Clinical Reasoning and Differential Diagnosis Guide

Hello, I'm Eric Strong from Stanford University and the Palo Alto VA Hospital. This is a three-part video series entitled A Guide to Clinical Reasoning or How to Create an Accurate Differential Diagnosis from a Patient's Presentation. The learning objectives of this video are first, to demonstrate a standardized approach to generating a focused differential diagnosis from a patient's presentation. Second, to create concise problem representations using semantic qualifiers and clinical syndromes. Next, to understand the types of frameworks to which the key features of a presentation should be applied. And finally, to know the categories of diagnoses which should be included in the differential. Unfortunately, literature and textbooks that discuss and attempt to teach clinical reasoning often fail because of dense terminology and a focus on abstract concepts that trainees find difficult to apply at the bedside with concrete examples. In this video, I'll be presenting an approach to clinical reasoning, focusing on the differential diagnosis that is practical at the bedside, accessible to students and other novice providers, and minimizes unfamiliar terminology. Part 1 will introduce the clinical reasoning process, and Parts 2 and 3 will go through examples, start to finish, of how to apply that process to a real-world patient. Part 2 will be an example at the level of a preclinical student, and Part 3 will be an example at the level of an intern. I'm going to start by defining the term clinical reasoning. Clinical reasoning is the process by which a healthcare provider takes objective data acquired from an actual patient and interprets it using factual knowledge from a textbook or the medical literature in order to either make a diagnosis or develop a treatment plan. It isn't a single individual skill, but rather a collection of related skills. It involves interpretation of subjective data, evaluation of the accuracy and validity of data, synthesis of individual pieces of data into higher order groups, determination of the relevance of scientific literature for a specific clinical situation, critical evaluation of the arguments for and against diagnoses, application of biostatistics, And finally, integration of different types of knowledge into a complete decision-making process. Although clinical reasoning isn't just about figuring out a diagnosis, that will be the focus of this particular set of videos. I consider there to be five steps to generating a differential diagnosis. The first is to acquire data, and to do this, I recommend that you use all available sources. That includes first and foremost the interview and examination of the patient, but also diagnostic tests and chart review focusing on the medication list, recent primary care notes, and recent discharge summaries. Also, depending on the circumstances, you may consider talking to close family members and friends for collateral information provided, of course, that the patient consents. As you gain experience with clinical reasoning, you may begin to find yourself skipping the five-step process and actually start developing differential diagnoses as you are talking to and examining the patient, that is, as you are acquiring the data. This skill dramatically helps with efficiency, and helps choose the most appropriate diagnostic tests. However, it does not necessarily result in a more accurate diagnosis in the end, as it sets one up for something called anchoring bias. Anchoring bias is the tendency to place too much weight on a single piece of information acquired early in the data acquisition process and a failure to update the differential diagnosis when conflicting information is later presented. For example, if a patient's HPI or history of present illness focuses on his shortness of breath and he happens to mention early on that it's worse when lying down, a specific symptom called orthopnea, that may lead a clinician to begin hypothesizing during the interview that the patient has heart failure. Additional investigation may then be appropriately directed towards confirming or refuting this hypothesis. If the clinician then learns the patient has a fever, has a focal decrease in breath sounds on exam, and a consolidation on chest x-ray, he or she should reconsider their original hypothesis. If the clinician fails to reconsider the initial leading diagnosis of heart failure, even after acquiring much data suggesting a different diagnosis, in this case pneumonia, he or she has committed anchoring bias. So here's the bottom line. For efficiency's sake, it's a good idea to consider the differential diagnosis in real time during the patient's HMP, but be aware of anchoring bias, and once the entire data set is collected, force yourself to return to the beginning of the clinical reasoning process so that each piece of data and element of the presentation is given its appropriate weight. Once you've acquired the data, the next step is to identify the key features. Key features are the individual elements of the presentation which are likely to help differentiate one diagnosis from another. For example, in a patient with episodic chest pain, whether or not the pain comes on with exercise will help to determine the likelihood it is from cardiac ischemia. Therefore, this is a key feature. In contrast, the severity of chest pain on a 1 to 10 scale is