Geriatrics teaches us that older adults with infections often present with non-specific symptoms rather than typical localizing symptoms of infection present in younger adults. Sometimes they present with fever, delirium, malaise, or fatigue. In today’s GeriPal/JAGS joint podcast, Jeff Caterino challenges this common teaching by examining the extent to which non-specific symptoms are predictive of infection for older adults presenting to the emergency department. Turns out – they’re not so predictive as you might think!
This is a hugely important issue – antibiotic use and misuse among older adults is rampant, and not only endangers these patients, it endangers others by creating antibiotic resistant organisms.
Eric: Welcome to the GeriPal PodCast! This is Eric Widera.
Alex: This is Alex Smith.
Eric: And Alex, who do we have as our guest today?
Alex: Today we have Jeff Caterino, who is professor of both emergency medicine and internal medicine at the Ohio State University in Ohio. Welcome to the GeriPal Podcast, Jeff.
Jeff: Thanks Alex, it’s great to be here.
Eric: We’re going to be talking about a recent paper you published in JAGS this Last month. On non-specific symptoms, and the lack of diagnostic accuracy for infection in older patients in the emergency department. But before we do, we always start off with a request for a song. Do you have a song for Alex to play?
Jeff: So I asked for something by Dire Straits, and I think he’s got Romeo and Juliet.
Alex: Yes, I think this’ll be our first Dire Straits.
Eric: Mixing it up!
Alex: Yeah, Dire Straits is an amazing band. Mark Knopfler, incredible guitarist. I will be a poor shade of that. We’ll see what happens. [Singing]
Eric: A little more, later on in the podcast.
Alex: Oh, it’s fun. I love Dire Straits. Actually, Mark Knopfler’s coming to Berkeley in the fall. I’m trying to convince my son, who loves him, to go see him. It’ll be his first major concert.
Eric: Is there a reason you picked that song?
Jeff: I think I’m stuck in anything I heard in middle school. I’m not quite sure, and I also know that you guys have done a lot of Beatles and a lot of other things like that. So I wanted something a little bit different.
Alex: Yeah. Yeah. It’s great.
Eric: It’s great. So let’s get into the topic at hand. These non-specific symptoms for bacterial infections. How good are they? But before we talk about that specifically, how did you get interested in this as a topic?
Jeff: So, it’s the same problem that we see in emergency departments that everybody else sees. We get older patients with fatigue or altered mental status, and we can’t find a reason for it. The difference is we’ve only got an hour or two to figure it out. We don’t have the luxury of waiting a few days to see what happens.
And we do what a lot of folks do, which is default to, “Well, oh, it must be a urinary infection, or a bacteremia that we haven’t picked up.” et cetera. And we know that’s not always the case. But we don’t have anything better. So we’re trying to look at some of the dogma about that altered mental status and fatigue and malaise and everything else. And see how true it is.
Eric: Yeah, that’s a tough one. Let’s imagine the 88-year-old female coming in from a nursing home, with some change in mental status, malaise for the last couple days. The easy thing is, check a UA. UA’s positive, treat her for UTI and send her back to the nursing home. But odds are that UA was positive a month ago, or three months ago. A year ago. Then it’s not really a UTI.
Jeff: And there’s a lot of layers to that. Because we also know that that something made that patient come to the emergency department instead of just staying in the nursing home. So, I think one of the challenges is, when you compare to the literature in the nursing home or the community-dwelling population and the physicians’ offices, we see a somewhat different population. And so, testing some of these algorithms and testing these ideas in the emergency department itself, I think, is important.
Alex: Yeah. And I like the way you’re challenging dogma here. Because it is sort of geriatrics dogma, that older patients may not mount a fever. Older patients may have non-specific symptoms like malaise, lethargy, altered mental status. And that may be the presentation of a serious infection. And that may be the clinician’s only clue as to what’s going on.
And so, is that in fact, helpful information? Those sorts of non-specific symptoms? Right?
Jeff: Yeah. Exactly. And I think we know that those sort of non-specific symptoms are more common in patients with infection. But the question is, does their presence … that’s kind of the backwards question. The question moving forward is, when you see the patient with those symptoms, how likely is it that they do have an infection?
And that’s really the diagnostic question that I think really hadn’t been answered before.
Eric: So do they?
Jeff: The answer is no. Well, what we actually did in the paper was we enrolled. And this was part of some NIH-funded research. But, we enrolled all comers over 65 to the emergency department. Because one issue is if we only enroll those diagnosed with infection, we don’t know who we missed.
