The publishable case report:
the filter
Interns don’t fail at case reports because they can’t write. They fail at the filter: the “rare” case that isn’t rare to the literature, the error they want to publish, the mystery with no ending. Every program watches the same three failures on repeat — and ninety minutes on the filter saves each intern a wasted month, and saves their first scholarly experience from ending in a desk rejection they take personally.
Why this block
The case report is the front door of scholarship — the smallest complete unit of the entire publication process, which is exactly why the bootcamp’s scholarship session put it on the ladder’s first rung. This block is the bouncer at that door. One facilitator requirement is non-negotiable: bring one of your own published case reports, in print — at the end of the session you will hold it up and walk it through the test, including the unglamorous parts: the consent conversation, the rejection, the revision. The room needs to see that a real published case is ordinary in process and rigorous in filter, not the reverse.
What interns leave able to do
- Run the week-zero screen before commitment hardens: attending told, a publication-experienced mentor identified, institutional policy checked through the required route.
- Run the two-stage literature reality check — the ten-minute triage, then the librarian-backed confirmation with a saved search log — and interpret what it returns.
- Apply the publishability test — novelty, completeness, a usable teaching point — and the six sharper questions reviewers actually ask.
- Classify a candidate case into the publishable categories — and recognize the three failure modes on sight.
- Explain why error-involved cases belong in protected quality channels, never in an intern’s manuscript.
- Explain why an incomplete workup or an unknown outcome kills a case report — and what would have rescued it.
The four cases
Disclosed one part at a time; the room commits before each reveal.
Thursday, 7:15 a.m. sign-out. Overnight you admitted a 58-year-old man with nausea, fatigue, and a glucose of 210 — mild, unimpressive — except his anion gap is 18 and his beta-hydroxybutyrate is sky-high. Euglycemic ketoacidosis, on an SGLT2 inhibitor. The workup was textbook, he responded to treatment, and your attending called it “a great catch.”
You are buzzing. This is the case. You have already mentally written the title. At sign-out you announce it: “we should publish this!”
Your senior doesn’t look up from the list. One question: “What did your literature search show?”
What is she asking for, exactly — and what do you expect it to show?
Two weeks later a co-intern corners you, shaken. Last month on their team, a patient was harmed by a significant medication error — a real one, with a real consequence, now in the event-review process. Your co-intern says: “I keep thinking about it. There’s so much learning in this case. Honestly, it would make an amazing case report — people need to hear what happened so it doesn’t happen somewhere else. Should we write it up?”
What do you tell them? What are the problems — and is there any right venue for the learning?
Another co-intern spent three weeks on a diagnostic odyssey: fevers, joint pain, a nondescript rash, negative autoimmune and infectious workups, a non-diagnostic biopsy. The patient stabilized on empiric steroids, declined further workup, was discharged with outpatient follow-up — and never came back. No diagnosis. No outcome.
“Journals love mystery cases,” your co-intern says. “Let’s publish it.”
Publishable? Why or why not — and what would have made it publishable?
Your patient: a 34-year-old woman, six months of fatigue and achy joints, attributed by two urgent-care visits to stress. On your service for an unrelated admission, you notice her skin looks faintly bronze despite winter, and ask about family history — her father has “some iron thing.” Ferritin 1,450; transferrin saturation 78%; genetic testing confirms hereditary hemochromatosis — the disease the textbooks file under middle-aged men — in a 34-year-old woman. She starts phlebotomy; at three-month follow-up her fatigue is gone and her ferritin is falling.
Your senior: “This one, you could publish.”
Why does this case pass when the others didn’t? Build the rule.
Part 1 — the literature reality check, in two stages
Rare ≠ reportable. The only question that matters: does this case teach the reader something the literature does not already cover well? Euglycemic ketoacidosis on an SGLT2 inhibitor was a novel report around 2015; a decade later there are hundreds of published cases, series, reviews, and guideline coverage. The case is now confirmatory, not novel — a journal’s reviewers will know it in thirty seconds, your senior knows it from experience, and you can know it in ten minutes, if you look.
Stage 1 — the ten-minute triage: PubMed, on your phone, right now — the condition, the twist, and “case report” as terms; scan the recent reviews and series. This is triage, not the verdict: the count alone decides nothing. Dozens of prior reports do not automatically kill a case that adds something genuinely new — a diagnostic pitfall, a population, a mechanism, a longitudinal outcome, a treatment response — and a single definitive prior report can end novelty on its own. What the count does is set the bar for your one-sentence “what this adds” claim, which is the actual verdict. Read the two or three most recent papers’ discussions — they state what is known and what remains unclear, and your sentence must speak to the unclear part.
