Dermatology Question Quality Review

Executive Summary

The 200-question candidate sample for Dermatology was reviewed across eight shards of 25 questions each. The findings are consistent and mutually reinforcing across shards: the dominant quality problem is a large volume of sub-threshold recall items that do not meet the clinical reasoning standard set by the benchmark and recent PYQ sets. Approximately 55–60% of the sampled questions are Bloom's Level 1–2 items with no clinical vignette, no meaningful distractors, and no discriminatory value at PG level. The remaining 40–45% contains a meaningful subset of well-constructed clinical questions, but these are diluted by structural defects, factual errors, topic mislabeling, and image-dependency failures that cut across the entire sample.

The Bloom's distribution confirms the problem quantitatively: 69 of 200 candidate questions are Bloom's 1, and 78 are Bloom's 2, meaning 73.5% of the sample sits at the two lowest cognitive levels. The benchmark standard — even its easier items — consistently embeds factual recall inside a clinical scenario. The candidate pool does not do this reliably.

Six distinct issue categories emerge from the evidence. They are ordered by operational urgency:

Bare-recall items with no clinical scaffolding — the largest single category, affecting roughly half the sample
Image-dependent questions with absent or unverifiable images — a structural integrity failure that makes affected questions undeployable regardless of content quality
Factual errors and answer-key defects — a smaller but high-risk cluster requiring expert correction before any use
Topic mislabeling and classification drift — a systemic tagging problem that distorts analytics and misroutes questions to wrong topic buckets
Structural format defects — compound options, "all of the above" keys, and incoherent negative stems that violate basic MCQ construction standards
Near-duplicate and redundant coverage — a deduplication problem concentrated in high-frequency topics (herpes zoster, molluscum contagiosum, scabies, vitiligo, slapped-cheek appearance)

Recommended disposition across the 200-question sample: approximately 35–40 questions are keep-worthy as-is or with minor edits; approximately 60–70 require substantive fixes before deployment; approximately 90–100 should be disabled, with priority given to bare-recall items where gold-standard coverage already exists in the PYQ and benchmark sets.

What Good Looks Like

The benchmark and PYQ sets establish a clear and consistent quality bar. The following features define it:

Clinical embedding of even simple facts. The benchmark question on Munro microabscesses (Q 5a7b778f) is Bloom's 1 in cognitive demand but is presented inside a full histopathology vignette with patient demographics, morphology, and biopsy findings. The candidate pool routinely strips this scaffolding and asks the same fact as a bare one-liner.

Multi-clue vignettes that require genuine differential reasoning. The benchmark question on scleroderma (Q 6e3ccb4a) provides skin thickening, Raynaud's phenomenon, telangiectasias, and restricted mouth opening, then asks for the pathophysiological mechanism — requiring the candidate to synthesise multiple findings rather than pattern-match a single keyword. The benchmark question on drug-induced lichen planus (Q 0be6885d) adds a medication history and asks for the management step that integrates both the diagnosis and the causative drug, making it a genuine Bloom's 3 management question.

Distractors that are clinically plausible and educationally meaningful. In the PYQ on bullous pemphigoid (Q 24dfb4ea), all four options are blistering diseases with subepidermal or intraepidermal clefts; the discriminating feature is the DIF pattern. In the PYQ on alopecia types (Q 12941e27), all four options are real alopecia subtypes and the question requires knowing which ones scar. Distractors in the candidate pool frequently include disease names listed as organisms, implausible alternatives, or "all of the above" constructions that collapse the question to a single fact.

Management questions that are anchored to a specific clinical context. The PYQ on herpes simplex (Q dc520b9f) uses a bedside test result (Tzanck smear implied) and a recurrence history to justify acyclovir. The PYQ on topical corticosteroid side effects (Q 573aad46) tests a specific exclusion (blue pigmentation is not a steroid side effect) rather than asking for a list.

Appropriate Bloom's calibration. The benchmark questions that carry a Bloom's 1 label (Q 5a7b778f) are acknowledged as such and are justified by their PYQ provenance. The candidate pool contains many questions labelled Bloom's 2 or 3 that function at Bloom's 1 because the stem contains every discriminating feature without any competing complexity.

The best questions in the candidate sample — Q d5f28049 (mycetoma in a farmer), Q abd26668 (Mohs surgery with multi-constraint clinical scenario), Q 2dbac800 (indeterminate leprosy vignette), Q e4cd51e5 (tuberculosis verrucosa cutis with occupational history), Q 114d0eab (melanoma management decision) — share all of these features and represent the ceiling the rest of the pool should be held to.

Main Issue Categories

1. Bare-Recall Items With No Clinical Scaffolding

Why this pattern is bad

Indian PG examinations at INICET and NEET-PG level have moved decisively toward clinical vignette formats. A question that asks "Tinea versicolor is caused by:" or "Exclamation mark hairs are seen in:" tests rote memorisation, not the diagnostic reasoning that the exam is designed to assess. These items also have near-zero discriminatory power: any candidate who has opened a textbook will answer correctly, and any candidate who has not will guess randomly. They add noise to score distributions without adding signal about clinical competence. When the same fact is already tested in a richer PYQ vignette, the bare-recall version is strictly inferior and should not coexist in the same pool.

