Home / Pharmacology
Verified packet scope

This published report is grounded in a randomized packet from a bank of 14472 questions: 200 validated generic candidates, 0 validated risky candidates, and 20 gold-reference items (8 benchmark, 12 PYQ), for 220 sampled items total.

Benchmarked against 8 benchmark questions and 12 recent PYQs.

Pharmacology Question Quality Review

Executive Summary

This report covers a random sample of 200 validated non-gold candidate questions from the Pharmacology pool of 14,472 items. The sample was analyzed across eight shards of 25 questions each. The findings are consistent and mutually reinforcing across shards: the candidate set is dominated by low-cognitive-demand recall items that fall well short of the benchmark standard, and a meaningful minority carry outright factual errors in their answer keys or option text.

The Blooms distribution tells the story plainly. Of the 200 candidate questions, 80 are Blooms-1 and 100 are Blooms-2, leaving only 20 questions at Blooms-3 or above. The eight benchmark items are uniformly Blooms-3 or Blooms-4, all embedded in clinical vignettes with multi-step reasoning. The gap between the candidate pool and the benchmark is not marginal — it is structural.

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

  1. Answer Key and Factual Accuracy Failures — a recurring pattern of wrong keys, inverted correct answers, and clinically incorrect statements marked as true, affecting an estimated 8–12% of the reviewed set. These require expert pharmacologist review before any deployment.
  2. Recall Overload: Blooms-1/2 Items Without Clinical Anchoring — the dominant volume problem, affecting roughly 70–75% of the candidate set. Most of these are not fixable by light editing; they require either vignette conversion or disabling.
  3. Underdeveloped Clinical Vignettes — a smaller but salvageable cluster of questions that attempt a clinical frame but lack the data density, decision node, or reasoning demand to reach benchmark standard. These are the highest-priority fix targets.
  4. Topic Mislabelling and Scope Contamination — a cross-cutting metadata problem that distorts topic-level analytics and occasionally surfaces questions from dental, nursing, or non-INICET/NEET-PG scopes.
  5. Outdated and Superseded Clinical Content — a content-currency risk concentrated in antimicrobials, TB regimens, and withdrawn drugs, where the factual basis of the question has changed since authoring.

Of the 200 questions reviewed, the estimated disposition is approximately: Keep as-is: ~25–30, Fix (vignette upgrade or key correction): ~50–60, Disable: ~110–120. The disable rate is high but justified — the pool is large enough that disabling low-quality items will not create coverage gaps, and the benchmark standard cannot be met by cosmetic edits to Blooms-1 recall items.

What Good Looks Like

The eight benchmark items and the twelve recent PYQs define the quality bar. Several features distinguish them from the candidate set and should be used as the operational standard for all keep/fix/disable decisions.

Clinical vignettes with decision-relevant data. Every benchmark item embeds the pharmacological concept inside a patient scenario that includes age, diagnosis, relevant comorbidities, and at least one quantitative data point (eGFR, serum potassium, blood pressure, drug level). The scenario is not decorative — it changes which answer is correct. For example, benchmark item c0d04d50 requires the candidate to calculate a half-life adjustment using eGFR 25 mL/min and the fraction renally excreted; the clinical details are load-bearing. Contrast this with the majority of candidate items, where a patient age and diagnosis are appended to a question that could be answered from the drug name alone.

Multi-step reasoning, not single-fact retrieval. Benchmark items require at least two inferential steps. In 3dafce52, the candidate must recognize that adding spironolactone to an ACE inhibitor in a patient with serum potassium already at 5.8 mEq/L creates a hyperkalemia risk, then identify the correct management response. In a8317e77, the candidate must know that phenytoin follows zero-order kinetics at therapeutic concentrations and apply that to explain a disproportionate rise in drug level. Neither question is answerable by recall alone.

Distractors that represent genuine clinical alternatives. In good items, wrong options are not obviously wrong — they are drugs or actions that would be correct in a slightly different clinical context. In 5a5f0f9e, febuxostat and sodium bicarbonate are both used in hyperuricemia management; the question tests whether the candidate understands why rasburicase is superior specifically for tumor lysis syndrome prophylaxis. In c6ad5785, losartan and ramipril are both antihypertensives; the question tests knowledge of clonidine's specific withdrawal syndrome.

Factual precision in keys and options. PYQ items like ebc88ca0 (sucralfate-phenytoin interaction, 120-minute separation) and 709c631b-ad2d (first-order kinetics, half-life unchanged when dose doubles) are precise enough that a well-prepared candidate can verify the answer from first principles. There is no ambiguity about what is being tested or why the correct answer is correct.

Appropriate use of Blooms-1 items when they are genuinely high-yield. PYQ items like 4397cf28 (clofazimine and skin hyperpigmentation) and 688ad586 (carbetocin dose) demonstrate that Blooms-1 questions are acceptable when they test a specific, frequently examined, and clinically consequential fact — not when they test a fact that any first-year student would know or that is deducible from the drug name.

