Episode 68: The History

Internal medicine physicians like to pride ourselves on our clinical reasoning – the ability to talk to any patient, pluck out seemingly random bits of information, and make a mystery diagnosis. But how does this actually work? In this episode, called The History, I’ll be joined by Gurpreet Dhaliwal as we explore the beginnings of our understanding on how clinical reasoning works – starting in the middle of the 19th century with polar tensions between two ways of approaching our patients that are still felt today. Along the way, we’ll talk about the American Civil War, Car Talk, Sherlock Holmes, and whether the practice of medicine can ever be considered a science.

Sign up for Digital Education 2022 here: https://cmecatalog.hms.harvard.edu/digital-education



This is Adam Rodman, and you’re listening to Bedside Rounds, a monthly podcast on the weird, wonderful, and intensely human stories that have shaped modern medicine, brought to you in partnership with the American College of Physicians. This episode is called The History, part of the continuing series on the history – no pun intended – of diagnosis.


What exactly is it that internists do? Media has a very clear answer – the prototypical  representation is the Holmesian genius, with some sort of personality deficiency – think House – with the ability to ask the right questions and pluck the right bit of otherwise unnoticed information to make the right diagnosis. The same is true for the real-life physicians that we respect as well. One of my heroes as a medical student was Faith Fitzgerald of UC Davis, who passed just a few months ago. She was able to combine humanism with a fierce interrogation to solve seemingly any problem: to quote one of her favorite aphorisms, “If you listen to the patient, they will tell you all you need to know.” By the way, one of the most influential journal articles I read as a medical student was by her – Curiosity. I’ve provided a link in the shownotes, and if you read just one thing today, let it be that (https://www.acpjournals.org/doi/full/10.7326/0003-4819-130-1-199901050-00015).


The study of the ability for expert physicians to see what others don’t – often in surprising ways – is called clinical reasoning. Clinical reasoning came of age in the 1970s along with behavioral psychology, when researchers really turned their attention to how the human mind works. And if you read any review papers on “clinical reasoning” you will see that they all effectively start there, in the 1970s.


Well, if you know me, you can imagine what I’m going to say next. That’s not true, not by a long shot. In the next few episodes, I’m going to trace the history of what we now call “clinical reasoning”, starting in the 19th century as traditional forms of reasoning were formalized to a new standard of pathologic anatomy. As the 20th century rolls by, these ideas become further developed as “scientific” diagnosis, until post-WW2 cybernetics movement, when the attempt to build a “robot doctor” really takes off. And this entire history has fundamentally shaped the medicine that we practice today – from the way we write our notes (and all the time doctors spend on documentation) to the structure of our electronic medical records, and to the current rage of machine learning and artificial intelligence to fulfill a century-old dream of building a diagnostic machine.


All of that is to say, this is a huge undertaking, and this is going to be a multi-part episode. To properly tell this story, I am joined by the most thoughtful diagnostician that I know – though as far as I know, without any serious personality deficiencies a la Gregory House – Gurpreet Dhaliwal.


Gurpreet Dhaliwal (00:33):

Yeah, my name’s Gurpreet I’m a clinician educator, uh, at the San Francisco VA medical center. I, um, see patients with students and residents in the hospital, in the clinic, in the emergency room. And I’m interested in how, uh, doctors think and in particularly how they arrive at diagnostic, uh, decisions and labels


Gurpreet would never describe himself this way, but he is probably the most prominent – and breathtaking – public diagnostician out there right now. If you ever get the chance to see him present, I highly recommend it. He not only explains his thoughts as he walks through a case, but his metacognition – that is, the processes behind these thoughts. Watching Gurpreet reason is better than any lecture you’ll ever get on physician reasoning. And he has been incredibly influential on me as both a physician and a thinker. I’m only playing this tape to show you just how much:


Adam Rodman (01:14):

Well, and the reason is, so this is, I am recording this. Uh, I have been like, I’ve been a fanboy <laugh> since I was a fourth year medical student, um, which makes this conversation very remarkable to me. You, you were a huge, you and I think faith Fitzgerald were like the, the intellectual, uh, influences on my, uh, on my nascent diagnostic career when I was a resident

Gurpreet Dhaliwal (01:38):

I’m Andre good play the small role. And I will just say that Dr. Fit, Gerald who recently passed, she, uh, was profoundly influential, uh, in me and even conceptualizing the idea that there is a, uh, form of diagnostic expertise. Hers was superhuman. Um, but just seeing that that’s like watching an Olympian, at least you start to get interested in like, what would be the steps to get to that level.


