Time Stamps

  • 02:04 Initial Presentation 
  • 03:03 The CoreIM team tackles the case 
  • 09:30 Social History 
  • 16:05 Physical Exam  
  • 22:47 An exercise in Bayesian inference 
  • 36:35 Pre-mortem
  • 43:37 Revealing the diagnosis
  • 48:57 How do we define diagnostic success? 

HumanDx Interactive Case

Show Notes

  • When considering the possibility of a travel-related illness in a patient, it is critical to define the potential incubation period as precisely as possible. Many, if not most, travel-related diseases can be excluded on this basis alone.
  • There is no uniform definition of pulse-temperature dissociation, but von Liebermeister’s rule is a historical rule-of-thumb stating that for every increase in temperature of 1C, the pulse should increase by 8-10 beats per minute. Relative bradycardia is a characteristic (but not consistently-observed) finding in infections caused by intracellular pathogens.
  • Bayes’ theorem implies that our suspicion for a diagnosis after receiving a test result should be determined by two considerations:
    • Our initial suspicion for the diagnosis prior to performing the test
    • The performance (i.e. sensitivity and specificity) or the test itself.
  • While this concept may be intuitive in the abstract, the practice of actually calculating posterior probabilities using Bayes’ theorem can sometimes illustrate in surprisingly tangible terms the importance of prior probability and test performance on the interpretation of a test result.
  • Hindsight bias refers to the tendency of clinicians, when retrospectively reviewing their cases, to overestimate their ability to have foreseen the outcomes of those cases. Hindsight bias complicates our efforts to critique our performance retrospectively via “cognitive autopsies”.
  • Pre-mortem examination is a meta-cognitive technique in which a clinician first assumes their diagnosis or management was erroneous, and then tries to prospectively identify causes for their failure. This approach can potentially heighten a clinician’s “diagnostic attentiveness” when dealing with seemingly mundane clinical scenarios, and lessens the clinician’s dependence on purely retrospective methods of case reflection.
  • Acquired cerebellar ataxia does not always indicate a focal cerebellar lesion. A parainfectious cerebellar syndrome is characteristic of certain systemic infections, including severe legionellosis. CSF studies are unimpressive, suggesting that direct invasion/infection of the CNS is not the cause.


JOHN Hey everyone, John Hwang here, with Cindy Fang, bringing you another episode of Hoofbeats, where we challenge you to solve diagnostically difficult, real-world cases alongside fellow clinicians. Notice I didn’t say “master diagnosticians” this time?

CINDY And that’s because the discussants for our episode this week are actually members of our own CoreIM team.

JOHN Even though none of us here at CoreIM are master diagnosticians — yet — Cindy and I thought interviewing doctors in our own relatively junior cohort could still shed some valuable light on the reasoning process.

CINDY At the very least, they’re easier targets.

JOHN But maybe in all seriousness we can trust them to be more forthright when dissecting their own thought processes, including when breakdowns occur. Let me tell you from experience: Having no reputation to protect can be tremendously liberating.

CINDY That sounds like me on my teaching rounds every morning. So we bribed and cajoled three of our colleagues to talk through this case with us. This was a patient John and I had on our service when we were interns. You’ll hear the case presented in discrete chunks of information, and you’ll hear our team’s reaction after each one.

JOHN As always, we want you to play discussant along with them, so take a moment after each bolus of information. Also, we’re teaming up again with our friends at HumanDx; they’re publishing an interactive version of this case on their platform for you to solve. Links in the show notes. This is also a good time to mention that HumanDx is running a weeklong challenge specially featuring cases of altered mental status and neurologic disease, so definitely check them out — remember, practice makes perfect.

CINDY Alright, so now that we’re on the same page, let’s move on to the case. John, do you want to go ahead and present the first chunk?

JOHN Ok, here we go. So this is a 23 year-old man brought in by EMS for confusion and disorientation. For the past four days, he’s been having flu-like symptoms. He reports fevers, chills, muscle aches, sore throat, a cough productive of yellow sputum, and nausea with an episode of nonbloody vomiting. His friends noticed during this time he’s been kind of “off”, not quite talking like himself, though this was subtle. However, on the day of admission, they found him lying on his bathroom floor disoriented and confused, unable to tell them what had happened. At that point he was brought to the hospital. First up at the plate is our very own Dr. Shira Sachs, fellow Hoofbeats writer and host, hospitalist at Cornell.

SHIRA: So I’m thinking mostly in an infectious category right now. With these four days of fever, myalgias and sore throat being relatively non-specific symptoms. I’m thinking about, is this a viral URI…?

And then in the altered mental status, with this disorientation for the last three days, I want to be able to tie potentially any of those viral infections to his disorientation. Is he dehydrated leading to a metabolic abnormality, be it uremia, hypo or hypernatremia? Does he have a meningoencephalitis directly related to an infection?

JANINE: …And I would say putting on my primary care hat, if this guy showed up in clinic or in the emergency room was like a first point of contact, he really perfectly fits my illness script for viral meningitis, because he’s a young, otherwise (I hope) healthy guy who just, he compensated really quickly, in a way that’s very unusual for a young guy — the syncopizing, the having an altered mental status just from a viral infection. So I would jump straight to that and then try to round out what else could I be missing? Is, is not meningitis at all? Is it not an infection at all? But that’s right where my mind went. And especially with the measles outbreak happening right now, it’s hard not to go straight to meningitis. But then I really liked the way Shira really laid out all the other systems that could be involved.

