Community Medicine Question Quality Review

Executive Summary

This review covers 200 validated non-gold candidate questions randomly sampled from the Community Medicine pool of approximately 10,989 items, analyzed across eight shards of 25 questions each. The benchmark set consists entirely of Bloom's Level 3–4 applied scenario questions requiring multi-step numerical reasoning or clinical epidemiological judgment. The candidate sample is heavily skewed in the opposite direction: 88 of 200 candidate questions are Bloom's Level 1, and a further 89 are Bloom's Level 2, leaving only 23 at Level 3 and 10 at Level 4. This means roughly 88% of the sampled pool operates below the cognitive level that the benchmark and recent PYQs consistently demand.

The problems in this subject are not randomly distributed. They cluster into six distinct, operationally separable issue categories, each with a different remediation path. The most pervasive is a subject-wide overproduction of bare-recall items that test single facts, definitions, or programme details with no clinical or analytical framing. A second major category involves structural breakage — questions that are literally unanswerable because a required image is absent. A third involves factual errors in answer keys, concentrated in national health programme items and communicable disease content. A fourth involves topic misclassification, where questions are filed under wrong topic names at a rate that distorts apparent coverage. A fifth involves a specific item-writing flaw — "All of the Above" or "None of the Above" as the keyed answer — that collapses discriminatory value. A sixth involves near-duplicate proliferation of conceptually adjacent recall items, particularly in epidemiological time intervals, vaccine classification, and study design.

The actionable bottom line: approximately 30–35% of the sampled items should be disabled outright (primarily bare-recall trivia, broken image-dependent items, and factually unsafe questions). A further 35–40% require targeted fixes before deployment. Only 25–30% are deployable as-is or with minor polish. The subject has a significant deficit of Bloom's 3–4 applied calculation and scenario-based items relative to the benchmark standard, and this gap should be addressed through new item development rather than attempted rescue of low-quality recall items.

What Good Looks Like

The benchmark set provides a clear and consistent quality standard. Every benchmark question shares the following characteristics:

Scenario-embedded numerical reasoning. Each question provides a data-rich clinical or epidemiological context — a cohort of 1,000 patients, a 2×2 table, a trial with defined power and significance parameters — and asks the candidate to compute or interpret a specific measure. The concept being tested (NNT, OR, RR, specificity, sample size) is never named in the stem; the candidate must identify and apply it.

Bloom's Level 3 or 4 as the floor. No benchmark question asks for a definition, a formula name, or a single memorised fact. Even the "easier" benchmark items (e.g., the stratified sampling PYQ 6939d4a0) embed the concept in a described scenario and ask for classification based on reasoning, not recognition.

Distractors that exploit real misconceptions. In the OR benchmark (9cd5b0d1), the distractors 3.0, 2.0, and 1.5 correspond to plausible arithmetic errors (wrong cell selection, inverted ratio). In the specificity benchmark (dadd54bf), the distractors 75%, 90%, and 60% correspond to sensitivity, PPV, and a common denominator error respectively. No benchmark question uses "All of the above," "None of the above," or algebraic expressions as options.

Clinical or public health anchoring. Even purely statistical questions (confidence interval interpretation 497bb3f2, RCT imbalance 531d719d) are framed around a clinical decision or a research scenario with named outcomes. The concept is never tested in a vacuum.

Unambiguous correct answers. Every benchmark item has a single defensible correct answer that can be verified arithmetically or by reference to a specific, current guideline. There are no "best of a bad set" answers and no answers that depend on which edition of a textbook is being used.

The recent PYQs extend this standard into applied epidemiology beyond biostatistics: the prevalence/incidence reasoning question (2ad50ccb), the dependency ratio calculation (7aaf29cd), and the Anaemia Mukt Bharat treatment protocol (69dd9e4d) all require the candidate to apply a concept to a defined situation rather than recall a fact. The Japanese encephalitis life cycle matching question (dddcade6) is the one PYQ that operates at pure recall (Bloom's 1), and it is the weakest item in the gold set — it represents the floor of acceptable quality, not the target.

Main Issue Categories

1. Bare-Recall Overload: The Dominant Volume Problem

Why this pattern is bad

The benchmark standard operates exclusively at Bloom's 3–4. A question that asks "What is the denominator of the maternal mortality ratio?" or "What is the acceptable noise level?" or "Who is the father of public health?" does not distinguish a candidate who understands epidemiology from one who has memorised a glossary. At PG entrance level, these items have near-zero discriminatory power: strong candidates answer them correctly in two seconds, weak candidates guess correctly at 25% base rate, and the item contributes nothing to rank ordering. They also crowd out higher-quality items in any test assembly that draws from this pool.

How it shows up

This is the broadest pattern in the reviewed set, appearing in every shard. It is not confined to one topic — it spans biostatistics definitions, communicable disease facts, national programme details, nutrition thresholds, environmental health limits, and demographic indicators. The common structural signature is a one-sentence stem with a single named concept and four options that are either the correct answer plus three implausible alternatives, or four numerical values where only one is the memorised figure.

