Anatomy Question Quality Review

Executive Summary

This report covers a random sample of 200 validated non-gold candidate questions drawn from the Anatomy subject pool of 13,876 items. The sample was analyzed across eight shards of 25 questions each. The findings are synthesized here into a single production-ready quality assessment for the content operations team.

The headline finding is that the Anatomy candidate pool has five distinct, operationally separable quality problems, ranging from outright subject contamination to a pervasive clinical-framing deficit. None of these are isolated incidents; each appears repeatedly across multiple shards and almost certainly scales into the full 13,876-question bank.

Bloom's distribution summary for the candidate sample:

Bloom's Level	Count (candidate 200)
1 (Recall)	113
2 (Comprehension)	70
3 (Application)	13
4 (Analysis)	3
5 (Synthesis/Evaluation)	1

The benchmark gold standard operates almost entirely at Bloom's 3 with clinical vignettes. The candidate pool is inverted relative to that standard: 56.5% of questions are pure recall (Bloom's 1), and only 6.5% reach Bloom's 3 or above. This is the single most consequential structural gap in the subject.

Rough disposition estimate across the 200-question sample:

Disposition	Approximate Count
Keep as-is	~35–40
Fix (vignette upgrade, stem repair, reclassification)	~80–90
Disable (wrong subject, wrong key, irredeemable trivia)	~65–75

The five issue categories, in order of operational priority, are:

Subject Contamination — Non-Anatomy Content in the Anatomy Pool
Factual Errors and Unsafe Answer Keys
Bare-Recall Overload — The Clinical Vignette Deficit
Low-Yield Numerical and Definitional Trivia
Structural Item-Writing Defects

Each is described in detail below.

What Good Looks Like

The benchmark and recent PYQ sets establish a clear quality bar for this subject. The following features define a strong Anatomy question at the Indian PG level:

Clinical anchoring at Bloom's 3. Every benchmark question presents a patient scenario — age, presentation, mechanism, or imaging finding — that forces the candidate to reason from clinical evidence to an anatomical conclusion. The anatomy fact is not asked directly; it is the answer to a clinical problem. See benchmark items 50707582 (tracheal biopsy histology), 1e6ce134 (motorcycle accident, deltoid flattening, C5 root), cf41a988 (stab wound, shoulder shrug loss, posterior triangle), and 02a9b480 (mastectomy, arm swelling, apical axillary nodes). Even the "easy" benchmark items are Bloom's 3.

Distractors that require genuine discrimination. In cf7a1044 (Circle of Willis collateral flow), all four options are anatomically real variants of the circle; the correct answer requires understanding which variant provides adequate cross-flow, not just recognizing a term. Distractors are drawn from the same conceptual neighborhood as the correct answer, not from unrelated topics.

Procedural and surface anatomy with clinical consequence. Items like e3db486e (horizontal fissure on chest X-ray) and 8fde9660 (T4 level posterior mediastinal hematoma) test anatomical knowledge in the context of a clinical decision — where to look, what is injured, what the finding means. This is the dominant format in recent INICET PYQs.

Embryology with developmental mechanism, not just timing. Benchmark item 4683904c (transposition of great arteries) asks for the developmental mechanism (failure of aorticopulmonary septum to spiral), not just the name of the defect. This is the standard for embryology questions.

Acceptable Bloom's 1 questions exist but are anchored to high-yield, exam-validated facts. PYQ items like c5a014a9 (calcium hydroxyapatite in bone matrix), a1c6b42b (Froment's sign, ulnar nerve), and 24686614 (superficial peroneal nerve → peroneus longus and brevis) are pure recall but are justified by their direct PYQ provenance and high exam frequency. Bare recall is tolerable when the fact is genuinely high-yield and exam-tested; it is not tolerable as the default format for the entire pool.

The candidate sample consistently falls short of this bar. The problems are described below.

Main Issue Categories

Issue Category 1: Subject Contamination — Non-Anatomy Content in the Anatomy Pool

Why this pattern is bad

Questions from Pathology, Physiology, Pharmacology, Immunology, Biochemistry, Surgery, Community Medicine, and Dentistry are filed under Anatomy topic tags. This corrupts topic-level analytics, misleads students practicing by topic, and wastes review bandwidth on items that should never have been in this subject. It also inflates the apparent size of the Anatomy bank without adding usable Anatomy content. The contamination is not random noise — it is concentrated under the "Neuroanatomy" topic tag, which appears to function as a catch-all or default label for miscellaneous questions that were not correctly classified at the time of entry.

