Skip to main content
All posts

Theory at the Bedside (2): Theories

Geoffrey Stetson, MD · November 16, 2026 · 12 min read
One beam lit part of the water. Two, aimed on purpose, show much more.

Two weeks ago, we left Mr. R in the emergency department with a one-liner, and we left my feedback problem with a one-liner of its own. Today is about the next thing every clinician does with a one-liner, which is to build the list of what might explain it.

Mr. R's differential diagnosis

To get a differential diagnosis started for Mr. R, I asked OpenEvidence, an AI tool a lot of clinicians use for clinical questions. I also like to think of MedEdMentor as an “OpenEvidence for HPE,” or perhaps OpenEvidence is a “MedEdMentor for clinical questions” ;). It came back sorted the way most of us were taught to sort: the diagnoses that are most likely, and the ones we cannot afford to miss.

Mr. R's differential, sorted the way we were taught: what is most likely, and what we cannot afford to miss.

I mostly agreed with the list that OpenEvidence built. Heart failure would definitely be on my differential given Mr. R’s history as a heavy smoker. Coronary artery disease is almost certainly present and it is only a matter of time before it shows up clinically. For me, leaning towards heart failure would hinge a lot on the physical exam as I look for things like signs of volume overload (lower extremity edema, crackles in the lung bases, elevated jugular venous pressure). Also, an elevated BNP (brain natriuretic peptide) blood test would be highly suggestive (BNP is released from heart muscle cells when they are under strain). For PE (pulmonary embolism, a blood clot in the lungs), I would want to know if there were any symptoms (new chest pain, unilateral leg swelling +/- redness), signs (characteristic EKG changes, right heart strain on echocardiogram, elevated D-Dimer), or history consistent (long sedentary period, any injuries to his lower extremities, new meds or changes to his meds) with the diagnosis. However, PEs can be subtle, especially in someone with such chronic illness, so he might be getting a CT scan with contrast of his chest (assuming his renal function is good) because it is a “can’t miss” diagnosis. (If you are not a clinician, the details here matter less than the pattern: each entry on the list tells me what to ask, what to look for, and what to test.)

Hypotheses, and the scripts underneath them

In the last post I separated two senses of the word “theory.” When I say at the bedside, "my theory is that this is heart failure," that is a hypothesis about this patient, and the theory in the scientific sense is the thing underneath it. Every entry on Mr. R's list is a hypothesis of that kind, and the thing underneath each one is what clinicians call an illness script.

An illness script includes epidemiology, presenting signs and symptoms, physical exam findings, laboratory findings, and findings on other tests, such as EKGs, echocardiograms, CT scans, etc. I also think about how someone responds to treatment as part of the illness script. Illness scripts are meant to grow over time as a clinician's understanding of a disease state and how it can present changes. For example, a medical student just learning about heart attacks will think about a middle-aged man with an elevated BMI with crushing left-sided chest pain that radiates to his neck and left arm and is accompanied by sweating and anxiety. An experienced clinician will include less-common variants like a middle-aged woman with type 1 diabetes presenting with nausea and/or shortness of breath. (The clinical reasoning literature describes scripts a little more formally, as enabling conditions, the fault, and the consequences, and my list is how I talk about them with my learners.)

So, a script tells me what to look for: who tends to get this, what the exam and the labs should show, and what should happen when we treat it. And my understanding of the script gets better over time. A theory in education does similar work for an educational hypothesis, in that it says what is supposed to be going on, and what to look for if it is. Similarly, we as educators and scholars get to know theories on deeper and deeper levels as we grow in this work.

Take the educational alliance from Telio and colleagues. Telio is a psychiatrist, so someone very familiar with the therapeutic alliance. She and her colleagues built the educational alliance on top of that construct. Its key components are relationships built on trust and respect, coming to a common understanding of goals, and formulating a plan to reach those goals. The way I teach it adds frequent conversations to check in on progress and on whether the goals or the strategy need to change. Of note, “coaching in HPE” is a similar framework.

The educational alliance suggests that problems are caused by a lack of meaningful relationships between supervisors and learners, or a lack of trust and respect between them, or not enough communication and collaboration about how the learner could improve. A lot of supervisors say that time for meaningful relationship building and frequent conversations about goals is difficult to come by, and they point to time as the factor preventing them from forming educational alliances.

But in my experience (and this part is my experience, not a claim from the paper), when people take the time to build educational alliances with their learners, the learners often get far more feedback, they're more engaged in it, the feedback is targeted at the areas that are important to the learner and the supervisor, and growth happens much more rapidly. It often IS more work and time for the supervisor, but it is also more rewarding for the supervisor and more meaningful to the learner. The educational alliance does for learners what the therapeutic alliance does for patients/clients; they just feel cared for and cared about. This feeling makes a huge difference, because it puts learners in a mindset where they can hear hard feedback, since they know the person working with them cares about their well-being and their progress, and has their best interests at heart.

