How to Use Anki During Clerkships and Shelf Exams

Disclosure: I build AnkiQuiz, the card-based quiz tool discussed below. AnkiQuiz is not affiliated with Anki, AnKing, AnkiHub, UWorld, AMBOSS, the NBME, or any medical school. It does not replace a validated question bank, and AI-generated questions can contain errors.

The Anki routine that worked during preclinical courses often breaks during clerkships. Your schedule is fragmented, the tested specialty changes every few weeks, and a review queue built for Step 1 can compete with patients, notes, question-bank blocks, and sleep.

The solution is not to maintain every card at any cost. During clinical rotations, Anki works best as a rotation-specific retention layer. Let the current clerkship and shelf exam define the active scope, let legitimate question-bank work reveal gaps, and use cards only when spaced retrieval rehearses the operation that failed.

A short quiz made from those focused cards can then check whether you can retrieve the lesson without the familiar cloze or card layout. It is an intermediate check, not a shelf-exam simulator or Step 2 score predictor.

If you need the broader learning loop first, start with the daily Anki workflow for medical school , then use this guide to narrow it for clinical rotations.

The short answer

Why the preclinical Anki routine breaks on rotations

Your available time becomes fragmented

A lecture week may offer predictable study blocks. A clinical day can include prerounding, rounds, notes, teaching, procedures, and an uncertain finish time. A plan that depends on one uninterrupted two-hour review session is fragile even when the card count looked reasonable in M2.

Use smaller, protected review blocks and a clear stopping rule. Normal Anki reviews can fit into short periods, but learning a management algorithm or reviewing a difficult question usually needs a more deliberate session.

The relevant scope changes with the service

During an internal-medicine rotation, cardiology, pulmonology, infectious disease, and inpatient management may deserve attention. The next clerkship may shift the priority to surgery, pediatrics, psychiatry, or obstetrics and gynecology.

If every prior rotation keeps introducing the same volume of new cards, the queue grows while its relevance to the next shelf exam falls. Clinical Anki therefore requires active scope decisions, not only scheduler settings.

Shelf questions ask for more than isolated facts

A familiar card may ask for one association. A shelf-style question can require you to identify the diagnosis, decide what information matters, and choose the next diagnostic or management step among plausible options.

Anki can preserve the facts and decision rules represented on your cards. It cannot replace clinical reasoning practice in validated questions. If you are unsure how to divide the work, the guide to Anki vs UWorld explains the difference between retention and exam-style application.

Your Step 1 queue does not disappear automatically

Passing Step 1 changes the value of some cards, but it does not change their due dates or decide which facts remain relevant to Step 2. Without a deliberate transition, old reviews continue arriving while a second clinical card stream is added on top.

Build one rotation-specific active set

1. Triage Step 1-only cards carefully

The current official AnKing rotations, shelf exams, and Step 2 lesson documents tags for Step 1-only material, Step 2-only material, and content shared by both exams. It recommends using #AK_Other::Only_Step_1 to help identify cards that may be suspended after Step 1 while preserving relevant Step 2 and overlap cards.

Treat those tags as a search aid, not a command to change thousands of cards unseen:

1. Confirm that the instructions match your installed AnKing version.
2. Preview the matched cards.
3. Preserve material still required by your school or a current weakness.
4. Make the change in a manageable batch.
5. Check the next few days of reviews before widening the scope.

Tag names and deck organization can change. Follow the current AnKing documentation rather than copying an old search string without inspecting the results.

2. Use the current rotation or shelf tags

The same AnKing lesson points clinical students to the Step 2 hierarchy and the !Shelf tags for each clerkship. Start with the rotation you are actually on. Preview the cards, learn the underlying topic, and then unsuspend the subset that supports your current curriculum and question-bank work.

Do not activate every card under a broad specialty merely because the tag exists. A premade deck contains more detail than any one student needs at the same moment. Each unsuspended card creates a future review obligation that will continue after the rotation ends.

3. Use a filtered deck for focus, not as a substitute for scope

Anki's filtered-deck documentation explains that a filtered deck temporarily gathers cards using searches or tags, and that cards retain a link to their home deck. It can be useful for reviewing one rotation, recent failures, or a weak topic before a shelf exam.