Stage 2 — the confirmation, with help: before real work begins, the search gets done properly — synonyms and controlled vocabulary, citation chaining from the papers you found, the drug label or guideline when the twist is an adverse effect — ideally with a medical librarian, who does this professionally, likes trainees, and whose ten minutes beat a month of yours. And keep a search log: database, date, exact query, filters, result count. It takes two minutes, it becomes the manuscript’s search statement, and it is the difference between “I looked” and evidence that you looked.
The week-zero screen, run in parallel with the search:
- Tell the supervising attending, and identify a publication-experienced mentor — part of the filter, not a later rescue.
- Confirm the case is complete and follow-up obtainable through appropriate clinical channels.
- Check your institution’s case-report, consent, and image policies through the required route — never self-determined.
- Write the one-sentence “what this adds” claim and a likely venue.
- Then decide: proceed, redirect, or stop.
The reframe that saves the enthusiasm: a non-publishable case is still publishable-adjacent — a morning-report teaching case, an institutional research-day poster where local norms welcome interesting-but-not-novel cases, or the spark for a real question: “why do some patients on these drugs get euglycemic ketoacidosis and others don’t?” is a research question, and block 3 is where it goes. And the timing discipline, boarded: the search happens before the announcement — before the writing, before the excitement hardens into commitment. Excitement is fuel; the search is steering.
Part 2 — the error case
The instinct is honorable — the learning is real and the motive is good. The venue is wrong, and the problems escalate:
- Consent. Many journals require written publication consent — requirements vary by journal, institution, and content — and a patient harmed by an error, possibly considering litigation, possibly traumatized, is the least likely person on earth to give it. Whether, when, and by whom that patient could even be approached is the institution’s call, never an intern’s alone.
- Permanence. A published case is public, permanent, and findable — and how it might interact with any later proceedings depends on jurisdiction and facts no intern can assess. Risk management and institutional counsel exist precisely to navigate this territory; an intern freelancing around them is compounding the original event.
- De-identification fails in unique cases. The more interesting the case, the more identifiable it is — to the family, the staff, and the plaintiff’s attorney. “Anonymous” in a textbook is fully recognizable in a hometown.
- It short-circuits the protected process. The event is already in review — the structured, shielded channel where analysis can be frank because it is protected. Publishing mid-process can compromise it.
The right venues, named: the event-reporting system, the morbidity-and-mortality conference, the root-cause analysis, and QI work — the bootcamp’s error session is the full treatment. And the exception, stated honestly: a patient-safety literature exists — carefully de-identified, institutionally approved analyses with the process improvement included — but that is an institution-approved, systems-focused project run through the safety infrastructure — typically after resolution, with experienced leadership, appropriate consent, and explicit ethical review. The resident rule: never on your own initiative — and not because of rank; the institutional process, not any individual, decides whether and how learning from an event is disseminated. If the learning deserves a wider audience, bring it to your mentor, program director, or QI leadership, and let that process decide. One more line, because it matters: the co-intern is shaken. Check on the person before the project idea — the second-victim lesson travels.
Part 3 — the mystery case
A case report is a teaching instrument: it lets a reader borrow your diagnostic reasoning and learn how the story ends. The mystery case offers neither — no diagnosis, so the differential was never tested against an answer; no outcome, so the empiric treatment’s wisdom is unknowable. It is a riddle with no solution page, and readers and reviewers bounce off it identically.
The “journals love mysteries” confusion, untangled: some venues do run clinical-mystery formats, staged reveals and expert discussion — and every one of them ends with the answer. The genre is mystery solved with flair, not mystery abandoned. What would have rescued this case: a definitive diagnosis — especially a rare one — reached by a documented, thorough workup, with follow-up and outcome. Even a negative ending can publish if the evaluation was exhaustive and the outcome documented: “extensive evaluation unrevealing; at twelve months she remains well” is an outcome. “Never came back” is not — and the discipline of getting the follow-up, the phone call that turns lost into documented, is itself a research skill.
The principle, generalized and boarded: the workup is the scholarship. With a thin differential and no follow-up, a case is a story; with a rigorous differential, a complete workup, and a known outcome, it is evidence-shaped teaching.
