Elsie Item-Writing Academy
About 45 minutes · Academy module: Flaw Clinic: Shelf / Clerkship — Ward-Level Item Discipline
Unofficial faculty-development resource. Not affiliated with, endorsed by, or representing the AAMC, NBME, USMLE, LCME, or NRMP. Supports faculty-development documentation.
Shelf and clerkship exams sit between the classroom and the wards: items should reward the reasoning a good clerk actually uses — pattern recognition, risk stratification, and knowing the next step. The flaw taxonomy applies in full, with clerkship-specific accents.
Grammatical cueing. Clerkship authors write fast, often adapting teaching cases, and agreement errors slip in: a stem ending in "…is an" with only one vowel-led option, or a plural verb that fits only the key. The habit to build is mechanical — read every option into the stem aloud before the item leaves your desk. It takes twenty seconds and catches the most embarrassing class of flaw.
Absolute terms. "Always" and "never" are especially corrosive in clerkship items because clinical teaching itself traffics in rules of thumb. An absolute in an option invites the experienced student to supply the exception they saw on the wards — and they may be right. Prefer "most likely," "most appropriate," and "best next step," which match how clinical decisions are actually framed.
"All of the above" / "none of the above." On a shelf exam, these options interact badly with partial knowledge: a student who knows two options are correct can select "all of the above" without evaluating the third, and a "none of the above" key makes the item unreviewable — you cannot tell from the statistics what the student believed. Both are retired in this track.
Two defensible answers. Clerkship content is full of reasonable alternatives (hold vs. reverse anticoagulation, repeat imaging vs. biopsy), and the vignette must contain the discriminating detail — stability, timing, a threshold value — that makes one option the best next step. If your content expert hesitates between two options, the item is not ready; the stem is.
Unfocused stems and window dressing. A clerkship vignette should read like a sign-out, not a novel: age, key history, the findings that matter, the question. Decorative detail adds reading load without discrimination and risks accidental cueing. End every stem with one explicit task.
Negative stems without emphasis, clueing, implausible distractors. The same discipline as other levels: emphasize any unavoidable negation (EXCEPT, NOT), keep the stem's vocabulary from echoing in the key alone, and ensure every distractor is a step a real clinician might consider — never a straw man. Build the review habit into your authoring workflow: a short, repeatable checklist run on every item before it enters the bank catches most of these flaws while they are still cheap to fix. All three drills below carry model rewrites verified by two independent cold solves.
---
Flawed item (intended key: B)
A 9-year-old boy with sickle cell disease develops fever of 39.2°C and appears ill. Blood cultures are drawn before antibiotics are started. The most likely cause of bacteremia in this patient is an:
A) Escherichia coli B) Streptococcus pneumoniae C) Neisseria meningitidis D) Staphylococcus aureus E) Pseudomonas aeruginosa
Flaw taxonomy label: Grammatical cueing.
Correction: The stem ends with "is an" — and only one option begins with a vowel sound. A student who knows nothing about functional asplenia can eliminate four options on grammar alone. This is the flaw that most embarrasses authors, because it is purely mechanical: the fix is to end the stem with a noun ("Which organism is the most likely cause of bacteremia in this patient?") so that every option attaches grammatically, and to build the twenty-second habit of reading each option back into the stem before the item ships.
Model rewrite — cold-solve verified: 2/2 (solver 1: B, solver 2: B)
A 9-year-old boy with sickle cell disease develops fever of 39.2°C and appears ill. Blood cultures are drawn before antibiotics are started. Which organism is the most likely cause of bacteremia in this patient?
A) Escherichia coli B) Streptococcus pneumoniae C) Neisseria meningitidis D) Staphylococcus aureus E) Pseudomonas aeruginosa
Key: B
Flawed item (intended key: A)
A 25-year-old woman has a positive home pregnancy test and reports her last menstrual period was 8 weeks ago. Transvaginal ultrasound shows a gestational sac with a yolk sac but no fetal pole. She has no pain or bleeding. Which of the following statements is always true?
A) Repeat ultrasound in 7–10 days is always required before any intervention. B) Dilation and curettage is never indicated in this situation. C) Methotrexate is always contraindicated. D) Progesterone supplementation never changes the outcome.
Flaw taxonomy label: Absolute terms ("always"/"never").
Correction: The stem asks what is "always true," and every option answers in absolutes — the item is grading certainty rather than clinical judgment. A student who has seen one exception to any option can argue the item rather than answer it. Clinical decisions are framed in likelihood and appropriateness, not universals. The rewrite drops the absolutes, asks the single next-step question the vignette earns, and replaces the universal claims with concrete management options — including one (vaginal progesterone) that is simply not indicated, rather than "never" anything.
Model rewrite — cold-solve verified: 2/2 (solver 1: A, solver 2: A)
A 25-year-old woman has a positive home pregnancy test and reports her last menstrual period was 8 weeks ago. Transvaginal ultrasound shows a gestational sac with a yolk sac but no fetal pole. She has no pain or bleeding. What is the most appropriate next step?
A) Repeat ultrasound in 7 to 10 days B) Proceed to dilation and curettage C) Administer methotrexate D) Begin vaginal progesterone supplementation E) Reassure and schedule routine prenatal care
Key: A
Flawed item (intended key: B)
A 72-year-old woman with atrial fibrillation treated with warfarin presents with melena and a hemoglobin of 7.8 g/dL. She is hemodynamically stable with no active large-volume bleeding. Which of the following is the most appropriate initial management?
A) Continue warfarin and transfuse packed red cells B) Hold warfarin, transfuse, and admit for evaluation C) Hold warfarin, give vitamin K, transfuse, and admit D) Proceed immediately to colonoscopy E) Start intravenous proton pump inhibitor and continue warfarin
Flaw taxonomy label: Two defensible answers.
Correction: Options B and C are both defensible for a stable warfarin-associated GI bleed — holding the anticoagulant with transfusion and admission is standard, and adding vitamin K is guideline-supported rather than wrong. The item as written has no key; it has a coin flip that will survive into scoring and generate an unresolvable item-analysis dispute. The repair is to make the options genuinely distinct in appropriateness: the rewrite keeps "hold, transfuse, admit" as the clear best step and replaces the near-duplicate with an option that is clearly wrong (PCC reversal plus discharge — reversal without admission for a stable bleed, then sending her home), so the vignette's discriminating detail (hemodynamic stability) does its job.
Model rewrite — cold-solve verified: 2/2 (solver 1: B, solver 2: B)
A 72-year-old woman with atrial fibrillation treated with warfarin presents with melena and a hemoglobin of 7.8 g/dL. She is hemodynamically stable with no active large-volume bleeding. Which of the following is the most appropriate initial management?
A) Continue warfarin and transfuse packed red cells B) Hold warfarin, transfuse, and admit for evaluation C) Proceed immediately to colonoscopy D) Reverse warfarin with prothrombin complex concentrate and discharge home E) Start intravenous proton pump inhibitor and continue warfarin
Key: B
---
Recording it adds the module to your Academy completion record on this device — module, track, level, and date, ready to download from your account page for faculty-development documentation.