Understand the scheduling option before studying. With rescheduling enabled, answers can update the cards' normal schedules; with it disabled, the deck acts more like a preview. Also remember that a filtered deck cannot pull suspended cards. If the relevant cards are still suspended, filtering alone will not activate them.

4. Keep cross-rotation cards only when they earn their workload

Some mature cards remain useful across medicine, surgery, pediatrics, and Step 2. Others test low-priority details from a completed block. Keep a card active when the knowledge is reusable and its review cost remains acceptable. Suspend, edit, or retire it when it is duplicated, outdated, ambiguous, or no longer worth scheduling.

This is triage, not failure. The goal is to preserve knowledge that supports clinical decisions while leaving enough time to practise those decisions.

A clinical-day Anki workflow

Before the clinical day: protect a bounded due-review block

Complete a focused set of due, rotation-relevant reviews before the day becomes unpredictable. Answer before revealing the back and grade the answer you actually produced. The Anki manual says to use Again when the answer was incorrect or could not be recalled; Hard is for a correct answer that was slow or uncertain.

If the queue is larger than the block, do not hide that fact by clicking quickly or grading forgotten cards as Hard. Reduce new cards, narrow the active scope, and plan a separate backlog block.

During downtime: use Anki for short retrieval, not patient documentation

Normal Anki reviews can fit into legitimate downtime when your clinical responsibilities and local policies allow it. Keep the session easy to stop. Do not enter details from a patient encounter while standing on the ward, and never let card work distract from care or supervision.

AnkiQuiz itself runs through the Anki desktop add-on, so save a changed-prompt quiz for a planned desktop session. It is not designed to become another task you must complete between patients.

After the clinical day: let questions drive repair

Use your course material and a legitimate question bank to practise the decisions required by the rotation. Review important incorrects, correct guesses, and unusually slow questions. Then decide whether the failure needs a card at all.

The useful sequence is:

question or clinical topic → diagnose the gap → relearn if needed → smallest effective card change → later transfer check

The detailed UWorld incorrects Anki workflow shows how to make that decision without copying protected question-bank content or turning every explanation into several new cards.

Once a week: audit the complete workload

Look beyond today's due number:

• How much time did reviews actually take?
• Did you complete the question-bank work you planned?
• Are new cards producing a rising queue?
• Are old rotations still dominating the session?
• Are you sleeping enough to learn and function safely?
• Which repeated misses are genuine retrieval gaps?

If Anki repeatedly crowds out questions, concept review, or sleep, pause new cards before the backlog becomes the problem. A smaller active deck that you can maintain is more useful than a comprehensive deck you avoid opening.

Turn each clinical miss into the right repair

Knowledge retrieval

You understood the task but could not recall a reusable fact, such as a diagnostic criterion, adverse effect, screening interval, or first-line treatment within the scope of your source.

Repair: search the existing deck, then unsuspend or create one focused card if spaced retrieval directly addresses the gap.

Mechanism or causal understanding

You remembered the endpoints but could not explain why the finding occurred or why the treatment worked.

Repair: return to a trusted source and reconstruct the mechanism. Add a concise “why” prompt only after the relationship makes sense.

Management sequence or threshold

You knew the diagnosis but not what should happen next, or you confused two branches of an algorithm.

Repair: identify the decision point. A useful prompt states enough context to choose one action and makes the condition or threshold explicit. Date or source guideline-dependent notes so they can be reviewed when recommendations change.

Discrimination

Two diagnoses, tests, or treatments both seemed plausible and you could not identify the decisive difference.

Repair: write one contrast prompt that asks which finding separates the alternatives and why. Do not paste the original vignette.

Interpretation, reasoning, or execution

You overlooked the time course, misread the final task, overvalued a nonspecific clue, or ran out of time.

Repair: practise the failed process with new, legitimate questions. Another fact card may add work without changing the behavior that caused the miss.

When normal reviews feel easy but changed questions do not, use the recognition, recall, and reasoning diagnostic to locate the exact layer that is failing.

Write clinical cards that are possible to grade

A card should have one clear retrieval target. Useful clinical relationships can include:

• finding or pattern → likely diagnosis;
• diagnosis plus relevant context → next step;
• intervention → indication or important contraindication;
• mechanism → expected clinical consequence;
• two plausible options → decisive differentiating feature;
• management branch → condition that activates it.

