How to Retain Medical Knowledge: What the Forgetting Studies Show
How much medical knowledge do doctors actually forget?
Less, and later, than the folklore says. In cross-sectional testing of basic science knowledge in medical students and doctors, unrehearsed material tends to hold for about 1.5 to 2 years after it was last used, then fades on a slowing curve. Knowledge you keep using holds much longer.
In a Dutch cross-sectional study that tested basic science recall in medical students and practicing doctors, students averaged about 40% correct, while doctors after many years of practice scored around 25 to 30%. The authors write, “The popular notion that most of basic science knowledge is forgotten shortly after graduation is not supported by our findings.” See Custers and ten Cate.
Controlling for rehearsal, the same study found little loss for 1.5 to 2 years after knowledge was last used, then a negatively accelerated, logarithmic decline. The pattern fit Ebbinghaus for meaningful material, except the downward slope began later than in most prior reports.
There was also a long tail. After 25 or more years, retention clustered around 15 to 20% on the same basic science test.
Zooming out, across general and medical education on basic science content, about two-thirds to three-quarters of knowledge is retained after one year, with retention slightly below 50% in the following year, based on a 2010 review.
These data are about basic science recall, not clinical judgment on the ward. The key variable is last use, because material you keep applying tends to stick. This is the starting point for thinking about how to retain medical knowledge.
Why rereading and cramming don't stick
Rereading and highlighting feel productive because the page starts to look familiar. That feeling is recognition, not recall, and it does not exercise the skill you need when the book is closed and a case is unfolding.
A 2013 review of study techniques rated practice testing and distributed practice as high utility, and rated highlighting, rereading, and summarization as low utility. "Most students report rereading and highlighting, yet these techniques do not consistently boost students' performance," and the authors suggest practice testing instead of rereading.
A test-enhanced learning review found that effortful-retrieval tests, like short answer, promote better retention than recognition tests, like multiple choice. Feedback enhances the benefit, and the gain from testing is greatest when practice is spaced. From what I can tell, that maps to real cases where cues are messy and you have to generate the answer.
Cramming is massed study, the opposite of distributed practice, which that 2013 review rated high utility.
Retrieval practice: test yourself, don't review
Retrieval practice means you pull the answer out of your head, then check it. It feels harder than rereading, and that difficulty is the learning signal you want.
In a randomized trial in pediatric and emergency medicine residents, two topics were taught, status epilepticus and myasthenia gravis. One topic was learned with repeated short answer tests with feedback, the other with repeated study of a review sheet, at teaching and then about every 2 weeks. On a final test over 6 months later, the tested topic averaged 39 percent vs 26 percent for the studied topic, a 13 point gap, with 19 and 21 residents completing, from a randomized trial in residents.
Feedback mattered. They got it after each test, which aligns with what a 2008 review reported, that feedback enhances the benefit of testing and that spacing strengthens it. That same review found effortful short answer prompts beat recognition tests like multiple choice for retention.
The raw scores in both arms were low. Honest, and useful. Six months later, residents remembered more of what they had repeatedly retrieved, but plenty had faded across the board.
Spacing: why a little, often, beats one big catch-up
The cleanest signal I’ve seen comes from Kerfoot's spaced education trial. Third-year students finished a one-week urology rotation, then got weekly emailed case scenarios with a multiple choice question, an answer, and a teaching point on two of four topics. Composite end-of-year scores were higher with spacing, P<0.001.
Ninety-five of 133 students (71%) completed the end-of-year test. The effect was largest in students who had finished urology 6 to 8 and 9 to 11 months earlier, with Cohen’s effect sizes 1.01 and 0.73.
From what I can tell, the drop-off bites months after you leave a service, not during the rotation. That’s exactly where case-shaped recall, little and often, earns its keep.
This was weekly, not daily, and it was medical students, not practicing doctors. Nobody has shown daily beats weekly here.
Why cases, and mixed cases, help clinical recall
Reasoning: a fact tied to a specific patient and a decision gives you retrieval cues you can recognize later, under pressure. The vignette, the competing diagnoses, the consequence of picking one path over another.
Kerfoot’s 2007 randomized trial with third-year students after a one-week urology rotation used weekly emailed case scenarios with a multiple choice question, an answer, and a teaching point on two of four topics. The spaced material was case shaped, not isolated facts or flashcards, which matters for how to retain medical knowledge.
On mixing, a classic signal comes from medical students learning ECG diagnosis. In an ECG learning study, mixed practice produced 46 percent diagnostic accuracy on new ECGs, compared with 30 percent after blocked practice, p<0.05. The outcome was performance on new tracings, not memorization of labeled examples.