How it shows up

This is the broadest pattern in the sample, appearing in every shard. It clusters most densely in Fungal Skin Infections, Basic Dermatology, Pediatric Dermatology, Hair and Nail Disorders, and Skin Tumors. The typical form is a one-sentence stem with a single keyword that maps directly to the correct answer, four options of which three are implausible to any prepared candidate, and no patient demographics, clinical history, or examination findings. Many carry both easy and blooms-1 quality flags, confirming the pattern is already partially detected by automated checks but not yet acted upon at scale.

A secondary form is the "thin vignette" — a question that wraps a single fact in a sentence of clinical colour without adding any genuine diagnostic complexity. Q 99e20061 (pityriasis alba in an atopic child) and Q 6f6bd4b2 (classic psoriasis on extensor surfaces) carry Bloom's 2–3 labels but resolve at Bloom's 1 because the stem contains every discriminating feature without any competing differential.

Example question IDs

Q e29431ed — "Tinea versicolor is caused by:" → Malassezia furfur. The organism name is the only content being tested. No vignette, no clinical context, no discriminating distractors. Bloom's 1, no tags.
Q be9ae193 — "Acquired symmetric hyperpigmentation associated with pregnancy and OCP use is called:" → Melasma. The entire diagnosis is embedded in the stem; the question answers itself before the options are read.
Q 8b94e4b8 — "A flat discolouration on the skin measuring 1 cm is called:" → Macule. One distractor is "Plague" (likely a typo for "Plaque"), indicating poor editorial quality on top of trivial content.
Q 20d6accf — "Exclamation mark hairs are seen in:" → Alopecia areata. Pure eponym recall, Bloom's 1, no clinical scenario.
Q ae47252c — "Lupus vulgaris is: [TB of skin]." Single definitional sentence, distractors are SCC, BCC, and lupus erythematosus — implausible to any prepared candidate.
Q 87958877 — "Tinea unguium involves which structure?" The Latin root unguis directly gives away the answer. Zero discriminatory value.
Q ab960295 — "Inheritance of tuberous sclerosis?" Bloom's 1, no clinical context, no ash-leaf macule, no seizure history, no family history.
Q fdca6569 — "Most common early finding in Tuberous Sclerosis? Ash leaf spot." Bloom's 1, easy, no tags, no template membership.
Q c31097b0 — "What is the treatment of choice for melanoma? Surgical excision." No staging, no margin discussion, no sentinel node consideration. Clinically oversimplified to the point of being misleading.
Q 6666919d — "Eczema herpeticum is typically seen with which viral infection?" → HSV. Single association, Bloom's 1, no atopic child with sudden worsening, no punched-out erosions.
Q 343b2cbb — "Fordyce spots location." Pure recall, Bloom's 1, no clinical context.
Q 9defafa8 — "Common warts caused by HPV." Universally known, zero discrimination.
Q 6b6706a8 — "Normal epidermal turnover time = 4 weeks." Isolated numerical fact, no clinical application.
Q 0e4ca797 — "Mycosis fungoides cell type." Single-fact recall (T cells), Bloom's 1.
Q 888376a2 — Donovanosis single-fact recall. Bloom's 1, easy, no tags.
Q 67a01c39 — Blaschko's lines definitional recall. Bloom's 1, easy, no tags.
Q ed1aa959 — Ehlers-Danlos hyperextensibility one-liner. Bloom's 1, easy, no tags.

Recommended disposition

Disable all items in this category where the same fact is already covered by a richer PYQ or benchmark item. For facts not yet covered at higher Bloom's levels, disable the bare-recall version and flag the concept for vignette reconstruction. Do not attempt to "fix" these by adding a single sentence of clinical colour — the reconstruction required is substantial enough that a new question should be written from scratch. Estimated volume in the 200-question sample: 90–100 items meet the disable threshold on this criterion alone.

2. Image-Dependent Questions With Absent or Unverifiable Images

Why this pattern is bad

A question that says "The image shows presence of…" or "A child has a pruritic rash as shown below" is structurally non-functional when the image is absent. Unlike a bare-recall item, which is at least answerable (if trivially), an orphaned image-dependent question is either unanswerable or degrades to a guessing exercise based on whatever clinical hint survives in the text. This is a deployment integrity problem: these questions cannot be used in any test without the image, and if the image is present but incorrectly linked, the question actively misleads candidates. The problem is distinct from low Bloom's level — some of the affected questions are well-constructed Bloom's 3–4 items that would be strong keeps if the image were confirmed.