How it shows up

Across all eight shards, the Neuroanatomy topic tag is the primary vector for contamination. Non-anatomy questions appearing under this tag include:

Pathology/Molecular Biology: p53 tumor suppressor gene (Q-276719b3), Lines of Zahn in thrombus (Q-5f932402), granuloma formation in acute infection (8c943c4a)
Genetics/Biochemistry: human genome gene count (Q-2d98ba17), phosphatidylserine role (e41bb051), Cyclin B / G2-M transition (ae498a96)
Physiology/Pharmacology: endothelin-1 actions (9872cde0), QT prolongation causes (97bd1125)
Immunology/Microbiology: mediators of neutrophils (96a3a674), Nitroblue tetrazolium test (b7f4d5a2)
Community Medicine/Pediatrics: age range for early adolescence (Q-6ab11f57)
Surgery/Vascular: most common site of large bowel ischemia — splenic flexure (55494168), deep burns causing hyperthermia (90c8e57a)
Dentistry/Prosthodontics: carbon marker survey line in prosthodontics (4f597cc0), cowhorn forceps #16 for mandibular third molars (Q885f8232)
Cardiology/Medicine: causes of early systolic murmur (129d2616)
Urology/Surgery: most common presenting sign of bladder carcinoma (Q89519773)

Beyond the Neuroanatomy tag, contamination also appears under Histology (97607e2d — urinary stone types filed as Histology) and Head (Q885f8232 — dental instrument filed as Head anatomy).

The pattern is observed in every shard reviewed. Given that the 200-question sample yields approximately 20–25 clearly out-of-scope questions, the contamination rate in the full 13,876-question pool is likely substantial.

Recommended disposition

Disable all confirmed out-of-scope items from the Anatomy pool. Do not attempt to fix them within Anatomy — they belong in other subjects or should be discarded. Specific disable calls:

Q-276719b3, Q-5f932402, Q-2d98ba17, Q-6ab11f57 (Shard 001)
96a3a674, 129d2616 (Shard 002)
9872cde0, 97bd1125, e41bb051 (Shard 003)
90c8e57a, 55494168, 4f597cc0, 3d87f637 (Shard 005)
b7f4d5a2, 8c943c4a, ae498a96 (Shard 006)
Q89519773, Q885f8232 (Shard 008)

Operational recommendation: Run a full audit of the Neuroanatomy topic tag across the 13,876-question pool. The tag is demonstrably being used as a miscellaneous bucket. Any question under Neuroanatomy that does not test a structure, pathway, or function of the nervous system should be flagged for reclassification or removal. This is a one-time cleanup task with high return on investment.

Issue Category 2: Factual Errors and Unsafe Answer Keys

Why this pattern is bad

A question with a wrong correct answer is worse than no question at all. It actively miseducates candidates, erodes trust in the platform, and — if deployed in a scored assessment — penalizes students who know the correct anatomy. This is the highest-severity quality defect in the sample. Unlike formatting problems or Bloom's level mismatches, a wrong answer key cannot be fixed by a non-expert editor; it requires SME verification against a primary reference (Gray's Anatomy, Snell's Clinical Anatomy, or equivalent).

How it shows up

The reviewed set contains confirmed or strongly suspected factual errors in answer keys across multiple topic areas:

Nerve root and dermatome errors:

Q-84d81034 (Shard 003): Herniated disc inferior to C6 compresses the C7 nerve root. The question marks "lateral surface of digit 5" as the correct dermatome — this is C8, not C7. C7 maps to the middle finger (digit 3). This is a hard factual error.
e4c5bed1 (Shard 007): Pudendal nerve roots listed as S2–S3 (marked correct). Standard teaching is S2–S4. The S4 contribution is clinically significant (perineal sensation, sphincter control) and is tested in PYQs.

Upper limb and regional anatomy errors:

9616532b (Shard 006): Marks "subscapular artery" as the primary blood supply to the mammary gland. The primary supply is the lateral thoracic artery and internal thoracic (mammary) perforators. The subscapular artery is not a primary mammary supplier.
12b9edf9 (Shard 006): Marks "it is the lateral bone of the forearm" as the only true statement about the radius, while simultaneously marking "it is a major contributor to the wrist joint" as false. The radius forms the primary articular surface of the radiocarpal joint — this is a foundational anatomical fact. The answer key is internally contradictory.
2dcff7df (Shard 002): Marks "internal oblique" as forming the anterior wall of the inguinal canal. The anterior wall is formed by the external oblique aponeurosis along its entire length, with the internal oblique contributing only to the lateral third. Marking internal oblique alone as correct is misleading at best and wrong at worst.