Main Issue Categories

1. Answer Key and Factual Accuracy Failures

Why this pattern is bad

A wrong answer key is the most dangerous defect in a question bank. It actively miseducates candidates, penalizes well-prepared students, and — if deployed in a scored test — produces invalid results. Unlike a low-quality question that merely wastes a slot, a wrong-keyed question causes harm. In pharmacology specifically, where the correct answer often has direct patient-safety implications (drug contraindications, toxicity management, mechanism-based interactions), factual errors in keys are not acceptable at any Blooms level.

This pattern appears as a recurring cluster across the reviewed set, not as isolated incidents. At least 10–12 questions across the eight shards carry confirmed or highly probable keying errors or clinically incorrect statements marked as correct. The errors are not uniformly obvious — several require expert pharmacological knowledge to detect, which means they are unlikely to be caught by non-specialist editorial review.

How it shows up

The errors fall into three sub-types:

Inverted correct answer: The factually correct statement is listed as a distractor and the factually incorrect statement is keyed as correct. This is the most dangerous sub-type.

Imprecise or incomplete correct answer: The keyed answer is partially true but omits the primary mechanism or overstates a secondary effect, misleading candidates about the pharmacological principle being tested.

Clinically incorrect correct answer: The keyed answer describes a clinical use, contraindication, or effect that is wrong by current standards or by standard pharmacology references.

Example question IDs with short explanations

  • d2d6c7a8 (Shard 002): Vigabatrin mechanism marked as "GABA-agonism." Vigabatrin is an irreversible inhibitor of GABA transaminase (GABA-T); it is not a GABA receptor agonist. The distinction is pharmacologically fundamental and clinically relevant (it explains the drug's irreversible visual field toxicity profile). The correct answer as keyed is wrong.

  • 0c9ffb52 (Shard 002): Stem asks about a "muscle relaxant with ganglion-blocking action" and marks halothane as correct. Halothane is a volatile inhalational anaesthetic, not a muscle relaxant. Trimethaphan — the actual ganglion blocker — is listed as a distractor. Both the stem framing and the key are wrong.

  • bcd6bcc7 (Shard 004): The stated correct answer ("Thiazides act by inhibiting sodium-chloride co-transporter") is factually TRUE, not an exception. The question asks for the FALSE statement; the key is inverted.

  • 89ad1340 (Shard 003): Mafenide properties question marks "Can penetrate eschars" as the incorrect option. Eschar penetration is Mafenide's defining pharmacological advantage over silver sulphadiazine and is explicitly stated as such in every standard pharmacology text. The key is inverted.

  • e6523f6f (Shard 005): Marks "Esmolol is used in left ventricular decompensation" as a true statement. Esmolol is a short-acting beta-1 blocker used in SVT and perioperative hypertension; it is generally contraindicated in decompensated heart failure. This is a clinically dangerous error.

  • Q72b7cbca (Shard 005): Marks "around the umbilicus" as NOT a preferred insulin injection site. The periumbilical abdomen is in fact a standard and commonly recommended insulin injection site. The key is factually inverted.

  • 546837ea (Shard 008): Marks "Digitalis" as the cause of peripheral neuropathy. The canonical answer is isoniazid (and thalidomide is a strong secondary answer). Digitalis does not cause peripheral neuropathy. Clear keying error.

  • 5f1e3e61 (Shard 008): Marks "It is administered orally" as a true statement about penicillin G. Penicillin G is acid-labile and is not reliably administered orally; penicillin V is the oral form. The keyed answer is factually wrong.

  • 51c0bd9e (Shard 004): Marks "increased peripheral glucose uptake" as the primary mechanism of metformin. The primary mechanism is inhibition of hepatic gluconeogenesis via AMPK activation. Peripheral glucose uptake is a secondary effect. The answer as keyed will mislead candidates about a high-frequency exam topic.

  • 727f72c5 (Shard 006): Marks "weight gain" as the most common SSRI side effect. Nausea and diarrhea are the most common early side effects per standard pharmacology texts; weight gain is a long-term effect. The key is contextually wrong without qualification.

  • 0df9f009 (Shard 003): Silver sulphadiazine marked as providing "broad-spectrum coverage against both Pseudomonas and MRSA." Silver sulphadiazine has limited MRSA activity; the MRSA claim overstates its spectrum and conflates it with other topical agents.

Recommended disposition

All items in this category: Fix (key correction) before any deployment. Items where the fix requires restructuring the entire option set (e.g., 0c9ffb52, bcd6bcc7) should be treated as Disable unless a pharmacology subject-matter expert can verify and rebuild the question. Items with a single correctable key error (546837ea, 5f1e3e61) should be fixed with expert review and re-verified before use. No item in this category should be deployed in its current state.

2. Recall Overload: Blooms-1/2 Items Without Clinical Anchoring

Why this pattern is bad

This is the dominant volume problem in the candidate set. Approximately 70–75% of the 200 reviewed questions are single-fact recall or simple mechanism questions with no clinical context. At PG entrance level (INICET, NEET-PG), the exam tests whether a candidate can apply pharmacological knowledge to a clinical decision — not whether they can retrieve a fact from memory. A question bank that is 75% Blooms-1/2 recall will not prepare candidates for the actual exam and will not discriminate between candidates who understand pharmacology and those who have merely memorized lists.

