Time Stamps

Show Notes

  1. “Listen to your patients, they are telling you the diagnoses.” (Osler-ism)
  2. Occam’s razor and Hickam’s dictum can be useful diagnostic tools, but they should be applied to the reasoning process with flexibility and caution.
    1. These rules are typically applied based on the clinical context. Occam’s razor tends to hold true in young, immunocompetent patients, whereas patients with immunocompromised state, advanced age or with lack of access to health care can have multiple unrelated diagnoses.
      • Occam’s razor suggests that there is such a thing as the one true diagnosis.
      • Hickam’s dictum suggests more often there are many true diagnoses. 
      • However, do NOT expect patients to follow them like clinical rules.  
    2. It can be very tempting to forget Crabtree’s Bludgeon and coerce unrelated data into a coherent theory. 
      • Crabtree’s Bludgeon questions whether any diagnosis can be safely considered “true”, since ad hoc explication is so tempting for physicians.
    3. On the other hand, physicians may discard a valid hypothesis prematurely due to the existence of outliers or when data appear to be unrelated at first glance.
  3. A problem representation is a mental model we construct that summarizes our understanding of the problem, it is not the same as summarizing all of a patient’s abnormal findings. 
    1. Data collection, interpretation, and abstraction need to take place before the generation of problem representation, and errors in each step can lead to a faulty PR.
    2. When we have difficulty solving the problem at hand, it is helpful to retrace the steps instead of being committed to solving the perceived problem. You can’t get to the right answer by trying to solve the wrong problem!
    3. Problem representation should be dynamic and flexible

Sponsor: Glass Health builds a medical knowledge management platform that helps physicians learn medicine faster and leverage their knowledge to provide better patient care. Use promo code COREIM for 1 month of free access to Glass Pro!

Attribution Link: https://glass.health/coreim
Promo Code: COREIM


Cindy: Hi everyone, Welcome back to another episode of Hoofbeats, where we challenge you to solve diagnostically difficult, real world cases alongside experienced clinicians. I’m Cindy Fang, a hospitalist at NYU

John: And I’m John Hwang, also NYU faculty working at Bellevue Hospital in New York City.

Cindy: Imagine you are the day team resident or the attending physician listening to a case on teaching rounds. An overnight resident is going to tell you about an admission.

Hey, this patient just rolled up half an hour ago, I’m still working on the admission but let me tell you about  her really quick. She is pretty hard of hearing and is Spanish speaking so I couldn’t really do a full interview with the interpreter phone yet, PLUS the daughter insisted on telling me what’s going on herself, so most of the information is from the ED and her daughter at bedside. But long story short, she’s here for melena and has a bit of AKI, ED has already called GI for us, oh she’s also got pneumonia too, ED also called pulmonary for us.

Cindy: What, ED didn’t call renal for us?

John: … can’t help but notice Cindy that ever since you took over writing Hoofbeats episodes that your on-air voice has gotten noticeably crabbier. How about we first just read their H&P, okay.

This is a 80 year old female who had

  • Coffee ground colored stool about three times last week 
  • Cough/ copious sputum production x at least 1-2 months, prescribed a course of azithro maybe 2 months ago, last week prescribed a course of levaquin x 7 days, maybe slightly better? The cough is still bothering her 
  • ROS: progressive fatigue over the last few months, maybe lost some weight too

PMH: bronchiectasis with MAI, RA only on NSAIDS

PSH: C section, appendectomy Meds: diclofenac

FH: daughter with asthma, another daughter with unknown type of cancer

SH: Lives in queens with her daughter, was lost to follow up one to two years ago because she had to go back to Ecuador to take care of a daughter with cancer, just returned and is now trying to re- establish care with prior doctors  

Physical exam is significant for tachycardia to 100s but stable BP, she appears fatigued, has hand joint deformities consistent with RA, breath sounds are decreased bi-basilarly. Labs are significant for hgb 5.4/20, plt 946, WBC 15. BUN/Cr 20/1.5 on BMP with baseline Cr 0.7 from a couple of years ago

John: So that’s it so far? Seems like a pretty typical medicine admission.