A differential for "nobody tells me how I'm doing"

So, just like I asked OpenEvidence for a differential diagnosis for Mr. R, I asked Uni, our AI mentor for health professions education, for a differential for my educational problem. I gave Uni the complaint the way students usually put it to me ("nobody tells me how I'm doing") and asked for the same two groups of hypotheses as Mr. R's list, “most likely” and “cannot miss.” Then I added one entry of my own at the end. One caution before the list: at the bedside, "most likely" rests on how common each disease is, and nobody has those numbers for feedback problems, so the order here is a judgment (Uni's first, and then mine).

The same complaint, run as a differential. Uni's list, plus one I would never leave off.

Number one is very common. Number two is super similar to number one, and the difference is worth spelling out. In the first, the attending did give feedback (a correction in the operating room, a quick "hold the retractor like this") and nobody, including the student, registered it as feedback. In the second, everyone knew it was feedback, but the student did not use the information to make meaningful behavior changes, either by choice, or because they didn’t know how to. Number three matters because, as we know, the relational component of feedback is really key to its efficacy (this hypothesis follows from the educational alliance). Number five is real too: there are a lot of times when attendings don't work with the students much, so they just don't have a lot of data to go off of, often because the students are mostly with the residents. And number seven I think is quite accurate, because faculty often report being afraid to give tough feedback for two reasons: to avoid hurting people's feelings or to avoid retribution on their own evaluations from the learners. The other half of number seven is feedback that is given unequally; in one study, women residents who struggled got conflicting feedback from different attendings, and men got consistent feedback.

The one I added is another cannot miss: the student isn't engaging in the work because something is going on at home, or they’re struggling with a mental health issue, or something else outside of work is occupying their attention. It is the one where the complaint is real, but the cause may not be feedback at all. It is also not mine to diagnose. What it asks of me is to check in with the student, get a sense of what might be going on, and to know where to send them for support.

Tests that sort, and conversations that add to the list

At the bedside, some of the workup sorts the list (a BNP pushes heart failure up or down), and some of it changes the list (asking Mr. R's daughter exactly what looked different to her). My feedback work had both kinds.

There's a lot of published material about medical students not feeling like they're getting feedback and attendings feeling like they are giving plenty of feedback. The numerical data that we gathered at our institution was very much in line with what's published in the literature. That was the test that sorts. It told me the mismatch was real here too, though it could not tell me which explanation was behind it. (Both things were true of our faculty, as far as I could tell: they agreed with the students that feedback at our school was not what it should be, and most of them felt that they personally were giving it.)

I also had the opportunity to do interviews with faculty and students, and to give lectures and workshops to a lot of different faculty members, and I learned a lot about how they think about feedback. One of the biggest things that I learned about was the lack of expectations that supervisors were setting for their learners. This included expectations around the quantity and quality of feedback. But other expectations that had to do with general workflow weren't being set either. This led to students "making mistakes" based on expectations that they had never known about (that is number six on the list).

Another interesting finding came from faculty who I describe as being “more on the old-school side,” many of whom happened to work in the surgical fields. Some of them argued that their most meaningful feedback experiences came through being humiliated, or embarrassed, or yelled at. So, I ran into folks who thought that a very aggressive and direct feedback approach was effective. This idea is not on the differential diagnosis. It is a new entry, about what people believe feedback is for and how it works, and it came out of a conversation and not a survey or a test, the same way Mr. R's story told by his daughter might change his list (foreshadowing!).

Committing to a leading diagnosis

At some point at the bedside, you have to pick your most likely diagnosis, treat it, and see what happens. My first approach to improving feedback at my institution focused on improving the skill set of the supervisors who were leading feedback conversations and giving them the knowledge and skills to be effective in their jobs. I committed to the theory (a theory in the bedside sense, which is to say a hypothesis) that faculty needed more skills and strategies for feedback based on a needs assessment I did with a group of faculty, students, and administrators. The data made it clear to me that they didn't feel confident in their ability to facilitate effective feedback conversations and that they wanted better tools at their disposal.

I knew going into this effort that faculty were only part of the equation, and that arguably the learners themselves were the more important part. However, it was hard to get time in the curriculum to work with all the medical students. I don't necessarily have data to say whether my first approach worked or didn't, because a feedback intervention takes years to have an effect, and for that effect to show up in graduation questionnaires. But I always knew that I was only getting to half the people involved and needed to get in front of the students. That is how I can say, as I did in the last post, that it took me about three years to get to the students: I held the second hypothesis the whole time and was not able to act on it. So, I think I always went in with two theories: one, that faculty didn't have the skills necessary to carry out effective feedback conversations, and two, that students were not equipped to solicit and engage with meaningful feedback.

Occam's Razor and Hickam's Dictum

Some clinicians may recognize the tension between these phrases. Occam's razor tells us to look for the one diagnosis that explains everything. Hickam's dictum (usually credited to a physician named John Hickam, though no one has found where he said it) answers that "a man can have as many diseases as he damn well pleases," and Mr. R could have a COPD exacerbation and new heart failure or a PE on the same night (also, fluid in the lungs from heart failure, or inflammation in the lungs from a PE can trigger a COPD exacerbation). The same goes for the differential diagnosis of my educational problem, where numbers one through six could all be occurring in the same clerkship at once.