surprisingly non-predictive of the eventual diagnosis and thus I would generally not consider its severity to be a key feature. Key features include both positive and negative findings. They may be from history, exam, labs, other tests, or chart review. The third step is to create a problem representation. This should use semantic qualifiers and should synthesize related findings into clinical syndromes. Some of these terms may be unfamiliar to you, so let me explain. What do I mean by the term problem representation? This is a one to two sentence summary, using precise medical terminology, of the most highly relevant aspects of the patient's history, exam, and diagnostic tests. Sometimes, problem representation is used synonymously with the terms summary statement as well as impression, the latter term commonly used in written notes in the United States. Semantic qualifiers are qualitative abstractions of the symptom of a case in which an opposing abstraction is either explicit or implied. They help to reframe a patient's symptom into terms more familiar to the clinician and easier to communicate to others. Common categories of qualifiers include the onset of symptom, for example, was it abrupt or progressive, and acute or chronic? Has the course of the symptom been continuous or episodic? Is the site unilateral or bilateral, proximal or distal, diffuse or localized? What is the symptom trigger, postprandial or exertional, or pleuritic or positional? and is the symptom associated with pain or is it painless? The use of semantic qualifiers when reframing a patient's symptom is thought to aid in accessing chunks of information stored in the clinician's memory and is associated with a higher likelihood of arriving at the correct diagnosis. Let's look at some examples of how to use semantic qualifiers to reframe symptoms. Imagine a patient in the ER states For the last 30 minutes, my chest has hurt whenever I take a deep breath. We would reframe this as acute pleuritic chest pain. Or another patient states, Over the past several months, both legs have been getting weaker and weaker. This becomes chronic progressive bilateral lower extremity weakness. A word of caution here. Some information is lost when the patient's presentation is translated into semantic qualifiers which can set one up for bias if the original history is never revisited. Now what about that phrase synthesize into clinical syndromes? A short elaboration on this is that a constellation of clearly related findings should be grouped together into a single clinical syndrome if possible. For example, if a patient has confusion, a fever to 39 degrees, heart rate of 120 beats per minute, blood pressure of 130 over 60, respiratory rate 24, white blood cell count 16,000, creatinine of 2.4, and positive blood cultures, you can synthesize that as severe sepsis. And if another patient has jaundice ascites, confusion asterixis, a total billy of 25, and an INR of 2, that can be summarized as hepatic failure. Another word of caution, don't mistake the clinical syndrome for the diagnosis. For example, it's great if you recognize a patient has severe sepsis. However, severe sepsis is not a final diagnosis, and your diagnostic reasoning should not end here. You must also determine what has caused the severe sepsis. Is it community-acquired pneumonia? urinary tract infection, or appendicitis. So now, using semantic qualifiers and clinical syndromes, how do we construct the problem representation? There are at least two approaches to this. The one I favor is to link four categories of information into a single sentence using a standardized order. Age and gender first, then highly relevant past medical history, followed by the primary symptom using semantic qualifiers, and ending with the highly relevant diagnostic data using clinical syndromes when possible. For example, a 60-year-old woman with a history of poorly controlled diabetes presents with chronic progressive exertional dyspnea, with exam and chest x-ray findings of volume overload, and with unremarkable routine labs and EKG. The other approach to the problem representation is to consider only this category. So for this patient, the problem representation becomes just chronic progressive exertional dyspnea. I personally don't like this approach as much because I think the upside to being more concise is outweighed by the downside of eliminating the additional information, but you certainly will come across the second approach from time to time. Alright, so let's get back to our five steps. That was a long one. Step 4 is to adopt a framework to better understand the patient's problem. This framework may be anatomic, physiologic, or some other type, and it's commonly adapted from a reference source. Frameworks typically take the form of a categorized, general differential diagnosis where the strategy for categorization depends upon the specific problem. What are some types of frameworks? A framework for acute renal failure may first divide diagnoses up into pre-renal, intrarenal, and post-renal, meaning is the problem before blood gets to the kidney, inside the kidney, or after urine leaves the kidney. Some of these categories can be further subdivided, so pre-renal etiologies can either be from dehydration or from low cardiac output, and intrarenal etiologies can either be glomerular, tubular, or interstitial. Another example of a framework, this time for