Cindy: Let’s see what our discussant thinks. Today we have Dr. Lauren Comisar, who is a hospitalist at the Manhattan VA hospital center here with us.

Dr. Lauren Comisar: This is a lady who is fairly sick. Um, she’s got a few serious problems that seem to be acute, uh, and she’s a little bit tachycardic. So she’s someone who, in terms of being stable versus unstable, she could quickly go to end the route of being unstable. Uh, that being said, uh, her problems seem like they all have good differentials and good paths to kind of go down to work them up. Has she had any difficulty eating or swallowing? Do we have any further details on the mycobacterial infection? Like when that was and if she was treated for it and what those symptoms might have been like?  Do we have a urinalysis? Iron studies uh, and hemolysis markers?

Cindy: Here she didn’t explicitly say it outloud, but you can tell Dr. Comisar is questioning a few things already. The refractory nature of the pneumonia, the validity of melena, and if the anemia is all from acute blood loss versus if it is multifactorial. Because we are sitting in the call room, we won’t have any more subjective information until we meet the patient, but we do have more laboratory data trickling in. UA: trace to small blood, trace LE, 39 RBC neg protein but pro/cr ratio 484 mg/g (normal <200). Full iron studies are not available but ferritin is in the 2000s, and let’s say there’s no s/sof  hemolysis. CXR: Severe extensive confluent bilateral ill-defined nodular and masslike consolidative opacities. The findings are suspicious for severe multifocal pneumonia and/or extensive airways mucoid impaction, in this patient with history of bronchiectasis. Other etiologies including malignancy/metastatic disease cannot be excluded. Further evaluation with CT of the chest is recommended.  CT: Multifocal airspace opacities and extensive peribronchovascular nodules and masslike opacities in all lobes. Some of the nodular/masslike opacities in both lungs have internal lucencies, consistent with cavities and/or blown out bronchiectatic airways.   

The leading differential is severe atypical mycobacterial infection (given prior CT findings consistent with chronic MAI, and documentation of MAI in prior sputum and BAL), however, other superimposed infections are also possible, including fungal, bacterial or septic emboli. Metastatic disease could also have a similar appearance (and the nodules are overall most pronounced in the mid to lower lungs, supportive of a hematogenous distribution). The presence of perihilar mixed ground glass and consolidative opacities and interlobular septa also raises the possibility of superimposed pulmonary edema. Alveolar hemorrhage could also account for the ground glass.

John: Cindy’s showing me the images now… and, wow. The read doesn’t do it justice. This looks really bad. It’s really, really extensive

Cindy: Listeners, let’s pause here. What do you think? What is your problem representation at this point, what additional data would you want to gather when you actually get to meet and examine the patient? What additional work up would you order at this point?

Dr. Lauren Comisar: This patient is an 80 year old woman with rheumatoid arthritis on NSAIDs, uh, who presents with acute dark stools as well as subacute cough, uh, and weight loss, uh, and is found to have significant anemia, renal insufficiency, as well as diffuse bilateral pulmonary lesions. So now I feel like we’re kind of at an interesting crossroads in the case, because I, I was not that excited about kind of the pulmonary symptoms she was describing at the beginning. Uh, now we’re faced with this pretty dramatic chest imaging. ……this is dramatic and it’s something in the lungs, but like, is that could have been something that’s chronic and something that’s slowly worsening. Um, so I don’t know if she had another chest CT, but like maybe her lungs have just always looked terrible. Maybe she’s had kind of a waxing and waning, cough, uh, presentation that led her to that diagnostic workup for there a while back. And that could be totally unrelated to what’s going on now. Um, or it could be related and, and we should kind of divert our work up. I think you could invoke some type of upper GI malignancy, uh, such as esophageal or stomach, uh, that was metastatic to the lungs, uh, and then is, is leading to bleeding, uh, bleeding to the anemia and then slowly over time, the weight loss and the renal failure.