Include enough context to make the answer unambiguous, but do not turn every card into a full vignette. Dense cards are slow to review and difficult to grade. Keep sources and supporting explanation in non-tested fields when your note type permits it.

Do not copy protected question-bank content

Do not paste UWorld or other protected stems, explanations, tables, screenshots, or images into Anki, AnkiQuiz, or another third-party tool unless the license explicitly permits it. UWorld's current terms and conditions are the authority for its content.

Write an independent learning target from your understanding and permitted sources. A question ID may help locate an existing tagged card where the current deck workflow permits it; it does not make the underlying question text yours to reproduce.

Keep patient information out of cards and AI tools

A memorable patient can make a concept easier to retrieve, but an identifiable patient must not become study content. Do not enter names, dates of birth, medical-record numbers, exact dates, contact details, room numbers, photographs, or a rare combination of details that could identify someone.

US students should review the HHS guidance on de-identifying protected health information; changing a name alone is not a complete de-identification method. Follow your institution's policy and the law that applies where you train.

AnkiQuiz sends the card fields you select to its server and an AI provider to generate questions. Never select fields containing protected health information or other sensitive data. Prefer generic teaching points and properly licensed educational cases.

Where AnkiQuiz fits

AnkiQuiz can add a changed-prompt retrieval check between normal reviews and validated shelf or Step 2 questions. It reads selected text fields from an Anki deck and generates single-choice, multiple-choice, fill-in-the-blank, or true/false questions with scores and explanations.

Use it as a focused weekly check:

1. Choose one narrow set, such as surgery postoperative complications or pediatrics respiratory illness.
2. Prefer cards you have already reviewed; immediately quizzing brand-new cards mostly measures short-term familiarity.
3. Include useful text fields and exclude tags, source IDs, links, and any sensitive material.
4. Provide more source cards than the number of questions requested.
5. Generate a short quiz instead of testing the entire rotation deck.
6. Use an instruction such as: “Focus on diagnostic discriminators, management decision points, and important contraindications. Stay within the source cards and do not invent missing patient details.”
7. Verify each answer against the source card and a current medical reference.
8. Repair the source card or relearn the topic when a changed prompt exposes a gap.

The card-to-quiz setup guide explains how to install the add-on, select fields, and generate a first quiz.

AnkiQuiz does not mark the source cards reviewed, change due dates, or update FSRS history. Its questions can be wrong or ambiguous. Ordinary flashcards also lack enough context to reliably produce validated shelf-style or NBME-style vignettes. Use the score to inspect retrieval from that card set, not to estimate exam readiness.

Adjust the workflow across a rotation

At the start

Review the clerkship's objectives, your school's shelf requirements, and the current deck documentation. Choose the relevant tag or subdeck, preview its size and content, and set a new-card intake that fits the actual clinical schedule.

Do not attempt to finish the whole deck in the first week. Learn the topic before activating its cards.

In the middle

Let patient care, teaching, and question-bank performance identify recurring topics. Continue due reviews, but use misses to narrow the next learning target. Run an occasional changed-prompt quiz on one weak area rather than generating a test from the whole specialty.

Before the shelf exam

Shift more time toward timed, legitimate questions and careful review. Reduce broad new-card intake. Use filtered decks or tags for documented weak areas, and keep normal reviews focused on knowledge you still expect to use.

Do not confuse clearing every due card with being ready for the shelf. Readiness requires performance in the kinds of clinical decisions the exam asks you to make.

After the shelf

Before the next rotation begins, decide what remains active:

• preserve high-value cards that cross specialties or support Step 2;
• suspend details with low future relevance;
• repair or retire persistent leeches;
• remove temporary filtered decks when they are no longer needed;
• set the next rotation's scope before adding new cards.

This transition prevents each clerkship from leaving its full review load behind.

Transition from shelf exams to Step 2 dedicated

During clerkships, the current specialty defines most of the card scope. During Step 2 dedicated, mixed question blocks and assessment results should increasingly define priorities.

Keep mature clinical cards that remain useful and affordable. Unsuspend cards for repeated, high-value knowledge gaps. Reduce or suspend low-priority material before it displaces timed blocks, explanation review, and official assessments.