The problem is compounded by the fact that many of these questions test facts that are either too basic (known to any first-year MBBS student), too peripheral (device specifications, administrative definitions, dental procedures), or too easily deducible from the drug name or class to have any discriminatory value.

This pattern appears as a broad, pervasive pattern across all eight shards, not as a cluster in any particular topic area. It affects Cardiovascular, Antimicrobial, CNS, Endocrine, Autonomic, and General Pharmacology topics equally.

How it shows up

The pattern manifests in several recognizable forms:

Bare mechanism questions: "What is the mechanism of action of X?" with no clinical context. Examples: a5c097eb (rosiglitazone → PPAR-gamma), fec6012c (chloramphenicol → protein synthesis), 3f721804 (aspirin → cyclooxygenase), b04e1d7a (beta-2 agonist effect), 0cc758ab (atropine effect).

Drug classification recall: "X belongs to which class?" or "Which of the following is a Y?" Examples: 50cf6d02 (Cidofovir is an antiviral), cd99abea (Enfuvirtide is a fusion inhibitor), a564fed6 (Ifosfamide class), 20756ce9 (Clevidipine class), f31decb1 (Baclofen class).

Drug of choice without clinical context: "DOC for condition X is?" Examples: 4343754b (Mycoplasma DOC), 0fd4b5e0 (ethosuximide for absence seizures), b432f5a7 (pregabalin for diabetic neuropathy).

Single adverse effect recall: "Which drug causes X?" or "X is a side effect of which drug?" with no patient scenario. Examples: 3707452f (hepatotoxic antineoplastic), 4c30e2cc (drugs causing hyperprolactinemia EXCEPT), 2d181dc3 (alcohol and hyperpyrexia).

Definitional recall: "What is the definition of X?" Examples: 58d74b3d (counterfeit drug definition), 247fb7d7 (pharmacodynamics definition), 1b61598e (renal clearance definition).

Example question IDs with short explanations

  • 3f721804: "Aspirin inhibits which enzyme?" — COX inhibition is first-year MBBS knowledge. Zero discriminatory value at PG level. The concept is worth testing only in a clinical context (e.g., why does aspirin cause irreversible platelet inhibition while ibuprofen does not?).

  • e928ec9d: "Which antidepressant is an SSRI?" with fluoxetine as the answer. Answerable by anyone who has heard the drug name. Flagged easy, Blooms-1.

  • cd99abea: "Enfuvirtide is a fusion inhibitor." Distractors are from entirely different antiretroviral classes, making elimination trivial. No reasoning required.

  • 676b3cdc: "Tranexamic acid is an antifibrinolytic." Distractors are thrombolytics and anticoagulants — eliminable by category alone.

  • 3b7a9dd4: "Cyclosporin is used as what type of drug?" Answer: immunosuppressant. Pre-MBBS level knowledge.

  • 1a803988: "Which neurotransmitter is associated with parasympathetic outflow?" Answer given as "Cholinergic" — which is a receptor type, not a neurotransmitter. Factually imprecise and Blooms-1.

  • bf0c86f5: Shortest insulin pen needle length in millimetres. Device/equipment trivia with no pharmacological reasoning content.

  • 58d74b3d: Definition of a counterfeit drug. Administrative/regulatory knowledge, not pharmacology. All three wrong options are partially correct, making "All of the above" the answer by elimination.

  • eefd1ed6: Preferred site for intramuscular injection. Nursing/basic sciences level, not appropriate for pharmacology PG assessment.

Recommended disposition

The majority of items in this category: Disable. The pool is large enough (14,472 questions) that disabling low-yield recall items will not create coverage gaps. For a small subset where the underlying concept is genuinely high-yield and the question is structurally sound except for the absence of a clinical frame, a Fix (vignette conversion) is appropriate — but only where the concept cannot be adequately covered by an existing benchmark or PYQ item. Light editing (adding a patient age and diagnosis) is not sufficient; a genuine vignette requires a clinical scenario that changes the reasoning path.

3. Underdeveloped Clinical Vignettes

Why this pattern is bad

This category is distinct from pure recall items and deserves separate treatment because the remediation path is different. These questions have a clinical scenario — a patient, a diagnosis, sometimes a drug — but the scenario is decorative rather than functional. The answer is retrievable from the drug name or a single fact without engaging with the clinical details. They are Blooms-1 questions wearing a Blooms-3 costume, and they are mislabelled accordingly in the metadata.

This matters operationally because these items are more likely to pass editorial review (they look like vignettes), they inflate the apparent Blooms-3 count in the pool, and they give candidates false practice at clinical reasoning. They are also the most fixable category — the clinical frame already exists and needs to be made load-bearing rather than decorative.