How it shows up

The pattern appears in every shard. The typical form is a stem that begins with "The image shows…", "As shown in the image below…", "The following skin condition is associated with:", or "Biopsy of lesion was performed and shown below." In text-only review, these questions are either completely unanswerable (Q 8b92bedf, Q c3b7d0a8, Q fd444ad6, Q 801db0c7) or partially answerable from clinical context alone but stripped of their intended Bloom's level (Q 5dde2db1, Q 6c261b91, Q e4cd51e5, Q fdf43fd4, Q f60b177c).

A secondary form is the question that references a "bedside test" or "Wood's lamp finding" without describing the result in text — the PYQ Q dc520b9f handles this correctly by providing enough clinical context (recurrent genital vesicles, oral ulcer) that the Tzanck smear result is implied; the candidate pool versions frequently omit this safety net.

Example question IDs

Q 8b92bedf — "The image shows presence of…" No clinical description whatsoever. Completely unanswerable as text.
Q c3b7d0a8 — "These lesions are seen in which of the following?" No clinical description. The stem is a demonstrative pronoun pointing at nothing.
Q fd444ad6 — "The following skin condition is associated with:" No disease named, no clinical description, no image. Unanswerable.
Q 801db0c7 — "What is true about this condition?" No image, no clinical context. The correct answer (hamartoma of naevo-melanocytes) suggests a melanocytic nevus question, but nothing in the stem confirms this.
Q 5dde2db1 — Wood's lamp examination shown. Clinical context (obese woman, axillary lesions) is present and adequate; question is partially salvageable if image is confirmed. Contingent keep.
Q f4ec5f6d — "Biopsy of lesion was performed and shown below." No clinical description of the lesion. Unanswerable without image.
Q 6c261b91 — "The immunofluorescence image is shown." Clinical context (blistering disease) is partially present but the DIF pattern — the entire point of the question — is in the image.
Q fdf43fd4 — Dermatomyositis image question. Stem says "presentation shown below" with only "sun exposure" as a clinical clue. Loses its Bloom's 3 character without the image.
Q c6a88ca9 — "A child has a pruritic rash as shown below." No description of the rash. Unanswerable.
Q f60b177c — Chancroid "except" question. PYQ-tagged, Bloom's 4, strong content — but references an image for the clinical finding. Contingent keep if image is confirmed.

Recommended disposition

Triage into two groups. Group A (no clinical description at all in the stem): disable immediately — Q 8b92bedf, Q c3b7d0a8, Q fd444ad6, Q 801db0c7, Q c6a88ca9, Q f4ec5f6d. These cannot be fixed without a complete rewrite. Group B (clinical context partially present, image adds discriminating detail): hold for image verification — Q 5dde2db1, Q 6c261b91, Q fdf43fd4, Q f60b177c. If the image is confirmed present and correctly linked, these are keep or fix candidates. If the image cannot be verified, disable. Do not deploy any image-dependent question until image integrity is confirmed.

3. Factual Errors and Answer-Key Defects

Why this pattern is bad

A question with an incorrect answer key or a factually wrong stem actively harms candidates by reinforcing incorrect knowledge. This is the highest-risk category from a content safety perspective. Even a single deployed question with an inverted key or a factually wrong correct answer can distort learning and, in a high-stakes exam context, generate candidate complaints and score challenges. The pattern is smaller in volume than the recall or image-dependency problems but requires the most urgent remediation because the harm is direct and immediate.

How it shows up

In this sample, factual errors and key defects appear in three sub-forms. First, inverted or miskeyed "except" questions where the stated correct answer is actually a true statement about the condition (Q 57ad13fa, Q 84c5e656). Second, organism/disease name conflation where a disease name is given as the answer to a question asking for a causative organism (Q 87ee6bfe, Q 8717c3cd). Third, outdated or guideline-discordant management answers where the keyed answer reflects older practice that has been superseded (Q 7a48fe93). A fourth sub-form — factually contested correct answers — appears in Q c376f990, Q e6b70ab4, Q dad30d9b, and Q 1006303f, where the keyed answer is defensible in some sources but contested in others, creating a reliability risk.