JOHN That’s Dr. Janine Knudsen, writer and producer at CoreIM’s Mind the Gap, and a primary care physician colleague of mine here at Bellevue. You hear her openly acknowledging her use of pattern recognition and early anchoring in her initial remarks.

STEVE: I mean, one episode of nausea and vomiting is kind of surprising to me. The fact that he was found collapsed in his bathroom, makes it does seem like it’s more of a severe viral illness than you would normally expect. If you told me that you don’t really know what’s going on with this guy, I’d probably figured he was probably having a lot of diarrhea or vomiting. If this was a Monday morning, some of this might have been from him partying too hard over the weekend. So it’s hard to say, but, I’d want to know more.

JOHN: Is projection a cognitive bias?

STEVE: I like to think of myself as the common man, John.

SHIRA: John, are you throwing shade to your discussants?

JOHN: Just to Steve.

CINDY And last of course, that is Dr. Steven Liu, writer and producer along with Janine at Mind the Gap

JOHN And — as he’s quick to remind us — suffering from the effects of a lingering cold, which apparently excuses him from having to justify any diagnostic errors. …So, let’s get deeper into this  case, before Cindy and I jump in and start picking everything apart. Cindy, do you want to give us the next chunk of information?

CINDY He has a history of Crohn’s disease diagnosed six months ago via endoscopic biopsy. He is on infliximab and azathioprine; his last infliximab infusion was within the past month. And he has not been having symptoms recently.

JANINE: Yeah. Well all of a sudden we have an immunosuppressed 23 year old, so I think that changes a lot. Definitely leans more towards infection, just to sort of state what I think we’re all thinking. But then you sort of wonder, autoimmune disease? Like is there anything else that could masquerade and look sort of like this, that’s autoimmune? I think I would turn to Google for this one.

JANINE: I would definitely look at autoimmune diseases associated with Crohn’s. That’s the first place to go, just not doubting that diagnosis to start. And then if I were to doubt that this is Crohn’s, and maybe there’s some other syndrome that could cause all these things together, I’m not thinking of it off the top of my head. So I might try to look up, you know, a vasculitis, because that’s an autoimmune disease that might be more likely to affect more organ systems. But now I’m really, really reaching; I think I’m still considering infection is the top, reason.

SHIRA: Yeah, I’d want to think about what infectious etiologies this patient could now have related to his azathioprine and infliximab exposure. And we always think about TNF-alpha inhibitors in TB, but I want to do a deeper dive. You know, if we end up going down the meningitis pathway. Should we be thinking about less typical bacterial infections in this population, like Listeria? Should we be thinking about, you know, viral agents I typically wouldn’t think of in this patient like VZV or something like that? Um, and then again, just like Janine, I was also wondering, you know, it would be nice to know a little bit more about his Crohn’s history. Is there a different vasculitis that can also have CNS manifestations and GI manifestations? Or is he just really unlucky and has a second autoimmune process that can cause CNS manifestations as well?

CINDY One thing I do want to point out, how reliable is his past medical, or any history we get,  from this guy coming in with ams? Did he omit any other important information?  Remember, plenty of diagnostic errors happen at the stage of data acquisition, even before the stage of data interpretation.

JOHN Yeah… Exploring that skill of data acquisition might be something we want to explore in the future. It’s going to require a new format, though.

Changing tack a little, something else I will say just listening to them talk. And this comes from knowing Janine and Steve and watching them firsthand on the wards — and I’m certain Shira is this way too: They’re very candid about what they don’t know. You hear them talking about not knowing off the top of their head all the diseases associations with Crohn’s. And I can tell you, that’s how they talk, whether we’re talking in a group of peers like this, or they’re teaching in front of students and residents. I bring this up because I think it takes guts, and goes counter to everything I encountered. Between MCATs, classes, STEPs, work and attending rounds, writeups, etc — when was I ever incentivized or encouraged to be wrong? But you hear these folks here, for lack of a better phrase, reverse pimping. Instead of asking questions that only they could answer, they make a point of deliberately asking questions whose answers they don’t know.

JOHN Moving onto our third data chunk. So he lives in the UK and is traveling here in the US on vacation. He arrived in NYC three weeks earlier, and has stayed in the city, not really pursuing any unusual activities. He’s staying at a hotel with his friends, and is frequenting bars and pubs, but hasn’t drunk since the onset of his illness four days earlier. Back in England, he lives in an urban environment, specifically the city of Sheffield, works as an electrician, and keeps no pets.

JANINE: I’m googling mad cow disease, just so you know, really reaching, really reaching.

SHIRA: No, I just, I just Google searched prion disease England to see what the stats are right now!

JOHN: I love how one of you typed in “mad cow disease” even though she’s a doctor.

SHIRA: Yeah. This information is good to know. You know, he’s a young guy. He’s involved in the bar scene like we mentioned earlier. Could there be any other toxic exposures that could explain his altered mental status? Even though we’re thinking more along the lines of an infection, you know what, anytime someone’s traveling, you really need to get a sexual history because people tend to have new sexual context when they’re not in their home environment. And, uh, as we already stated, you know, acute HIV is certainly on our mind. So I think this just kind of continues to bring up other things we might want to be thinking about, but I don’t think it sways us too far in any direction. Any other thoughts team?

