ENT Question Quality Review
Executive Summary
The ENT candidate sample of 200 validated non-gold questions was reviewed across eight shards of 25 questions each. The findings are consistent and convergent across shards: the sample is structurally skewed toward low-order recall, with Blooms-1 items accounting for 80 of the 200 candidate questions (40%) and Blooms-2 accounting for a further 97 (48.5%). Only 23 questions (11.5%) operate at Blooms-3 or above — the level at which the benchmark and recent PYQ sets are predominantly pitched.
The dominant problem is not a single category of error but a layered accumulation of quality deficits: a large volume of single-fact recall items that should never have entered a PG-level bank, a structurally broken subset of image-dependent questions whose images are absent, a cluster of near-duplicate questions on the same narrow topics, a smaller but operationally serious group of questions with contestable or incorrect answer keys, and a recurring pattern of pseudo-clinical vignettes that use patient framing but resolve to a single memorised fact.
Across the 200 questions reviewed, the estimated disposition breakdown is approximately: Keep as-is: ~40–45 questions (20–22%); Fix before use: ~55–60 questions (27–30%); Disable: ~95–105 questions (47–52%). The disable rate is high but justified — the benchmark and recent PYQ sets already provide strong coverage of the high-yield concepts that these low-quality items nominally address, and speculative rewrites of Blooms-1 trivia items rarely produce questions that match the clinical depth of purpose-built vignettes.
What Good Looks Like
The benchmark and recent PYQ items establish a clear quality standard for this subject. The following features characterise the best items in the gold set and should serve as the operational target for the candidate pool:
Clinical anchoring even at Blooms-2. The BPPV question (acafcb97) gives a 50-year-old female with positional vertigo, no hearing loss, and no prior illness — the candidate must reason through a differential, not retrieve a single fact. The piriform recess foreign body question (4d94eca8) requires understanding of which cranial nerve supplies that region before the answer becomes accessible. Even the olfaction test question (cbba76d5), which is Blooms-1, is acceptable because it tests a genuinely exam-tested fact (Arnold stick test is not an olfaction test) with plausible distractors drawn from real clinical tools.
Distractors that represent genuine misconceptions. The JNA question (b66df1e8) works because "biopsy is contraindicated" is a true statement that candidates must distinguish from the false statement about cervical lymph nodes — both options are clinically plausible and require knowledge to separate. The cochlear implant candidacy question (4463b4ad) works because ossicular fixation and otosclerosis are treatable by stapedectomy, not implantation — a distinction that requires applied understanding, not list recall.
Management decisions in context. The post-FESS revision scenario (d127f61b) is the clearest example: patent ostia on endoscopy after prior surgery means the problem is not mechanical obstruction, so the next step is tissue biopsy to exclude fungal or neoplastic disease. This is Blooms-4 reasoning embedded in a realistic clinical scenario. The perilymphatic fistula case (8d6609dd) similarly requires the candidate to connect a post-myringoplasty timeline to a specific complication rather than pattern-match on a symptom list.
Negative/exception format used only when the exception requires reasoning. The "least appropriate intervention" question for bilateral SNHL in a child (da7e715d) works because adenoidectomy with grommet insertion is the treatment for conductive hearing loss from OME, not sensorineural loss — the candidate must understand the mechanism of each intervention to identify the mismatch.
The candidate sample should be measured against this standard throughout.
Main Issue Categories
1. Blooms-1 Recall Excess: Single-Fact Items With No Clinical Reasoning Demand
Why this pattern is bad
Indian PG entrance examinations — INICET, NEET-PG — have moved decisively toward applied clinical reasoning at Blooms-2 through Blooms-4. A question that asks "What is the focal length of a head mirror?" or "Glossoplegia is defined as paralysis of which structure?" does not discriminate between a candidate who understands ENT and one who has memorised a single line from a textbook. These items inflate the apparent size of the question bank without adding any assessment value. They also crowd out higher-order items in practice tests and daily plans, reducing the overall learning signal delivered to candidates.
How it shows up
This is the broadest and most pervasive pattern in the sample. Across all eight shards, approximately 80 of the 200 candidate questions are formally tagged Blooms-1, and a further substantial proportion of Blooms-2 items are functionally Blooms-1 because they present a clinical label without any reasoning requirement. The items cluster around four recognisable sub-types:
- Eponym-to-definition mapping: "Cart Wheel sign in ASOM" (a32a6a5c), "Gutzmann's pressure test is done for which condition?" (63d04154), "Glycerol test is used for diagnosis of?" (f54eea2f), "Plunging ranula = retention cyst" (e99321cd), "Schwartz sign definition" (91172624), "Melkersson Rosenthal syndrome" (95563a3e), "Eagle's syndrome" (aa8c8d00), "Citelli's angle" (5d9b5005, 70edef67), "Woodruff's plexus location" (84bcdb58), "Kallmann syndrome and anosmia" (2208181b).