Example question IDs

Q 57ad13fa — "Which statement about candidiasis is false?" → "Involves nails" is keyed as the correct exception. Candidiasis does involve nails (onychomycosis is a well-recognised form). The answer key appears inverted. High-priority fix or disable.
Q 84c5e656 — "Which is NOT a clinical form of lichen planus?" → Verrucous lichen planus is keyed as the correct answer (i.e., not a real form). Verrucous/hypertrophic lichen planus is a well-recognised variant in all standard dermatology texts. Factual error; disable or rekey.
Q 87ee6bfe — "Causative organism for kerion" → "Tinea capitis" is keyed as correct. Tinea capitis is the disease, not the organism. The causative organisms are dermatophytes (Trichophyton spp., Microsporum spp.). Content-accuracy defect requiring correction.
Q 8717c3cd — "Causative agent of Favus" → Tinea schoenleinii is keyed. The correct organism name is Trichophyton schoenleinii; "Tinea schoenleinii" is not a valid organism name. Nomenclature error.
Q 7a48fe93 — "Mainstay of treatment for vitiligo" → PUVA is keyed. NB-UVB is now the preferred first-line phototherapy in current guidelines; PUVA is no longer considered the unambiguous mainstay. Risk of teaching outdated management.
Q db0915b7 — "Contraindications of potassium iodide EXCEPT" → "Pregnant women" is keyed as the exception (i.e., not contraindicated). Potassium iodide is contraindicated in pregnancy due to fetal thyroid suppression. Likely answer-key error; disable pending expert review.
Q e6b70ab4 — "Long contact with tuberculoid leprosy can transmit" is keyed as a characteristic finding. Tuberculoid leprosy is the least infectious form of leprosy; this statement is factually imprecise and potentially misleading.
Q 61e481b7 — Erythema annulare centrifugum described as having "central scarring." EAC does not typically scar; the correct descriptor is "central clearing." Teaching an inaccurate clinical feature.
Q dad30d9b — Necrolytic migratory erythema "except" with Hepatitis C as the correct exception. HCV association with NME is reported in some literature; the factual basis for this exclusion is debatable.
Q c376f990 — "Punched-out/inverted saucer" lesion attributed to lepromatous leprosy. This morphology is more classically described in borderline leprosy; the attribution is contested.

Recommended disposition

Disable immediately without further review: Q 57ad13fa, Q 84c5e656, Q db0915b7 — these have clear answer-key errors that would actively mislead candidates. Fix with expert content review before redeployment: Q 87ee6bfe, Q 8717c3cd, Q 7a48fe93, Q 61e481b7 — the underlying concept is worth testing but the factual content needs correction. Flag for expert adjudication: Q e6b70ab4, Q dad30d9b, Q c376f990, Q 1006303f — the correct answer is defensible in some sources but contested in others; these should not be deployed until a subject-matter expert confirms the keyed answer against a named current reference.

4. Topic Mislabeling and Classification Drift

Why this pattern is bad

Incorrect topic tags corrupt topic-level analytics, cause questions to appear in wrong topic-filtered test sets, and make it impossible to audit coverage gaps or over-coverage by topic. When a leprosy question appears under "Bacterial Skin Infections" with a factually imprecise stem, or a lichen planus question appears under "Psoriasis," the content team cannot trust topic-level quality reports. The problem is operationally distinct from content quality: a well-constructed question with a wrong topic tag needs only a reclassification, while a poorly constructed question with a correct tag needs a rewrite. Mixing these remediation paths wastes effort.

How it shows up

Topic mislabeling appears in every shard and clusters around three specific patterns. First, questions about lichen planus are repeatedly filed under "Psoriasis" or "Dermatitis and Eczema" (Q cc0cacc5, Q 49d0cf66, Q 4af42166). Second, questions about porphyria, prurigo nodularis, and other systemic/metabolic conditions are filed under "Phototherapy and Photobiology" (Q 8c40fe95, Q a90a2726). Third, questions about vascular, paraneoplastic, or systemic dermatology entities are filed under "Basic Dermatology" or "Autoimmune Skin Diseases" without a dedicated category (Q dad30d9b, Q bcd63a45). A fourth pattern is the "Fungal Skin Infections" bucket absorbing questions about pityriasis rosea (Q 54dc0e51) and other non-fungal papulosquamous conditions.

Example question IDs

Q cc0cacc5 — Lichen planus question filed under "Psoriasis." Content is factually acceptable; fix is reclassification only.
Q 49d0cf66 — Lichen planus histology filed under "Dermatitis and Eczema." Same reclassification fix.
Q 4af42166 — Pityriasis rosea vignette (collarette of scales, herald patch) filed under "Psoriasis." The clinical vignette and correct answer both point unambiguously to pityriasis rosea.
Q 8c40fe95 — Acute intermittent porphyria (abdominal pain, psychiatric symptoms, hallucinations) filed under "Phototherapy and Photobiology." Porphyria is a metabolic disorder; the question also lacks the photosensitivity finding that would make it dermatologically relevant.
Q a90a2726 — Prurigo nodularis with dupilumab as the answer, filed under "Phototherapy and Photobiology." High-quality Bloom's 5 question; fix is reclassification to "Dermatological Pharmacology."
Q 54dc0e51 — "Hanging curtain sign" (pityriasis rosea) filed under "Fungal Skin Infections."
Q bcd63a45 — Gorham syndrome filed under "Autoimmune Skin Diseases." Gorham disease is a vascular/lymphatic bone disorder; the topic label is wrong and the question is low-yield for Indian PG.
Q dad30d9b — Necrolytic migratory erythema filed under "Basic Dermatology." Belongs to paraneoplastic/systemic dermatology.
Q d5625520 — Donovanosis filed under "Viral Skin Infections." Donovanosis is a bacterial STI caused by Klebsiella granulomatis.