JANINE: No, I completely agree. I think, you know, England’s not a country that I associate with particularly rare or strange infections except, prion disease,  which I think is mad cow specifically. Which I think is probably just a factor of like weird facts I learned in medical school and probably less relevant to this patient. But it doesn’t really set off any alarm bells.

JANINE: …the travel happened three weeks ago and I’m just trying to think of any infection that has an incubation period of three weeks, but that it gets so acutely worse all of a sudden and I, there’s not a lot of that comes to mind so it makes me really wonder if the travel is actually associated with his current presentation or whether this is all something he got once he actually arrived here. You know, just cause you know, infections on airplanes obviously very common, but I think for him it’s probably more local.

STEVE: … Is there a MERS outbreak that’s occurring right now too?

JANINE: Yeah, I was going to make a joke about him making a connection through Dubai, but I didn’t think that was appropriate. [laughter] I don’t even know where MERS is now, but…

SHIRA: That would be a very circuitous route from England to New York. Yeah. [laughter]

JANINE: Yeah. Bad geography knowledge.

SHIRA: He took the SpiritAir of England to get here…

JOHN Thanks, Shira, that’s one potential sponsorship for CoreIM we can remove from our wish list. …Janine’s point, though, is worth reiterating, right Cindy? When we consider the possibility a patient has a diagnosis related to travel, whether it’s a febrile illness, or diarrhea, or what have you, the single most important thing is to define the potential incubation period as precisely as possible. That has tremendous discriminative power.

CINDY Right. In this case, it’s been three weeks since between when he traveled and when he arrived. There aren’t a ton of diseases that take that long to get going.

JOHN I bet we can even name them all here. Certain parasites. Like falciparum malaria incubates for a month, and other species of malaria can go for even longer, like vivax. Visceral leishmaniasis, schisto.

CINDY Viruses. Like HIV and most of the hepatitis viruses.

JOHN Right… TB, of course. This kind of a list is, at least to me, much more manageable than trying to remember — “er… which countries is leishmaniasis endemic in?”

CINDY None of these diagnoses really comes into play in urban England, except possibly for HIV, as Shira says. So the team puts aside this data point.

CINDY Alright. Fourth data chunk. On examination, he is febrile, to 104.6. He’s tachycardic to 128, has a blood pressure of 134/88, and tachypneic to 22, a normal oxygen saturation of 98% on room air.

JANINE: Yikes.

SHIRA: He’s sick. Pay attention.

SHIRA: I mean I see those vitals and I need to see this patient right away. I might need to collect thoughts or look more stuff up later, but I need to go evaluate this patient immediately.

JANINE: Yeah, exactly. I don’t know. Not, this is hopefully not going to make it into the recording, but I’m trying to remember if viral infections are actually proven to lead to higher fevers than bacterial? Like if the numbers here would really lead us in one direction or the other, but I’m not really remembering.

SHIRA: I mean I guess there’s certain infectious agents that cause that like temperature – pulse dissociation, which this patient doesn’t have. [JANINE: Yeah…] I don’t typically think of it necessarily helping us one way or another about the actual ID bucket.

JOHN I always wondered what actually defined pulse-temperature association, though. Like, is his moderately elevated HR of 128 disproportionately low relative to his very high fever of 105? As it turns out, there isn’t necessarily a uniform definition across case studies of this phenomenon. There was a 19th century German physician who specialized in research into fever, and has a semi-quantitative rule of thumb named after him — von Leibermeister’s rule: For every increase in temperature of 1C, the pulse should increase by 8-10 beats per minute. This patient has a temperature of 40.6 C, 4 degrees hotter than the normal body temp of 37, so his heart rate of 128 is, according to this rule, to be expected.

CINDY You know this is a clinical reasoning podcast right?

JOHN Yes, yes, yes. First of all, you were the one… Sometimes I can’t help indulging in trivia. I will say, sometimes trivia is a useful scaffold for… …Let’s just move on. [laughter]

JOHN Data point number five. His neurologic exam is abnormal. He’s awake, oriented, and answers questions, but appears tired and listless. He cannot perform serial 7s past 93, or spell words backward. His speech is scanning; in other words, slow and halting, are there are noticeable pauses in between syllables within a word. He struggles with rapid alternating motions and exhibits dysmetria bilaterally — in other words, he overshoots when asked to point. He is unsteady on his feet, struggling to get out of bed, and when he does, his stance is wide.

JANINE: Definitely am more worried about something in the central nervous system, specifically his cerebellar dysfunction. I’m not sure I would necessarily say this is just a cerebellar infection. I would still just be worried about meningitis in general, but this is very concerning.

SHIRA: Yeah. There’s some global features of his encephalopathy, his inattention, his speech difficulty… but there are specific cerebellar signs, you know. I think we’ll probably get head imaging early on. We want to make sure there’s no mass lesion, but also going down sort of the meningitis category, we want to think about things that can cause a vasculitic-like picture and stroke, like symptoms, like something like VZV. 

JANINE: Um, Google is great. Am I allowed to tell you what I found?

STEVE: A case report with John Hwang?

JANINE: Well, no, not, not that good. [laughter] I went there and just googled infection, cerebellar infection. And there’s a great article I found summarizing the differential in this journal. What journal is it… neuro neurologic clinic journal? And the pathogens most frequently affecting the cerebellum are Listeria, VZV, JC virus, and CJD. I feel validated. Um, and I mean there’s, there’s a few others that they list, some other viral infections, but I think, uh, I think that’s somewhat helpful. But again, like Shira, I don’t want to put all my eggs in one basket, only anchor on that.