anemia, divides the etiologies into hypoproliferative and hyperproliferative. Hypoproliferative anemias can be from nutritional deficiencies, bone marrow failure, kidney disease or chronic disease and or inflammation. Hyperproliferative anemia can be from acute blood loss or hemolysis. There is not just one acceptable framework for a specific patient problem. For example, Let's consider the patient we just mentioned a minute ago. A 60-year-old woman with a history of poorly controlled diabetes who presents with chronic progressive exertional dyspnea, with examined chest x-ray findings of volume overload, and with unremarkable routine labs and EKG. What type of framework might we choose to adopt here? In my experience, the most likely framework a clinician would choose for this problem is an anatomic one, also referred to as organ-based. So for this woman with dyspnea, that means that it could be a cardiac problem, a pulmonary problem, or a heme problem. As we just saw briefly with the kidneys, all of our organs can be further subdivided into functional components. So a problem in the heart could be in the myocardium, coronary vessels, valves, conduction system, or pericardium. A problem in the lungs can be located in the airways, alveoli, pulmonary vessels, interstitium, or pleura. And finally, hematologic issues can involve any of the individual cell lines or coagulation problems or issues with paraproteins. However, with dyspnea as the chief complaint, the major hematologic concern is of course anemia. My thoroughness with these frameworks is limited by the minimum font size I want to use in the diagrams here, but if I was being more thorough, I would also list the renal system here with subcategories for the renal arteries, glomeruli, tubules, interstitium, and collecting system. So that's the anatomic framework, which is just one way of categorizing the differential diagnosis for this problem. Another completely acceptable framework that's based more on physiology might ask, what are the pathophysiologic mechanisms that can trigger dyspnea? There is hypoxia, which can be from VQ mismatch, impaired diffusion, or shunt. There is hypercapnia, which can be from obstructive lung disease, central hypoventilation, neuromuscular disease, or decreased respiratory compliance. Less commonly is dyspnea from acidemia, either from pathologic acid production or from poor acid elimination. Finally, input from the cerebral cortex from anxiety and pain can also lead to the subjective sensation of dyspnea. This framework isn't necessarily better or worse than the anatomic one, just different. It's likely that some people will naturally gravitate towards one and some to the other. One other type of framework that I feel both obligated and reluctant to mention is the mnemonic framework. I feel obligated because many medical schools still teach and expect their students to use it. I am reluctant because I think it is vastly inferior to other types of frameworks. But nevertheless, here it is. One such mnemonic that's taught is vindicate. The V stands for vascular, the I for inflammatory, the N for neoplastic, D is degenerative, the second I is idiopathic, C is congenital, A is autoimmune, T for traumatic, and lastly E for endocrine. Okay, one more step to go. Apply the key features to the framework, which will generate the preliminary differential diagnosis. When applying the key features, the clinician should use their presence or absence to estimate the likelihood of the diseases or pathophysiologic states that are suggested by the framework. As a brief example, imagine a 55-year-old man with a history of alcoholism and depression, presents with chronic progressive bilateral lower extremity edema with an exam notable for anisarca and a serum albumin of 1.5 grams per deciliter. One of the several frameworks one might employ for this case might be a physiologic one, where there are four categories for the four major mechanisms of hypoalbuminemia. There is impaired intake of protein, seen in malnutrition, impaired utilization and synthesis of protein, seen in liver disease, excessive glomerular filtration of protein, seen in the nephrotic syndrome, and excessive GI loss of protein, a syndrome known as protein-losing enteropathy. The key features for such a patient might include a history of depression, how does depression affect the probability of any of the categories in the framework, It's not linked to liver disease independent of substance abuse, and it's not associated with either nephrotic syndrome or protein-losing enteropathy, however it does increase the chance of malnutrition as the patient may not be eating properly in his depressed state. If key features also included the presence of spider angiomata and splenomegaly on exam, that would dramatically increase the likelihood of liver disease. If the patient has no history of GI symptoms, particularly diarrhea, That would dramatically decrease the probability of protein losing enteropathy, as all pathologies that cause this general diagnosis also lead to diarrhea and a host of other symptoms. And if the key features included a UA without any proteinuria, that would be definitive enough evidence as to completely rule out nephrotic syndrome from the differential altogether. Estimating to what degree individual key features impact which components of the framework and or differential. It's a skill that requires both textbook