- Instrument and examination technique trivia: "Focal length of head mirror" (47edf02a), "Head mirror applied to which eye" (3058bbcb), "What is a myringotomy?" (7d842b86), "Which type of hearing aid is shown below?" (4a16a7a0 — also image-broken), "Nodes of Rouvier" (713f317a).
- Basic anatomy recall with no clinical hook: "Eustachian tube opens into which part?" (7df6c775 — with "Larynx" as a distractor), "Elastic cartilage in which structure?" (16c38192), "Scala tympani vs. scala vestibuli" (8aed38e9), "Reinke's space location" (3ceb7797), "Tegmen separates middle ear from middle cranial fossa" (d355d352 — the question answers itself in the stem), "Auricle nerve supply EXCEPT" (7b071df8), "Promontory distance from tympanic membrane" (7bd72c73), "Scala media filled with endolymph" (650b45ea), "Bilateral post-pointing due to defect in which system?" (182ce9fc), "Scutum = bony outer attic wall" (ba147723), "Kiesselbach's plexus contributors" (ffaafbdd), "Caloric test assesses which structure?" (98be1b16).
- Causative organism / classification recall: "Malignant otitis externa is caused by?" (9b3c8d6d), "Rhinoscleroma is caused by which type of infection?" (05c6d11f), "Commonest complication of CSOM?" (cc49537c), "Most common salivary gland tumor in children" (2e5239be — also factually contested).
Recommended disposition
The instrument trivia sub-cluster (47edf02a, 3058bbcb, 7d842b86, 4a16a7a0) should be disabled without exception — these facts are not tested in any benchmark or recent PYQ and have no clinical application. The anatomy recall sub-cluster (7df6c775, 16c38192, 8aed38e9, 3ceb7797, d355d352, 7b071df8, 7bd72c73, 650b45ea, 182ce9fc, ba147723, ffaafbdd, 98be1b16) should be disabled unless a specific item can be shown to have PYQ precedent and can be embedded in a clinical vignette without rewriting the core concept entirely. The eponym sub-cluster should be evaluated individually: eponyms with PYQ precedent (e.g., Hennebert's sign in 7fcf106f, which is already a Keep) should be retained in upgraded form; eponyms with no PYQ precedent and no clinical application (a32a6a5c, 63d04154, f54eea2f, 95563a3e, 5d9b5005, 84bcdb58, 2208181b) should be disabled. The causative organism sub-cluster (9b3c8d6d, 05c6d11f, cc49537c) should be disabled as standalone items; the underlying concepts (malignant otitis externa, rhinoscleroma, CSOM complications) are worth testing but only in clinical vignette format.
2. Pseudo-Clinical Vignettes: Patient Framing That Resolves to Single-Fact Recall
Why this pattern is bad
This is a distinct and operationally important category, separate from pure Blooms-1 recall. These questions open with a patient age, symptom, or brief scenario but then require no differential reasoning — the answer is immediately deducible from a single feature in the stem without engaging the clinical context. They create a false impression of clinical depth in the question bank and mislead candidates about the reasoning level required in actual examinations. They also inflate the apparent Blooms-2/3 count without delivering the discriminatory value those levels imply.
How it shows up
The clearest examples in this sample are:
- f7878e52: "70-year-old man with tinnitus → most probable diagnosis = acoustic neuroma." Tinnitus alone in a 70-year-old is far more likely to represent presbycusis or noise-induced hearing loss. The clinical framing is present but the answer key is both clinically questionable and not the product of differential reasoning — it is a pattern-match on "tinnitus + acoustic neuroma" without engaging age, laterality, or audiometric findings.
- ff68db49: "Unilateral SNHL + MRI shown → acoustic neuroma." The MRI image is absent (see Category 3), but even if present, the question resolves to a single pattern-recognition step.
- d9366638: Parapharyngeal mass with facial nerve palsy and chondromyxoid stroma on biopsy — the clinical framing is richer, but the correct answer ("most common salivary gland tumor") is a generic statement rather than a specific management or diagnostic conclusion. The vignette does not require the candidate to reason through the differential; it requires them to recall that pleomorphic adenoma is most common.
- 94a293ad and 0dd1e48a (shard 003): Both present patient contexts but are flagged easy, suggesting the clinical framing adds no discriminatory load.
- bce5a5ce (shard 008): "Unilateral non-suppurative otitis media in adults → malignancy." This is a clinically important red-flag concept but the question is a one-step pattern-match rather than a reasoning exercise.
The pseudo-clinical pattern is particularly common in the Diseases of the Ear and Rhinology topics, where a brief symptom opener is used to dress up what is essentially a recall item.
Recommended disposition
Items in this category are generally fixable rather than disabled, because the underlying concept is often clinically relevant and the vignette structure is already partially in place. The fix path is to add specificity to the stem (laterality, duration, associated features, audiometric or imaging findings) and to ensure the correct answer requires the candidate to reason through at least one plausible alternative. f7878e52 requires both a stem fix (add unilateral SNHL, imbalance, and audiometric rollover) and an answer key review. d9366638 requires the correct answer to be made more specific (name the tumor, not just the category). bce5a5ce is acceptable as a Blooms-2 red-flag recognition item but should add one or two clinical details (adult male, no prior ear infection, no response to antibiotics) to make the reasoning explicit.