JOHN: Steve, you’re handicapping yourself by speaking last each time you do know that…

JANINE: I think he’s googling, we know what he’s doing. [laughter]

STEVE: I mean I wish I had something more interesting to say. 

STEVE: I’m honestly so one thing I get caught up, and this is why I don’t leave the hospital on time often, is that I get like an interesting idea in my head and then I will just try to stretch the case reports to read about it. So earlier when we were talking, I guess when, I guess I can backtrack a little bit, when we were talking about the cerebellar findings, I was wondering to myself, well, one infections do I know that go to the base of the brain. And so I was conflating base with cerebellar, so that I was thinking of things like HSV meningitis, and I was reading more about that to see what I could find in terms of clinical findings. Um, the other two folks shared much more, I think, accurate description of what you’d find, what kind of viruses you’d find that go to the cerebellum. And so at that point I kind of gave up on that search and now I’m reading about PCP in infliximab patients for no particular reason except for, I just felt like it.

JOHN: I’m sorry that we’re boring you.

STEVE: No, you asked me what I was doing, so I was telling you…! [laughter]

CINDY So now we hear these folks trying to incorporate cerebellar dysfunction into their thinking. And there seems to be agreement the cerebellar dysfunction sounds unusual enough that it probably adds specificity to this case, but no one’s pattern recognition is being activated yet — you see Janine reaching again for her external google drive brain.

JOHN You can also see how, in the absence of a fast answer, they are instead mapping out the possible syndromic diagnoses here. What I mean by that is, they’re proposing different mechanisms that might connect the infection they’re assuming he has to his cerebellar dysfunction. Shira and Janine both talk about infectious cerebellitis, direct viral invasion of the cerebellum.

CINDY And Shira brings up a mass lesion, like an abscess sitting on the vermis that could be causing cerebellar ataxia.

JOHN Steve considers the more broad category of infections that go to the base of the brain, like HSV encephalitis.

CINDY And they’re still considering that this could just be an infectious meningo-encephalitis that just happens to have prominent cerebellar exam findings.

JOHN Another category that isn’t explicitly brought up here is parainfectious cerebellitis — a cerebellitis caused not by direct infection of the cerebellum, but an indirect immune-mediated syndrome that occurs in the aftermath of an infection. Not to foreshadow too much, of course. Ready for the next data point?

CINDY Alright. It’s a lumbar puncture. Obviously this was preceded by a head CT, which was read as normal. In terms of the LP, the CSF looks normal, colorless and clear, with a normal opening pressure. The WBC is slightly elevated, with 13 cells, all lymphs and monos. Protein is 19, glucose is 78, both normal. Grain stain is negative, cultures don’t grow, and PCR for HSV and VZV are negative.

JANINE: So the first thing I want this to be is a false negative, but it probably isn’t, because that would just be so easy. Um, and the next thing I’m thinking is this does not lead me away from infection. Like I’m still super worried about a meningitis. So what are the meningitides that cause, that can still have a normal CSF. Um, and so I would go down that route next.

SHIRA: I agree. I mean, there are 13 cells in there, so that’s certainly not totally normal. And I’d want to kind of think about, sure, if this isn’t a typical bacterial meningitis, we talked about listeria, we already know our HSV and VZV are negative. I’m going to expand my database and look online about what other meningoencephalitides could this be, and continue to think about other things that could cause this presentation at the same time.

STEVE: I agree. I think that you’ve moved away from bacterial clauses obviously, but the you’d want to see how often do you could get a false negative while you have somebody on infliximab.

JOHN I think it’s interesting how Janine says that this mostly-normal LP does not exclude bacterial meningitis for her. …Cindy, when was the last time you used Bayes’ theorem to actually manage one of your patients?

CINDY What do you mean, like actually calculated probabilities of disease based on test results? [John: Yeah.] To be honest? Almost never.

JOHN I don’t think many of us do. We learn it in medical school, and personally, I used to think it was enough to just know the principle. But I think that actually doing simple Bayesian calculations from time to time can really make this idea tangible. I think Janine’s question about the LP is a good opportunity to do this. 

So the basic idea of Bayesian inference is that how we interpret a test result, like a negative LP, depends on two things: How likely we thought the disease was initially, and how good the test itself is.

So let’s say Janine’s initial suspicion for bacterial meningitis is 10%. And let’s say the sensitivity of LP for bacterial meningitis is 88%. I’m getting that number from a 2004 NEJM study. Let’s also assume for the sake of convenience that the LP is also 88% specific for bacterial meningitis. Which, I think they are.

So, if Janine’s pretest probability is 10%, and she goes online and plug these numbers into a Bayesian calculator, it tells her that the probability of bacterial meningitis following a negative LP falls to about 1.5%.

CINDY That’s pretty low. I have a feeling many if not most clinicians, in the absence of other worrisome indicators, forgo empiric meningitis treatment.

JOHN  Agreed. But let’s say that her suspicion is higher. Let’s say she thinks it’s more likely than not, say 60%. If we plug in the pre-test probability as 60%, along with the same sensitivities and specificities of 88%, into that Bayesian calculator, you get a post-test probability of 17%. A 17% posterior probability this patient has bacterial meningitis despite a negative LP. That number is much more discomfiting, wouldn’t you agree?