Recommended disposition

Reclassification is the primary remediation for this category and does not require content rewriting in most cases. Recommended reclassifications: lichen planus questions → "Autoimmune Skin Diseases" or a new "Papulosquamous Disorders" topic; pityriasis rosea questions → "Basic Dermatology" or "Papulosquamous Disorders"; porphyria questions → "Metabolic and Systemic Dermatology" or move to Medicine; prurigo nodularis/dupilumab → "Dermatological Pharmacology"; donovanosis → "Bacterial Skin Infections" or "STIs." Questions that are both mislabeled and low-quality (Q bcd63a45, Q dad30d9b) should be disabled rather than reclassified, since reclassification alone does not fix the content problem.

The "Phototherapy and Photobiology" topic bucket should be audited in full across the 3,237-question pool; the shard evidence suggests it is being used as a catch-all for questions that do not fit elsewhere, which will make it unreliable for topic-filtered test construction.

5. Structural Format Defects

Why this pattern is bad

Certain MCQ construction errors are not content problems — they are format problems that make a question invalid regardless of whether the underlying fact is correct. Compound options ("Options A and C are correct"), "all of the above" as the keyed answer, logically incoherent negative stems, and truncated option text are all format defects that violate standard MCQ construction principles used by INICET, NEET-PG, and USMLE. These formats reduce discriminatory power, allow partial-knowledge guessing, and in some cases make the question logically self-contradictory. They are also a signal of inadequate editorial review at the point of question creation.

How it shows up

The pattern appears across multiple shards and takes four distinct forms in this sample.

Compound options: Q 78a10a7f keys "Options A and C are correct" as the answer. This format is explicitly disallowed in quality MCQ design because it rewards candidates who know only one of the two facts and guess the compound.

"All of the above" as the correct answer: Q 140b0aa1 (primary pyodermas), Q 291deeb2 (fungi affecting hair), Q ce4139ae (ectopic sebaceous glands), Q cde653df (dermatophytes affect hair/nail/scalp). This format collapses the question to a single fact (if you know any one option is correct and "all of the above" is available, you can select it without knowing the others) and is universally discouraged in high-quality item banks.

Logically incoherent negative stems: Q 1c70746a keys "All of the above" as the answer to a "which is NOT a differential" question — logically, if all three listed options are not differentials, the question is testing three independent facts simultaneously with a single answer, which is incoherent. Q 43250b43 asks which finding is "characteristically seen in dermatomyositis" but three of four options are genuine dermatomyositis findings, making the question internally inconsistent.

Truncated or incomplete option text: Q 68ceb751 has options that are incomplete sentences ("It is a form of Tinea capitis" — a form of what?), suggesting a copy-paste truncation error during question entry.

"Both/None" option format: Q 416cdedf uses "Both" and "None" as options, a low-quality distractor structure that does not appear in any benchmark or PYQ item.

Example question IDs

Q 78a10a7f — Pyogenic granuloma; keyed answer is "Options A and C are correct." Compound-option format; must be rewritten as single-best-answer.
Q 140b0aa1 — Primary pyodermas; correct answer is "All of the above." Bloom's 1 recall with a universally discouraged option structure. Disable.
Q 1c70746a — "Which is NOT a differential for fever with vesicular rash" keyed as "All of the above." Logically incoherent. Rewrite as a focused single-exclusion question.
Q 43250b43 — Dermatomyositis findings; three of four options are genuine findings. Internally inconsistent. Disable or rewrite to ask about the most specific finding.
Q 68ceb751 — Tinea capitis/Favus; options are truncated fragments. Likely copy-paste error. Disable and rewrite from scratch.
Q 416cdedf — Facial hemiatrophy; uses "Both" and "None" as options. Disable.
Q 291deeb2 — "Which fungus affects hair? Both of the above." Compound option plus Bloom's 1 recall. Disable.
Q ce4139ae — Ectopic sebaceous glands; "all of the above" is correct. Disable.
Q cde653df — Dermatophytes affect hair/nail/scalp; "all of the above" is correct. Disable.
Q bde6c85e — "Which of the following is true about the specified skin disease?" The disease is never named in the stem. Unanswerable without external context; fix by naming the disease explicitly.

Recommended disposition

Disable all "all of the above" keyed items and all compound-option items without exception — these formats are not fixable by minor editing; the question must be rewritten from scratch. For logically incoherent negative stems (Q 1c70746a, Q 43250b43), disable and rewrite. For truncated option text (Q 68ceb751), disable and rewrite. For the unnamed-disease stem (Q bde6c85e), fix by naming the disease and replacing "all of the above" with a specific false distractor. This category is smaller in volume than the recall cluster but has zero tolerance for deployment as-is.

6. Near-Duplicate and Redundant Coverage

Why this pattern is bad

Duplicate or near-duplicate questions in the same pool inflate apparent coverage of a topic while providing no additional discriminatory information. In a test-construction context, duplicates increase the probability that a candidate who knows one item will correctly answer a second item testing the same fact, artificially inflating reliability estimates. They also crowd out coverage of genuinely undertested concepts. The problem is particularly acute when the duplicate is a lower-quality version of a PYQ that is already in the benchmark set — in that case, the candidate pool version adds no value and the PYQ version is strictly superior.