3. Image-Dependent Questions With Absent or Unverifiable Images
Why this pattern is bad
Questions that reference images without embedding them are non-functional in their current state. Depending on the image, the question is either entirely unanswerable from text alone (making it a broken item that will generate candidate complaints and incorrect scoring) or trivially answerable without the image (making the image reference misleading about the question's actual difficulty and Blooms level). Both outcomes are unacceptable for a deployed question bank.
How it shows up
This pattern appears across multiple shards and topics, suggesting it is not isolated to a single import batch:
- 798b53fd (shard 002): "Name the procedure being done" — entirely unanswerable without the image.
- ff68db49 (shard 002): "MRI findings are as shown" — answerable from clinical context alone, making the image reference misleading.
- f4326e02 (shard 005): PYQ (NEET-PG 2016–17) identifying indirect laryngoscopy and spatula test — image absent, question non-functional.
- 8c3efb90 (shard 005): "Identify the line shown in the given image" — entirely non-functional without image.
- 0f748dca (shard 005): Cochlear implant image question — image absent and also misclassified under Diseases of the Nose and Paranasal Sinuses.
- 9d6074a0 (shard 004): "Which of the following studies is shown below?" — image absent.
- 4a16a7a0 (shard 008): "Which type of hearing aid is shown below?" — image absent, and the question is Blooms-1 even with the image.
- c44d98e1 (shard 006): JNA histopathology reference — image verification needed before deployment.
- a916068a (shard 008): Post-SMR septal perforation with image reference — PYQ-tagged and otherwise well-constructed; image must be confirmed attached.
The pattern is likely more widespread than the eight confirmed instances above, given that image-dependent questions are common in ENT (otoscopy, audiograms, imaging, endoscopy) and the import process appears to have separated some images from their stems.
Recommended disposition
Questions where the image is the sole basis for answering (798b53fd, 8c3efb90, 9d6074a0) should be disabled immediately unless the image can be confirmed attached and functional. Questions where the image adds clinical value but the question is partially answerable from text (ff68db49, f4326e02) should be fixed by either embedding the confirmed image or rewriting the stem as a descriptive vignette. Questions where the image is referenced but the question is otherwise Blooms-1 (4a16a7a0) should be disabled regardless of image status. The PYQ-tagged item a916068a should be kept pending image confirmation. A pool-wide image audit is strongly recommended given the frequency of this pattern in the sample.
4. Contestable or Incorrect Answer Keys
Why this pattern is bad
A question with a wrong or ambiguous correct answer is worse than no question at all — it actively miseducates candidates and, if deployed in scored assessments, produces unfair outcomes. In ENT, where management guidelines and investigation hierarchies have evolved (e.g., MRI vs. CT for JNA, combined vs. antibiotic-only management of necrotising otitis externa), answer keys that reflect older or minority-source positions are a specific risk. The "All of the above" and "Any of the above" answer key pattern is a related structural problem: it avoids committing to a testable fact and rewards uncertainty rather than knowledge.
How it shows up
Several distinct sub-types appear in this sample:
- Factually incorrect or outdated answer keys: Q-3a543c04 marks CT scan as the investigation of choice for angiofibroma; current consensus favours contrast-enhanced MRI for soft-tissue and intracranial extent. Q-af9ca007 (Ramsay Hunt syndrome) lists "involves Vth cranial nerve" as a true statement, which is non-standard (RHS involves VII and VIII). Q-e99321cd states plunging ranula is a retention cyst; most current sources classify it as an extravasation pseudocyst. Q-6d1f0ed0 states parotid is the most common H&N site for lymphoepithelioma; nasopharynx is the standard answer. Q-f7878e52 marks acoustic neuroma as the most probable diagnosis for isolated tinnitus in a 70-year-old.
- "All of the above" / "Any of the above" as the correct answer: eb355c2e ("most common organism in acute sinusitis" → "All of the above"), 838838df (nasal polyp associations → "All of the above"), fbefa989 (hyperacusis causes → "Any of the above"), d21ba0c1 (premalignant lesions → "All of the above"). This format is a known quality flag because it rewards candidates who cannot identify the single best answer and penalises those who know that one option is more correct than others.
- Contestable answer keys requiring reference anchoring: 9c2d88f1 (nasal cycle timing "6–8 hours" — literature cites 2–6 or 4–12 hours depending on source), bda6ecdd (rheumatic tonsillitis marked as correct indication over recurrent acute tonsillitis — the framing is ambiguous without specifying absolute vs. relative indication), c384d2c2 (otitic barotrauma "not true" — occurs during ascent is debatable since descent is the classic mechanism but ascent can also cause reverse squeeze), e3cf05a2 (malignant otitis externa management → "antibiotic therapy" alone, understating the role of surgical debridement).