This pattern appears as a narrower cluster, affecting perhaps 15–20% of the reviewed set. It is concentrated in Cardiovascular, Antimicrobial, and CNS pharmacology topics.

How it shows up

The typical pattern is: a one-sentence patient description (age, diagnosis, drug name) followed by a question that is answerable from the drug name alone. The clinical details — comorbidities, lab values, timeline, contraindications — that would make the scenario a genuine reasoning task are absent.

A secondary form is the "thin vignette" where a clinical scenario is present but the question asks for a mechanism or classification rather than a management decision. The scenario implies a reasoning task but the question reduces to recall.

Example question IDs with short explanations

  • 70ccf38c (Shard 004): A septic shock patient on dopamine with epinephrine added — the scenario implies haemodynamic reasoning, but the question asks only "which receptor mediates vasoconstriction?" The clinical details are irrelevant to answering the question. A genuine Blooms-3 version would ask about the expected haemodynamic consequence of adding epinephrine or about dose-dependent receptor selectivity.

  • 61d35a2f (Shard 007): Organophosphate poisoning patient — the answer (atropine) is immediately obvious from the diagnosis. The distractors (neostigmine, physostigmine) are meaningful foils, but the question does not require the candidate to reason about dose, monitoring, or the role of pralidoxime. Acceptable as a medium-difficulty item but not at benchmark level.

  • 7284d46a (Shard 007): Ketoconazole + terfenadine interaction — a brief scenario is present but the answer is immediately obvious from the drug pair. No reasoning about mechanism, consequence, or management is required.

  • f511c1f1 (Shard 001): Clomipramine + escitalopram → serotonin syndrome monitoring. The drug interaction is correctly identified and the answer is unambiguous, but the question lacks patient details (age, presenting symptoms, timeline) that would make it a genuine clinical reasoning task. This is the closest item in Shard 001 to benchmark quality and is the best candidate for vignette upgrade.

  • afa6069b (Shard 002): ACE inhibitor in heart failure — the clinical frame is present but the question tests a pure renal physiology mechanism rather than a management decision. Adding serum creatinine, potassium, and a question about monitoring or dose adjustment would bring it to benchmark standard.

  • aaa0e94c (Shard 004): COPD maintenance therapy — the vignette correctly contextualises the drug class choice and the distractors are educationally meaningful, but the answer (salmeterol) is straightforward from the diagnosis alone. A genuine Blooms-3 version would add a comorbidity (e.g., concurrent cardiovascular disease, or a patient already on tiotropium) that requires a more nuanced selection.

  • 9f415153 (Shard 008): Clinical vignette present but the diagnosis is unambiguous from the stem alone; the question resolves at pattern recognition rather than management decision.

Recommended disposition

Fix (vignette upgrade) for items where the underlying concept is high-yield and the clinical frame is already partially constructed. The fix requires: (a) adding at least one quantitative data point (lab value, vital sign, drug level) that is relevant to the answer, (b) reframing the question as a management decision or consequence prediction rather than a mechanism recall, and (c) ensuring that the clinical details change which answer is correct. Items where the concept is low-yield or already covered by a stronger benchmark/PYQ item should be Disabled rather than upgraded.

4. Topic Mislabelling and Scope Contamination

Why this pattern is bad

Topic mislabelling has two distinct operational consequences. First, it distorts topic-level analytics — if drug interaction questions are filed under Pharmacokinetics, or CNS drugs are filed under Autonomic Nervous System, then topic-level quality reports and coverage gap analyses will be wrong. Second, it affects candidate experience in topic-specific practice sets, where a question about HAART ocular side effects appearing in an Analgesics set is confusing and undermines trust in the platform.

Scope contamination — questions targeting dental, veterinary, or non-INICET/NEET-PG audiences — is a related but distinct problem. These questions are not merely mislabelled; they are wrong-audience items that should not exist in a medical PG preparation pool at all.

This pattern appears as a moderate-frequency cluster across the reviewed set, with mislabelling observed in at least 8–10 questions and scope contamination in at least 2–3 questions in the 200-question sample. Given the pool size of 14,472, the absolute number of affected items is likely substantial.

How it shows up

Topic mislabelling: Drug interaction questions filed under Pharmacokinetics; CNS drugs filed under Autonomic Nervous System; pregnancy pharmacology filed under Endocrine Pharmacology; antiepileptic drug-specific facts filed under Pharmacokinetics and Pharmacodynamics.

Scope contamination: Questions targeting dental procedures, nursing protocols, or veterinary pharmacology appearing in a medical PG pool with no tags, no template membership, and no INICET/NEET-PG relevance.

Example question IDs with short explanations

  • c0eafafe (Shard 001): HAART ocular side effect (immune reconstitution uveitis) filed under "Analgesics and Anti-inflammatory Drugs." Should be under Antimicrobial Agents or Clinical Pharmacology and Drug Toxicity.

  • f528a465 (Shard 001): Rifampicin's effect on oral contraceptives filed under "Pharmacokinetics and Pharmacodynamics." This is a drug interaction question and belongs under Drug Interactions. The difficulty field also contains a string value rather than a numeric value — a data quality issue.