So when you take the entire population of older adults coming to the emergency department, if you look just specifically at altered mental status, fatigue, all those sort of non-specific things, they really do not help you decide if there’s an infection present or not.
Alex: Now, I want to just drill into this a little bit more. You use some … First of all, the population was the emergency department in your emergency department in Ohio, is that right?
Jeff: Yep, that’s correct. So tertiary care, urban emergency department.
Alex: And big volume. Lots of folks coming through.
Jeff: 70 or 80,000, which is in the higher half. But not the biggest.
Alex: Not the biggest.
Alex: Got it. And of those, how many had non-specific symptoms?
Jeff: So we enrolled a bit over 400 patients. Out of those, about half had malaise or lethargy. And somewhere, depending on how you measured it, around 10 to 15% had altered mental status. And when I say “depending on how you measured it,” we looked at both patient report of their own symptoms, we looked at CAM-ICU tests that we did, and we also looked at whether the emergency physician charted that there was altered mental status.
Alex: Yeah. So there’s a 424 patients who have bacterial infection.
Jeff: No, no, no. These are 424 patients; 18% of them had bacterial infections.
Alex: Oh, okay, got it. 77-
Alex: … 18% of them had bacterial infection.
Jeff: Correct. So out of a general ED population, just about 20% will be infected. But out of that same general population, something like 10 or 15% will have an altered mental status, and almost half will have malaise, lethargy, fatigue, just sort of generally feeling unwell.
Alex: Yeah. That’s a high proportion.
Eric: So to be included in this, you had to be 65 years and older, any other … Obviously you go to the emergency department. Any other inclusion criteria?
Jeff: You had to be English-speaking, able to fill out, either you or a legally authorized representative, able to give consent, which means that there was a small proportion of this sort of “the sickest of the sick” that we missed.
But to be honest, the diagnostic decision making in those patients is generally pretty easy. If they’re too ill to participate in a brief survey, they’re usually going to be admitted anyway.
And there were a few exclusion criterias. We excluded those who were evaluated by our trauma team. Sort of the major traumas. If they presented with suicidal ideation, that sort of thing.
Eric: So you have these individuals, older individuals, 65 and older, come into the ED. You assess for these non-specific symptoms. And then you also assessed whether or not they had a bacterial infection. Is that right?
Jeff: Yep, that’s right.
Eric: And how are we defining … because I guess that’s the thing, that’s the crux of the question. How do we define what a bacterial infection is?
Jeff: So, this is incredibly difficult. And this may be another dogma that I challenge a little bit. The traditional definition of infection … take urinary infection. Includes the presence of urinary symptoms, plus a positive culture. Plus usually something else, often something else systemic.
The problem is, we know if you look in like the bacteremic UTI literature, where people have E. coli in their urine, and their blood. You know they have a urinary tract infection. Over half of them will not have symptoms. These are geriatric patients.
So, I was concerned about how we truly define infection. And what we decided was to steal something from the radiology, and actually the dermatology literature. Where we use three experts to look for agreement. They review the entire chart, including the ED, the cultures, everything that happened afterwards. And made the determination if infection was present or not. And that’s how we got our 18%, our 77 infections.
Eric: Do they have any criteria they were using specifically for, let’s say, pneumonias or UTIs? Or was this just their best judgment?
Jeff: These are sort of experienced infectious disease geriatrics and emergency medicine docs, who are familiar with the criteria. But they were also able to use their best judgment as well. So they did not have to stringently follow the criteria. And the interesting thing is the agreement, they agreed over 95% of the time, as to whether acute infection is present.
One of the other issues, just again, if you take an example of UTIs, there’s a large proportion of geriatric patients who have dysuria just from vaginal dryness, urethral irritation, et cetera. So, the sort of infection control definitions aren’t always 100% accurate.
We did some additional analyses if we used more specific criteria, those done by IDSA and CDC. And it really didn’t change the results. But I really think that if three physicians with expertise in this area, geriatrics infectious disease/EM, agree that this was probably an acute infection, it was probably an acute infection.
Alex: Now, you did something interesting. Is there more than you want to say on that one? Maybe we’ll save that for later. I know that Eric’s been listening to some Tom [Finucane] in preparation for this podcast. So I suspect there’s more there.