- Tautological stems: 861a059b asks "Which anatomical region is primarily involved in Nasopharyngeal Carcinoma?" with the answer "Nasopharyngeal cavity" — the question answers itself in the stem.
Recommended disposition
Items with clearly incorrect answer keys (3a543c04, af9ca007, e99321cd, 6d1f0ed0, f7878e52) should be fixed with expert review against a named reference (Scott-Brown, Dhingra, or the relevant INICET PYQ source) before any deployment. Items using "All/Any of the above" as the correct answer (eb355c2e, 838838df, fbefa989, d21ba0c1) should be fixed by restructuring to commit to a single best answer, or disabled if restructuring would require a complete rewrite. Items with contestable answer keys (9c2d88f1, bda6ecdd, c384d2c2, e3cf05a2) should be fixed with explicit reference citation added to the rationale. The tautological item (861a059b) should be disabled.
5. Near-Duplicate and Redundant Topic Clustering
Why this pattern is bad
When two or more questions in the bank test the same narrow fact with the same correct answer and similar distractors, they add no incremental assessment value and consume slots in practice tests and daily plans that could be occupied by higher-yield items. In a bank of 4,280 questions, redundancy at the topic level is expected, but redundancy at the specific-fact level — where two questions are functionally interchangeable — is a quality problem. It also signals that certain topics have been over-mined for easy recall items while higher-order clinical scenarios in the same topic remain absent.
How it shows up
Several confirmed near-duplicate pairs and clusters appear in this sample:
- Cholesteatoma: c4493f51 and 579d9e08 (shard 008) both ask "What is true about cholesteatoma?" with bone erosion as the correct answer and nearly identical distractors.
- Laryngomalacia: aa961777 and 6fae6224 (shard 006) both ask "Which statement about laryngomalacia is FALSE?" with overlapping distractors.
- Malignant otitis externa: e3cf05a2 (management), 8b209c44 (diagnosis), and e4b813bf (underlying condition) in shard 006 constitute near-redundant coverage of the same narrow fact cluster in a diabetic patient — all three are answerable from the same single fact (Pseudomonas in a diabetic).
- Oral premalignant lesions: 4aab239f ("most common precancerous lesion") and d21ba0c1 ("which are premalignant lesions") in shard 005 cover the same leukoplakia/erythroplakia/OSMF territory at Blooms-1.
- Caldwell-Luc / infraorbital nerve: 1ec62530 and 71d35960 (shard 008) both address infraorbital nerve injury after Caldwell-Luc, one as a complication scenario and one as direct recall.
- JNA: The benchmark set already contains two JNA questions (b66df1e8 and f384eb19) that are themselves near-duplicates of each other; the candidate pool adds further JNA recall items that overlap with this coverage.
- Antrochoanal polyp: ec56c585 (shard 003) and at least one further item in shard 005 cover overlapping antrochoanal polyp territory at Blooms-1/2.
The salivary gland / oral cavity sub-topic appears disproportionately across multiple shards (387022c4, d9366638, 72cfd62d, da3ac8fa in shard 002; e99321cd, 5b27f349, b7884786 in shard 007; 2e5239be in shard 008), suggesting this topic is over-represented relative to higher-yield ENT domains such as vestibular disorders, CSOM complications, and laryngology.
Recommended disposition
For confirmed near-duplicate pairs, keep the higher-quality item (the one with better clinical framing, higher Blooms level, or PYQ tagging) and disable the weaker duplicate. For the malignant otitis externa cluster, consolidate into one well-constructed clinical vignette (diabetic patient, granulation tissue at EAC floor, management decision) and disable the two bare-recall items. For the oral premalignant lesion pair, disable d21ba0c1 (the "All of the above" item) and consider upgrading 4aab239f to a clinical scenario. For the Caldwell-Luc pair, merge into one clinical scenario and disable the recall version. A deduplication pass by topic + correct answer across the full 4,280-question pool is recommended.
6. "Except/NOT True" Format Applied to Low-Yield or Poorly Constructed Content
Why this pattern is bad
The negative/exception format is a legitimate and high-yield question type when used correctly — the benchmark JNA question (b66df1e8) and the cochlear implant candidacy question (4463b4ad) both use it effectively because the exception requires genuine clinical reasoning to identify. The problem in this sample is that the format is frequently applied to content that is either too easy (making the exception trivially obvious), too obscure (making the exception untestable without memorising a complete list), or accompanied by implausible distractors that trivialise the question. When the format is used at Blooms-1 on a list of four items where three are obviously true and one is obviously false, it adds no discriminatory value over a direct recall question.
How it shows up
- Implausible distractors that trivialise the question: 074d8ee1 (Meniere's disease NOT true) includes "leads to diplopia" as a distractor — diplopia is not a recognised feature of Meniere's disease and no candidate would seriously consider it, making the question trivially easy regardless of the correct answer. 7df6c775 (Eustachian tube opens into which part) includes "Larynx" as a distractor — anatomically absurd.