How it shows up

The pattern is concentrated in five topic areas in this sample. Herpes zoster dermatomal vignettes appear at least three times in the candidate pool (Q 104a697c and others) and are near-identical to PYQs Q 22e22e8f and Q 57e216c6 already in the benchmark set. Slapped-cheek appearance (Parvovirus B19) appears as Q 2376d4bb in the candidate pool and is a direct duplicate of PYQ Q f2f1596e. Molluscum contagiosum is tested twice in the same shard (Q f4bea92c and Q 750e55d4) using near-identical "NOT true/FALSE" stems. Scabies "not a feature" questions appear twice in the same shard (Q 9395b294 and Q 9ca61e11). Munro microabscesses recall (Q 194ea8ad) duplicates the content of benchmark Q 5a7b778f without the clinical vignette wrapper.

A secondary form of redundancy is the cluster of Tzanck smear questions: Q cef40cb4 (HSV → multinucleated giant cells) and Q e9b06a16 (pemphigus vulgaris → acantholytic cells) test the same investigation at the same cognitive level in the same shard, even though the correct answers differ.

Example question IDs

Q 2376d4bb — Slapped-cheek appearance → Parvovirus B19. Direct duplicate of PYQ Q f2f1596e (same question, same correct answer, same distractors reordered). Disable.
Q 104a697c — Dermatomal blisters on T3 trunk → VZV. Near-identical in structure and content to PYQs Q 22e22e8f and Q 57e216c6. Disable or reframe to test a distinguishing feature (Ramsay Hunt, post-herpetic neuralgia, immunocompromised host).
Q f4bea92c / Q 750e55d4 — Molluscum contagiosum pair. Both test inclusion body location/lab confirmation using "NOT true/FALSE" stems. Retain Q 750e55d4 (cleaner distractors); disable or substantially rewrite Q f4bea92c to test a different aspect (treatment, immunocompromised behaviour).
Q 9395b294 / Q 9ca61e11 — Near-duplicate scabies "not a feature" questions in the same shard. Retain Q 9395b294 (slightly better distractor construction); disable Q 9ca61e11.
Q 194ea8ad — Munro microabscesses recall. Functionally identical in content to benchmark Q 5a7b778f, which tests the same fact in a richer histopathology vignette. Disable as redundant.
Q e9b06a16 — Tzanck smear in pemphigus vulgaris. Near-duplicate of Q cef40cb4 in cognitive demand. Disable in favour of the better-framed Q cef40cb4.

Recommended disposition

Disable all confirmed direct duplicates of PYQ items without exception. For near-duplicates that test the same concept at the same Bloom's level, retain the higher-quality version and disable the lower-quality version. For topics where multiple near-duplicates exist (herpes zoster, molluscum, scabies), the retained question should be upgraded to test a higher-order aspect of the same concept — complications, management in special populations, or distinguishing features — rather than the same basic diagnostic fact. A deduplication pass across the full 3,237-question pool is warranted for at minimum: herpes zoster, molluscum contagiosum, scabies, vitiligo subtypes, slapped-cheek/Parvovirus B19, and Munro microabscesses.

Prioritization

The six issue categories are not equally urgent. The following prioritization reflects both the volume of affected questions and the severity of harm from deployment.

Immediate action (before any deployment):

Factual errors and answer-key defects (Category 3) — small volume but highest harm. Questions Q 57ad13fa, Q 84c5e656, Q db0915b7 have clear inverted or incorrect keys and must be disabled before any use. Questions Q 87ee6bfe, Q 8717c3cd, Q 7a48fe93, Q 61e481b7 require expert content correction. Questions Q e6b70ab4, Q dad30d9b, Q c376f990 require expert adjudication against a named current reference.
Image-dependent questions with absent images (Category 2) — Group A items (Q 8b92bedf, Q c3b7d0a8, Q fd444ad6, Q 801db0c7, Q c6a88ca9, Q f4ec5f6d) are unanswerable as text and must be disabled immediately. Group B items must be held pending image verification.
Structural format defects (Category 5) — "all of the above" keyed items, compound options, and logically incoherent stems are not deployable in any form. Disable Q 78a10a7f, Q 140b0aa1, Q 1c70746a, Q 43250b43, Q 68ceb751, Q 416cdedf, Q 291deeb2, Q ce4139ae, Q cde653df.

High priority (content operations sprint):

Near-duplicate and redundant coverage (Category 6) — confirmed duplicates of PYQs (Q 2376d4bb, Q 194ea8ad) should be disabled in the same sprint as the immediate actions. Near-duplicates (herpes zoster cluster, molluscum pair, scabies pair) should be resolved in a deduplication pass across the full pool.
Topic mislabeling (Category 4) — reclassification is low-effort for well-constructed questions (Q cc0cacc5, Q 49d0cf66, Q 4af42166, Q a90a2726, Q 54dc0e51). The "Phototherapy and Photobiology" bucket should be audited in full. Questions that are both mislabeled and low-quality should be disabled rather than reclassified.