Part 4 — the publishability test, and the categories
Let the group build why the hemochromatosis case passes, then name the three parts: novelty — it teaches what the literature covers poorly: a common disease presenting in a demographic where it is systematically missed, with a live teaching point about anchoring; completeness — the noticed skin, the family history asked, the confirmatory testing, the treatment, and the three-month outcome; and a usable teaching point — a hospitalist in another state could act on it Monday. State the lesson at the size one case can support: not a screening rule, which no single case can establish, but a pattern — unexplained fatigue and arthralgia with a clue attached, the bronze tint, the father’s “iron thing,” should put iron studies on the differential, in women too.
Sharpened into the six questions reviewers actually ask:
- What precisely is new or newly useful?
- Is the diagnosis sufficiently supported, with reasonable alternatives addressed?
- Is the timeline complete, through treatment and outcome?
- Could the patient or their community recognize the case despite removed identifiers?
- Does the lesson transfer without overstating one observation?
- Is there a right format and venue for it?
Any “no” is a to-do or a stop, not a detail. And then run the block’s own discipline on its own example — make the room attack the hemochromatosis case: would it actually survive stage 2? What would reviewers demand? A model case that gets a pass for being memorable teaches the wrong filter.
| Publishable category | The shape |
|---|---|
| Unusual presentation of a common disease | Common disease, atypical face — the demographic, symptom pattern, or imaging that breaks the taught pattern. The hemochromatosis case lives here. |
| Common presentation of an unusual disease | An ordinary-looking syndrome with a rare cause — the zebra hiding in a herd of horses, and how it was found. |
| Treatment variation or novel approach | A therapy used in a new context with a mechanistic rationale and a documented outcome — including instructive failures. |
| Rare-but-known adverse effect with teaching | A known toxicity rarely seen, where the teaching is the mechanism, the management, or what should have triggered earlier recognition. |
And the boundary case: the genuinely novel report — a first-or-nearly-first association with a plausible mechanism. It is real and it is rare, and the Part 1 search is what proves you have one; if your case survives the search, the bar becomes mechanism and rigor rather than excitement. The whole filter compresses to one sentence, worth pocketing: would a hospitalist in another state change something they do after reading this case? If yes, write it. If “maybe, if they hadn’t read the two hundred prior reports” — block 3 awaits.
AI and case material — the boundaries
The scenario to run in the room, because it happens weekly now: an intern wants to paste the discharge summary — “I’ll delete the name” — into a public chatbot to “check if it’s publishable.” The rules, in order of severity: never enter PHI, dates, images, genetic details, or a rare identifiable narrative into an unapproved tool — removing the name does not de-identify a story this unusual, and an unapproved system is a disclosure, not a drafting aid. AI does not get the novelty verdict — it misses literature and fabricates citations, so the two-stage search stands, and every citation any tool suggests gets opened and verified against the source. What AI may legitimately do, within your institution’s policy and approved tools: help generate candidate search terms, explain an unfamiliar method, or reorganize already de-identified text you wrote. Generated or altered clinical images are never presented as patient evidence. Disclose any material use the way the target journal requires, keep a simple tool/version/date/purpose note — and the accountability for accuracy, originality, privacy, and consent stays with the humans on the byline. The full curriculum-wide rules live on the hub.
Pocket card
- Week zero: attending told, mentor named, policy checked through the required route — never self-determined.
- Search in two stages: phone triage, then librarian-backed confirmation with a saved log. The verdict is your “what this adds” sentence — not the count.
- Publishable shapes: unusual face of common · common face of unusual · treatment variation · rare adverse effect with teaching.
- Complete or don’t bother: history, differential, workup, diagnosis, outcome.
- A mystery with no answer and no follow-up is a story, not a case report.
- Error cases → event review, M&M, QI. Never an intern’s solo manuscript.
- The test: would a hospitalist elsewhere change practice after reading it? Six questions, no free passes.
- Nothing identifiable into unapproved AI — and AI never gets the novelty verdict. Citations opened, verified.
- Leave with: the triage sheet — claim sentence, saved search, flags, mentor contact, disposition, next action dated within a week.
Notes
Every case in this block is a fictional composite. The next block is the pipeline.
This page is a teaching scenario for facilitated small-group education. Every patient in it is fictional. Consent requirements, review-board policies, and publication norms vary by institution and journal — verify locally before starting any project. Last reviewed July 2026.