  • abda7f41 (Shard 003): Lamotrigine half-life filed under Pharmacokinetics and Pharmacodynamics. This is a drug-specific recall fact about an antiepileptic and belongs under CNS/Antiepileptic Drugs. The question tests a single numerical fact with no PK reasoning.

  • 330870b2 (Shard 002): Drug safe in pregnancy (heparin vs. warfarin) filed under "Endocrine Pharmacology." Belongs under Drugs Affecting Blood and Blood Formation or Clinical Pharmacology.

  • ff0a30ff (Shard 002): Anticholinesterase for Alzheimer's filed under "Autonomic Nervous System Drugs." Belongs under Central Nervous System Pharmacology.

  • 3bba1a2c (Shard 008): Sibutramine (anti-obesity/CNS drug) filed under "Antimicrobial Agents." Clear mislabelling.

  • d0c5a4f4 (Shard 003): Antisialogogue in orthodontic bonding (Banthine). Dental pharmacology question with no INICET/NEET-PG relevance, no tags, no template membership. This is a scope contamination item.

  • b1320be8 (Shard 003): Cycloserine neuropsychiatric effects filed under Antimicrobial Agents, but the distractor "Ethosuximide" is an antiepileptic, creating cross-topic confusion without clinical grounding.

Recommended disposition

Topic mislabelling: Fix (reclassify) — these are metadata corrections that do not require content changes. However, reclassification should be accompanied by a content review to confirm the question is appropriate for the target topic. A systematic topic taxonomy audit is recommended for the full pool, not just the sampled items.

Scope contamination: Disable — questions targeting dental, nursing, or non-medical-PG audiences should be removed from the INICET/NEET-PG pool entirely. They cannot be fixed by reclassification because there is no appropriate topic for them in a medical PG bank.

5. Outdated and Superseded Clinical Content

Why this pattern is bad

Pharmacology is a rapidly evolving subject. Drug approvals, guideline updates, and market withdrawals change the correct answer to questions that were accurate when written. A question with a stale correct answer is worse than a question with no correct answer — it actively teaches the wrong fact. For a PG exam preparation platform, content currency is a patient-safety issue as well as an exam-validity issue.

This pattern appears as a narrower cluster in the reviewed set, concentrated in antimicrobials (TB regimens, antiretrovirals), analgesics (withdrawn drugs), and endocrine pharmacology (newer agents). The absolute number in the 200-question sample is modest (5–7 items), but the risk is higher in the full pool given that older questions are less likely to have been reviewed recently.

How it shows up

Withdrawn drugs presented as current treatment options: Propoxyphene (FDA withdrawal 2010), sibutramine (cardiovascular risk withdrawal). Questions that list these as correct answers or as plausible distractors without noting their withdrawn status.

Superseded treatment guidelines: RNTCP Category II DOTS regimen questions that do not reflect the current NTEP (National TB Elimination Programme) framework using universal drug susceptibility testing.

Stale "recently approved" language: Questions using the phrase "recently approved" for drugs that received approval years ago (e.g., bedaquiline, approved 2012, described as "recently approved" with no date context).

Contestable "drug of choice" designations: Questions that designate a single DOC for conditions where current guidelines list multiple equivalent first-line options (e.g., Mycoplasma pneumonia — doxycycline vs. azithromycin; diabetic neuropathy — pregabalin vs. gabapentin).

Example question IDs with short explanations

  • be7c9fe7 (Shard 007): Marks propoxyphene as the recommended analgesic for headache in peptic ulcer disease. Propoxyphene was withdrawn from the US market in 2010 and from most other markets due to cardiac toxicity. The correct answer should be paracetamol. This is both a factual error and a content-currency failure.

  • 3bba1a2c (Shard 008): Sibutramine question does not note that sibutramine has been withdrawn from most markets due to cardiovascular risk. As written, it implies sibutramine is a current treatment option.

  • 8a07a734 (Shard 008): RNTCP Category II DOTS regimen question. This regimen has been superseded by NTEP guidelines. As written, it could mislead candidates preparing for current exams.

  • 5a88c0b7 (Shard 003): "Recently approved drug for MDR-TB" with bedaquiline as the answer. Bedaquiline received FDA approval in 2012 and has been in Indian guidelines for years. The word "recently" is now factually stale and misleading. The question also does not reflect the current BPaL (bedaquiline-pretomanid-linezolid) regimen context.

  • 4343754b (Shard 003): DOC for Mycoplasma infection marked as doxycycline. Azithromycin is equally accepted as first-line in most current guidelines, making the single correct answer contestable.

  • b432f5a7 (Shard 008): DOC for diabetic neuropathy — pregabalin vs. gabapentin. Both are guideline-recommended; the distinction between them as "drug of choice" is guideline-dependent and the question does not specify which guideline is being referenced.