Ongoing quality improvement (systematic):

Bare-recall items with no clinical scaffolding (Category 1) — the largest category by volume, requiring a systematic disable-and-rebuild programme. Priority should be given to topics where gold-standard PYQ coverage already exists (herpes zoster, psoriasis histopathology, vitiligo subtypes, alopecia types). Topics where high-quality vignettes are sparse — leprosy, photodermatoses, pregnancy dermatoses, biologics/advanced therapeutics — should be flagged for new question development at Bloom's 3–4 rather than recall-level reconstruction.

Bloom's distribution target: The current candidate distribution (Bloom's 1: 35%, Bloom's 2: 39%, Bloom's 3–5: 26%) should move toward a target of Bloom's 1: ≤15%, Bloom's 2: ≤25%, Bloom's 3–5: ≥60% to align with the benchmark and recent PYQ standard.

Example Keep / Fix / Disable Calls

The following table summarises representative disposition calls drawn from across the eight shards. These are illustrative of the categories above, not exhaustive.

KEEP — as-is or with minor tag/wording polish

Question ID	Topic	Reason to Keep
`d5f28049`	Fungal/Tropical	Farmer with indurated ulcer, sinuses, discharging granules → mycetoma. Clean Bloom's 3 vignette, good distractors, PYQ-tagged.
`10f1be84`	Pediatric Derm	Post-weaning infant, perioral/perianal lesions not responding to clotrimazole → acrodermatitis enteropathica vs. Chédiak-Higashi. Bloom's 4, genuine differential reasoning.
`4b487146`	Autoimmune	Young female, fever, shin nodule, foamy histiocytes, no vasculitis → erythema nodosum. Appropriate clinical framing, good distractors.
`abd26668`	Skin Tumors	Infiltrative BCC on anticoagulation with keloid history → Mohs surgery. Only Bloom's 5 item in its shard; multi-constraint clinical scenario.
`2dbac800`	Leprosy	Hypopigmented atrophic patch + diminished sensation → indeterminate leprosy. Bloom's 4, plausible distractors including pityriasis alba and borderline leprosy.
`e4cd51e5`	Bacterial/TB	Occupational exposure, 2-year history, Mantoux positive, KOH negative, biopsy reference → tuberculosis verrucosa cutis. Bloom's 4, matches benchmark standard.
`114d0eab`	Skin Tumors	Flat pigmented thigh lesion, management decision between excision biopsy vs. wide excision vs. FNAC. Bloom's 5, tests surgical judgment.
`20e08fd9`	Photodermatoses	Hydroa vacciniforme vignette: age-appropriate, sun-exposure trigger, hemorrhagic vesicles healing with scars. Bloom's 3, no image dependency.
`7087fa87`	Pregnancy Derm	Impetigo herpetiformis in pregnancy: gestational context, pustular ring morphology, management question (corticosteroids vs. methotrexate contraindicated). Bloom's 3.
`65c1504f`	Blistering Diseases	Chest wall blisters, "except" for diagnostic modalities (DFA, PCR, Tzanck vs. LDH). Bloom's 4, genuine discriminatory value.
`12bf97d8`	Blistering Diseases	Bullous pemphigoid vignette (85-year-old, negative Nikolsky's, relapsing blisters). Clinically grounded, age-appropriate. Bloom's 3.
`8161b68d`	Blistering Diseases	Dermatitis herpetiformis: IgA deposition, anti-epidermal transglutaminase, clinically meaningful distractors. Aligns with PYQ benchmark `24dfb4ea`.
`40ac3af5`	Bacterial	Lupus vulgaris vignette: age, duration, annular plaque, peripheral crusting, central scarring. Distractors include tinea corporis, granuloma annulare, borderline leprosy.
`44cc63fa`	Derm Pharmacology	Biologic agents in psoriasis: which is not anti-TNF-α? Alefacept (CD2/LFA-3 fusion protein) is the correct outlier. Meaningful pharmacological classification.
`a90a2726`	Derm Pharmacology	Prurigo nodularis with dupilumab. Bloom's 5, high-quality; fix is reclassification from "Phototherapy" to "Dermatological Pharmacology."
`ffb1a3c3`	Fungal	Pityriasis versicolor "all except" (NEET PYQ): Wood's lamp, KOH morphology, causative organism, drug of choice tested in one item. Keyed exception (oral terbinafine not DOC) is a genuine high-yield discriminator.