I think scholars believe there is a single right theory because of their clinical backgrounds and their positivist approach to the world (a term the next post will unpack). When we're in a noon conference, a morning report, or an M&M (morbidity and mortality conference), and we're discussing a difficult case, the objective is typically to arrive at the "right answer." So, I think it's just our makeup as clinicians and biomedical thinkers that leads us to seek the right answer, but in HPE, it may not exist. Teaching, learning, and feedback are relational, psychological, and very humanistic endeavors that create systems that are complicated in ways very different from the human body, making “just one right answer” quite rare.

The mentors I interviewed for the Abstracting-Contextualizing Model paper described both sides of this, and a few of their words are in the paper's published supplement. One described the Occam failure, people who come up with a framework and "try and make it fit in every single instance," and said of a forced fit, "you can make it work, but it's just not right." Another described a mentee who built "a laundry list of potentially applicable" theories and, in that theory-shopping, "learned about how people have theorized, talked about, or considered that phenomenon," which is about as close to a differential as a mentor can provide. And a third put the Hickam version this way:

This is also a place where the analogy strains, which I promised to point out. Mr. R can have two diseases at once because both are really there in his body (a PE is a clot in the pulmonary arteries, whether or not anyone looks for it). The things a theory describes are real too: a student and a supervisor either trust each other or they don't, and they either share goals and a plan or they don't. But the educational alliance itself is not sitting inside the relationship waiting to be found. It is a lens people built for looking at that relationship, and psychological safety is another lens for examining the same relationship. So, choosing one of them, or both, is a decision about how to look, and not the discovery of something that is there.

Closing

So, what I tell people is that there is no one right theory, and you don't have to use just one theory, and you don't have to use the entire theory. Some fit better than others (the mentor above is right that a forced fit is "just not right"), and pieces from theories built on very different assumptions do not always sit well together, which is where the paradigms post (post three) will pick up. But you do need to settle on some combination of theories, frameworks, and models (I do not differentiate between these, though I know some people do) that helps you better understand the phenomenon you care about. And it has to be something you can explain to other people, because an approach that only makes sense to you, as the teacher or the scholar, is not much help to anyone else.

Now, what to do with this information? My suggestion is to take the “one-liner” you wrote from the first “Theory at the Bedside” post about problems and phenomena, and come up with three hypotheses for what might be happening that could explain your phenomenon of interest. Of those three, take your favorite one, and input it into the Theory Suggester on MedEdMentor. Or, better yet, have a conversation about it with Uni if you have access. See if there is a real theory out there that closely aligns with the hypothesis that you favor. And, if there are multiple, see if there is a way to combine them that makes sense to you. I’d love to hear how this exercise goes for you. As always, geoffstetson@mededmentor.org.

In post one, I said that you would meet “the team” in this post. That wasn’t accurate. My apologies. However, I promise you that you will meet “the team” in the next post, where we will discuss paradigms. Yay!!!

Next week is a holiday in the U.S., so there will be no post. I will be back with something new on November 30th. I haven’t decided whether that will be post three for this series or if I’ll give you all the materials for a new workshop. We will all just have to wait to find out!

Thanks for reading!

References

  1. Telio S, Ajjawi R, Regehr G. The "educational alliance" as a framework for reconceptualizing feedback in medical education. Acad Med. 2015;90(5):609-614.
  2. Custers EJFM. Thirty years of illness scripts: theoretical origins and practical applications. Med Teach. 2015;37(5):457-462.
  3. Stetson GV, Ow GM, O'Brien B. Navigating problems, phenomena, and theories: the Abstracting-Contextualizing Model of Mentoring for theory use in medical education. Acad Med. 2026. https://doi.org/10.1093/acamed/wvag103. Quotes from Supplementary Material 3.
  4. Miller WT. Letter from the editor: Occam versus Hickam. Semin Roentgenol. 1998;33(3):213.
  5. Stetson G. Theory at the Bedside: Problems and Phenomena. MedEdMentor blog. November 2, 2026. https://mededmentor.org/blog/theory-at-the-bedside-problems-and-phenomena/
  6. Mueller AS, Jenkins TM, Osborne M, Dayal A, O'Connor DM, Arora VM. Gender differences in attending physicians' feedback to residents: a qualitative analysis. J Grad Med Educ. 2017;9(5):577-585.
  7. Carless D, Boud D. The development of student feedback literacy: enabling uptake of feedback. Assess Eval High Educ. 2018;43(8):1315-1325.
  8. Bok HGJ, Teunissen PW, Spruijt A, et al. Clarifying students' feedback-seeking behaviour in clinical clerkships. Med Educ. 2013;47(3):282-291.
  9. Edmondson A. Psychological safety and learning behavior in work teams. Adm Sci Q. 1999;44(2):350-383.

Enjoyed this article?

Create a free MedEdMentor account to access our full library of resources.

Get started free