Abdominal anatomy errors:

19efdd36 (Shard 002): An EXCEPT question where the marked correct answer (appendices epiploicae are seen on the external surface of the large intestine) is actually a true feature of the large intestine — it cannot be the exception. The answer key is inverted.
0ff7d348 (Shard 002): Marks T12 vertebra as the posterior boundary of the epiploic foramen. The posterior boundary is the inferior vena cava covered by peritoneum, not a vertebral body.

Neuroanatomy errors:

Q33036fcc (Shard 008): Marks the thalamostriate vein as forming the floor of the body of the lateral ventricle. The floor is formed by the thalamus medially and the body of the caudate nucleus laterally; the thalamostriate vein runs in the groove between them and is not a structural floor component.

Sensory receptor error:

4b8f035c (Shard 005): Marks Meissner's corpuscles as the sensory receptor in the epidermis. Meissner's corpuscles are located in the papillary dermis, not the epidermis. Merkel discs (Merkel's tactile discs) are the correct epidermal mechanoreceptors.

Ossicle/temporal bone confusion:

76acbf07 (Shard 007): Asks which ossicle is "not present at birth" and marks "petrous part of temporal bone" as correct. The petrous temporal bone is not an ossicle. The question conflates ossicular development with temporal bone ossification.

Recommended disposition

Disable all confirmed wrong-key items immediately pending SME correction. Do not deploy in any scored or practice context until the answer key has been verified against a primary reference. Specific disable calls:

Q-84d81034, 4b8f035c (dermatome/sensory receptor errors — high-frequency exam topics, high risk of student harm)
9616532b, 12b9edf9, 2dcff7df (upper limb/regional anatomy errors)
19efdd36, 0ff7d348 (abdominal anatomy errors)
Q33036fcc (neuroanatomy ventricular anatomy error)
e4c5bed1, 76acbf07 (nerve root and ossicle errors)

Operational recommendation: Dermatome/myotome mapping questions and nerve root level questions are a known high-error-rate category in Indian PG question banks. A targeted SME fact-check pass on all questions in this category — not just the ones identified here — is warranted before any further deployment.

Issue Category 3: Bare-Recall Overload — The Clinical Vignette Deficit

Why this pattern is bad

This is the most pervasive quality gap in the sample, appearing in every shard. The benchmark standard for this subject is unambiguous: even "easy" questions are presented as clinical vignettes at Bloom's 3. The candidate pool is dominated by single-line recall questions — "Which nerve supplies X?", "What is the length of Y?", "Which structure forms the Z wall?" — with no patient context, no clinical reasoning demand, and no discriminating scenario. These questions test whether a candidate has memorized a fact, not whether they can use anatomical knowledge in a clinical situation. INICET and NEET-PG have moved decisively toward applied, scenario-based anatomy; a bank dominated by bare recall is misaligned with the current exam standard.

The Bloom's distribution makes this concrete: 113 of 200 candidate questions (56.5%) are Bloom's 1. Only 13 (6.5%) reach Bloom's 3. The benchmark set is 100% Bloom's 3. This is not a marginal gap — it is a structural inversion.

How it shows up

The pattern is present across all topic areas but is especially dense in:

Neuroanatomy: Q-672b359b (CSF between arachnoid and pia), 9ba0e669 (nerve supply of pupillary sphincter), 89e3bc87 (laryngopharynx nerve supply), 66e23d4b (medial geniculate body function — the stem describes the structure so precisely that the answer is self-evident), 3319e47b (CN III supplies inferior oblique)
Embryology: e496cd84 (amnion surface), 2477e275 (Node of Ranvier definition), 2d54fbaf (embryonic period ends at 8 weeks), b1f1432b (first stage of lung development)
Histology/Microscopic Anatomy: cbb4426c (pseudounipolar neurons), 41c45d5d (collagen type in skin), b831324a (intercalated discs)
Regional anatomy: f22a755d (C1 has no body or spinous process), 458bba04 (posterior vaginal wall length), ca6a5557 (primary function of deltoid)

A secondary variant of this problem — observed in Shards 005 and 008 — is the non-functional clinical vignette: a patient scenario is present in the stem, but the clinical details contribute nothing to the reasoning. In Q98fbcd8c (Shard 005), a 54-year-old unconscious woman is described, but the question simply asks for the afferent level of the biceps reflex — the unconscious state is irrelevant. In Q091109a5 (Shard 008), a patient with vertigo and tinnitus is described, and the answer is the vestibulocochlear nerve — the vignette is just a symptom definition. These questions inflate the apparent Bloom's 3 count without genuine application demand.