Specific examples from the reviewed set:

db156e0d — "What is the term for a rate adjusted to allow for age distribution?" (answer: age-standardized rate). Pure definitional recall, no scenario, no reasoning.
88f9d692 — "Minamata disease is caused by toxicity of which element?" Single environmental health fact.
a6ff7993 — "What is the denominator used in maternal mortality ratio?" The benchmark PYQ 7aaf29cd shows how this concept should be tested: provide population data and ask the candidate to calculate the ratio.
ff9c19bd — "What is the significant value of p?" Tests a threshold every student memorises in week one.
87e9b0bd — "What is the monthly pension amount under IGNDPS?" Administrative trivia with an outdated figure.
a956953a — "ROME scheme was recommended by which committee?" Committee-name recall with no applied reasoning.
be5ff5a1 — "Father of public health = John Snow." Historical trivia.
8da69f2c — "Exact value within ±2 SD in a standard normal curve?" Constant recall.
9a026b8d — "Which source contains the highest amount of Vitamin D?" Nutrition trivia.
3e56bcbd — "What is the acceptable noise level?" Single threshold fact.
d52d633a — "WHO defines adolescence as which age range?" Definitional lookup.
3d17d253 — "Which healthcare worker is stationed at a sub-centre?" Programme structure recall.
96f6ec08 — "What is the reservoir of hookworm?" Single-fact parasitology.
695d546c — "Target age group for pulse polio immunisation?" Trivially known.
49551d41 — "Effective literacy rate is calculated from which age group?" Census definition trivia.
dad4a1d2 — "Definition of relative risk." The benchmark (bf118aa8) shows this concept requires a cohort scenario with actual numbers.
9575b5ba — "Primary goal of dietary modification to prevent CVD = primary prevention." Trivially obvious.
be2c77ee — "Which disease is communicable? — Amoebiasis vs. Rickets/Diabetes/Cancer." No reasoning required.

Recommended disposition

Items in this category that test concepts already covered by benchmark or PYQ items at Bloom's 3–4 should be disabled — there is no value in retaining a Bloom's 1 version of a concept when a superior version exists. Items testing concepts not yet covered at higher Bloom's levels should be fixed by conversion to a scenario-based format (provide data, ask for calculation or interpretation) rather than retained as recall items. The conversion pathway is clear for most of these: the MMR denominator question becomes a calculation scenario; the noise level question becomes an occupational exposure scenario; the relative risk definition becomes a cohort data interpretation question. However, given the volume of items in this category and the availability of benchmark-quality coverage for the most important concepts, the default recommendation for items in this category is disable unless the concept is genuinely underrepresented at higher Bloom's levels.

2. Structurally Broken Image-Dependent Questions

Why this pattern is bad

These questions are not merely weak — they are non-functional. A question that says "Interpret the following graph" or "Identify the image below" or "The formula shown is:" with no image present is unanswerable from text alone. Deploying such an item in a test produces a question that either gives away the answer through elimination (if three options are clearly wrong) or is effectively random for all candidates. This is a distinct and more urgent problem than low Bloom's level, because it represents a delivery failure rather than a quality failure.

How it shows up

At least five confirmed broken image-dependent items were identified across the reviewed shards:

bfe60bcb — "The formula shown is:" with no formula present. The question is about the Pearl index but is completely non-functional as text.
964d6a66 — "Which of the following can be monitored using this tape?" with no tape image. Intended to test MUAC measurement.
411cb845 — "The following instrument is used for:" with no instrument image.
2a7bf4b7 — "Which of the following is true about the box plot shown?" with no box plot. This is a PYQ (AIIMS 2018) and the question is worth retaining if the image is restored.
817cc642 — "Interpret the statistical graph shown below" with no graph attached.
94f9cc22 — Scatter plot correlation question (Bloom's 4) where the referenced scatter plot must be verified as rendering correctly before deployment.

The PYQ set also contains image-dependent items (657ff7fa — box plot interpretation, 0fa92647 — multivariate table) that function correctly because their images are intact. The problem is not the format — image-based questions are appropriate and appear in actual exams — but the absence of the image asset.

Recommended disposition

All confirmed broken image-dependent items should be disabled immediately. Items where the image is a PYQ asset worth preserving (2a7bf4b7, bfe60bcb) should be placed in a fix queue for image restoration and re-verification before re-enabling. Items where the image is not recoverable or where the text-only version is meaningless should be disabled permanently. A subject-level audit of all image-tagged items is warranted given that at least five broken items appeared in a sample of 200 — the full pool of ~10,989 likely contains a proportionally larger number.