CINDY Yeah, I think most of us would feel uncomfortable discontinuing abx when there’s a one in six chance of missing bacterial meningitis.

JOHN I’d like to think that it’s enough to understand the implications of Bayes theorem in the abstract. But truthfully, this happens time and again — when I take the time to do these calculations, I’m always surprised by how wildly the post-test probability can swing, depending on what my pre-test suspicion for the disease, and how well that test performs. I mean, a 1% chance of meningitis with a normal LP versus a 15% chance. That’s huge. I don’t think my gut really “believes” in Bayes, even if my head does. And so I think this is a good argument for why we should make it a habit, at least from time to time, to practice these kinds of Bayesian calculations.

JOHN So the next chunk of data. Routine admission labs. He’s hyponatremic to 131. His transaminases are moderately elevated; AST 357, ALT 111, with otherwise normal LFTs. And his cell counts are deranged. He’s leukopenic to 2.6 (74% neutrophils, 20% lymphs, 5% monos), his Hgb 10.9 (MCV 81), and his platelet count is 87.

JANINE: So this is what I was worried about that if we got the full set of labs and we saw abnormalities… I was thinking there’ll be some liver involvement and then I was thinking maybe there would be electrolyte derangements and I don’t think this was, I didn’t think this would help me and I still don’t think it sort of changes the next steps in terms of going straight for the MRI, still focusing on the brain., which is frustrating, but it is what it is.

STEVE: I think that’s probably, I mean, so like you might expect some inflammation, but at the same time he’s on infliximab. So him having pancytopenia is probably not the most surprising thing in the world and it fits along with our whole idea that he’s immunocompromised. I feel like that was the least surprising slide you’ve showed us so far.

JANINE: I don’t know though. I mean, I, you know, not everyone on infliximab is pancytopenic, so I would still be worried that something else is going on. The first thing I’ll think of his marrow suppression. But it definitely makes me think, Oof, I don’t want to miss something else that’s a lot bigger or different than an infection that I haven’t already thought about.

JOHN: Shira? You’re leaning back in your chair, Shira.

JANINE: Shira is the deep thinker of the group, clearly. [laughter]

SHIRA: Yeah. I mean, I think similar to what Janine said, I’m curious, you know, is this bone marrow suppression from whatever primary infectious process is causing his other symptoms and signs. And Janine, can I ask you to specify what you meant when you don’t want to miss something else?

JANINE: I mean, I think, you know, we talked at the very beginning about other autoimmune conditions or vasculitides that could masquerade as infection or could be contributing to the picture and so I’m back in that category again, just thinking what else could there be, but nothing immediately pops into my mind for that.

CINDY And we are down to our last data point. The patient has a chest X-ray done. We had our discussants look at the image without interpretation, but for the benefit of you folks listening out there, I’ll say that it showed a left upper lobe consolidation which was corroborated on a non-contrast chest CT.

JANINE: Oof… [long pause] I don’t like it.

JOHN: Why not? Is it, is it over penetrated?

JANINE: I was really hoping there would be nothing in the lungs. And maybe I’m over reading the left apex…?

STEVE: No, you’re not over-reading the left apex.

JANINE: I was like, if I read that I need to go back to my radiology skills. I was really hoping there’d be nothing in the lungs because the oxygenation status is normal. I know he had a slight cough, but I was hoping that’d be something in his throat. So now we’re, you know, tuberculosis is back up there. Other, you know, pneumonias, they’re back on the list so it broadens the differential.

STEVE: …So you’ve given me a guy with a pneumonia. He’s got cerebellar findings, as everybody’s pointed out, and he is coming in relatively septic. So I would try to link those three findings together. And what I usually end up doing is probably building differentials based off of one or two of my key points, probably what’s going to cause pneumonia and cerebellar things and build from there. : My guess is, I might start off with causes of an acute pneumonia that might then spread to the brain. And so that’s what I’ve been thinking about recently.

STEVE: But I mean, otherwise, what I think of is like strep pneumonia, going to brain and causing strep meningitis.

SHIRA: Yeah. Along those lines, I also want to make sure I think about any atypical like lung plus brain infections. And for me, like nocardia, actinomycetes, but I think particularly nocardia. I want to think about and look up if it could be consistent with this patient’s presentation.

STEVE: I think of actinomyctes and nocardia as being more slow.

SHIRA: Yeah. Agreed. But the same with TB.

STEVE: Definitely.

JOHN: Be careful what you tell TB to do. That’s a Bellevue lesson right there.

STEVE: I think that’s an anywhere lesson dude. [laughter]

CINDY So at this point, we start pushing our team for closing thoughts, much to their chagrin. And on what they would do diagnostically and therapeutically.

JANINE: I would say we have a young, a young man on an immunosuppressant medication coming in with a few days of global infectious symptoms, found to have altered mental status with cerebellar signs, possible pneumonia and signs of systemic inflammation.

SHIRA: Yeah. I said we have a young immunocompromised patient presenting with altered mental status found on exam to have fever and cerebellar dysfunction with prominent lab abnormalities including pancytopenia and hepatocellular transaminitis with a left upper lobe consolidation, aka WTF.

JOHN: I’m not familiar with that syndrome…

SHIRA: Cut that out! Please cut that out, John!

JOHN: Steve, do you use problem representation? [laughter]

STEVE: Only when specifically asked, but…

JOHN: I actually genuinely value you saying that. I’m curious, uh, that there’s a segment of people out there, I think that would argue that these things are entirely artificial constructs.