So that is the set up, and this is probably my longest introduction ever. Let’s get started! One of my favorite Internetisms is the phrase TL;DR – too long; didn’t read. It originated in the early days of the internet to ridicule so-called “walls of text” that proliferated on forum posts, but then morphed into a summation sentence, where someone might write a thoughtful essay, but then at the end write something like, “TL;DR – clouds are really interesting” or something of that sort. And for the record, a 13 page thoughtful essay on why clouds are cool is the sort of thing that’s right up my alley. In fact, the TL;DR ethos has taken over the internet, whether it’s on Twitter where we have what we think are deep debates 280 characters at a time, or with, say, Axios, which is taking off with a goal to convert the entire news industry into TL;DRs. In many ways, this podcast is the anti-TL;DR, where I get a chance to expound almost an hour at a time on subjects that I never really thought people would much care about.


All of that is to say, because this episode is part on an ongoing series on the history of diagnosis, I could recommend that you first listen to “Cry of the Suffering Organs,” “Signals,” and a “Vicious Circle” but that would take, like, three hours. So a TL;DR – from the 18th through the early 20th century, a “classical” view of diagnosis came into being based on the idea of pathological anatomy – that diseases existed in discrete places in the human body, and that the job of the physician was to use history and diagnostics – originally the physical exam, but later all sorts of studies including radiographic imaging, biopsies, microbiology, electrophysiologic, and laboratory studies to identify these diseases. Cracks in this model started to appear in the early 20th century, when researchers working on syphilis diagnostics realized that different assays had different abilities to detect whether or not someone had syphilis. Spurred by wartime advances in statistics to detect enemy aircraft, the statistician Jacob Yerushalmy realized that ALL diagnostics for a disease had in fact their own “test characteristics” – and fundamentally redefined the act of diagnosis in terms of uncertainty (that would be that “vicious circle”.)


A lot of my discussion has been about the history of diagnosis relies on the development of diagnostic tests, whether those were physical exam findings, or into the 20th century all sorts of laboratory and imaging finds. In what I’m calling the “classical” period of diagnosis, diagnostics tests were used to find signs, as opposed to symptoms, and the field of what we would now call physical diagnosis was called semiotics. But as every physician knows, this is only a part of the diagnostic process; arguably the more important skill is the history – actually talking to our patients and finding out what is going on.


Taking a history is one of the oldest skills in the field of medicine, presumably predating the written record, since our earliest written sources – say, the Edwin Smith papyrus, clearly demonstrate an understanding that a chronological representation of illness is related to diagnosis. But first a brief aside on the use of the word history. It always throws people off – there’s a common trope that if a patient is confused or somehow otherwise unable to answer questions about their presentation – say, for example, from delirium caused by a severe infection – that they are a “poor historian”. And there’s, of course, a counter trope pointing out that the historian is the one taking the history (in the sense that I, for example, am a historian) – and the only poor historian here is the physician. It goes without saying that this is pedantic – the word historian has only meant a professional studying history for several hundred years, and prior to that would have meant ANYONE given an account of what happened, making both the patient AND the physician historians in this setting. And just as when there is a “poor history” – usually there’s a combination of factors in play, though certainly balancing in one direction more than another.


That being said, the traditional word for history I think is more telling – anamnesis, meaning “the remembering of things” – the patient’s recollection of their own disease, accepting all of the inaccuracies and quirks of human remembering. We stopped using anamnesis in the United States in the early 20th century, but the term is still used in many countries across the world.