knowledge, familiarity with scientific literature, and experience. In my opinion, it is the single aspect of clinical reasoning that more than any other differentiates novice clinicians from the experts. Perhaps I should have started this video with a discussion of the next issue, but although most of you are likely familiar with the term differential diagnosis, let me define it specifically so that we are all on the same page with how I'm using it. A differential diagnosis, often abbreviated as just the differential, is a list of possible diagnoses which may explain the patient's presentation. It should include those diagnoses in which either its likelihood is high enough or the danger if it should be missed is high enough in order to warrant additional testing to investigate that specific diagnosis. It may or may not include additional diagnoses whose likelihood is low enough to not warrant immediate testing. but which have not been completely ruled out. The differential diagnosis should be prioritized in descending order of likelihood. A solid focused differential diagnosis should include the following. The one diagnosis that you believe to be the most likely. This is known as the working diagnosis or provisional diagnosis. Two to four diagnoses that are very common in general for which this patient's case could be either a typical or an atypical presentation, any diagnoses which are rapidly fatal if untreated, of which this patient's case could plausibly be the result, This is often known as a quote, don't miss diagnosis. Finally, any diagnoses which are specifically suggested by standout features of the patient's history, including unusual hobbies or job and recent travel to an exotic location. Most of these unusual standout features will be what we refer to as red herrings. A red herring is an unusual element of the presentation that falsely appears quite relevant but ultimately distracts the clinician away from the true diagnosis. It seems common for medical trainees to be instructed to keep their differential diagnoses broad. What this is meant to mean is that the trainee should not prematurely jump to conclusions regarding the culprit organ system or determine that a single diagnosis is the only one worth considering. Unfortunately, the recommendation to keep the differential broad is frequently misunderstood to mean that the differential should be very long, and span every organ system in the body. A long differential is more problematic than a focused one, even in training, because it actually displays less thought and it can be difficult to formulate a diagnostic plan if there are 15 different conditions simultaneously under consideration. For the typical internal medicine admission, I would consider 4 to 6 diagnoses a good ballpark range to aim for when developing a practical differential. My last words of caution. First, the framework and differential diagnosis are not the same thing. The framework may be adopted directly from a reference source and is not specific for your patient. The differential diagnosis, on the other hand, includes only those diagnoses relevant to the patient in question. A differential diagnosis that has not been made specific to the patient is nearly worthless. When prioritizing the differential and establishing the provisional diagnosis, in general, An atypical presentation of a common disease is more likely than a typical presentation of a rare disease. Finally, the true typical presentation of a disease does not always match the textbook description of a disease or its historically taught presentation. For example, it's usually taught in medical schools and reinforced in suboptimally researched textbooks that spontaneous bacterial peritonitis, or SBP, usually presents with abdominal pain, abdominal tenderness, and so-called peritoneal signs on physical exam. In reality, the most common presentation of SPP in a patient with known liver disease is altered mental status, with or without a fever, and without any abdominal signs at all. As another example, it's frequently believed that patients with pericardial tamponade either usually or always present with hypotension. In reality, when studied, it's been shown that the majority of patients with proven tamponade are actually not hypotensive at presentation. There are many many more examples of discrepancies between how an inaccurate classic presentation of a disease negatively impacts diagnostic reasoning. Most patients with heart attacks don't have crushing substernal chest pain. Most patients with migraine headaches don't experience a visual aura before headache onset. The list goes on and on. So that's it for my five steps to a differential diagnosis. Once again, they are first, acquire data. Second, identify the key features. Third, create a problem representation using semantic qualifiers and clinical syndromes. Fourth, adopt the framework. And last, apply the key features to the framework in order to generate the differential diagnosis. That concludes part one of this guide on clinical reasoning, focusing on how to create an accurate and focused differential diagnosis. As I mentioned at the beginning, in parts two and three, I'll go through examples from start to finish of how to apply this approach to a real-world patient. As you listen to parts two and three, I'd consider pausing intermittently as you go in order to practice working through the case on your own.