What good looks like for this category

The fix is not to add a patient name and age to an existing bare-recall stem. The clinical details must drive the reasoning. Compare:

Bare recall (current): "Which nerve supplies the deltoid muscle?"
Non-functional vignette (also current): "A 30-year-old male presents with shoulder pain. Which nerve supplies the deltoid muscle?"
Benchmark standard: "A 28-year-old male presents following a motorcycle accident with inability to abduct his arm beyond 15 degrees and loss of sensation over the lateral upper arm. The deltoid appears flattened. Which nerve root is most likely injured?" (Benchmark item 1e6ce134)

The benchmark version requires the candidate to integrate motor deficit, sensory loss, and anatomical localization. The bare-recall version requires only memory retrieval.

Recommended disposition

Questions in this category fall into two sub-groups with different remediation paths:

Fix (vignette upgrade): Questions that test a genuinely high-yield, exam-relevant anatomical fact but are currently bare recall. These are worth upgrading because the underlying concept is sound. Examples: d9e21f8d (PCL function — add dashboard injury scenario), 0776f145 (pubofemoral ligament limits abduction — add hip examination scenario), 2177ef6b (branchial cyst — add lateral neck swelling scenario), 38861ff2 (Edinger-Westphal nucleus — add posterior communicating artery aneurysm scenario), 8cc317d7 (oculomotor nerve and intracranial aneurysm — add ptosis/mydriasis scenario), c5c8385c (median nerve at wrist — add laceration scenario). The upgrade rule: the clinical details must make the anatomy the answer to a clinical question, not just decoration.

Disable: Questions that test a low-yield or trivially easy fact that, even with a vignette, would not reach the quality bar for PG-level assessment. Examples: e496cd84, f22a755d, 9ba0e669, 89e3bc87, 3319e47b, 2d54fbaf, de5a1a27. If strong gold-standard coverage already exists for the underlying concept (e.g., CN III palsy is covered by benchmark item 1e6ce134), prefer disable over speculative rewrite.

Issue Category 4: Low-Yield Numerical and Definitional Trivia

Why this pattern is bad

A distinct sub-cluster within the bare-recall problem deserves its own category because it implies a different remediation path. These are questions that ask for a specific measurement, count, or textbook definition — the length of the Eustachian tube, the capacity of the renal pelvis, the length of the distal convoluted tubule, the length of the vaginal wall — with no clinical anchoring and no presence in any recent PYQ or benchmark set. Even if a clinical vignette were added, the underlying fact is not the kind of knowledge that drives clinical decisions or that INICET/NEET-PG tests. These questions should be disabled, not fixed, because the concept itself is not worth a question slot at this level.

This is operationally distinct from the vignette-upgrade category (Issue 3) because the problem is not the format — it is the choice of fact. A question asking for the length of the cartilaginous Eustachian tube (25 mm) cannot be made into a good PG question by adding a patient scenario, because no clinical decision hinges on knowing that specific measurement.

How it shows up

Observed in this sample:

3eee6595 (Shard 003): Length of the cartilaginous Eustachian tube — 25 mm. No PYQ provenance, no clinical relevance.
9dd065d9 (Shard 004): Length of the distal convoluted tubule — 5 mm. Distractors (2, 8, 12 mm) are arbitrary. Not tested in any major exam.
458bba04 (Shard 004): Posterior vaginal wall length vs. anterior. Bare anatomical measurement, no clinical context.
8df17827 (Shard 008): Normal renal pelvis capacity — 7 ml. No clinical relevance to PG decision-making.
2477e275 (Shard 004): Node of Ranvier — textbook definition question. One distractor ("indentation of a Schwann cell due to synapsing of an adjacent unipolar neuron") is anatomically nonsensical, indicating poor distractor construction.
66e23d4b (Shard 004): Medial geniculate body function — the stem describes the structure so precisely that the answer is self-evident from the question itself. This is a self-answering question, a specific form of definitional trivia.