History is incredibly important, and taking one is a complex process. I’ve mentioned this before in a very, very early episode of Bedside Rounds – but one of the best examples that always comes to my mind is Car Talk. Car Talk, which has been off the air for years now except in reruns from the 1980s, was an NPR talk show by Tom and Ray Magliozzi, Click and Clack, the Tappett brothers,  where they dissect various cases about malfunctioning automobiles, often by making their bemused listeners make car noises with their mouths – which isn’t that different from what I do when I describe cardiac murmurs. Gurpreet actually wrote an article over a decade ago in JAMA showing how their shows models the history taking/diagnostic process – and in fact, points out that a retrospective analysis shows that they made the correct diagnosis 71% of the time, which is similar to the old medical aphorism about how often you make the diagnosis on history alone. It’s one of my favorite medical articles – and the fact that I listened to Car Talk as a kid with my grandfather probably plays a large part – and I have it linked in the show notes.


So what exactly is clinical reasoning then? What does it mean to make a diagnosis? Since I have him here, I naturally Gurpreet:


Gurpreet Dhaliwal (03:40):

Super unfair question. Cuz it’s super a tough question. There’s a, um, I think a simple way, I always think about it is, um, acquiring and processing da data to reach a conclusion that allows you to do something actionable for a patient, but there is a philosophical discussion around it. Like what are the boundary conditions? There’s a paper from 2018 or so by Meredith young and colleagues and they basically grappled with where this clinical reasoning end. Like, is it also our communication? Is it evidence based medicine? Um, is it the physical exam or is it physical exam, something different? So I think of it in a rather simple cognitive, uh, framework, like what’s going on in the brain’s mind to process information and reach a conclusion, but I just wanted to acknowledge that people are uncertain where the boundaries are of that.


A potentially unbounded epistemological construct that involves the acquisition and processing of data! This is the stuff that keeps me up at night! We are going to take a step back and think about what it means to collect all of this data – what does it mean to take a history? And how has this changed over time?


As far as I know, the definitive work on this subject has not been done – though I would love to read it! To summarize briefly from my own forays, over the past several thousand years, a pragmatic tradition of history taking had bubbled up in traditional Western/Arabic medicine. Even as nosologies changed – as we morphed from humoral medicine to a medicine of symptoms to a medicine of flows and blockages and finally to pathological anatomy, the inquisitorial methods persisted. This was taught and passed down as what we now call “clinical reasoning” – and we still largely do the same today. As pathological anatomy became ascendent – and like I talked about in Cry of the Suffering Organs, gave a “target” with which to confirm diagnoses in the autopsy – clinical reasoning started to become more formalized. To give you an idea where things stood in the middle of the 19th century – in the US, really the dawn of pathological anatomy – I’m going to summarize the most important clinical reasoning textbook in the United States in the 19th century – Jacob Mendez Da Costa’s Medical Diagnosis.


First, some brief biographical details, because Da Costa was a very interesting figure. He served as a military surgeon in the American Civil War, and had a deep abiding belief in the combination of art – by which he meant history, philosophy, and ethics – and science in the practice of medicine. He has one occasionally-in-use eponym in the 21st century that reflects this human understanding of patients – Da Costa’s syndrome is a functional anxiety disorder of soldiers who have suffered combat trauma.  After the war, he became a professor at Jefferson, where he continued to push for humanistic medicine just as scientific medicine is starting its ascent. This is from a valedictory address near the end of his career:


“I think that the cultivation of the humane letters has the most distinct bearing on the cultivation and appreciation of science. Science is nothing without imagination; and imagination is most readily kept fresh by literature. What little good there is a mere descriptive person, and in the small facts which with painful toil he accumulates. But let these facts be welded together by thought, their bearing traced by imagination, experiments devised by the mind projecting itself in advance of them, and the plodder is likely to become the great discoverer.”


All of this to say, I totally would have followed him on Twitter if he were around today.


So let’s go to Da Costa’s Medical Diagnosis, published in 1864. I very briefly summarized Da Costa all the way back in the first episode of this series, but I want to really focus on the model that he gives for how to take a history, and how we use that information to make a diagnosis.


This process of clinical reasoning Da Costa fundamentally sees as an art form, a necessary human attempt to create the wonders of Nature:


“Nature does not limit herself in her irregularities any more than in her rules. The text must therefore, be looked upon as treating only of general laws and of their most notable infractions; in fact, but as a series of etching, with here and there a prominent figure shaded, but not as an attempt to reproduce the colors of an original whose varied hues could not be closely copied, even by the hand of a master.”