Second, to create concise problem representations using semantic qualifiers and clinical syndromes. Next, to understand the types of frameworks to which the key features of a presentation should be applied. And finally, to know the categories of diagnoses which should be included in the differential.

Unfortunately, literature and textbooks that discuss and attempt to teach clinical reasoning often fail because of dense terminology and a focus on abstract concepts that trainees find difficult to apply at the bedside with concrete examples. In this video, I'll be presenting an approach to clinical reasoning, focusing on the differential diagnosis that is practical at the bedside, accessible to students and other novice providers, and minimizes unfamiliar terminology. Part 1 will introduce the clinical reasoning process, and Parts 2 and 3 will go through examples, start to finish, of how to apply that process to a real-world patient.

Part 2 will be an example at the level of a preclinical student, and Part 3 will be an example at the level of an intern. I'm going to start by defining the term clinical reasoning. Clinical reasoning is the process by which a healthcare provider takes objective data acquired from an actual patient and interprets it using factual knowledge from a textbook or the medical literature in order to either make a diagnosis or develop a treatment plan.

It isn't a single individual skill, but rather a collection of related skills. It involves interpretation of subjective data, evaluation of the accuracy and validity of data, synthesis of individual pieces of data into higher order groups, determination of the relevance of scientific literature for a specific clinical situation, critical evaluation of the arguments for and against diagnoses, application of biostatistics, And finally, integration of different types of knowledge into a complete decision-making process. Although clinical reasoning isn't just about figuring out a diagnosis, that will be the focus of this particular set of videos.

I consider there to be five steps to generating a differential diagnosis. The first is to acquire data, and to do this, I recommend that you use all available sources. That includes first and foremost the interview and examination of the patient, but also diagnostic tests and chart review focusing on the medication list, recent primary care notes, and recent discharge summaries.

Also, depending on the circumstances, you may consider talking to close family members and friends for collateral information provided, of course, that the patient consents. As you gain experience with clinical reasoning, you may begin to find yourself skipping the five-step process and actually start developing differential diagnoses as you are talking to and examining the patient, that is, as you are acquiring the data. This skill dramatically helps with efficiency, and helps choose the most appropriate diagnostic tests.

However, it does not necessarily result in a more accurate diagnosis in the end, as it sets one up for something called anchoring bias. Anchoring bias is the tendency to place too much weight on a single piece of information acquired early in the data acquisition process and a failure to update the differential diagnosis when conflicting information is later presented. For example, if a patient's HPI or history of present illness focuses on his shortness of breath and he happens to mention early on that it's worse when lying down, a specific symptom called orthopnea, that may lead a clinician to begin hypothesizing during the interview that the patient has heart failure.

Additional investigation may then be appropriately directed towards confirming or refuting this hypothesis. If the clinician then learns the patient has a fever, has a focal decrease in breath sounds on exam, and a consolidation on chest x-ray, he or she should reconsider their original hypothesis. If the clinician fails to reconsider the initial leading diagnosis of heart failure, even after acquiring much data suggesting a different diagnosis, in this case pneumonia, he or she has committed anchoring bias. So here's the bottom line. For efficiency's sake, it's a good idea to consider the differential diagnosis in real time during the patient's HMP, but be aware of anchoring bias, and once the entire data set is collected, force yourself to return to the beginning of the clinical reasoning process so that each piece of data and element of the presentation is given its appropriate weight.

Once you've acquired the data, the next step is to identify the key features. Key features are the individual elements of the presentation which are likely to help differentiate one diagnosis from another. For example, in a patient with episodic chest pain, whether or not the pain comes on with exercise will help to determine the likelihood it is from cardiac ischemia. Therefore, this is a key feature.

In contrast, the severity of chest pain on a 1 to 10 scale is surprisingly non-predictive of the eventual diagnosis and thus I would generally not consider its severity to be a key feature. Key features include both positive and negative findings. They may be from history, exam, labs, other tests, or chart review.