Cindy: So when I first heard of this case from my colleague, there’s a good explanation for every acute problem she has. The melena can be caused by NSAID use, AKI could be in the setting of GIB, the recurrent pneumonia seems odd but that can be explained by untreated bronchiectasis and resistant organisms. Like Dr. Comisar said moments ago, all the common problems have good differentials.

John: Yeah, I noticed that. And I assume that was a big reason you chose this case. Because I have a sinking feeling that in the “real world”, nine times out of ten I would just follow the path of least cognitive resistance: See a bunch of familiar clinical problems, do the familiar workup, go down the familiar management tree for each of them.  Meanwhile, our discussant is already trying to come up with potentially-unifying diagnoses. And I do wonder, to what extent is that how her mind works in reality, versus is this an artifact of the case exercise?

How we are going to the bedside to meet the patient as a team. While awaiting the interpreter phone, we notice that the patient’s eyes are red and she seems to be bothered by them.  

John: I love how in these cases, a red eye always has huge significance. Like if I get a red eye, whatever, Visene. But even though this is Hoofbeats, if I were you I’d pause and lock in your diagnoses now, because I’m willing to bet our discussant is going to pick up on something here.

Dr. Lauren Comisar: This elderly woman with rheumatoid arthritis on diclofenac coming in with several months of malaise, cough, shortness of breath, uh, and weight loss now with a significant anemia and renal failure found to have these diffuse extensive pulmonary lesions, uh, as well as conjunctival injection. That’s honestly the part that’s stumping me a little bit. Uh, and it’s leading me to think maybe this is more of a systemic inflammatory or infectious type process, um, rather than what I was originally thinking, when you told me about the conjunctival injection, my first thought was leptospirosis, uh, which I know can cause kind of multi-organ failure and, and can be kind of a dramatic presentation. Um, I would have to revisit what, uh, if it can cause GI bleeding, um, as one of its manifestations, uh, I was also starting to think about, uh, more autoimmune inflammatory type of processes, uh, now that we’re getting, uh, this sort of conjunctivitis. Um, and then we also, our renal failure, we were thinking could be pre-renal, but we also have some hematuria, we have some protein area. Um, so you could start thinking down a path of, uh, like ANCA associated vasculitis or other, other autoimmune condition. But then again, that makes it a little harder to tie in the GI bleeding… has she had any stools that we’ve been able to catch a glimpse of since she’s been here?

Cindy: So for Dr. Comisar, the chief complaint melena curiously becomes an outlier here. Nonetheless she asks for a pulmonary and ophthal consultation, and is working up a systemic inflammatory process, specifically anca associated vasculitis. Lets see what happens when we actually get to interview the patient with an interpreter phone. 

On interview the patient confirms her 20lb wt loss in 3 months, also reports months of progressive hearing loss and painful red eye with some vision impairment for four months. She reports bothersome sinus symptoms for a few weeks, for which she had been referred to ENT a week ago but had not been able to make the appointment yet.  She also started having significant nosebleeds to the point where she had been swallowing the blood, which coincided with the time when she started having melena. John: I think you know where this is going. Work up for ANCA vasculitis was sent and revealed elevated at proteinase antibody elevated at 5.7 (normal <1 AI), the patient underwent a VATS (video assisted thorascopic lung biopsy) which confirmed granulomatosis with polyangiitis (GPA). Ophthal exam also revealed necrotizing scleritis. She was started on pulse steroids and later transitioned to maintenance rituximab for granulomatosis polyangiitis with upper and lower respiratory, renal, ocular and auditory involvement. Of note, she did receive an endoscopy after discharge, which did not reveal any upper GI source of bleeding.

John: So Cindy presented this case to me, you know, before she wrote this episode. And I remember at the end just folding my face into my hands, because I completely forgot that most basic thing we learn in med school about melena: that it’s not always GI. Cindy, you were able to interview the doctor who actually took care of this patient, right?

Cindy: Yes, I would like to introduce Dr. Jonah Zarezky, a hospitalist here with me at NYU, to talk about how he diagnosed this patient in real life, specifically what he thought of the melena.