STEVE: Well, almost everything’s an artificial construct. What do you mean? [laughter]

JOHN: Thanks Steve, that includes our friendship.

STEVE: I mean like these are just different, but I mean I think these are different ways of thinking, right? So it’s different modalities that you use.

SHIRA: Uh, yeah, I’ll try to explain what, where I’m trying to wrap my head around, you know. Clearly we’re worried about underlying infection with sepsis, pneumonia and then we have to tie together has altered mental status, pancytopenia transaminitis. We talked about the viral things I still want to investigate include HIV, which we haven’t done. I totally agree with Jeanine. The EBV and CMV remain on my list. I again, I don’t know. Yes, it’s a primary left upper lobe consolidation but they both can certainly cause transaminitis and pancytopenia. And so I’d probably get PCRs and serologies for both and get HIV fourth generation antigen antibody testing.

SHIRA: And then this altered mental status. You know, is there a lesion we didn’t see on the CT scan? Are there small microabscesses or something that we we didn’t see on CT? Again, direct CNS invasion or is this just global encephalopathy from his underlying sepsis?

SHIRA: And I would definitely get an MRI with contrast, blood, urine and sputum cultures. I’d send the other noninvasive workup just because… [John: What, what, what would that be?] Strep, legionella urine antigens. But I think in this patient, you know, if our cultures don’t grow anything, we’re going to have to pursue further invasive testing, whether that’s a bronch or something like that to try to get an underlying etiology. And again, I’d send testing specific testing for HIV, EBV, CMV at this point. And  I probably do broad spectrum antibiotics. You already told us the HSV PCR was negative so I probably would hold on the acyclovir for now. I’d probably give him vanc, cefepime, and I think the ampicillin for meningitis at this point. Like we already talked about how a CSF is not really so consistent with the bacterial meningitis, but I would probably go broad spectrum gram positive gram negative coverage. I don’t know that I’d pulled the trigger on, on antifungal coverage just quite yet.

JOHN: Vancomycin, cefepime and ampicillin?

SHIRA: I honestly think I would just

JOHN: Yeah, vancopime. Steve?

STEVE: Vanc, ceftriaxone, azithro make sense to me. Um, I don’t know…

JOHN: Vanc, ceftriaxone, azithro at pneumonia dosing?

STEVE: Meningitis dosing. I think it’s reasonable to cover initially for now, I think. You know, even if this ends up not being bacterial meningitis, the real harm I think given that you’re likely to start similar antibiotics anyway is probably relatively limited if you just go ahead and go full out there. Um, I don’t know that, I wouldn’t necessarily think you need the broad spectrum gram negative. I don’t think anybody would fault you for doing it. So vanco-pime is just as good as vanc/ceftriaxone. I would say.

JOHN To sum up at this point, our group is in general agreement about the essential features and the essential problems that this case poses — what can cause lobar pneumonia and encephalopathy with prominent cerebellar features in a young patient on a TNF-alpha inhibitor, associated with a variety of lab abnormalities? They don’t have an exact diagnosis, but they know what they’re going to read up on.

CINDY And they agree generally on a treatment plan: continuing empiric treatment for meningitis as well as lobar pneumonia, and pursuing investigation of the CNS and lung processes in parallel.

JOHN Before we move on to reveal the diagnosis, let’s stop. Take a moment to lock in a diagnosis. And I’m curious — did you find this case challenging? Do you suspect your answer is incorrect? If, assuming that your diagnosis is ultimately incorrect, or that you failed to arrive at one altogether, what do you think might have been the reason why? We asked our discussants this, before revealing this answer.

JOHN: I mean also my other question for you is, is to do a bit of a premortem here and just if, if you think that you’re not going to get the diagnosis in this case, why do you think that that is? How you diagnose the problem ahead of time?

JANINE: Um, yeah, I would say two things for me. One is that he’s immunocompromised, so I have really broadened my differential and hone in on him, but also honed in infection, so both broad and narrow at the same time. And then the other piece is that I really anchored on it being, an encephalitis or meningitis and we have the CSF findings that are not leading or are actually excluding a lot of it, the diagnoses I had. And so that sort of narrowed me into a corner that I feel stuck in. Based on that.

JOHN: Do you mean, are you implying, Janine, that it’s hard for you to move on from your initial framing of the case? Even when I give him the evidence to the contrary?

JANINE: Um, yeah, I think so. It makes me think that I need to change my problem representation, um, to something else.

SHIRA: I mean I think initially we were appropriately concerned for a meningoencephalitis and I don’t think it’s wrong to be concerned early on about things that you never want to miss. And then I think, you know, we got the physical exam and we’re trying to tie it to that initial suspicion with his cerebellar findings that as we got the CSF, we were trying to backtrack and explain our thought process. And then, you know, we got more and more systems being involved very rapidly and then got the lung imaging at the very end. So I think it was kind of slipping how we were thinking about the case once we had already really generated a lot of disease processes and ideas in our mind and realizing that maybe what we were emphasizing, we were looking at like the wrong, the wrong piece of the puzzle as the centerpiece.

SHIRA: Yeah. I think instead of thinking ID, under ID, meningoencephalitis, we’re now couching it in ID with sepsis and an infiltrate in an immunocompromised patient, and is the altered mental status more encephalopathy related to that underlying infection. Whereas previously we were really trying to tie together infection and altered mental status as the primary process instead of as a secondary phenomenon.