Recommended disposition

Withdrawn drugs: Disable unless the question can be rewritten to explicitly frame the drug as withdrawn and test knowledge of the reason for withdrawal (which is itself a high-yield fact). A question that presents propoxyphene as a current treatment option should not be deployed in any form.

Superseded guidelines: Fix if the concept remains high-yield and the question can be updated to reflect current guidelines with minimal restructuring. Disable if the update would require a complete rewrite (e.g., the entire premise of the question is based on a superseded regimen).

Stale "recently approved" language: Fix by removing the temporal qualifier and replacing with a specific clinical context that makes the drug's role testable without relying on recency.

Contestable DOC designations: Fix by either specifying the guideline being referenced, adding a clinical context that makes one option clearly superior, or reframing as a mechanism/comparison question rather than a DOC question.

6. Intra-Pool Redundancy and Near-Duplicate Clustering

Why this pattern is bad

Redundancy within the candidate set wastes question slots, inflates apparent topic coverage, and — when near-duplicates appear in the same test — reduces the discriminatory power of the assessment. In a pool of 14,472 questions, redundancy is expected, but when near-duplicates appear within a 25-question shard (i.e., within 0.17% of the pool), the local density is high enough to suggest systematic over-authoring of certain topics.

This pattern appears as a moderate-frequency cluster in the reviewed set, with near-duplicate pairs observed in at least four shards. The affected topics are digoxin indications, antiretroviral drug classification, oral contraceptive side effects, clinical trial phases, and thyroid drug safety in pregnancy.

How it shows up

Direct near-duplicates: Two questions testing the same fact from the same angle, with only minor wording differences. Examples: digoxin indication pair (a9d1edc0, c68e5fc3), antiretroviral nevirapine pair (3e9f2c85, 5577ba02), NK1 antagonist pair (3c1d9951, d3d443ee).

Conceptual near-duplicates: Two questions testing the same underlying concept from slightly different angles, both at Blooms-1/2, with no incremental learning value. Examples: OCP side effects pair (a1607ecf, 8f24e774), thyroid drug in pregnancy pair (79a49120, 2df77a20), clinical trial phases cluster (f924a3fa, 71ce25a8, c473c075).

Example question IDs with short explanations

  • a9d1edc0 + c68e5fc3 (Shard 004): Both cover digoxin indications in CHF/AF at Blooms-2, both flagged easy. Direct intra-shard duplication with no incremental value. One should be disabled; the surviving item should be upgraded to a clinical scenario with contraindication reasoning.

  • 3e9f2c85 + 5577ba02 (Shard 004): Both converge on nevirapine as the answer — one asks which drug prevents mother-to-child HIV transmission, the other asks which drug is NOT in the ATRIPLA regimen. Both are Blooms-1. Together they represent thin, overlapping coverage of a single ARV fact.

  • 3c1d9951 + d3d443ee (Shard 007): Both test aprepitant/NK1 receptor antagonism at Blooms-1. One asks the MOA of aprepitant, the other asks which NK receptor antagonist prevents vomiting. Identical cognitive demand, identical answer.

  • 79a49120 + 2df77a20 (Shard 003): Near-duplicate pair on PTU/carbimazole in pregnancy. Both test the same concept (PTU preferred in pregnancy over carbimazole/methimazole) from opposite stems. The stronger item (2df77a20) should be retained; the other disabled.

  • f924a3fa + 71ce25a8 + c473c075 (Shard 005): Three questions on clinical trial phases in a single 25-question shard. All are Blooms-1/2 definitional recall. The cluster suggests template-filling rather than deliberate topic coverage.

Recommended disposition

For direct near-duplicate pairs: Disable the weaker item (lower Blooms level, fewer tags, less precise distractors). For the surviving item, consider a Fix (vignette upgrade) if the concept is high-yield. For conceptual near-duplicate clusters (e.g., three clinical trial phase questions): Disable all but the strongest item, and consider whether the concept warrants a single high-quality applied question rather than multiple recall items.

Prioritization

The five issue categories are ordered below by operational urgency — the combination of harm potential, volume, and remediation complexity.

Tier 1 — Immediate action required before any deployment

Answer Key and Factual Accuracy Failures (Category 1) must be resolved first. These items are actively harmful. The estimated 10–12 confirmed or probable keying errors in the 200-question sample extrapolate to a potentially large number of wrong-keyed items in the full pool of 14,472. A targeted expert pharmacologist review of all Blooms-1/2 items — particularly in antimicrobials, CNS pharmacology, cardiovascular drugs, and endocrine pharmacology — should be completed before any batch deployment. Items with confirmed keying errors should be immediately flagged as non-deployable.

Tier 2 — High volume, high impact on pool quality

Recall Overload: Blooms-1/2 Items Without Clinical Anchoring (Category 2) is the dominant volume problem. Approximately 110–120 of the 200 reviewed items fall into this category and should be disabled. This is not a fix-at-scale problem — converting 110 recall items to genuine Blooms-3 vignettes requires expert authoring effort that is not justified when the pool already contains 14,472 questions. The priority action is systematic disabling of sub-threshold items, with selective vignette conversion reserved for the 15–20 items where the concept is high-yield and the question is structurally close to salvageable.