- Exception format at Blooms-1 on anatomy lists: 7b071df8 (auricle nerve supply EXCEPT — lesser petrosal nerve), 60f49c25 (Gradenigo's triad EXCEPT — Blooms-1, easy), e860bfca (Gradeneigo syndrome except — reasonable content but no clinical stem).
- Exception format on management lists without clinical context: 7e3b1485 (superior laryngeal nerve palsy except — tagged Blooms-1 despite the format), 7a9888de (vestibulotoxic drugs — list recall), 09da7b5a (septoplasty — list recall).
- Exception format with "All of the above" as the correct answer in the positive version: 838838df (nasal polyp associations) — the exception format's value is entirely negated when the positive version would have "All of the above" as correct.
Recommended disposition
Items where the exception format is applied to Blooms-1 anatomy or eponym lists (7b071df8, 60f49c25, 7a9888de, 09da7b5a) should be disabled or fixed by adding a clinical stem that makes the exception clinically meaningful. Items with implausible distractors (074d8ee1, 7df6c775) should be fixed by replacing the implausible distractor with a clinically plausible false statement. Items where the format is used appropriately but lacks a clinical stem (e860bfca, 7e3b1485) should be fixed by adding a brief vignette. The general principle for this format: if the exception cannot be identified by clinical reasoning and requires only list memorisation, the question should be restructured or disabled.
7. Structurally Broken Questions (Corrupt Stems, Tautological Stems, Misclassified Topics)
Why this pattern is bad
These are questions that cannot be used in their current state regardless of the underlying concept's merit. They represent different failure modes — data import corruption, self-answering stems, and topic misassignment — but share the common feature that no amount of content editing will make them usable without a structural intervention. They are operationally distinct from the other categories because the fix path is either immediate disabling or a complete rebuild, not a targeted content edit.
How it shows up
- Corrupt import artefact: f02712f7 (shard 008) has a stem reading "The volume of the balloons shown in epistaxis balloon is: (Recent NEET Pattern 2016-17)" and options that are answer-key line numbers ("72. (d)", "73. (b)"). This is not a question; it is a data-import fragment. It is entirely non-functional.
- Tautological stems: 861a059b (shard 005) asks "Which anatomical region is primarily involved in Nasopharyngeal Carcinoma?" with the answer "Nasopharyngeal cavity" — the answer is contained in the question. d355d352 (shard 003) similarly asks about the tegmen in a way that answers itself.
- Topic misclassification: 47edf02a (focal length of head mirror) is filed under "Neurotology" — a clear misassignment. 0f748dca (cochlear implant image question) is filed under "Diseases of the Nose and Paranasal Sinuses." 303bd178 (radicular cyst / non-vital tooth) in shard 008 is a dental/oral surgery concept with no ENT-specific hook, misplaced in the ENT bank. b4da54b4 (shard 008) uses the non-standard term "auditory neurotherapy" for a Meniere's disease management question — either a terminology error or a misclassification of a non-ENT concept.
- Non-standard terminology: b4da54b4 uses "auditory neurotherapy," which does not appear in mainstream ENT references (Scott-Brown, Dhingra, Cummings). If the intended answer is intratympanic gentamicin or betahistine, the question needs to be rewritten with correct terminology.
Recommended disposition
f02712f7 should be disabled immediately — it is not a question. 861a059b and d355d352 should be disabled as tautological. 47edf02a and 0f748dca should be reclassified to their correct topics as a minimum fix, but given that both are also Blooms-1 items with other quality problems, disabling is the more efficient path. 303bd178 should be disabled from the ENT bank (it may belong in a Dental/Oral Surgery bank if one exists). b4da54b4 should be disabled pending expert review of the terminology and intended answer.
8. Clinically Applicable Concepts Presented as Bare Recall (Upgrade-Eligible Items)
Why this pattern is bad
This category is distinct from the Blooms-1 recall excess category because the underlying concepts here are genuinely high-yield and exam-relevant — they appear in PYQs, they test important clinical decision points, and they are worth having in the bank. The problem is purely one of execution: the question presents a management fact, a diagnostic criterion, or a surgical decision as a direct recall item rather than embedding it in a clinical scenario. These items are not worthless; they are underperforming relative to their potential. They represent the most productive fix target in the sample because a relatively small stem addition can elevate them to benchmark quality.
How it shows up
- 940c64bb (shard 001): Radical mastoidectomy obliterating Eustachian tube communication — factually correct, clinically important, but presented as bare recall. A brief post-operative complication scenario would make this Blooms-3.
- 22ea0966 (shard 001): Chronic retropharyngeal abscess caused by Pott's disease — correct and high-yield, but a one-liner. A vignette with neck stiffness, dysphagia, and vertebral destruction on X-ray would make this a strong Blooms-3 item.
- d1db1ec0 (shard 003): Unilateral secretory otitis media in an adult → nasopharyngeal carcinoma — the red-flag concept is exactly right, but the question is a bare one-liner. Adding an adult male patient with unilateral middle ear effusion and no prior infection history would make this a benchmark-quality item.