STEVE: I just think that the part that throws me off is that with encephalopathy, it makes sense that he would have a lot of the waxing waning mental status, but there do seem to be some slightly more focalized neurologic findings so that I wouldn’t necessarily expect that. So yeah, I would love to just say that you guys gave us really bad CAP. [laughter] That’d be like awesome. [laughter] Because you guys all know Neal Shapiro has been wanting that for CPC since he was like, I don’t know, just graduated..

STEVE: But I think it’s, I think what’s limiting me right now is that I don’t know that I have a good explanation for that. So it’s hard for me too, as tempting as it would be — I mean, in reality, probably what I would do in this case would initially see how much he would improve just with initial treatment. Um, like, cause you know, you get a patient and then they kind of sit overnight, uh, and then you obviously try to keep them stable, but you want to see what it goes onto the back, what happens the next day. Um, see if things do improve. If those cerebellar findings go away, then great. It’s CAP. Did we have to go further? But um, I dunno, I can’t resolve that very easily for myself right now. And so I think leaves me pretty confused as to what I think is actually going on.

CINDY Some common themes emerging here. Shira and Janine both imply that their thinking may have been constrained by the way the case was framed, even though they’re consciously aware of this possibility. And Steve points out the differential remains prohibitively wide because they aren’t confident about the  significance of his cerebellar findings..

JOHN This technique, assuming that one has failed and trying to diagnose where things went wrong ahead of time, is called pre-mortem examination. It’s a strategy commonly used in business management to identify weaknesses and opportunities in projects, but has been transplanted into clinical decision making.

CINDY This is typically the opposite of what we do. The more common method is to analyze our mistakes after they happen, and no doubt that’s an important process, the so called cognitive autopsy or post-mortem. But retrospective review has the potential to be distorted by hindsight bias: The tendency for ourselves to rationalize, “Well, I knew that all along.”

JOHN There’s a famous experiment that illustrated this type of bias, published in 1972 by researchers Baruch Fischoff and Ruth Beyth. At that time, Richard Nixon was president and scheduled to visit China. The visit was unexpected. The US and China hadn’t had formal diplomatic relations since the communist regime was established in 1949, and Nixon had a reputation as an ardent anti-communist. No one knew what would happen. The authors took this opportunity and gave subjects a list of hypothetical outcomes, such as “The US will establish a permanent diplomatic mission in China,” “The US will refrain from formally recognizing the Chinese government.” Fischoff and Beyth asked the subjects to rate how probable they thought these outcomes were from 0 to 100%. Then, months after Nixon’s visit had concluded, the subjects were asked to recall how likely they thought each outcome was. The authors found that, for events that actually transpired, subjects would consistently overestimate — they would say they had thought it was 80% likely formal relations would be established, when initially they had only put down 40%. The opposite was observed for events that didn’t happen. Fischoff and Beyth interpreted this as evidence that humans tend to anchor in their present reality, and we have difficulty retrieving their past states of mind with any degree of accuracy.

CINDY And that’s why pre-mortem examination comes in useful. Imagining something has gone wrong with your patient in the real world is one way to structure a diagnostic timeout. I often ask on rounds, If this patient with a cellulitis does not improve on the abx you are about to start, why and what do you want to do? Is there an alternative diagnosis, such as an underlying abscess or DVT, or is it a management error, such as choosing the wrong abx that does not cover the right pathogen, or subtherapeutic abx due to drug drug interaction? I find this useful in preventing medical errors too – I am actually  known for asking my residents the same questions almost everyday – Of all your patients, who is the most likely to die on you today? How? What can you do now to lower the chance of the likely bad outcome or at least prepare yourself for the rapid response?

CINDY So, I think the time has come to discuss the diagnosis.

JOHN Yes. I want to state for the record that the treating team’s thought process — at least if we’re remembering correctly — closely resembled the path that our CoreIM discussants took. Everyone was confused, and the patient’s mental status abnormalities loomed largest in our minds.

CINDY With regards to some of the tests the CoreIM team wanted, or you may want: The HIV test was negative. Hepatitis serologies were negative. Herpesviral serologies were negative. Blood and urine cultures were negative. He was indeed put on antibiotics to cover both meningitis and community-acquired pneumonia — vancomycin, ceftriaxone and azithromycin — and within 72 hours steady improvement was apparent. The brain MRI/MRA, like his CT head, was normal.

JOHN So what were the diagnostic tests? Two. A urine antigen assay which returned positive for Legionella, and a sputum culture which grew L. pneumophilia. This was a severe case of legionellosis.

CINDY He ultimately made a full recovery, and returned home safely.

JOHN This case has stood out to me for the past seven years, I think, because of how challenging the diagnosis was to make prospectively, yet how it made perfect sense retrospectively. Every piece of data is consistent with the illness script for legionellosis. You have the patient himself, a young man who is immunosuppressed — and more specifically, who has impaired cell-mediated immunity while on a TNF-alpha inhibitor — of specific relevance to intracellular organisms like Legionellosis. He’s staying in a hotel, which, along with hospitals and pools are one of the settings commonly associated with legionellosis — since aquatic ameba and protozoans that live in water storage tanks serve that are the natural reservoir for legionella.

CINDY And you have the illness itself. A lobar pneumonia, acute onset with florid systemic inflammation, and plenty of extrapulmonary manifestations: GI symptoms of nausea and vomiting, transaminase elevations, hyponatremia, cytopenias.