The third step is to create a problem representation. This should use semantic qualifiers and should synthesize related findings into clinical syndromes. Some of these terms may be unfamiliar to you, so let me explain. What do I mean by the term problem representation? This is a one to two sentence summary, using precise medical terminology, of the most highly relevant aspects of the patient's history, exam, and diagnostic tests.

Sometimes, problem representation is used synonymously with the terms summary statement as well as impression, the latter term commonly used in written notes in the United States. Semantic qualifiers are qualitative abstractions of the symptom of a case in which an opposing abstraction is either explicit or implied. They help to reframe a patient's symptom into terms more familiar to the clinician and easier to communicate to others. Common categories of qualifiers include the onset of symptom, for example, was it abrupt or progressive, and acute or chronic? Has the course of the symptom been continuous or episodic?

Is the site unilateral or bilateral, proximal or distal, diffuse or localized? What is the symptom trigger, postprandial or exertional, or pleuritic or positional? and is the symptom associated with pain or is it painless? The use of semantic qualifiers when reframing a patient's symptom is thought to aid in accessing chunks of information stored in the clinician's memory and is associated with a higher likelihood of arriving at the correct diagnosis.

Let's look at some examples of how to use semantic qualifiers to reframe symptoms. Imagine a patient in the ER states For the last 30 minutes, my chest has hurt whenever I take a deep breath. We would reframe this as acute pleuritic chest pain.

Or another patient states, Over the past several months, both legs have been getting weaker and weaker. This becomes chronic progressive bilateral lower extremity weakness. A word of caution here. Some information is lost when the patient's presentation is translated into semantic qualifiers which can set one up for bias if the original history is never revisited.

Now what about that phrase synthesize into clinical syndromes? A short elaboration on this is that a constellation of clearly related findings should be grouped together into a single clinical syndrome if possible. For example, if a patient has confusion, a fever to 39 degrees, heart rate of 120 beats per minute, blood pressure of 130 over 60, respiratory rate 24, white blood cell count 16,000, creatinine of 2.4, and positive blood cultures, you can synthesize that as severe sepsis.

And if another patient has jaundice ascites, confusion asterixis, a total billy of 25, and an INR of 2, that can be summarized as hepatic failure. Another word of caution, don't mistake the clinical syndrome for the diagnosis. For example, it's great if you recognize a patient has severe sepsis. However, severe sepsis is not a final diagnosis, and your diagnostic reasoning should not end here.

You must also determine what has caused the severe sepsis. Is it community-acquired pneumonia? urinary tract infection, or appendicitis. So now, using semantic qualifiers and clinical syndromes, how do we construct the problem representation? There are at least two approaches to this.

The one I favor is to link four categories of information into a single sentence using a standardized order. Age and gender first, then highly relevant past medical history, followed by the primary symptom using semantic qualifiers, and ending with the highly relevant diagnostic data using clinical syndromes when possible. For example, a 60-year-old woman with a history of poorly controlled diabetes presents with chronic progressive exertional dyspnea, with exam and chest x-ray findings of volume overload, and with unremarkable routine labs and EKG.

The other approach to the problem representation is to consider only this category. So for this patient, the problem representation becomes just chronic progressive exertional dyspnea. I personally don't like this approach as much because I think the upside to being more concise is outweighed by the downside of eliminating the additional information, but you certainly will come across the second approach from time to time.

Alright, so let's get back to our five steps. That was a long one. Step 4 is to adopt a framework to better understand the patient's problem. This framework may be anatomic, physiologic, or some other type, and it's commonly adapted from a reference source. Frameworks typically take the form of a categorized, general differential diagnosis where the strategy for categorization depends upon the specific problem.

What are some types of frameworks? A framework for acute renal failure may first divide diagnoses up into pre-renal, intrarenal, and post-renal, meaning is the problem before blood gets to the kidney, inside the kidney, or after urine leaves the kidney. Some of these categories can be further subdivided, so pre-renal etiologies can either be from dehydration or from low cardiac output, and intrarenal etiologies can either be glomerular, tubular, or interstitial.

Another example of a framework, this time for anemia, divides the etiologies into hypoproliferative and hyperproliferative. Hypoproliferative anemias can be from nutritional deficiencies, bone marrow failure, kidney disease or chronic disease and or inflammation. Hyperproliferative anemia can be from acute blood loss or hemolysis.