His entire textbook is rooted in the new and exciting field of pathological anatomy – that the goal of the physician is not only observation for observation’s sake, but to think about how these observations might give insights to hidden diseases. But unlike prominent members of the Paris Clinical School, who valued new diagnostics over the more traditional history, Da Costa fundamentally thought that the patient’s own story was the most important.


“In a study of this kind, an investigation of symptoms plays unavoidably a prominent part. In truth, the detection of disease is the product of close observation of the symptoms, and of correct deduction from these symptoms.”


To ask the patient their story is not enough. The interviewer needs to not only have skill and experience, but also organization. After all, everyone has seen the novice attempt to take a history:


“He wanders in his search from one part of the body to another, attracted by different symptoms in turn, pointless question succeeds to pointless question, and a conclusion, almost certainly erroneous, is finally jumped at, or an acknowledgement made of inability to arrive at any.”


Da Costa strikes on a fundamental truth here – though one that he merely alludes to, not having the benefit of modern psychology or, dare I saw, post-structuralist philosophy – that our method of organization in interviewing ultimately affects the way that we think. And he gives to models to organize our questioning and our thinking: what he calls the synthetical method and the analytical method.


In the synthetical method, as the name suggests, the goal is to collect all of the patient’s information, as thoroughly and systematically as possible. The physician starts NOT by asking about the patient’s complaint, but by essentially creating a database about the patient’s history. Da Costa recommends, among other questions – age, occupation, childhood diseases, familial hereditary conditions. Only after this is done, does the physician ask about the disease, though taking a similar comprehensive approach, going through every single organ or organ system in their detailed questioning, similar to how a review of systems works today. Only after this comprehensive process is complete does the physician move on to the remainder of the diagnostic process – the exam, any studies, and the cognitive processes.


This is to be contrasted by the analytical method. The order here is flipped, and starts with the patient’s experience, and their own telling of the story. The physician then asks further questions – again, working through a similar structure, based on the information that the patient has given them, and narrowing or broadening based on the physician’s own thought process. The goal here is the melding of diagnostic thinking with the taking of the history.


Da Costa actually gives an example of what this looks like – and since I’m a rather magnanimous podcast host, out of the blue I decided to give Gurpreet the case to see how similar his reasoning was to Da Costa’s analytical method.

Adam Rodman (23:57):

So this, this is great. I’m I’m going to read the case. I wanna know your thoughts and then I’ll see what he says. Um, this is, this is a proto, this is clinical reasoning from the 1860s. Okay. A person consults us for a cough brought on by exposure. He has been able for four or five days, having previously been in good health, we notice on examining him that his breathing is hurried and that he is a fever. The lower portion of one side of his chest is dough on percussion. And the respiration there is wanting the action sounds of the heart are normal. Okay. So what are your first thoughts when you, when you hear this case presentation?

Gurpreet Dhaliwal (24:28):

So I was influenced by what was said first, right? The order of sequence. So I think you said he has a fever and a cough and DYS me if I caught, if it was DYS. So right away, I sort of limited myself. I, the entire medical knowledge, I know I sort of, um, locked in a little bit on pulmonary diseases. And then there were some things like an exposure, which were vague. I, I didn’t know if that that’s sort of probably our equivalent to sick contact, sick

Adam Rodman (24:51):

Contacts. Yes. See, it was around someone sick four or five days.

Gurpreet Dhaliwal (24:53):

Right. But when it’s used that that has to be, has to be explored. Right? Some people everyone’s been around someone with a cold at some period, versus like I’m living with someone in my house who has COVID right. That there’s various degrees of sick context, but I take for what it’s worth that there might be some exposure. Um, but then in terms of trying to take the entire universe of maybe pulmonary diseases, the fevers associated with it, the physical exam had a lot of specificity there. That is the decreased breast sound and decreased air movement. And I’m wondering, is there a really dense consolidation there? Or is there a, a large plural effusion and then I can’t help, but start to go to diseases that might cause those two combinations, like, is this a very bad lobar bacterial lobar pneumonia. Um, is this a malignant plural effusion with a lung mass underneath it? Those are just ideas that came to mind. Or, and then if I can just pivot, I might be like, maybe I’ve been fooled hook, line and sinker. And this is a, and I know sometimes lung exams, you can have nonfocal findings, although these sounded quite definitive and abnormal.