But, you did something really interesting here. The main way in which you describe your results is in terms of likelihood ratios. Most of our listeners are practicing clinicians, not researchers. So I wonder if there’s a way in plain English that you could explain what a likelihood ratio is. And why that might be the best approach in this particular case.
Jeff: Well, it’s much less cool if I can do it in plain English, but I’ll try. What a likelihood ratio does is it takes the sensitivity and specificity, and it tells you the post-test probability of disease, given the presence of the characteristic or the test.
So, you get both a positive and a negative likelihood ratio for each of these symptoms in this case. You could do it just as much for white blood cell count or anything else. And if the likelihood ratios are in the appropriate range, they would significantly alter your post-test probability of disease. Either that it’s present or that it’s absent.
And really, this is the way we use diagnostic tests, sort of intuitively. We have pre-test probability of how likely we think something’s present. And then we order a test. And it’s enough to either rule in or rule out, or it doesn’t help us at all.
I think that the best, most commonly used thing would be D-dimer for pulmonary embolism. If we decide that there’s a low clinical suspicion, a negative D-dimer gives us enough post-test probability that there’s nothing there, and we can move on. So that’s, I think, the clearest example that I think most of us are using in daily practice.
Alex: Mm-hmm (affirmative).
Alex: No, go ahead.
Jeff: Oh, I was just going to say, the other nice thing about likelihood ratios is they don’t depend upon the sort of prevalence of the condition of the population. Which is really important. Sometimes you see very high predictive values, positive and negative predictive values. And they’re really just a function of how common the disease in the population we’re studying.
Alex: Right. Right. So the likelihood ratio … give you an example. Say somebody comes to the emergency department. You think maybe 50% chance they have a urinary tract infection. And so then what does the presence of non-specific symptoms … like lethargy, or altered mental status … add, or subtract, to that likelihood ratio? Does the presence of lethargy increase it by a substantial proportion, so that instead of 50%, after counting for that, you think, “Oh, maybe 85% chance they have a urinary tract infection”? That would be a really helpful-
Alex: symptom, or does the absence of it? Okay, they don’t have altered mental status. And now, instead of 50%, now maybe 45%. So it doesn’t help that much. It didn’t really change the likelihood ratio. That would be a not very helpful negative likelihood ratio. Does that sound right?
Jeff: That’s exactly how you do it. So the example from the paper, for altered mental status. And again, we tried to measure it a couple of different ways, looking for any signal we could get. Is it based upon the patient telling us, is it based on what the physician saw and charted, et cetera.
But for altered mental status, the positive likelihood ratios were around 2. And that increases your post-test probability of disease by about 15%. So, in your example, instead of saying, “Oh, it’s 50/50,” now maybe there’s a 60% chance that they have an infection. Which really isn’t enough, that going from 50 to 60%, or 50 down to 40%, isn’t really enough to change, to guide what you’re going to do.
When you push likelihood ratios up to 5 or 8 or 10, then you’re talking about 30 to 40 or 45% changes. So, a strong likelihood ratio, a positive likelihood ratio of 5 or 10, would take you from 50% chance of disease to 80 or 90% chance of disease. In which case you would then obviously go ahead and treat.
And that’s really what we saw. For altered mental status, it basically changed both the likelihood for or against disease, by about 5 to 15%. And the same thing for malaise and lethargy. Was really in the 5 to 10% range, as far as a change in likelihood of disease. Which isn’t enough to do anything to help you make a diagnosis.
Alex: Eric, were you going to say something?
Eric: Was there anything that you did find that was helpful?
Jeff: Fever. Fever is strongly predictive of infection in older adults. And I think we intuitively knew that. Interestingly, the sensitivities are poorer. Most older adults with fever don’t have … I’m sorry. Most older adults with infection don’t have a fever. But if one is present, the likelihood ratios were up in the 5 to 10 range. So they really increase the likelihood that infection was there.
Alex: So fever may be useful. But altered mental status and lethargy, fatigue, not so useful. Is that sort of the big takeaway?
Jeff: I think that’s the big takeaway. And I think probably we should talk about the one thing that we didn’t do, because of sample size in this paper, was we only looked at these things in isolation. That’s one of the main limitations, that we didn’t really combine this with other clinical findings, or fever versus not fever. So, it’s limited in that way.
But I think if you just have altered mental status, and don’t have anything else to hang your hat on, I think what we found is that it doesn’t help you say there’s infection or not an infection.