- 212b4062 (shard 004): Trotter's triad — valid content, but presented as direct recall. A brief clinical stem (conductive hearing loss, ipsilateral soft palate immobility, trigeminal neuralgia) would convert this to Blooms-3.
- 12afc64e (shard 007): T1N0M0 glottic carcinoma treatment — factually acceptable but bare recall. A 55-year-old smoker with hoarseness and a lesion confined to one vocal cord with normal mobility would make this a strong applied question.
- 83880782 (shard 008): Most common extracranial complication of ASOM → subperiosteal abscess — factually sound but Blooms-1/easy. A brief clinical lead-in (tender post-auricular swelling, displaced pinna) would reach Blooms-3.
- 9a0e5d84 (shard 008): VEMP assesses saccule via inferior vestibular nerve — correct but bare recall. A suspected superior semicircular canal dehiscence referral scenario would make this Blooms-3.
- c0577510 (shard 001): OSA "not true" question — factually sound but distractors are too easily eliminated. Replacing one distractor with "polysomnography is diagnostic" (a plausible wrong option) would increase discrimination.
Recommended disposition
All items in this category are fix candidates. The fix path is consistent: add a brief clinical stem (2–3 sentences, patient age/sex/presentation) that makes the correct answer the product of reasoning rather than recall. The answer key and distractors typically do not need to change. Priority should be given to items where the underlying concept has PYQ precedent or appears in the benchmark set (d1db1ec0, 12afc64e, 83880782) — these are the most likely to appear in future examinations and the most valuable to have at Blooms-3 quality.
Prioritization
The eight issue categories identified above are not equally urgent. The following prioritization reflects both the volume of affected questions and the operational risk of deploying affected items.
Immediate action required (before any deployment):
Category 7 (Structurally Broken Questions): f02712f7 must be disabled immediately. Tautological and misclassified items (861a059b, d355d352, 303bd178, b4da54b4) should be disabled in the same pass. This is a small group (~8–10 items in the sample) but the risk of deploying a corrupt question in a scored assessment is unacceptable.
Category 3 (Image-Dependent Questions With Absent Images): A pool-wide image audit should be triggered. In the sample, at least 7–8 confirmed instances were identified. Items where the image is the sole basis for answering (798b53fd, 8c3efb90, 9d6074a0, 4a16a7a0) should be disabled immediately. PYQ-tagged items with image references (f4326e02, a916068a) should be held pending image confirmation.
Category 4 (Contestable or Incorrect Answer Keys): Items with clearly incorrect answer keys (3a543c04, af9ca007, e99321cd, 6d1f0ed0, f7878e52) should be flagged for expert clinical review before any deployment. "All of the above" items (eb355c2e, 838838df, fbefa989, d21ba0c1) should be restructured or disabled.
High-volume remediation (content operations sprint):
Category 1 (Blooms-1 Recall Excess): This is the largest single category by volume (~80 confirmed Blooms-1 items in the candidate sample, with a further substantial proportion of functionally Blooms-1 items among the Blooms-2 count). The instrument trivia and basic anatomy sub-clusters should be disabled in bulk. The eponym sub-cluster should be evaluated individually against PYQ precedent. This is the highest-volume action item and will have the greatest impact on the overall quality distribution of the bank.
Category 5 (Near-Duplicate and Redundant Topic Clustering): A deduplication pass by topic + correct answer should be run across the full 4,280-question pool. Confirmed near-duplicate pairs in the sample (c4493f51/579d9e08, aa961777/6fae6224, 1ec62530/71d35960, 4aab239f/d21ba0c1) should be resolved immediately.
Targeted content editing (fix queue):
Category 8 (Upgrade-Eligible Items): The ~15–20 items identified as clinically applicable concepts in bare-recall format represent the highest-return fix investment. Each requires only a 2–3 sentence stem addition. Priority order: d1db1ec0, 12afc64e, 83880782, 940c64bb, 22ea0966, 212b4062, 9a0e5d84.
Category 2 (Pseudo-Clinical Vignettes): f7878e52 requires both stem fix and answer key review. d9366638 requires answer specificity. bce5a5ce requires additional clinical detail. These are individually fixable but require content expertise to execute correctly.
Category 6 (Exception Format Misuse): Items with implausible distractors (074d8ee1) should be fixed by distractor replacement. Items using the format at Blooms-1 on anatomy lists should be disabled. This is a moderate-volume category (~10–12 items in the sample) with a clear fix path.
Example Keep / Fix / Disable Calls
The following table summarises representative disposition calls drawn from across the eight shards. These are illustrative of the patterns described above and should be used as calibration examples for the content operations team.