Adam Rodman (25:55):

So let me go on with what he reads so we can get some more information here. Right. So, um, he exactly like you he’s like, this is an acute pulmonary affectation based on what we’re hearing four or five days, he’s clearly got what appeared to be pulmonary symptoms. So he builds his differential. He says, um, in all occur fever, cough, and disordered breathing. Is it acute pneumonia? Uh, could it be acute hypothesis? That’s tuberculosis, acute plurality, which, uh, or acute bronchitis. These are like 1860s terms. Yeah. But literally every single thing that you said there, and then he goes on and he talks about how he investigates it more. And he’s saying there’s dullness on percussion. The dullness is associated with the blowing respiration. Whereas, uh, in the case before us, no respiration is heard. So there’s dullness. And when he listens, he can’t hear anything at all.

Adam Rodman (26:43):

Um, and let’s see, he said, let us look at the sputum and see if it’s tenacious and rusty colored. It is not, it is thin Androy do you like this Isaac, but, but let’s hypothesis test acute plurality, the patient two, when question states that at the onset, he had a sharp pain in his side, and this, we are aware takes place in plurality, the vocal vibrations. This is Fous likewise are noticed to be absent on the affected side of the chest, which when measured is found to be enlarged. So what, what would you, based on this additional information, you know, it’s the 1860s, what do you think the diagnosis is

Gurpreet Dhaliwal (27:14):

We’re getting? So he has, um, uh, you know, he’s provided us with more information. Um, I think, you know, we’re trying to differentiate between all these, uh, different possibilities. And now we have this sort of complete reduction of sounds by tactile sense. And if memory serves me, correct, I think that leans us towards, uh, um, uh, an fusion as being the, uh, the cause. And so you, but it doesn’t really change too much, you know, I don’t know what their conceptualization of plurality is. We sometimes use that term almost to refer to like you, you know, some other pulmonary process and the plurals inflamed, or the plurals in inflamed, in isolation, like there can be a viral cause like Coxsackie or something of that nature. But I think this starts to become more concerning for, um, that there may be a, a bacteria or, uh, infection, maybe tuberculosis, the timeframe is short, but I think now the effusion is the dominant feature that they’re contending

Adam Rodman (28:03):

With. And so I just want you to know that is exactly, uh, the, the diagnosis that he makes in the end, that this is an effusion associated with a plurality and understand this is the 1860s. Germ theory doesn’t exist. Plurality is the initial insult that would’ve led to an effusion. But what, you know, what’s great about this is de Costa runs through the exact same differential using 1860s terminology that you just did.

Gurpreet Dhaliwal (28:26):

I, I love, I think there’s an orderliness to it that the modern physician could identify with right there. There’s nothing in that line of reasoning, even though the language is a little different and the technology is limited to the stethoscope. Um, and the hand there’s, if you started to add in CRPS and x-rays and CT scans, it’s the same sort of school of thought or mode of thought. Yeah. Which is that I’m, I’m following the lead of the patient and the clues that are in front of me, but I’m not, uh, cataloging the entire universe or database that I

Adam Rodman (28:55):

Could. And to me, that struck me. I mean, he even talks about probabilistic diagnoses that often were just, you know, ordering things in terms of probabilities, again, a very seemingly modern idea, and this is all 160 years ago.


Now back to Da Costa, while he clearly prefers the analytical method, he points out that there is not a priori a preferred way to take a history – it is based on the needs of the patient and the problem at hand: “The synthetical method is the more purely scientific; but it is too full, and calls for too much labor, to meet the requirements of ordinary professional life. It is much better adapted for recording cases in the pursuit simply of pathological knowledge, and decidedly the best where the history is obscure and symptoms are ill defined.” The synthetical system, therefore, is best when writing up a case description for publication, or in those cases where the physician does not know what is going on, such as in an “idiopathic fever”.


I think it’s fair to say that Gurpreet is a 21st century Da Costa. So I asked him that he made of Da Costa’s approach to history taking.