Alex: And this is a really important point. The beauty of likelihood ratios is that they’re sort of, what we’d call, researchers get a little lingo here, Bayesian. Meaning that they take into account in the real world, you’re not doing these tests without having some prior thought or hypothesis about how likely it is that this person has this condition. Has an infection or not.
In the real world, you have some sense, going into doing the test, of whether you think the patient has, or does not have. On the other hand, as you point out here, the likelihood ratio is only about the addition of that specific test.
Whereas clinicians tend to think more in terms of pattern recognition, right? Like, “I see somebody’s older who has this, that, and the other. And this lab finding, that to me is X condition.” Right?
Jeff: Exactly. And this gets into some of the problems with what are those other factors? And so we all look at urinalysis results. But we know that those are misleading in older adults, as I think you mentioned before. Chest X-rays, if you ask multiple radiologists, if there’s an infiltrate on a chest X-ray, you may get multiple answers.
So, this is I think one piece of the puzzle. And I definitely think what we’ve shown is that in isolation, these are not helpful. We’re going to need to do more work in the emergency department setting, specifically, to look at various combinations.
And actually we’re right now about halfway done with enrolling a large study of over 500 patients, specifically with suspected urinary tract infection. Over the age of 65, and I think we’ll be able to do some of the things that you were just talking about, really look at the combination of these factors.
Jeff: And start drilling down more on that.
Alex: That’s great. All right, we’ll have you on again.
Eric: Does this change how you practice in the emergency room? When you have, let’s say, an elderly individual coming in with these non-specific symptoms, maybe some altered mental status, maybe some malaise.
It seems like one important point is just because they have this, it doesn’t mean that they have, they’re delirium is secondary to a bacterial infection. It could be a whole host of other issues. Are there other key takeaways that should change how we practice from this study?
Jeff: I think so. And I think you also have to remember sort of the limitations of what can happen in the emergency department. And, that lack of time course, I think, is critical.
What this is doing for me is a couple of things. It’s warning about premature closure, about just writing off that altered mental status to infection, instead of looking for other causes. Although we tend to be pretty broad when we get these folks.
The second thing is, and I think this in combination with some of the other thought in the field, is particularly for those patients who are clinically stable. Stable blood pressure, not tachycardic, not septic basically, or at least not with systemic inflammatory response. Holding off on antibiotics might be reasonable.
Some other criteria include no elevated white blood cell count. I’ve seen, there’s more talk about this. And in those who are sort of rock-solid stable, if you hold on antibiotics, then it’s kind of finding a good plan. You may still wind up keeping them in the hospital for observation. You may be able to get them in with their physician the next day, which can be dicey, when it’s 8 o’clock on Thursday night, or Friday night.
But the idea that you don’t have to make that diagnosis immediately in the currently stable patient, I think, is something that has to slowly percolate through the emergency medicine community. And probably other acute care settings. Physicians’ offices, et cetera.
Eric: Yeah, I was just listening too. So there’s a great podcast as this came out on The Curbsiders, with Tom Finucane. Did I pronounce that right?
And part of what he also said was the idea of “I’m just going to prescribe antibiotics to be on the safe side,” is the completely wrong idea, because antibiotics are not just safe. There are risks associated with prescribing these medications. Risks to the individual. Another study from JAGS, or Journal … Medicine, showing that it also increases risks for other nursing home residents, not just the individual when we have a lot of prescribing of antibiotics.
And that actually probably the better part of valor is hold off on prescribing antibiotics unless they’re really sick, and that you really feel that they need to be on them, until the cultures come back and you can stop them.
Jeff: Yeah. And I haven’t heard his podcast but I’ve read some of his stuff, including I think in JAGS, about a year or so ago, he had a really nice article. And I agree to a point. I definitely think if you look at a lot of, and I’m not sure how to pronounce his name either, so I apologize for butchering it. But if you look, a lot of his work, it’s very clear that there’s a difference, especially in the nursing home population, between the very clinically stable patient, who has one finding, and those who may exhibit other signs of illness.
And again, that could be fever, it could be softer blood pressures, it could be elevated white counts, other signs of infection. Actually, I highlighted some things as I was preparing for this in his article. There’s lots of things, lots of little phrases like “those not sick”, “those who are not acutely ill”, et cetera.
And I think that criteria hasn’t been well defined yet. And I think we want to be careful about overly broadly applying it in some of the acute care settings. Particularly, the ED or the inpatient setting. So that’s the cautionary word.