The pattern is less dense than the vignette-deficit problem but is consistently present across shards. It likely reflects a historical question-writing practice of converting textbook tables (measurements, capacities, lengths) directly into MCQs without asking whether the fact is exam-relevant.

Recommended disposition

Disable all confirmed numerical trivia and self-answering definitional questions. Do not attempt to fix by adding a vignette — the underlying fact is not worth a question slot. Specific disable calls: 3eee6595, 9dd065d9, 458bba04, 8df17827, 2477e275, 66e23d4b.

Operational recommendation: Any question whose stem or correct answer contains a specific anatomical measurement (in mm, ml, cm) with no clinical decision attached should be flagged for SME review before deployment. This is a filterable pattern.

Issue Category 5: Structural Item-Writing Defects

Why this pattern is bad

This category covers questions where the anatomy content may be correct but the item construction is broken in a way that makes the question unfair, uninterpretable, or actively misleading. These defects are operationally distinct from factual errors (Issue 2) and from the vignette deficit (Issue 3) because they require different fixes: stem rewriting, distractor replacement, or image verification — not just fact-checking or vignette addition. Several sub-types appear in this sample.

How it shows up

Sub-type A: Image-dependent stems without a functional image

Questions that reference a visual element ("the point marked X," "the artery shown here," "identify the structure labeled") are non-functional as standalone text if the image is absent or unverified. In this sample:

Q-3e9fd2ac (Shard 001): References "the point marked $X$" in a testis descent diagram. The image is absent. The stem also embeds "Recent NEET Pattern 2016-17" as in-text attribution — exam source citations must not appear in question stems.
Q-46b90d3e (Shard 001): References "the artery shown here" for paracentesis. Image dependency unverified.

These questions cannot be evaluated for correctness without the image. They must be verified to have a functional, attached image before deployment. If the image cannot be confirmed, disable.

Sub-type B: Compound-answer items (two facts bundled in one correct option)

cbe25703 (Shard 007): CSF rhinorrhea question where the correct option bundles both "which structure is fractured" AND "what is the clinical risk" into a single answer. This tests two facts simultaneously; a candidate who knows one but not the other cannot be correctly assessed. The item-writing rule is one testable point per question.

Sub-type C: Cross-nerve distractor contamination

Several questions use distractors borrowed from a different nerve injury syndrome, which can actively mislead well-prepared candidates:

c5c8385c (Shard 006): Median nerve at wrist question includes "Policeman's tip deformity" as a distractor — this is a sign of Erb's palsy (upper trunk brachial plexus injury), not median nerve injury. A candidate who correctly knows Policeman's tip is not a median nerve finding may be confused by its presence as a distractor in this context.
4a068300 (Shard 006): Hypoglossal LMN lesion question includes "taste would be lost" as a distractor — taste is a CN VII/IX function, not CN XII. While this is technically a plausible wrong answer, it should be phrased carefully to avoid implying CN XII has taste function.

Sub-type D: "All of the above" and "None of the above" as correct answers or distractors

64a3e7e5 (Shard 002): "All of the above" is the correct answer, and the option set is internally redundant (the same fact restated at two vertebral levels). This format is poor exam design and is explicitly discouraged in modern item-writing guidelines.
ba28a1ec (Shard 003): "None of the options" is used as a distractor — a non-discriminatory option that adds no information.
27518901 (Shard 006): "All of the above" is a distractor that is arguably defensible as correct, undermining the single correct answer.

Sub-type E: Bloom's level mislabeling

330e49d7 (Shard 006): Tagged Bloom's 5 (Evaluate/Create) for a straightforward recall question about transversalis fascia and inguinal hernia. This is Bloom's 1–2 at most.
85ea0714 (Shard 005): Tagged Bloom's 3 for a pure recall question (optic vesicle at 30 days IUL).
98fbcd8c (Shard 005): Tagged Bloom's 3 with a clinical vignette that adds no reasoning demand — the vignette is decorative.

Bloom's mislabeling is operationally significant because it corrupts test assembly logic and difficulty calibration. A test built to include a certain proportion of Bloom's 3 questions will be miscalibrated if many of those questions are actually Bloom's 1.