There is not just one acceptable framework for a specific patient problem. For example, Let's consider the patient we just mentioned a minute ago. A 60-year-old woman with a history of poorly controlled diabetes who presents with chronic progressive exertional dyspnea, with examined chest x-ray findings of volume overload, and with unremarkable routine labs and EKG. What type of framework might we choose to adopt here?

In my experience, the most likely framework a clinician would choose for this problem is an anatomic one, also referred to as organ-based. So for this woman with dyspnea, that means that it could be a cardiac problem, a pulmonary problem, or a heme problem. As we just saw briefly with the kidneys, all of our organs can be further subdivided into functional components. So a problem in the heart could be in the myocardium, coronary vessels, valves, conduction system, or pericardium.

A problem in the lungs can be located in the airways, alveoli, pulmonary vessels, interstitium, or pleura. And finally, hematologic issues can involve any of the individual cell lines or coagulation problems or issues with paraproteins. However, with dyspnea as the chief complaint, the major hematologic concern is of course anemia.

My thoroughness with these frameworks is limited by the minimum font size I want to use in the diagrams here, but if I was being more thorough, I would also list the renal system here with subcategories for the renal arteries, glomeruli, tubules, interstitium, and collecting system. So that's the anatomic framework, which is just one way of categorizing the differential diagnosis for this problem. Another completely acceptable framework that's based more on physiology might ask, what are the pathophysiologic mechanisms that can trigger dyspnea? There is hypoxia, which can be from VQ mismatch, impaired diffusion, or shunt.

There is hypercapnia, which can be from obstructive lung disease, central hypoventilation, neuromuscular disease, or decreased respiratory compliance. Less commonly is dyspnea from acidemia, either from pathologic acid production or from poor acid elimination. Finally, input from the cerebral cortex from anxiety and pain can also lead to the subjective sensation of dyspnea.

This framework isn't necessarily better or worse than the anatomic one, just different. It's likely that some people will naturally gravitate towards one and some to the other. One other type of framework that I feel both obligated and reluctant to mention is the mnemonic framework.

I feel obligated because many medical schools still teach and expect their students to use it. I am reluctant because I think it is vastly inferior to other types of frameworks. But nevertheless, here it is. One such mnemonic that's taught is vindicate.

The V stands for vascular, the I for inflammatory, the N for neoplastic, D is degenerative, the second I is idiopathic, C is congenital, A is autoimmune, T for traumatic, and lastly E for endocrine. Okay, one more step to go. Apply the key features to the framework, which will generate the preliminary differential diagnosis. When applying the key features, the clinician should use their presence or absence to estimate the likelihood of the diseases or pathophysiologic states that are suggested by the framework.

As a brief example, imagine a 55-year-old man with a history of alcoholism and depression, presents with chronic progressive bilateral lower extremity edema with an exam notable for anisarca and a serum albumin of 1.5 grams per deciliter. One of the several frameworks one might employ for this case might be a physiologic one, where there are four categories for the four major mechanisms of hypoalbuminemia. There is impaired intake of protein, seen in malnutrition, impaired utilization and synthesis of protein, seen in liver disease, excessive glomerular filtration of protein, seen in the nephrotic syndrome, and excessive GI loss of protein, a syndrome known as protein-losing enteropathy. The key features for such a patient might include a history of depression, how does depression affect the probability of any of the categories in the framework, It's not linked to liver disease independent of substance abuse, and it's not associated with either nephrotic syndrome or protein-losing enteropathy, however it does increase the chance of malnutrition as the patient may not be eating properly in his depressed state. If key features also included the presence of spider angiomata and splenomegaly on exam, that would dramatically increase the likelihood of liver disease.

If the patient has no history of GI symptoms, particularly diarrhea, That would dramatically decrease the probability of protein losing enteropathy, as all pathologies that cause this general diagnosis also lead to diarrhea and a host of other symptoms. And if the key features included a UA without any proteinuria, that would be definitive enough evidence as to completely rule out nephrotic syndrome from the differential altogether. Estimating to what degree individual key features impact which components of the framework and or differential.