At the same time, I think there’s some fundamental truth there, that in those who are the most stable with the best follow-up, with the least comorbidities, I tend to agree with him. Is that good enough hedging?
Eric: I think that’s great hedging, ’cause the other thing, I will include a link to Tom’s article on our GeriPal web site. But he encouraged the use of air quotes whenever we use the term “UTI”.
Eric: I think that’s great. The important part too, is what you mentioned. Early closure of a diagnosis. Is that we rest on this idea that maybe their altered mental status is from the UTI. We call it a UTI, which I’m going to now use air quotes for, whenever I say the word “UTI”.
But it may be something else. It may be the medication that they just started a couple of weeks ago that’s causing their altered mental status. And maybe another reason that they’re having fevers, that it’s much more worrisome than from the urinary tract source. So really, I love the idea of using these air quotes as a way to show that I am unsure of my diagnosis, whenever I’m using the word UTI.
Jeff: I think that’s reasonable in certain circumstances. Someone else from up in Minneapolis has kind of coined the term “bacteriura” or “pyuria”, of clinically undetermined significance. Which I think is kind of interesting. And probably gets to what we’re saying. It looks like it was James Johnson up in the Minneapolis VA system, who wrote an article in American Journal of Medicine a year or so ago.
And I think you’re right. I think, though, we have to have an open mind, and not swing the pendulum too far the other way. Where we have such an incredibly strict definition of infection, that we miss the, what I think are true atypical presentations. And to me, the middle ground is a recognition that there’s a portion of the population that are, like I said, so very stable that deferring and close, watchful waiting is appropriate.
There are going to be others where if you’re unsure, you don’t want to get too far behind. Because we know how important in patients who become septic early on antibiotics are, et cetera. And maybe that study hasn’t been done yet. And I think that may be where things are going over the next five or 10 years.
Alex: Right. And I think it’s interesting, you critique some of the, like the AGS Choosing Wisely. What is it called … a statement? Or recommendation. About avoiding antibiotics and asymptomatic bacteria. By noting that they’re, this is largely based on evidence from the outpatient setting. And that when patients are presenting acutely in the emergency department, we don’t have as much evidence.
Jeff: Exactly. And I think we’ve seen this in a lot of conditions. If you look at the re-stratifications of patients presenting with chest pain. The literature is, I think, pretty clear that patients with chest pain presenting to their doctor’s office are different than the cohort presenting to the emergency department.
At the same time, over the last decade, we’ve swung the pendulum way too far, to testing everybody, stress testing everybody. And most recently, things are coming bock to emergency department validated scores like the HARP score, and for 30-day cardiac events and other re-stratification instruments that are now in our guidelines, where I think we found that middle ground. And I think we’ll see practice change in that over the next several years. So, I think you do have to do the work in the acute care setting, because there is something different about that population.
Alex: Terrific. Is there anything else that you want to address in this article?
Jeff: I don’t think so. I think the one thing I’d mention is that the article is primarily about infection in general. We do have some tables in the article looking specifically at pneumonia and urinary tract infection. Those are more preliminary because of the just small numbers of individual infection types in the study. So, I ask any reader to take those with a bit of a grain of salt.
Although there’s some interesting preliminary findings, for UTI in particular, the percentage of patients with urinary symptoms that our experts thought had UTI were low. And, in the sort of 50% to 60% range, that had specific urinary symptoms.
But when you look at the people who they chose as having UTI, it’s pretty clear. 100,000 colonies of a single organism got better with antibiotics, had a fever, et cetera. Just didn’t have the geno urinary symptoms.
So I think there’s some preliminary word there, but it’s certainly not the last word at the end of the individual infection level.
Alex: Right, right. Maybe there’ll be more coming out of your current study, which is focused on UTI.
Jeff: I think so. That’s the hope.
Alex: Nice. Well, terrific. Thank you so much, Jeff. This was terrific.
Jeff: Well, I really appreciate the opportunity, guys. This was great.
Eric: And how about we end with a little bit more of that song, Alex?
Alex: Okay, I’ll give it a try.
Eric: And with that, we’d like to thank all our listeners for joining us this week. If you have a second, please rate us on your favorite podcasting app. And we’ll see you next week.
Alex: Till next week, thanks Jeff!
Eric: Thank you, Jeff.
Jeff: Thanks, guys.