Sub-type F: Exam source citation embedded in stem

Q-3e9fd2ac (Shard 001): The stem includes "Recent NEET Pattern 2016-17" as in-text attribution. Exam source citations must never appear in question stems — they reveal the answer's provenance, bias candidate reasoning, and create copyright exposure.

Recommended disposition

Image-dependent questions (Q-3e9fd2ac, Q-46b90d3e): Verify image attachment before any deployment. If image cannot be confirmed functional, disable. Remove exam source citation from Q-3e9fd2ac stem regardless.
Compound-answer items (cbe25703): Fix by splitting into a single testable point. Keep the concept — it is clinically strong.
Cross-nerve distractor contamination (c5c8385c, 4a068300): Fix by replacing contaminating distractors with options drawn from the same nerve injury spectrum.
"All/None of the above" format (64a3e7e5, ba28a1ec, 27518901): Fix by replacing with specific anatomical alternatives. "All of the above" and "None of the above" should be retired as answer options across the subject.
Bloom's mislabeling (330e49d7, 85ea0714, 98fbcd8c): Fix Bloom's tags as part of a subject-level recalibration pass. Rule of thumb: no question without a clinical scenario or multi-step reasoning demand should be tagged above Bloom's 2.

Prioritization

The five issue categories are not equal in urgency. The following priority order is recommended for the content operations team:

Priority 1 — Immediate action required (before any deployment): Issue Category 2 (Factual Errors and Unsafe Answer Keys). Questions with wrong correct answers must be disabled immediately. They are actively harmful in any scored or practice context. The confirmed list is short (approximately 10–12 items in this sample) but the impact per item is high. A targeted SME sweep of dermatome/nerve root questions and upper limb vascular anatomy questions is the highest-return immediate action.

Priority 2 — High volume, high impact (next sprint): Issue Category 1 (Subject Contamination). Approximately 20–25 questions in this 200-item sample are from the wrong subject entirely. Scaled to 13,876 questions, this likely represents hundreds of contaminating items. The Neuroanatomy topic tag audit is the single highest-leverage cleanup task in the subject. It is largely mechanical (does this question test a nervous system structure or function?) and can be executed by a trained content reviewer without SME involvement for most cases.

Priority 3 — Structural quality improvement (ongoing): Issue Category 3 (Bare-Recall Overload). This is the largest category by volume and the most consequential for exam alignment. It cannot be fixed in a single sprint — it requires a systematic vignette-upgrade pass across the pool. The recommended approach is to prioritize upgrade of questions on high-frequency exam topics (nerve injuries, brachial plexus, triangles of neck, pelvic floor, embryological derivatives) where the underlying concept is strong but the format is weak. Low-yield bare-recall items on the same topics should be disabled rather than upgraded, to avoid redundancy.

Priority 4 — Targeted disable pass (can be batched): Issue Category 4 (Numerical and Definitional Trivia). These are easy to identify (measurement in mm/ml with no clinical decision attached, or self-answering definitional stems) and should be disabled in a single batch pass. Volume is moderate (estimated 15–20 items in the full sample).

Priority 5 — Item-writing quality pass (lower urgency, higher craft): Issue Category 5 (Structural Item-Writing Defects). Sub-types B, C, D, and E (compound answers, cross-nerve distractors, all/none options, Bloom's mislabeling) require individual item-level attention and are lower urgency than the above. Sub-type A (image dependency) and Sub-type F (exam source citation in stem) should be addressed in Priority 2 alongside the contamination audit, since they are quick to identify and fix.

Example Keep / Fix / Disable Calls

The following table summarizes representative disposition calls drawn from the reviewed sample. These are illustrative of the patterns described above, not an exhaustive list.