Tier 3 — Highest return on fix investment

Underdeveloped Clinical Vignettes (Category 3) represents the best fix opportunity. These 30–40 items already have a clinical frame; they need a data point, a decision node, and a management-level question stem. A well-executed fix on these items produces benchmark-quality questions at lower authoring cost than writing new items from scratch. The content team should prioritize these for vignette upgrade, using the benchmark items as the explicit template.

Tier 4 — Metadata and content currency

Topic Mislabelling and Scope Contamination (Category 4) and Outdated and Superseded Clinical Content (Category 5) are important but do not require the same urgency as Categories 1–3. Topic mislabelling is a metadata fix that can be batched. Scope contamination items should be disabled in the next review cycle. Outdated content (withdrawn drugs, superseded guidelines) should be flagged for a content-currency audit, with priority given to antimicrobial and TB regimen questions where the clinical stakes of a wrong answer are highest.

Tier 5 — Pool hygiene

Intra-Pool Redundancy (Category 6) is a pool hygiene issue that can be addressed in a scheduled deduplication pass. It does not require immediate action but should be included in the next topic-level coverage audit.

Example Keep / Fix / Disable Calls

The following table summarises representative disposition calls from across the reviewed set. These are illustrative of the patterns described above and should be used as calibration examples for the content operations team.

KEEP — No changes required

Question ID Shard Rationale
43ae0c58 007 Loading dose calculation (Vd × Ct / bioavailability). Genuine numerical reasoning, Blooms-3, PYQ-tagged. Matches benchmark standard.
4969f26b 006 Angiotensin II pharmacodynamics vignette. Blooms-4, well-constructed distractors, clinically grounded. One of the strongest items in the reviewed set.
fc94c96d 006 CYP450 inhibition: atorvastatin + itraconazole toxicity. Blooms-3, PYQ-tagged (JIPMER 2019), tests applied drug interaction knowledge with directionally meaningful distractors.
ecf81031 003 Absolute contraindication to combined OCP (thrombo-embolism). PYQ-tagged, Blooms-3, plausible distractors. Meets benchmark standard for a moderate-difficulty item.
2af330a6 004 Phenol poisoning vignette with green urine. Blooms-4 diagnostic reasoning, specific clinical details, discriminating distractors.
Qe3b914a9 005 Antibiotic resistance mechanisms EXCEPT quinolone. Blooms-5, NEET-PG tagged, tests a conceptually important distinction between resistance mechanism types.
Qd92dff14 005 Coagulation monitoring for lepirudin. PYQ-tagged (NEET-PG 2018), tests a practical clinical pharmacology distinction among anticoagulant classes.
6693325a 002 Sirolimus vs. cyclosporine — thrombocytopenia. Comparative adverse-effect question with genuinely discriminating distractors.
b91cdefa 006 Dexrazoxane for doxorubicin cardiotoxicity. Blooms-3, NEET-PG tagged, clinically relevant.
2e2a3bed 001 Amiodarone blocking delayed rectifier K⁺ channels. Clean mechanism question with appropriate distractors (other antiarrhythmics with different channel targets).

FIX — Specific remediation required

Question ID Shard Required Fix
d2d6c7a8 002 Key correction (urgent). Change correct answer from "GABA-agonism" to "Irreversible inhibition of GABA transaminase." Revise distractors accordingly.
0c9ffb52 002 Full revision (urgent). Reframe stem as "which agent possesses ganglion-blocking action?" and make trimethaphan the correct answer. Remove halothane from options.
bcd6bcc7 004 Key correction (urgent). The keyed answer is factually TRUE, not an exception. Identify the actual false statement among the options and correct the key.
89ad1340 003 Key correction (urgent). "Can penetrate eschars" is TRUE for Mafenide — this is its defining advantage. Restructure with a genuinely false statement as the keyed answer.
e6523f6f 005 Key correction (urgent). Esmolol is contraindicated in decompensated LV failure. Correct the key to reflect cardioselectivity and ultra-short half-life as the true statements.
Q72b7cbca 005 Key correction (urgent). Periumbilical abdomen IS a preferred insulin injection site. Key is factually inverted. Correct or disable.
546837ea 008 Key correction (urgent). Digitalis does not cause peripheral neuropathy. Correct answer is isoniazid. Revise key and verify all distractors.
5f1e3e61 008 Full revision (urgent). Penicillin G is not orally administered. Correct the factual error and review all options for accuracy.
f511c1f1 001 Vignette upgrade. Add patient age, presenting symptoms (agitation, hyperthermia, clonus), and timeline to convert from a bare drug-interaction question to a Blooms-3 serotonin syndrome recognition item.
70ccf38c 004 Vignette upgrade. Reframe question from "which receptor mediates vasoconstriction" to "what is the expected haemodynamic consequence of adding epinephrine at this dose?" to engage the clinical scenario.
afa6069b 002 Vignette upgrade. Add serum creatinine, potassium, and a question about monitoring or dose adjustment to reach benchmark standard (cf. 3dafce52).
cffab4f7 001 Option correction. Option B states ARBs can be safely used in bilateral renal artery stenosis — this is incorrect; ARBs share the same contraindication as ACE inhibitors. Replace with a genuinely distinguishing feature.
8a07a734 008 Content update. RNTCP Category II DOTS regimen is superseded by NTEP. Update to current NTEP regimen or flag as historical-only.
be7c9fe7 007 Key correction + content update. Change correct answer to paracetamol. Remove propoxyphene (withdrawn drug) from options or note its withdrawn status explicitly.
c616aaf8 007 Construction defect. Two options are identical ("10-20"). Differentiate all four options and standardise units.
727f72c5 006 Key correction or stem qualification. Either correct the key to nausea (most common early SSRI side effect) or reframe the stem to specify "long-term" side effect to make weight gain defensible.
abda7f41 003 Reclassify + upgrade. Move from Pharmacokinetics to Antiepileptic Drugs. Elevate to Blooms-3 by embedding in a drug-interaction scenario (e.g., effect of valproate co-administration on lamotrigine half-life).
5a88c0b7 003 Content update. Remove "recently approved" language. Reframe in current BPaL regimen context or replace with a question about bedaquiline's mechanism and monitoring requirements.
c0eafafe 001 Reclassify. Move from Analgesics to Antimicrobial Agents or Clinical Pharmacology. Add clinical context (patient restarting HAART after gap, developing eye pain) to make the teaching point testable.