Gurpreet Dhaliwal (09:55):

So I think that, uh, those terms were new to me, but the concept is not because I think if I even reflect on myself, I thinking that a physician is scientific. When you first enter the field, I thought it would reflect this very thorough collection of data. And through all obtaining all the data, you know, if a patient could report every symptom and I could get data on every, uh, um, system that they have, I could just do a waiting of all the data and come to a conclusion about what it is. I think it takes a very short, uh, order of time to recognize that, um, we work by the analytical method overwhelmingly where the patient’s story is the entry point. Um, and we start to filter and put on blinders pretty quickly in that process and to the best of our knowledge, it works well. Um, but there are risks with it.

Adam Rodman (10:41):

You mean with the analytical method?

Gurpreet Dhaliwal (10:43):

Yeah. Yeah. I think that the analytical method, it, it works because it starts from the patient, it’s a patient-centered approach, right. That patient comes in with what concerns them, or maybe we have a piece of data with what concerns them. So it definitely orients us towards solving that problem, like their swollen knee or the jaundice that they’ve come in with or the itch that’s bothering them. So it’s very patient oriented. Um, and then that I think is one of its intuitive appeals. Um, the challenge is the patient’s initial concern may not be their most pressing issue. Um, or the patient’s initial concern, uh, may be something that we can’t solve, but we’re distracted by something else along the process. And we turn towards that said and be like, you know, I, I can’t figure out your, um, itchiness, but you have, uh, I found that you have anemia. I’d like to pivot to that. And the patient’s like, what about the original problem I came in with?

Adam Rodman (11:30):

Right? I mean, patients are all edits is a lesson. You, we teach all of our residents. If you don’t manage the patient’s chief concern, what they came in with, even if you save their life, they’re not going to be happy if you don’t take care of the thing that they came to see you with for the first time.

Gurpreet Dhaliwal (11:43):

And that’s understandable. I mean, it totally makes sense.

Adam Rodman (11:46):

Uh, but so you, I mean, you mentioned the synthetic method and one of the interesting things about DeCosta is he clearly comes down on the side of the analytical method. It sounds like you, despite, you know, recognizing some of the drawbacks would also are clearly to use this like 19th century language coming down on the analytical method. What, what are the problems then with a synthetic method?

Gurpreet Dhaliwal (12:08):

I think, uh, if I was to say one word, two words can summarize it time and incidentaloma. So I think we, um, the time part is pretty obvious to people. If we did a complete thorough inventory on every single patient, meaning full history, full physical, and you can even take that by logical extension to full lab testing and full imaging, if you wanted, um, it would be tremendously costly and time consuming. Um, and I don’t think I need to say any more than that. I think everyone recognized that, but I think the other, um, concept that we’d really have to contend with, even if we just tried to be as thorough as the review of systems, that’s to say cataloging someone’s symptoms from start to finish, that was one of de CASA’s earlier thoughts, um, is that you are bound to pick up a concept. We now know from radiology, which is the incidentaloma. So if a patient has come in for visual changes and then they, um, tell me that their right knee hurts, uh, I can try to create some narrative, like, you know, maybe this is endocarditis with an embolic phenomenon. Um, but I’m far more likely to have opened up a second front that wasn’t related at all to the patient’s, um, initial concern and maybe primary concern and will consume either their resources or our resources in doing it.

Gurpreet Dhaliwal (14:33):

Right? The idea that anything that you pick up would not be useful to be fair. It’s a lesson we all learn though, when we go through training, right? Like I think when we enter you, um, you know, any sign you pick up should be useful. I remember once I, um, seeing a patient who had, I think it was a rash or something, something you’d rather nondescript, but it sticks in my mind because somehow I got myself doing a neurologic exam and their Babinski test was positive on one side. Um, and I wasn’t attending, so I’m not even citing me in my training. And I, I struggled with it forever. I was like, wow, the Babinski is equivocally positive, which rarely happens by the way, in a lot of cases where you can’t tell what’s going with the toe and it’s like, it is unequivocally positive and he’s here for something, no way at all, that I could relate to the neurologic system. And I struggled, ultimately I did a CT scan of the head, cause I just didn’t know how to resolve the uncertainty. And years later, I learned including a paper that came from U C S F that the Babinski test itself has a receiver operating