KEEP — No action required
| Question ID | Topic | Rationale |
|---|---|---|
| 7579cf96 | Diseases of the Ear | Paediatric OME vignette, 2-step reasoning chain, Blooms-3, clinically grounded |
| 63ed6e55 | Diseases of the Larynx | Infant subglottic haemangioma, EXCEPT format with clinical trap, Blooms-3 |
| 2318d9fd | Diseases of the Ear | Temporal bone transverse fracture EXCEPT, factually sound, Blooms-2 |
| 5d10f9e9 | ENT Oncology | Laryngeal cancer subtypes, comparative knowledge, Blooms-4, difficulty-3 |
| 369c9ddd | Neurotology | Vestibular schwannoma audiological features, rollover vs. recruitment, Blooms-2 |
| 60f5bbb3 | Neurotology | Processus cochleariformis as facial nerve decompression landmark, surgical anatomy, Blooms-2 |
| 7fcf106f | Neurotology | Hennebert's sign / false positive fistula test in congenital syphilis, PYQ-adjacent, Blooms-2 |
| c76c47b2 | Diseases of the Ear | Otosclerosis via PTA vignette, Blooms-4, difficulty-3, benchmark standard |
| 2b48fb56 | Neurotology | Glomus jugulare EXCEPT, anatomical nuance, Blooms-4, difficulty-3 |
| db287647 | Diseases of the Pharynx | Peritonsillar abscess clinical vignette, PYQ-tagged, Blooms-3 |
| 695b9ca3 | Diseases of the Ear | Pars flaccida perforation → cholesteatoma → surgical decision, Blooms-3 |
| eb32ab5c | Rhinology | Intranasal steroids in nasal polyposis NOT true, Blooms-4, PYQ-tagged |
| 837793bb | Diseases of the Larynx | Paediatric foreign body with stridor, Heimlich manoeuvre, Blooms-3 |
| 85756af0 | ENT Oncology | Laryngeal TB mouse-nibbled appearance, PYQ-tagged, Blooms-2, clean distractors |
| 01d97b73 | Rhinology | Onodi cells and optic nerve injury in FESS, Blooms-2, clinically relevant |
| d05c5fb2 | Diseases of the Pharynx | Oesophageal coin in asymptomatic 3-year-old, management timing, Blooms-3 |
| a916068a | Rhinology | Post-SMR septal perforation, PYQ-tagged (NEET-PG 2024), Blooms-4 — pending image confirmation |
FIX — Targeted edit required before deployment
| Question ID | Issue | Fix Action |
|---|---|---|
| f7878e52 | Pseudo-clinical + incorrect answer key (acoustic neuroma for isolated tinnitus in 70-year-old) | Add unilateral SNHL, imbalance, audiometric rollover to stem; review answer key |
| eb355c2e | "All of the above" as correct answer for most common organism in acute sinusitis | Restructure to single best answer (S. pneumoniae) or reframe as combination question |
| 074d8ee1 | Implausible distractor ("leads to diplopia") trivialises Meniere's EXCEPT question | Replace diplopia distractor with clinically plausible false statement |
| d1db1ec0 | Clinically important red-flag concept (adult unilateral SOM → NPC) presented as bare recall | Add adult male patient vignette; fix difficulty string to numeric |
| 212b4062 | Trotter's triad presented as direct recall | Add clinical stem (CHL, soft palate immobility, trigeminal neuralgia) |
| 12afc64e | T1N0M0 glottic carcinoma treatment as bare recall | Add 55-year-old smoker with hoarseness and normal cord mobility vignette |
| 7c92c13f | Malignant otitis externa options are taxonomic categories, not organisms | Reframe with specific organism options (Pseudomonas, Aspergillus, Staph, Candida) |
| c33cd341 | Internal laterality inconsistency in maxillary sinusitis stem (left tooth pain, right maxilla tapping) | Correct laterality throughout stem |
| bda6ecdd | Rheumatic tonsillitis marked correct over recurrent acute tonsillitis without specifying absolute vs. relative indication | Rewrite to specify absolute indication context |
| 9c2d88f1 | Nasal cycle timing answer key (6–8 hours) contestable against literature | Verify against Dhingra/Scott-Brown; correct or reframe with explicit source |
| e3cf05a2 | Malignant otitis externa management answer ("antibiotic therapy" alone) understates surgical debridement | Revise answer key to reflect combined therapy or reframe to ask about primary pharmacological agent |
| 3a543c04 | CT scan marked as investigation of choice for angiofibroma; MRI is current consensus | Correct answer key to MRI or reframe stem to specify "bony erosion assessment" |
| af9ca007 | Ramsay Hunt syndrome distractor "involves Vth cranial nerve" is factually incorrect as a true statement | Replace Vth nerve distractor with accurate clinical feature |
| 9a0e5d84 | VEMP/saccule fact presented as bare recall | Add suspected SSCD referral scenario to stem |