It's a skill that requires both textbook knowledge, familiarity with scientific literature, and experience. In my opinion, it is the single aspect of clinical reasoning that more than any other differentiates novice clinicians from the experts. Perhaps I should have started this video with a discussion of the next issue, but although most of you are likely familiar with the term differential diagnosis, let me define it specifically so that we are all on the same page with how I'm using it.

A differential diagnosis, often abbreviated as just the differential, is a list of possible diagnoses which may explain the patient's presentation. It should include those diagnoses in which either its likelihood is high enough or the danger if it should be missed is high enough in order to warrant additional testing to investigate that specific diagnosis. It may or may not include additional diagnoses whose likelihood is low enough to not warrant immediate testing. but which have not been completely ruled out. The differential diagnosis should be prioritized in descending order of likelihood.

A solid focused differential diagnosis should include the following. The one diagnosis that you believe to be the most likely. This is known as the working diagnosis or provisional diagnosis. Two to four diagnoses that are very common in general for which this patient's case could be either a typical or an atypical presentation, any diagnoses which are rapidly fatal if untreated, of which this patient's case could plausibly be the result, This is often known as a quote, don't miss diagnosis.

Finally, any diagnoses which are specifically suggested by standout features of the patient's history, including unusual hobbies or job and recent travel to an exotic location. Most of these unusual standout features will be what we refer to as red herrings. A red herring is an unusual element of the presentation that falsely appears quite relevant but ultimately distracts the clinician away from the true diagnosis. It seems common for medical trainees to be instructed to keep their differential diagnoses broad.

What this is meant to mean is that the trainee should not prematurely jump to conclusions regarding the culprit organ system or determine that a single diagnosis is the only one worth considering. Unfortunately, the recommendation to keep the differential broad is frequently misunderstood to mean that the differential should be very long, and span every organ system in the body. A long differential is more problematic than a focused one, even in training, because it actually displays less thought and it can be difficult to formulate a diagnostic plan if there are 15 different conditions simultaneously under consideration. For the typical internal medicine admission, I would consider 4 to 6 diagnoses a good ballpark range to aim for when developing a practical differential.

My last words of caution. First, the framework and differential diagnosis are not the same thing. The framework may be adopted directly from a reference source and is not specific for your patient.

The differential diagnosis, on the other hand, includes only those diagnoses relevant to the patient in question. A differential diagnosis that has not been made specific to the patient is nearly worthless. When prioritizing the differential and establishing the provisional diagnosis, in general, An atypical presentation of a common disease is more likely than a typical presentation of a rare disease. Finally, the true typical presentation of a disease does not always match the textbook description of a disease or its historically taught presentation.

For example, it's usually taught in medical schools and reinforced in suboptimally researched textbooks that spontaneous bacterial peritonitis, or SBP, usually presents with abdominal pain, abdominal tenderness, and so-called peritoneal signs on physical exam. In reality, the most common presentation of SPP in a patient with known liver disease is altered mental status, with or without a fever, and without any abdominal signs at all. As another example, it's frequently believed that patients with pericardial tamponade either usually or always present with hypotension. In reality, when studied, it's been shown that the majority of patients with proven tamponade are actually not hypotensive at presentation. There are many many more examples of discrepancies between how an inaccurate classic presentation of a disease negatively impacts diagnostic reasoning.

Most patients with heart attacks don't have crushing substernal chest pain. Most patients with migraine headaches don't experience a visual aura before headache onset. The list goes on and on.

So that's it for my five steps to a differential diagnosis. Once again, they are first, acquire data. Second, identify the key features.

Third, create a problem representation using semantic qualifiers and clinical syndromes. Fourth, adopt the framework. And last, apply the key features to the framework in order to generate the differential diagnosis.

That concludes part one of this guide on clinical reasoning, focusing on how to create an accurate and focused differential diagnosis. As I mentioned at the beginning, in parts two and three, I'll go through examples from start to finish of how to apply this approach to a real-world patient. As you listen to parts two and three, I'd consider pausing intermittently as you go in order to practice working through the case on your own.

Transcript for:Clinical Reasoning and Differential Diagnosis Guide

Transcript for:
Clinical Reasoning and Differential Diagnosis Guide