FIX — substantive edit required before deployment

Question ID	Issue	Recommended Fix
`57ad13fa`	Inverted answer key (candidiasis does involve nails)	Verify and correct the key, or rewrite the stem to make the intended exception unambiguous. Priority fix.
`87ee6bfe`	Disease name given as organism (Tinea capitis as causative agent of kerion)	Replace correct answer with a specific organism (Trichophyton violaceum or Microsporum canis).
`7a48fe93`	Outdated management (PUVA as mainstay for vitiligo)	Update to NB-UVB or add context qualifier ("before the era of NB-UVB") to avoid teaching superseded practice.
`8c40fe95`	Mislabeled topic (AIP under Phototherapy) + missing dermatological relevance	Add photosensitivity finding to stem or move to Medicine; reclassify topic.
`4af42166`	Mislabeled topic (pityriasis rosea under Psoriasis)	Reclassify to Basic Dermatology or Papulosquamous Disorders. Content is otherwise sound.
`bde6c85e`	Disease never named in stem; "all of the above" keyed	Name the disease (psoriasis) explicitly; replace "all of the above" with a specific false distractor.
`78a10a7f`	Compound-option format ("Options A and C are correct")	Rewrite as single-best-answer with four independent factual statements. Underlying content (pyogenic granuloma) is high-yield.
`4c2497bc`	Sparse stem (no demographics), non-standard difficulty field ("medium")	Add patient demographics and one additional clinical detail; standardise difficulty to numeric scale.
`032fb2fa`	PHACE syndrome stem omits defining feature (infantile hemangioma)	Rewrite to include large segmental facial infantile hemangioma and ask about associated anomaly.
`61e481b7`	EAC described as having "central scarring" (should be "central clearing")	Revise clinical descriptor to avoid teaching an inaccurate feature.
`4357eb6d`	Leprosy vignette lacks any clinical discriminator (no sensory loss, no nerve thickening)	Add at least one clinical sign to make the diagnosis defensible over pityriasis alba.
`d7acd586`	Lichen planus treatment answer valid only for localised disease	Add extent/severity qualifier to stem or revise options to specify "topical steroids for localised disease."
`c3b7d0a8`	Image absent; stem is demonstrative pronoun pointing at nothing	Embed clinical description (tender erythematous nodules on shins, bilateral) to replace missing image reference.
`11b87ed2`	Kerion/tinea capitis vignette lacks surface features to justify Bloom's 3	Add one clinical detail (boggy swelling, broken hair stubs) to make diagnostic reasoning explicit.
`e9fe3416`	"All of the options" answer for non-cicatricial alopecia is factually imprecise (SLE can scar; kerion can scar)	Replace SLE with telogen effluvium; remove "all of the above" option.

DISABLE — remove from active pool

Question ID	Reason
`8b94e4b8`	Pure morphology definition (macule = 1 cm flat discolouration); one distractor is "Plague" (typo for Plaque); Bloom's 1, no clinical context.
`be9ae193`	Diagnosis embedded in stem (melasma); question answers itself; Bloom's 1.
`6666919d`	Single-association recall (eczema herpeticum → HSV); Bloom's 1; no clinical scenario.
`43250b43`	Three of four options are genuine dermatomyositis findings; internally inconsistent; tests only name recognition.
`e29431ed`	Tinea versicolor causative organism; Bloom's 1; no vignette; trivially recalled.
`2376d4bb`	Direct duplicate of PYQ `f2f1596e` (slapped-cheek → Parvovirus B19).
`194ea8ad`	Functionally identical to benchmark `5a7b778f` (Munro microabscesses) but without the clinical vignette wrapper.
`84c5e656`	Factual error: verrucous lichen planus is a recognised variant; keyed as "not a real form."
`db0915b7`	Likely inverted key: potassium iodide is contraindicated in pregnancy; keyed as "not contraindicated."
`9fdcd7e8`	Langerhans cells as APCs; "dendritic cells" is a defensible alternative answer (Langerhans cells are a dendritic cell subset); conceptually flawed distractors + trivial recall.
`4b4ff1b8`	Rhinophyma is a type of rosacea; single-association recall; Bloom's 1; no clinical context.
`140b0aa1`	Primary pyodermas; correct answer is "all of the above"; Bloom's 1 recall with universally discouraged option structure.
`ae47252c`	Lupus vulgaris = TB of skin; single definitional sentence; distractors implausible to any prepared candidate.
`87958877`	Tinea unguium involves nails; Latin root gives away the answer; zero discriminatory value.
`ab960295`	Inheritance of tuberous sclerosis; Bloom's 1; no clinical context.
`416cdedf`	Facial hemiatrophy; uses "Both" and "None" as options; low-quality distractor structure not seen in any benchmark or PYQ.
`343b2cbb`	Fordyce spots location; pure recall; Bloom's 1; no clinical context.
`6986c084`	Quincke's disease = angioneurotic oedema; eponym synonym recall; Bloom's 1.
`9defafa8`	Common warts caused by HPV; universally known; zero discrimination.
`6b6706a8`	Normal epidermal turnover time = 4 weeks; isolated numerical fact; no clinical application.
`0e4ca797`	Mycosis fungoides cell type (T cells); single-fact recall; Bloom's 1.