Adam Rodman (15:23):

Curve. Exactly. I’m a very, it’s

Gurpreet Dhaliwal (15:25):

So far a field from the definitiveness that I learned of it as a upper, more than neuron

Adam Rodman (15:28):

Science. Right. And when you look at, in fact, when you look at the neurologic exam and you look at the data behind the neurologic exam, you realize, cause you think it’s. So, I mean, it is good. They are good tests, but they are not perfect tests. I have a similar story on the most beautiful clubbing I had ever seen on a patient who came into me in, uh, in clinic. And I became, they had no other symptoms. I don’t even remember what they were, they were coming to see me for. And I, I initiated this incredible, like pulmonary workup. And in the end I found nothing. And the answer is that like the, the identification of clubbing also has like, well, first of all, it’s not, it’s not, progno, it’s not path monic. Other other things can cause clubbing. And sometimes it doesn’t mean anything.

Gurpreet Dhaliwal (16:06):

I think if I saw clubbing too, that it’d be hard for me to resist getting a chest x-ray uh, if that was negative, I don’t know how far down the road I’d go. It’d probably depend on smoking history and other things, but that’s such a great example, opening up the capacity. We’ve learned it over time that just acquiring more and more data. Oftentimes like we said, costs time, money, and even psychological distress, right? You start raising my patient, I’m telling him you might have a stroke or a tumor in your brain cuz of cuz of how I touched your toe and you’re telling your patient, I looked at your fingernail and now you might have lung cancer.

Adam Rodman (16:34):

Exactly, exactly. Or, or you see those lines and you’re like, well, something bad is going on.

Gurpreet Dhaliwal (16:40):

Right. The flip side is we’re all imbued with these stories of the incidental pickup, right? Like,

Adam Rodman (16:45):

Oh yes. I know

Gurpreet Dhaliwal (16:46):

The PA the patient, you know, patient looked like they had a cold and someone got, had a glance at their fingers and they saw splinter hemorrhage and low and behold, the cold was just the beginning of their subacute bacterial endocarditis. And were it not for that great diagnostician who does a survey from head to toe? It never, would’ve been picked up that kind of story, um, looms in your mind as well.

Adam Rodman (17:05):

And I think one of the reasons, so this gets to something that I’m fascinated about is the medical obsession with Sherlock Holmes and mystery stories, right? Because mystery story Sherlock Holmes in, in particular, explicitly, uh, modeled after clinical reasoning. But in those stories, often the picking up of something that no one else can see almost incidentally is what makes the case,

Gurpreet Dhaliwal (17:26):

Right? That, and we talked earlier about expertise and sometimes expertise, um, is the perception of a clue that other people don’t pick up in the environment, right? That the homes had that trait and great diagnosticians have that trait too. One of the things we might think about there’s no stories are written on the times when homes does think the dog not barking at night is a, is a big deal. And it’s not right.

Adam Rodman (17:47):

<laugh> right. Exactly about everything that happens every other day of the week, where he

Gurpreet Dhaliwal (17:51):

Is wrong. And that most splinter hemorrhages are people, you know, banging their fingers on their hammer or a door or something else.


When it comes to making a diagnosis, Da Costa does not differentiate between the history, the exam, or diagnostics – and it’s 1864, so we’re talking about a stethoscope, laryngoscope, thermometer, and a specific gravity bottle – when it comes to the analytic method. All of these are weighed together in coming up with potential diagnoses. Quoting him, “Knowledge and, above all, the exercise of the reasoning faculties are now indispensable.


The act of being a physician is not just KNOWING a lot of stuff, a lot of facts about disease, say. It is the reasoning process itself – and using this reasoning process to ask certain questions and perform certain diagnostics.


“Were it otherwise, the science of diagnosis would be simply a matter of memory. It is, however, this very analysis of symptoms and the lengthy process of induction attending it which make medical diagnosis so difficult and so unattractive to the beginner. He sees that by reflecting and reasoning on what are frequently but indirect manifestations he must find the seat and nature of disorders hidden from his view. Nor is it reasoning on the ascertained facts alone that is required, the premises may be but probabilities; for in truth, diagnosis deals at times with the logic of probabilities as much as with the logic of patient facts.”