A caution line runs through this literature. Dunlosky’s 2013 review rated interleaved practice as moderate utility, with evidence still limited. These are student data, and effects can shrink or wobble in different settings.
Practically, mix presentations instead of doing one topic at a time. The next patient does not arrive labeled. Commit to a diagnosis among competing ones, see where you were wrong, then cycle. Here is a primer on committing to a differential.
What a two-minute daily case looks like
"I've suddenly gone short of breath with chest pain."
Martin Rowe, 55, runs a small joinery firm and is usually fit. He is an AI patient, fictional, and he is worried it is a heart attack.
- Two days ago, sudden sharp right-sided chest pain, worse breathing in. A learner weighs acute coronary syndrome, pneumonia, pneumothorax, or pericarditis.
- Vitals now: HR 112, RR 24, BP 120/78, SpO2 91%, temp 37.1 C. Acute coronary syndrome, pneumonia, pneumothorax, and pericarditis are still in play.
- Coughed up small amounts of blood twice, streaks in spit. Pulmonary embolism edges up, though acute coronary syndrome, pneumonia, and pneumothorax remain considerations.
- About a week of a swollen, tight, tender, warm left calf. Pulmonary embolism rises, with deep vein thrombosis as the source under consideration.
- Keyhole surgery on the left knee about three weeks ago, less mobile since, and a long-haul flight from Australia about five days ago. With the calf plus the surgery and flight, pulmonary embolism becomes the leading answer.
Calling this just a deep vein thrombosis names the source but misses the dangerous part.
Diagnosica's quick case works like this, patient says one line, five clues one at a time, you guess after each, then you see the answer and a one-line teaching point; it keeps a daily streak. A two-minute mini built from this case would feel like what you read above.
Nobody's studied whether this format improves retention. From what I can tell, the evidence above supports spaced, effortful, case-shaped retrieval, and if you care about how to retain medical knowledge, this is one convenient way to do it. Every case is written and signed off by a doctor. If you're comparing tools, see our clinical reasoning app roundup, and for a feel of the daily cadence, see a Duolingo for medicine.
Try a two minute quick case There is a free tier, no card needed. A quick case is a patient line and five clues, you guess as you go. Start a quick case
How to build the habit as an intern or resident
If you’re wondering how to retain medical knowledge on a schedule that never sits still, anchor recall to a moment you already have. The commute, the lull before handoff, waiting for a scan, standing outside a room while transport arrives.
Keep it tiny so it survives a bad week. One prompt, one diagnosis, one what-else-could-this-be thought, then back to the work in front of you.
Use your own patients for prompts without any identifiers. After you close the chart, ask yourself, what else could this have been, and what would change my mind. Name two findings that would push you off your leading diagnosis.
Teach a student or the off-service intern for two minutes. Explaining a concept out loud is retrieval. A single clean example beats a whiteboard of theory.
Lower friction. Keep a phone note with a short list of prompts you recycle on autopilot. Same questions, different cases. You don’t need novelty, you need a groove.
Spacing doesn’t require a perfect streak. Missed days aren’t a reset of what you know. Pick it up at the next gap, then keep going.
If the wheels are coming off this month, triage the life stuff first, then build back a tiny anchor. Some practical ideas live here, in getting through intern year. A short daily case on your phone is one way to keep the habit alive.
What to stop doing
- Stop rereading and highlighting, because the 2013 review by Dunlosky and colleagues rated highlighting, rereading, and summarization low utility and suggested practice testing instead. Across student samples, these methods did not consistently improve performance.
- Rewatching lecture videos as review is rereading with audio, and in my experience it feels busy but adds little retrieval. Short recall prompts beat passive playbacks.
- Stop collecting resources you never open, since hoarding apps, decks, and PDFs without a plan is avoidance disguised as prep. Pick a small set you’ll actually use for retrieval, and if you want ideas see AI study tools worth trying.
- Stop saving all practice for one weekend block, because the same 2013 review by Dunlosky and colleagues rated distributed practice high utility, so cramming into a single block is the wrong shape. Spread recall across short bouts, and if you like quick prompts see diagnosis games that teach.
Run a full AI patient case Talk or type to an AI patient, order investigations, commit to a diagnosis, and get a scorecard. Free to start, no card; iPhone, Android, and web. Try a full case
Educational use only, not medical advice. AI-generated; verify clinically against primary sources.
Clinical review pending.