JOHN In the real world, our surprise at the diagnosis I think was driven by in part by a knowledge gap, which is that neurologic abnormalities are very commonly observed in severe legionellosis, and that acute cerebellar ataxia is a classic neurologic syndrome seen with legionellosis. Some of the earliest case descriptions of legionellosis, dating to the late 1970s, just after the disease was discovered, make specific note this cerebellar syndrome.

CINDY The cerebellar ataxia syndrome seems to be an indirect, probably immune-mediated consequence of legionella infection. LPs are typically what was seen in this case — normal or almost entirely normal, and cultures don’t grow anything. So this does not represent direct invasion of the CNS or meningo-encephalitis.

JOHN This case helped me refine my mini-diagnostic schema for pulmonary-CNS illnesses. Hematogenous spread and metastatic disease from a primary pulmonary infection to the brain, as seen in pneumococcal pneumonia, opportunistic mycoses like Crypto, and tuberculosis figure prominently. But now I’ll always remember that legionella and mycoplasma, two organisms notorious for extrapulmonary weirdness, sometimes cause parainfectious CNS abnormalities, which can sometimes really dominate the clinical picture.

CINDY Our CoreIM discussants  sounded disappointed, though I don’t know why. They formulated a problem representation that essentially matched the diagnosis’ illness script, and they selected appropriate empiric treatment as well as ordering the correct diagnostic test.

JOHN Yeah, I have to say, they didn’t seem happy with us. Steve, though, said something which I think deserves mention.

STEVE: But I, I guess my thought would be at any point all of us, we’re going to make sure that we covered this guy for CAP. Right? We didn’t necessarily miss the diagnosis per se, um, in the sense that we were going to cover for it. But what I heard a lot of was I don’t want to miss something more serious.

STEVE: We all want to think that the added value that we bring to our patient care is our brilliance in diagnosis or ability to figure out a clever way of doing something. But you know, you, you talked to me about liking to give antibiotic talks and this is like the old phrase, the enemy of good is better, right? Trying to go for an elegant solution is often the stupid solution because the reality is like 95% of doctors would have ended up treating this appropriately because they would have treated for 99% would have treated for CAP. And actually probably doing the Legionella outbreak in 2012 which we were all residents during, you would have tested for this reflexively if that would have been the right call to make, and I would say a lot of New York doctors now during the summertime do that, especially when it’s been wet. Right? And so yeah, when we were talking, you guys were talking about let’s get blood cultures, let’s get sputum cultures. We said that multiple times. And so if our subconscious thinking activates it and it’s automatic, sometimes that is good enough.

CINDY Steve is saying even though they didn’t explicitly submit the correct diagnosis — they didn’t recognize that this patient had a classic legionellosis syndrome — they identified the clinical problem with enough accuracy that they would have treated the patient properly in the real world — sending off the right tests to make the diagnosis, and starting the right antibiotics up front.

JOHN I rib Steve alot, but I will say, I do not think this is not just him trying to save face for his group. How do we define diagnostic success? Many times, maybe even most times, it isn’t in convincing yourself or others you’ve got the right final diagnosis a priori. It’s in ensuring the well-being of your patient. Success for this patient in the real world, I would argue, would be in quickly excluding life-threatening CNS infection, in thoroughly evaluating for both usual and unusual causes of pneumonia, and ultimately, in ensuring he received early and appropriately broad antimicrobial coverage — including the macrolide or quinolone that would ultimately cure his legionellosis. .

CINDY The diagnostic strategy with the highest accuracy isn’t necessarily the optimal strategy for a doctor to use. Think about it: Let’s say every single PE scan you ordered came back positive for PE. You don’t get to brag that you’re a master at diagnosing PEs. This just means you’re not scanning enough people for PEs!



  • van de Beek D, de Gans J, Spanjaard L, Weisfelt M, Reitsma JB, Vermeulen M. Clinical features and prognostic factors in adults with bacterial meningitis. The New England journal of medicine 2004;351:1849-59.
  • Fischhoff B, Beyth R. I knew it would happen. Organizational Behavior and Human Performance 1975;13(1):1–16.
  • Johnson JD, Raff MJ, Van Arsdall JA. Neurologic manifestations of Legionnaires’ disease. Medicine 1984;63:303-10.
  • Kassirer JP, Wong JB, Kopelman RI. Learning clinical reasoning. Baltimore, MD: Wolters Kluwer/Lippincott Williams & Wilkins; 2010.
  • Klein G. Performing a Project Premortem. IEEE Engineering Management Review 2008;36(2):103–4.
  • Nigro G, Pastoris MC, Fantasia MM, Midulla M. Acute cerebellar ataxia in pediatric legionellosis. Pediatrics 1983;72:847-9.
  • Travelers’ Health [Internet]. Centers for Disease Control and Prevention. 2017 [cited 2019 May 21];Available from: https://wwwnc.cdc.gov/travel/yellowbook/2018/post-travel-evaluation/fever-in-returned-travelers
  • Weir AI, Bone I, Kennedy DH. Neurological involvement in legionellosis. Journal of neurology, neurosurgery, and psychiatry 1982;45:603-8.
  • Ye F, Hatahet M, Youniss MA, Toklu HZ, Mazza JJ, Yale S. The Clinical Significance of Relative Bradycardia. WMJ : official publication of the State Medical Society of Wisconsin 2018;117:73-8.


Tags: , , ,