Because knowing morbid signs – that is, diagnostic findings, is not the whole or indicative of diagnosis. “We are thus arriving, step by step, at the explanation of the morbid appearances, the starting-point in deduction always being what is known of the affection the presence of which is suspected, and the symptoms of which we are contrasting with those before us.”


If you can tell I’m getting excited here, you’re right! Da Costa is making a very clear statement about the nature of clinical reasoning. Sherlock Holmes has come up a lot so far; Holmes, if you will remember, was modeled after Conan Doyle’s mentor Joseph Bell, who was by all accounts an expert diagnostician. Now you see this reasoning spelled out clearly. He lays out two mental processes – inductive reasoning, going from specific observations to general principle – say, for example, using historical and exam findings of a specific patient to figure out a pathological process, and then deductive reasoning, which is the exact opposite – going from general principles to specific findings via logical reasoning. But he also points out that these solutions must fundamentally be thought of in probabilities.


This is a wonderful example of a very intelligent person trying to explain a cognitive process that has not yet been formalized or named. What he is describing is what we would now call “abductive” reasoning, where specific observations lead to multiple probable explanations, which can be further tested and questioned, going through a similar abductive process, until you have a most likely explanation. The end result, however, is what he termed a differential diagnosis, as opposed to a direct diagnosis: “But in truth, it is often what is called differential:  that is, it takes cognizance of and dwells on the essential signs by which one disease can be discriminated from another resembling it. “


Clearly I’m very excited by Da Costa – his insights into the nature of clinical reasoning and diagnostics far surpass his contemporaries and even those writing decades later like William Osler. I asked Gurpreet, as an expert of modern clinical reasoning literature, what he thought:


Gurpreet Dhaliwal (18:34):

Well, I liked, uh, you know, in the summary, I like that he recognized there was these two polar tensions. It really is reminiscent to me of a, um, or at least he’s raising these tensions of this synthetic versus analytical model. It, it does remind me of a passage in Katherine Montgomery’s book, how Dr. Stink, which you and I are both huge fans of. And I have referred to countless times where she says medicine has these sort of contradicting axioms, and we can’t get away from them. And we, you know, we can’t rectify them like, you know, trust the patient. They’re always telling you the, you know, the diagnosis kind of thing. And you can find the diagnosis 80% of the time in the history. And then these contradictory tales, like, you know, people may downplay certain things that they do and you have to, um, come at it in different ways to get it and, and medicines full of those.

Gurpreet Dhaliwal (19:17):

So this has the same thing, both are true. Um, but eh, to the extent that the Costa steers us in the direction of the analytical method, I think it’s more compatible with what our practice is. You know, um, we have, what we really operate under is sort of, uh, SIS, which is this practical wisdom. Like people argue about whether medicine is a science or whether it’s an art. Um, and again, back to Montgomery’s book, she just makes a great argument. It’s neither, it doesn’t have many elements of science to it, and it doesn’t have any elements of art to it. It’s really sort of a practical trade. Um, that happens to be infused by science, but you really are starting from the ground up. You have this patient in front of you and your job when all is said and done is to bring every ounce of knowledge. You have human biology, clinical trials, the health system, how humans operate to solve their problem. And it’s hard to believe that taking a full inventory of just their biomedical profile will accomplish all those things.

Gurpreet is referring to the book How Doctors Think by Kathryn Montgomery. It is, in my opinion, the single book that I’ve read that has influenced me the most as a physician, and it’s the book that I always recommend to my residents and students if they want deeper insights into physician metacognition. Montgomery is not physician – her PhD is in English literature if I recall – but she spent decades as a “licensed trespasser on clinical territory” at Northwestern University as the director of the medical humanities program, and she expertly describes the ACTUAL – not the normative – processes that we use to think about our patients. And she ends up making the very persuasive observation – and observation that we will get into in the next episode – that the practice of medicine, the actual intellectual processes behind medicine – are about as far from a science as you can get, and rather a type of practical reasoning that she calls “phronesis.” But all of that is for next time!