Dr. Jonah Zaresky: All right. Well, this, this was, uh, a woman in her seventies who, who was actually sent into the hospital for anemia and in the emergency room, they found that she did have melena and ended up calling GI. They also did a workup for some respiratory symptoms and found multifocal opacities on, on imaging. And they started her on a treatment for pneumonia. Her hemoglobin was stable in the emergency room. And so GI didn’t do any intervention like endoscopy. Um, and the patient was, uh, sent up to the floor. This was a, a case where I think we could have just continued to treat with, with antibiotics and seen what she did over several days. But when I interviewed her, I got a little, a little more depth to the story, found out she was having other more subacute symptoms, changes in hearing changes and vision. Um, and I sort of took a step back and reset my, uh, thinking on, on the case. I thought about it a little before I thought about it when she had already started talking about, um, you know, visual symptoms and changes in her hearing, uh, and that with the lung opacities, I was trying to unify those as a diagnosis. Um, and so Wegner was already sort of on the differential. And I wondered if the melena could be explained and, and I asked her, uh, about whether she had had a nose bleed and she says she had been getting them frequently. So that sort of, of, uh, reinforced that, that one possibility that I was, uh, uh, you know, considering when I got more of a history.

John: So he took a deeper history. And now I feel bad for our discussant. We asked her to diagnose a patient with an unusual illness, without many of the crucial details with no ability to ask follow up questions. If I wrote a case for CPC this way, I’d get review-bombed by angry students.

Cindy: I know, I am evil. However, the way we presented the case is how things unfolded in real life – because of the language barrier, hearing difficulty, and other system problems such as long ED wait time, Dr. Zaretsky was actually the first person who got to sit down and properly interview the patient with an interpreter phone, and that was not until hospital day 2 or 3. Real life is unfair and patients don’t present with well written HPI in CPC formats.

John: “Life is unfair”? You have gotten crabbier, Cindy. Though your point is fair. Clinical reasoning is interesting to talk about, but it can only be as sound as the data we acquire. Like any machine or model, if you put garbage in, you’ll get garbage out, as the saying goes.

Cindy: So I asked both Dr. Zaretsky and Dr. Comisar after the fact how they think they solved the case.

Dr. Lauren Comisar: I think you should always try to challenge yourself and see whether you can come up with one explanation that would explain everything. Um, because then otherwise then you’re saying it’s coincidental that multiple of these things are going on at once. Um, that being said, this is a lady who was removed from care when she was abroad and now she’s back. And so now she’s probably gonna list off everything that’s been bothering her. So it may well be that she want, some of these things are more chronic, the weight loss, the, the cough, um, and it is just incidental. Maybe she now has CKD that we just didn’t know about. Um, so you shouldn’t sort of marry yourself to the idea that it’s all gonna be explained by one thing. Uh, but I think if someone has multiple serious things that could be acute, you sort of have to at least go through the exercise of thinking about what could explain all or at least many of them simultaneously.

Dr. Jonah Zaresky: Often when patients come at you with, you know, many, many complaints, the, the first impulse is to say, hold on a second stop.I’m just gonna focus on sort of problem number one, and on an inpatient basis, some of these other problems you could focus on out outpatient basis, which is often a completely reasonable approach. Um, but I think there’s also times when you have to say, can this give, can these other problems actually be part of a bigger picture? And do I need to, uh, do, do I need to consider them when I’m considering diagnostic workup? I think isn’t always, um, sometimes we get the, an Occams razor, I think more times than not, uh, we don’t, but I think it gets to the point of, you have to always look at things in, in different ways. Um, uh, sometimes, sometimes it is the one unifying diagnosis. Sometimes it’s a patient with lots of diagnoses, but, um, you know, until you, you talk to the patient at length of the bedside, you, uh, you might not know which one it is.

Cindy: Since both Dr. Comisar and Dr. Zaretsky attributed their success to  Occam’s razor, let us quickly refresh on the concept.  Occam’s razor is really a law of parsimony, where the famous medieval philosopher William from the village of Ockham supposedly came up with the idea that “Plurality should not be posited without necessity”. Although all that says is that we should shave away superfluous explanations with a razor, oftentimes people in medicine interpret Occam’s razor as having one unifying diagnosis that explains multisystem findings.