DISABLE — Remove from deployable pool

Question ID Shard Rationale
58d74b3d 001 Counterfeit drug definition. Administrative knowledge, not pharmacology. All wrong options are partially correct.
bf0c86f5 001 Insulin pen needle length. Device trivia, no pharmacological reasoning.
50cf6d02 001 Cidofovir is an antiviral. Answerable by name recognition alone.
3f721804 001 Aspirin inhibits COX. First-year MBBS knowledge, zero PG discriminatory value.
fec6012c 002 Chloramphenicol MOA. Bare Blooms-1 recall, no clinical context, covered by better items.
cd99abea 002 Enfuvirtide as fusion inhibitor. Distractors from entirely different classes, trivially eliminable.
676b3cdc 002 Tranexamic acid as antifibrinolytic. Distractors eliminable by category alone.
13f167db 002 Magnesium sulfate is not an antacid. Trivial naming distinction, no clinical relevance.
d0c5a4f4 003 Antisialogogue in orthodontic bonding. Dental pharmacology, no INICET/NEET-PG relevance.
4343754b 003 DOC for Mycoplasma — contestable single correct answer (doxycycline vs. azithromycin equally accepted). Blooms-1, no clinical context.
3707452f 003 Which antineoplastic causes hepatotoxicity. Lacks specificity, multiple correct answers possible, no clinical scenario.
a564fed6 004 Ifosfamide drug class. Pure classification recall, no clinical value.
1b61598e 004 Definition of renal clearance. Definitional recall, unsuitable for PG assessment.
e928ec9d 004 Which antidepressant is an SSRI — fluoxetine. Answerable by anyone who has heard the drug name.
3e9f2c85 004 Which drug prevents mother-to-child HIV transmission. Blooms-1 recall, overlaps with 5577ba02.
5577ba02 004 Which drug is NOT in ATRIPLA. Blooms-1 drug-name recall, overlaps with 3e9f2c85.
Q71ce25a8 005 Phase IV clinical trials definition. Weakest of three clinical-trial-phase questions in the same shard.
Q90b5eb4f 005 Mebendazole is not an azole antifungal. Nomenclature trick, not pharmacology. Educationally misleading.
Q7c83e8d6 005 NNRTIs EXCEPT lamivudine. Pure drug-class list recall, overlaps with Qb4656b2f.
3b7a9dd4 006 Cyclosporin is an immunosuppressant. Pre-MBBS level knowledge.
1a803988 006 Which neurotransmitter is associated with parasympathetic outflow. Factually imprecise answer ("Cholinergic" is a receptor type), Blooms-1.
cf934c77 006 Protease-activated receptors EXCEPT. Factual basis of correct answer is erroneous (PARs are a family of four, not three).
b63e942e 006 Sildenafil indicated for which condition. Ignores primary indication (erectile dysfunction), Blooms-1, misleading framing.
3b359f47 006 Insulin receptor type. Standard first-year pharmacology recall.
31d2d467 007 Which drug is a CYP450 inhibitor — ketoconazole. Pure Blooms-1 recall, covered by better items.
a5c097eb 007 Rosiglitazone MOA — PPAR-gamma. Blooms-1, no clinical context.
2ad14d07 007 Which drug is effective against dermatophytes — griseofulvin. Blooms-1, trivial recall.
a31bf55a 007 Filgrastim used for neutropenia. Single-fact recall, Blooms-1.
20756ce9 008 Clevidipine drug class. Pure classification recall, no tags, no template