Anki remains the retention layer. A short AnkiQuiz session can check one weak card set with new wording. Neither should become a substitute for mixed clinical questions or a validated readiness measure.

What to do after a long shift or missed day

Do not compensate for one missed day by adding the planned new cards on top of an already larger review queue.

1. Pause new cards temporarily.
2. Protect newly due, rotation-relevant reviews.
3. Add a bounded catch-up block for older cards.
4. Preserve essential question-bank and learning time.
5. Resume fewer new cards only when the queue is stable.

For a larger queue, follow the complete medical-school Anki backlog recovery plan . It explains when a simple regular review is enough and when filtered decks or retrievability-based ordering may help.

Seven clerkship Anki mistakes to avoid

Keeping every Step 1 card active by default

Some knowledge carries forward, but maintaining every low-priority detail can consume the time needed for clinical learning.

Unsuspending an entire shelf tag without previewing it

A tag is a candidate set. Inspect it and activate material as you learn it.

Making several cards from every question-bank explanation

Capture the gap that changed the answer. Interesting details are not automatically worth scheduling.

Using Anki to solve a reasoning or timing problem

Cards preserve selected knowledge. They do not directly fix stem interpretation, choice comparison, pacing, or test-day decisions.

Grading forgotten cards as Hard

This makes the queue look cleaner while giving the scheduler inaccurate feedback. Use Again for failed recall, then fix workload at the scope and new-card level.

Putting patient or protected question-bank content into an AI workflow

Only process card text you are permitted to use, and exclude all sensitive or identifying information.

Treating a generated quiz score as a shelf prediction

A quiz made from your own cards measures a narrow sample of those cards. It is not standardized, validated, or representative enough to estimate a shelf or Step 2 score.

Questions medical students commonly ask

Should I keep doing Anki during clinical rotations?

Yes, when it remains a sustainable way to retain high-value knowledge. Narrow the active scope to the current rotation and reusable Step 2 material. Reduce card work when it displaces patient care, sleep, learning, or question-bank practice.

Which AnKing cards should I use for shelf exams?

Use the current AnKing documentation for your installed deck version. Its clinical workflow points to the Step 2 hierarchy and !Shelf tags by rotation. Preview cards and activate only the material you have learned and intend to maintain.

How many new Anki cards should I do per day during clerkships?

There is no universal number. Start from your available review time, current due load, question-bank plan, and clinical schedule. Lower or pause new cards as soon as future reviews stop fitting. Another student's card count does not include your rotation hours or existing queue.

Should I do Anki or UWorld first after a shift?

Protect the task most likely to be displaced. If Anki regularly consumes the evening, schedule the question block first and bound the review session. If a small due queue is predictable, clearing it first may work. The important rule is that one tool must not silently eliminate the other.

Should I make a card for every UWorld incorrect?

No. Use a card for an important, reusable knowledge or decision-rule gap. Relearn mechanisms and practise interpretation, reasoning, and timing errors in appropriate questions.

Does AnkiQuiz work with AnKing Step 2 cards?

Yes. Select a focused deck or subdeck in Anki and include the text fields that contain useful study content. AnkiQuiz does not alter the original AnKing cards or their schedules.

Can AnkiQuiz generate shelf-style questions?

It can reword facts and relationships present in selected cards, but ordinary cards often lack the clinical context needed for a valid shelf-style vignette. Generated questions are not standardized or validated. Use them to test retrieval from your cards, not to replace UWorld, AMBOSS, NBME materials, or your school's assessments.

Can I make cards from patients I see?

Use generic learning points without identifying details and follow your institution's rules. Do not put protected health information into Anki or submit it to AnkiQuiz. When in doubt, use a public or institution-approved educational source instead of the patient encounter.

Build a rotation-sized learning loop

The sustainable clerkship workflow is smaller than the entire clinical deck. Learn what the current rotation requires, activate a focused card set, review it honestly, and let question practice identify the next repair. Use changed prompts occasionally to expose cue-dependent recall, then return to validated questions for clinical application.

Start with the rotation you are on now. Inspect its current AnKing shelf tag, choose one question-bank block, and count how many misses truly require a card. That rotation-sized loop is more useful than carrying every previous course into tomorrow's rounds.