John: And this being a medical podcast, of course you can’t talk about William of Occam without mentioning Hickam (…of Atlanta, I think?). My preferred formulation is the profane one. “A man can have as many diseases as he damn well pleases”.

Cindy: Without a doubt, Occam’s razor and Hickam’s dictum ARE useful diagnostic tools. For example, when I have a patient with CD4 count of 0 presents with diarrhea, Hickam reminds me that I have to rule out TB and CMV even though the cryptosporidium already comes back positive. Or when an otherwise healthy 24 year old with a chronic rash, subacute anemia thought to be from menorrhagia presents with pleuritic chest pain, the inclination is to focus on the acute pulmonary symptom, but William of Occam would urge me to dig further and ultimately diagnose lupus in this patient.

John: Obviously old age, chronicity and multiple comorbidities aren’t going to guarantee Hickam is right. Nor do youth, health and acute illness guarantee he is wrong. These maxima are examples of what are called heuristics: methods of problem solving that are neither perfect nor optimal, but simple and cognitively efficient.

Cindy: As you can see who is an 80 year old lady who was lost to follow up for years clearly is not following the rule of Hickam’s dictum.

John: So something I struggle with: Avoiding treating what’s just supposed to be a useful tool as a hard-and-fast rule, when in fact patients are under no obligation to do so.

Cindy: There are a couple more mistakes you can commit when applying these principles. The most famous being Crabtree’s bludgeon, where we force everything to fit together even if that takes inventing a convoluted, complicated theory.

John: We see how Dr. Comisar avoided this mistake, although she was actively looking for an unifying diagnosis, she examined each piece of data closely before forcing them into a narrative. Remember how she was asking out loud multiple times, “was pneumonia part of the whole picture? “is the pulmonary finding related to the whole picture or separate”, “which of these findings are related?”

Cindy: She (and Dr. Zaretsky) also avoided another pitfall. The thematic opposite of Crabtree’s bludgeon would be to throw out a viable diagnostic hypothesis because of an outlier or a piece of data that seem unrelated with the rest. (Unfortunately this doesn’t have an eponym name.)

John: At a certain point they both had become suspicious of a systemic inflammatory process, but this seemingly could not explain the melena, which could be easily interpreted as GIB.  So I would think that idea is dead on arrival. Melena is the chief complaint, it is the most severe symptom, AND the most acute symptom. It really should carry the most weight; it should be the pivot point for this case. I think it would be very difficult for me to tolerate melena as the unexplained outlier. So I found it interesting that both our discussants kept this hypothesis active when they walked into the patient’s room. I would guess this at least partly explains how they recognized as soon as hearing it that the red eyes or the nose bleed were signals rather than noise. 

Cindy: Dr. Comisar and Zaretsky showed us that Occam’s razor & Hickams Dictum are powerful diagnostic tools IF used properly- that is,  do not expect your patients to follow them like rules, and if you are going to apply them, do it with caution and flexibility. I actually have overheard Dr. Zaretsky share this case in the office with a few colleagues, and I did observe that it is very easy to phrase the PR for this case as “elderly female with subacute unresolved pneumonia, presenting with upper GI bleed and AKI”. When you think of the case as an upper GIB problem, it is difficult to arrive at the diagnosis of vasculitis, right?  

John: Now I know this isn’t the first time problem representation has come up, but just in case for our listeners, do you want to briefly review what PR means?

Cindy: Outside of medicine, a problem representation is a mental model we construct that summarizes our understanding of the problem. Cognitive psychologists believe that when we encounter a complicated problem, before you start solution planning, step one would be defining the problem. 

John: Based on this original definition, problem representation really can be a graph, an equation, an image, however the person conceptualizes the problem they are trying to solve.

Cindy: In medicine, or at least in clinical reasoning lectures, there seems to be a strong focus on the semantics and language of PR. Problem representation also took on a few additional functions, such as a communication tool or a teaching tool used to assess learner understanding of the clinical scenario.