Question ID	Shard	Topic	Issue	Disposition	Rationale
`fb438a6f`	002	Back	None significant	Keep	Clinical vignette, cervical disc herniation, Bloom's 3, matches benchmark style
`d2368019`	002	Abdomen	Minor stem polish needed	Keep	Seminoma lymphatic drainage, clinical vignette, correct answer, good distractors
`a2e29669`	007	Lower Limb	None significant	Keep	Obturator nerve compression by bladder carcinoma, Bloom's 3, clean vignette
`ba0988f7`	004	Thorax	None significant	Keep	Tension pneumothorax needle insertion, applied intercostal anatomy, Bloom's 3
`25938e23`	007	Thorax	Minor stem polish	Keep	Pulmonary vein tear, pericardial layers, trauma scenario, discriminating distractors
`Q5cc7612d`	008	Pelvis	None significant	Keep	Blood supply of rectum NOT true, PYQ-validated, plausible distractors
`457aad4d`	007	Neuroanatomy	Add vignette to reach Bloom's 3	Keep / Fix	Lateral medullary syndrome / PICA, high-yield, factually sound
`db483268`	001	Neck	Minor stem polish	Keep	Parathyroid embryology, 3rd vs 4th pouch, plausible distractors, Bloom's 2
`cbe25703`	007	Clinical Anatomy	Compound-answer flaw	Fix	CSF rhinorrhea / cribriform plate — split into single testable point
`c5c8385c`	006	Upper Limb	Cross-nerve distractor contamination	Fix	Median nerve at wrist — replace Policeman's tip distractor with median nerve spectrum option
`d9e21f8d`	002	Lower Limb	Bare recall — add vignette	Fix	PCL function — add dashboard injury scenario to reach Bloom's 3
`38861ff2`	004	Neuroanatomy	Bare recall — add vignette	Fix	Edinger-Westphal nucleus — add PComm aneurysm scenario
`330e49d7`	006	Abdomen	Bloom's 5 mislabeled; bare recall	Fix	Transversalis fascia / inguinal hernia — correct Bloom's tag to 1–2, add clinical scenario
`Q-3e9fd2ac`	001	Embryology	Image absent; exam citation in stem	Fix / Disable	Testis descent diagram — verify image; remove "Recent NEET Pattern 2016-17" from stem; if image unconfirmable, disable
`e4c5bed1`	007	Neuroanatomy	Wrong nerve root (S2–S3 vs S2–S4)	Disable (pending SME correction)	Pudendal nerve roots — factual error in correct answer
`Q-84d81034`	003	Neuroanatomy	Wrong dermatome (digit 5 vs digit 3 for C7)	Disable (pending SME correction)	C7 dermatome error — high-frequency exam topic, high harm potential
`9616532b`	006	Upper Limb	Wrong answer key (mammary supply)	Disable (pending SME correction)	Subscapular artery not primary mammary supplier
`19efdd36`	002	Abdomen	Inverted answer key	Disable (pending correction)	Appendices epiploicae EXCEPT — answer key is factually inverted
`Q33036fcc`	008	Neuroanatomy	Wrong answer key (ventricular floor)	Disable (pending SME correction)	Thalamostriate vein is not the floor of lateral ventricle
`4b8f035c`	005	Histology	Wrong answer key (Meissner's vs Merkel)	Disable (pending correction)	Meissner's corpuscles are in dermis, not epidermis
`Q-276719b3`	001	Neuroanatomy (mislabeled)	Wrong subject — Pathology	Disable	p53 tumor suppressor — not anatomy
`9872cde0`	003	Neuroanatomy (mislabeled)	Wrong subject — Physiology	Disable	Endothelin-1 actions — not anatomy
`4f597cc0`	005	Neuroanatomy (mislabeled)	Wrong subject — Dentistry	Disable	Carbon marker survey line — prosthodontics, not anatomy
`ae498a96`	006	Neuroanatomy (mislabeled)	Wrong subject — Cell Biology	Disable	Cyclin B / G2-M transition — not anatomy
`9dd065d9`	004	Microscopic Anatomy	Numerical trivia	Disable	DCT length 5 mm — not exam-relevant, arbitrary distractors
`8df17827`	008	Abdomen	Numerical trivia	Disable	Renal pelvis capacity 7 ml — not exam-relevant
`66e23d4b`	004	Neuroanatomy	Self-answering stem	Disable	Medial geniculate body — stem describes the answer
`8ab3886b`	002	Head	Fabricated distractor language	Disable	"Epitheliocytipilon" is a nonsense term; item is unreliable
`27518901`	006	Head	Ambiguous stem, defensible "all of the above"	Disable	Tongue anomaly question — no single defensible answer
`e496cd84`	007	Embryology	Bare recall, below PG threshold	Disable	Amnion surface — trivially easy, no clinical framing, no tags

This report covers the 200-question candidate sample analyzed across eight shards. Findings should be interpreted as representative of patterns in the sample; the full 13,876-question pool will require systematic tooling to apply these categories at scale. The five issue categories identified here — subject contamination, wrong answer keys, bare-recall overload, numerical trivia, and structural item-writing defects — are mutually distinct in their remediation paths and should be tracked as separate work streams.