How to Apply Anki to Complex Diagrams: Breaking Down Differential Diagnoses

clinical problem solvers podcast mullen memory alex mullen anki complex diagrams algorithms differential diagnosis ddx

Generating a differential diagnosis—a list of potential causes for a patient’s presentation—can be loads of fun. The “ddx” might be my favorite part of medicine. It’s what guides workup and gets the physician to the final answer. The fact that diagnostic radiologists run differentials all day was a big part of Cathy’s and my decisions to pursue radiology. Here’s the problem: I’ve always struggled with differentials. If I don’t have a consistent approach for handling a given problem—shortness of breath, chest pain, altered mental status, etc—I miss things. In all the commotion, I forget obvious etiologies. I fall prey to availability bias, anchoring on diagnoses I happen to have seen recently. That’s why I’ve become a big fan of the podcast Clinical Problem Solvers. (OnlineMedEd’s intern content does a similar thing.) On @CPSolvers, the hosts work through clinical cases to develop systematic differentials for common medical issues, whether patient complaints (shortness of breath), lab abnormalities (hyponatremia), imaging abnormalities (bone lesions), or processes (hemolysis). Here’s an example of their diagnostic approach to shortness of breath (SOB):

Shared with permission from the Clinical Problem Solvers

Of course, not everything is worth committing to long-term memory. I forwent memorizing their approach to bone lesions, for example. I can always look that up. However, I’d argue this SOB diagram makes the cut. SOB is such a common, often acute issue in medicine. It pays to be able to mentally review the most common causes—and be confident I haven’t missed obvious ones—without taking time to pull out my phone. Additionally, having certain schema memorized allows me to make clinical reasoning judgments in real time I might have otherwise overlooked. Memorizing this schema via spaced retrieval practice is a stepping stone to developing an intuitive conceptual framework for SOB. (The utility of retrieval practice in fostering deep understanding is underscored by a famous 2011 study by Karpicke and Blunt. Retrieval practice outperformed concept mapping in preparation of … drawing a concept map! In other words, retrieval practice beat concept mapping—generally considered a deep learning activity—at its own game.)

So, what might be the optimal way of getting this diagram into long-term memory? If I’m aiming to apply evidence-based learning practices (namely, spaced retrieval practice via Anki) to this problem, I’m wondering: What’s the best way to translate this approach into Anki cards?

In my opinion, putting “Approach to SOB?” on the front with the diagram screenshot on the back is a no-go. If there were one golden rule of Anki card creation, it might be: Don’t make long cards. There’s minimal research investigating flashcard best practices, so we’re left to our experience and intuition. Long cards make reviewing a slog and force you to waste time regurgitating information you already know. Take the extreme: imagine repeating this “Approach to SOB?” card because you forgot “kyphoscoliosis.” That’s inefficient.

The solution? Break it down. I try to start from the view at 10,000 feet and work my way down until each card is concise, with certain caveats I’ll describe. I add “broad ddx” to my first card (or whenever a card asks for overarching categories), prompting me to recall the major buckets. As the broad ddx guides workup, I often tie the two together. Card 1 therefore becomes: 

dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis

The unique 2D spatial layout helps me recall these six buckets. Cathy describes 2D layout as a recall aid in this article.

Here’s card 2 (the red pen is for clarity; I don’t actually use it when making my cards):

dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis

I’m bordering on breaking my “Don’t make long cards” rule here. In this case, however, I find it useful to think of the cardiovascular causes of SOB as one chunk—a triplet spatially arranged as in the diagram. Also, as this card demonstrates, I often forgo typing out the info, since I want to retain the 2D layout of the concept map as much as possible. 

Here’s card 3:

dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis

Again, bordering on breaking the golden rule. Here, however, I utilize an anatomical approach, and I prefer to be able to fully run through it without breaking it into disconnected chunks. My personal approach—a form of body palace—differs slightly from the organization described in the diagram. I start at the top of the right lung (pneumothorax), before sliding down the pleura to the base (effusion) then entering the lung parenchyma, specifically the interstitium (ILD). I then move into the alveolus, where I think of the five substances that might fill it: water (I think secretions), pus (pneumonia), blood (diffuse alveolar hemorrhage), protein (ARDS), cells (cancer). I  move into the airways, where I consider obstructive causes such as asthma, COPD, and bronchiectasis. Finally, I jump into the vasculature to examine the final four causes: PE, pulmonary hypertension, hepatopulmonary syndrome, and vasculitis. That’s a lot, but it’s helpful for me to keep that journey intact. Because of the approach’s intuitive, anatomical nature, learners should be able to run through that “palace” in just a few seconds. 

Finally, the last three cards:

dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis
dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis
dyspnea shortness of breath clinical problem solvers cpsolvers diagnostic schema differential diagnosis

I often apply acronyms when I struggle to recall these chunks. As an example, taking the last bucket, “Other” (which I translate to “forced”), I might use “DATA” to recall deconditioned, acidosis, T4 (hyperthyroid), and anxiety. I’ll write this on the back in my mnemonic field. 

So, SOB → six cards. I’ve applied this break-it-down strategy—which includes several techniques we discuss on this site: chunking, 2D spatial memory and concept maps, body palaces, acronyms—to learn several dozen of these schema through CPSolvers and OnlineMedEd. They tend to cement in long-term memory once I’ve applied them to actual patients a few times. I’ve found them incredibly valuable for systematizing my thinking for common medical issues. I’ve already picked up many diagnoses I otherwise would have missed!

What do you think? What strategies do you use when learning complex diagrams or algorithms? Do you learn or apply diagnostic schema for differential diagnoses? Share a comment below!