John: This is probably a result of research done in the 1980s-90s. Cognitive researchers would listen to expert clinicians and medical students think aloud during a case, and one of the things they noticed is that experts seemed to be forming their problem representations early in their process. They proposed that by instantiating key clinical features and pathophysiology into one’s working memory, PR served as a link between raw data (signs and symptoms) and the final diagnosis; it would cue retrieval of illness scripts.

Cindy: In the most classic example that was used in these studies, two to three days of right knee pain becomes acute unilateral monoarthritis in a large joint, which should help retrieve the illness scripts for gout versus septic arthritis.

John: If you construct your problem representation with this function in mind, then that makes sense why there’s a rather artificial construct (or formula) that we are asked to use nowadays, right? The society for general internal medicine’s definition: one-sentence summary that highlights the defining features of a case that has three components: relevant patient demographic & risk factors, the temporal pattern of illness, and the clinical syndrome with signs and symptoms.  This mimics how we have been asked to construct illness scripts.

Cindy: In a prior episode on illness script, we talked about the concept of script activation. So, my next question is, does constructing a better problem representation help you arrive at the correct diagnosis better and faster by efficient script activation?

John: To my knowledge, no study has ever proven that. Though that would be a difficult thing to study, I think. And, as importantly, what do we mean by “better problem representation”, exactly?

Cindy: No one knows. Copying the official answer, “A good problem representation captures the relevant information while excluding extraneous information, while the use of semantic qualifiers allows for conceptualization and abstraction of clinical findings”. So in general, we are taught that “the more abstract the better”, but that seems to be the only rule I know of.

John: It’s interesting how abstracting melena into GI bleed is almost always correct. It doesn’t even register as a step for me, it’s practically involuntary, instantaneous. But in this case, it leads one astray from the correct diagnosis of vasculitis. Is it wrong to do this?

Cindy: There’s nothing wrong with abstracting melena into UGIB, but there is something wrong with committing to that abstraction when it was made too early into the reasoning process. If we think about the process of PR generation, there’s data collection, interpretation, abstraction, before you form your problem representation, and each step is a potential point for error. When we are stuck with a problem that seems impossible to solve, instead of thinking harder (which rarely helps), it might make sense to take a step back and ask: am I solving the wrong problem?  What could be wrong with my PR: was my abstraction of data erroneous? Was my interpretation of the data incorrect?  Or going back even further, did I start with incomplete or incorrect data, do I need to go back and talk to the patient again? Which Dr. Comisar definitely did during the interview multiple times by asking us, is that melena real?

Dr. Lauren Comisar: has she had any stools that we’ve been able to catch a glimpse of since she’s been here?  …not to cast out on things that people say, but, um, sometimes what we know as a clinician that looks like bleeding…So you wanna to see it with your own eyes.

John: Problem representation should be dynamic and flexible. Diagnosis is going back and forth, it’s not necessarily linear I completely agree with you there. At the end of the day, there’s the age old question regarding PR: Does a better PR make me a better diagnostician? My takeaway from this case is, good PR is helpful but does not necessarily guarantee success (esp when there’s a knowledge deficiency or if you don’t have a good illness script already in place for that PR to help activate), but a faulty PR that misidentifies the problem at hand will for sure lead us to failure.

Cindy: Alright listenders, that should do it for this episode. Once again, I would like to thank Dr. Zaretsky and Dr. Comisar again for joining us on this episode. As always, let us know what you think by visiting our website at www.coreimpodcast.com or send us an email at hello@coreimpodcast.com. We are also on Facebook and Twitter, at @CoreIMpodcast.

John: Special thanks to Drs. Amy Ou, Shreya Trivedi, and Marty Fried. Special thanks to our audio editors for this episode along with our other CoreIM colleagues.

Cindy: … And an honorable mention, as always, to Dr. Steven Liu.Opinions expressed in this podcast are our own, and do not represent the opinions of other affiliated institutions, nor should they be construed as medical advice.

John: Thank you for joining us. With CoreIM, I’m John Hwang.

Cindy: And I’m Cindy Fang. See you next time.



Tags: , , , ,