel that had clot on it.
This story is important. We need to use the best available evidence in all aspects of clinical practice. The Rational Clinical Examination series helps separate fact from fiction when considering the various signs and symptoms and how they relate to disease. A note about CME you can get CME for having listened to this podcast by going to JAMA.com, looking up the original article by Dr. Srygley published in the March 14, 2012, issue of JAMA, reviewing the charts, tables, figures, and other materials.
Click the CME tab, take the test, and you get CME. More information about GI bleeding is available in the Rational Clinical Examination Series, and on our website JAMAevidence.com where you can listen to our entire roster of podcasts. I'm Ed Livingston and I'll be back with you soon for another addition of JAMAevidence.
Form Title Download
Media
Transcript
Form Title Bookmark
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Share
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Accessibility
The cadmore media player is a professional video and audio player experience designed to deliver media as well as relevant transcript, segmentation, and metadata. On the screen is a traditional html audio player with buttons such as play, pause, forward, captioning, language and play speed selection and more. Adjacent to the audio player are generally tabs, one of which shows a scrolling transcript of the audio and the other shows any segmentation of the media, sometimes referred to as chapters. At the top of the audio player is a menu bar containing an Explore button and Tools button. The Explore button allows you to dig deeper into the audio by opening up an explore dialog which contains an about tab which contains rich metadata about the media, a segments tab which contains further information about the audio, and a related tab which contains links to other media related to this media. The tools button opens a menu with options for sharing the media on social media, creating bookmarking links, creating embed codes for your website, generating citations, and more.
Keyboard Shortcuts
Keyboard shortcuts only work while you are in forms mode. By entering forms mode, you will be able to quickly access portions of the media player with single key commands.
Spacebar: Play/Pause Toggle
E: Explore Menu,
O: Tools Menu
T: Transcript Tab,
S: Segments Tab
F: Forward 15,
R: Rewind 15
L: Languages,
P: Playspeed
M: Mute Toggle,
V: Volume Bar
N: Video Only Toggle,
B: Full Screen Toggle
C: Captions Toggle
Escape: Closes dialogs and menus
Z: Next Transcript Chapter
X: Previous Transcript Chapter
Q: Next Transcript Paragraph
W: Previous Transcript Paragraph
Form Title Embed Media
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Embed Size
-
You Size - Responsive
-
We Size - Responsive
-
Picture Overlay Popup
-
400 x 225
-
512 x 288
-
400 x 700
-
768 x 1024
-
1024 x 512
-
1280 x 608
Form Title Cite
No citation provided.
Info
Segments
Related
Thumbnails
Accessibility
Using forms mode, you can utilize keyboard shortcuts to make usability more efficient.
Space bar will play and pause the media.
You can press A to open the Accessibility Dialog.
This describes many other keyboard shortcuts for media controls.
Exit Clip Mode
The video is currently set-up to show just a clip from 0:00 to video end. You have requested to see a portion of the media outside of the clip.
Click OK to switch to the full media or click if you would like to stay inside the clip. You can always go back to the clip with Tools and Clip Mode