| 83880782 | Most common extracranial complication of ASOM as bare recall | Add post-auricular swelling clinical lead-in |
| 7f8bd9b2 | NPC prophylactic RT stem conflates prophylactic and therapeutic intent | Rewrite stem to clarify elective nodal irradiation rationale |
| 0f748dca | Cochlear implant question misclassified under Diseases of the Nose | Reclassify to Audiology/Otology; confirm image attachment |
| f4326e02 | PYQ (NEET-PG 2016–17) image absent | Embed confirmed image or rewrite as descriptive vignette |
DISABLE — Remove from active bank
| Question ID | Reason |
|---|---|
| f02712f7 | Corrupt import artefact — options are answer-key line numbers, not clinical choices |
| 47edf02a | Instrument trivia (focal length of head mirror), Blooms-1, misclassified under Neurotology, no PYQ precedent |
| 3058bbcb | Instrument trivia (head mirror applied to right eye), Blooms-1, no clinical utility |
| 861a059b | Tautological stem — answer contained in question ("Nasopharyngeal Carcinoma" → "Nasopharyngeal cavity") |
| d355d352 | Tautological stem — tegmen question answers itself |
| a32a6a5c | Cart Wheel sign in ASOM, Blooms-1, eponym trivia, no PYQ precedent |
| caeeb561 | "Glossoplegia = paralysis of tongue," pure vocabulary definition, Blooms-1 |
| afa14fa8 | Ashley's flap eponym recall, Blooms-1, no PYQ precedent |
| c6ef2253 | Alkaline nasal douche composition, Blooms-1, rote pharmacology trivia |
| 63d04154 | Gutzmann's pressure test, Blooms-1, pure eponym recall |
| a02e3f09 | Kiesselbach's plexus location, Blooms-1, first-year anatomy |
| b2144776 | "Another name for glue ear," Blooms-1, terminology recall |
| 9b3c8d6d | Malignant otitis externa causative organism, Blooms-1 (concept worth testing only in vignette format) |
| 05c6d11f | Rhinoscleroma bacterial/viral/fungal classification, Blooms-1, lowest discriminatory level |
| cc49537c | Commonest complication of CSOM, Blooms-1, rote knowledge |
| f54eea2f | Glycerol test eponym-to-disease mapping, Blooms-1 |
| f44fa7cd | Ciliary transport velocity in antrum (numerical recall), Blooms-1, no PYQ precedent |
| aa8c8d00 | Eagle's syndrome definition, Blooms-1 |
| 416a1d38 | Patey's operation eponym, Blooms-1, no PYQ precedent |
| 7df6c775 | Eustachian tube opens into which part — "Larynx" as distractor, trivially easy |
| ba147723 | Scutum = bony outer attic wall, Blooms-1 |
| 3ea95f6b | Puberphonia definition, Blooms-1 |
| 182ce9fc | Bilateral post-pointing, Blooms-1, no clinical context |
| 16c38192 | Elastic cartilage in epiglottis, Blooms-1, pure anatomy |
| 9d6074a0 | Image-dependent question, image absent, entirely non-functional |
| 8c3efb90 | Image-only question ("Identify the line shown"), image absent, non-functional |
| 4a16a7a0 | Image-dependent hearing aid question, image absent, Blooms-1 even with image |
| 650b45ea | Scala media filled with endolymph, Blooms-1, no clinical hook |
| 3edb8bfe | Nasal myiasis presenting symptom, Blooms-1, factually uncertain |
| 11f3a399 | Acute rhinitis nasal drainage colour, nursing-level question |
| 7bd72c73 | Promontory distance from tympanic membrane (2 mm), pure anatomical measurement |
| 6fae6224 | Laryngomalacia FALSE — direct duplicate of aa961777 |
| 60f49c25 | Gradenigo's triad EXCEPT, Blooms-1, easy (concept worth testing only in clinical vignette) |
| 5d9b5005 | Citelli's angle = sinodural angle, Blooms-1, eponym trivia |
| 84bcdb58 | Woodruff's plexus location, Blooms-1 |
| ffaafbdd | Artery not contributing to Kiesselbach's plexus, Blooms-1 |
| 98be1b16 | Caloric test assesses which structure, Blooms-1 |
| 2208181b | Kallmann syndrome and anosmia, Blooms-1 |
| 80b8b342 | Drug of choice in laryngeal stenosis, Blooms-1 bare pharmacology |
| 5b27f349 | Most common site for salivary calculi, Blooms-1 |
| e99321cd | Plunging ranula = retention cyst, Blooms-1 + factually contested |
| 713f317a | Nodes of Rouvier nomenclature, Blooms-1 |
| 7d842b86 | "What is a myringotomy?" textbook definition, Blooms-1 |
| 2e5239be | Most common salivary gland tumor in children, Blooms-1 + factually ambiguous |
| 303bd178 | Radicular cyst / non-vital tooth — dental concept, misplaced in ENT bank |
| b4da54b4 | "Auditory neurotherapy" — non-standard terminology, factual reliability uncertain |
| d21ba0c1 | Premalignant lesions "All of the above," Blooms-1, near-duplicate of 4aab239f |
| 8946c5bb | Cochlear implant energy conversion, Blooms-1, pure definitional recall |
| 95563a3e | Mel |