3. Factual Errors and Unsafe Answer Keys

Why this pattern is bad

A question with a wrong answer key is worse than no question at all. It actively misteaches candidates, penalises those who know the correct answer, and rewards those who have memorised the wrong fact. In Community Medicine, where national programme guidelines are updated periodically and some facts are genuinely contested in the literature, this risk is elevated. The reviewed set contains both clear factual errors (where the keyed answer is demonstrably wrong) and contested answers (where the keyed answer is defensible in one source but contradicted by another or by current guidelines).

How it shows up

Several distinct subtypes were identified:

Clear factual errors:

71053b22 — Marks "Life expectancy at birth" as NOT included in the Human Development Index. Life expectancy is a foundational HDI dimension. This is a straightforward factual error; the item must be disabled immediately.
e70433b8 — Marks "Oral polio vaccine" as NOT a live attenuated vaccine. OPV (Sabin) is definitionally a live attenuated vaccine. The intended answer was likely the injectable inactivated influenza vaccine. The answer key is wrong.
b3808c8a — Coefficient of variation calculation uses "median" as the denominator (CV = SD/median × 100). CV is defined as SD/mean × 100. The formula in the stem is incorrect, making the calculated answer misleading.
c0e71363 — "All of the following are nominal scale except?" lists socioeconomic status as a nominal variable, but socioeconomic status is ordinal. This creates two potentially correct answers (body weight = ratio scale; socioeconomic status = ordinal, not nominal).

Outdated programme content:

21505e0a — References DOTS Category I/II/III classification, which has been replaced under the current NTP India framework. The question is not wrong about the old system but is misleading in a current exam context.
2ba056f7 — States that RNTCP's primary change was "therapy not based on DOTS," which is misleading — NTEP still uses daily observed therapy; the change was from intermittent to daily regimen and introduction of universal DST.
87e9b0bd — IGNDPS pension amount of Rs. 200 is outdated; the scheme amounts have been revised.
25b427a9 — References "100% immunization by 2000 A.D." as a historical MCH target, which may confuse candidates unfamiliar with the historical context.

Contested or ambiguous keys:

ab74392c — Keys "Family planning" as NOT a major strategy of RCH-II. Family planning is a core pillar of RCH-II; this appears to be a keying error requiring expert review.
cb4f4b6c — Keys "Type-2 only" for VAPP. Post-bOPV switch, most VAPP is attributed to Type 3; Type 2 was removed precisely because of VAPP risk. This is a nuanced area requiring expert review.
d471d9ec — Keys as false the statement that dengue spread has been affected by ambient temperature changes in India over the last 30 years. This statement is epidemiologically supported; the answer key appears inverted.
8cc29534 — Keys "Vaccination is given at 6 months" as correct for measles outbreaks without the qualifier that this applies only in outbreak/high-risk settings, not the routine schedule.
08c47e47 — Keys "Random number table" as the best method of randomization over computer-generated randomization, which is contestable in contemporary practice.

Recommended disposition

Clear factual errors (71053b22, e70433b8, b3808c8a, c0e71363) should be disabled immediately pending correction — they are actively harmful. Outdated programme items (21505e0a, 2ba056f7, 87e9b0bd) should be disabled unless updated to reflect current guidelines. Contested keys (ab74392c, cb4f4b6c, d471d9ec, 8cc29534) require expert content review before any disposition decision; they should be held out of deployment until reviewed. The volume of factual accuracy concerns in national programme content suggests a targeted expert audit of all programme-specific items (RNTCP/NTEP, RCH, DOTS, IGNDPS, Anaemia Mukt Bharat) is warranted across the full subject pool.

4. Topic Misclassification Distorting Coverage Maps

Why this pattern is bad

When questions are filed under the wrong topic, two problems occur simultaneously: the topic they are filed under appears to have more coverage than it does, and the topic they belong to appears to have less. For a subject like Community Medicine with many sub-topics (Epidemiology, Biostatistics, Immunization, NCD, MCH, Environmental Health, Health Administration), systematic misclassification makes it impossible to identify genuine coverage gaps or to build balanced test templates. It also means candidates studying a specific topic may encounter irrelevant questions, and topic-level quality audits miss items that belong to them.

How it shows up

Misclassification was observed at a rate of approximately 4–6 items per shard, suggesting it is a systemic rather than isolated problem. The most common patterns in the reviewed set:

Communicable disease items filed under Non-Communicable Diseases:

6df166a0 — IPV immunization schedule filed under "Non-Communicable Diseases." Should be under Immunization.
aceca88a — Cholera outbreak control filed under "Non-Communicable Diseases." Should be under Infectious/Communicable Diseases.
f883f43f — bOPV strains filed under "Non-Communicable Diseases." Should be under Immunization.
ae07728d — Dracunculosis eradication filed under "Epidemiology" rather than Infectious Diseases or Immunization.

Clinical medicine questions filed under Community Medicine topics:

2f42243b — Red eye after malaria (clinical differential: viral keratitis vs. uveitis) filed under Non-Communicable Diseases. This is an Ophthalmology question with a questionable answer key; it does not belong in Community Medicine at all.
a66b90e1 — Wernicke's encephalopathy in a chronic alcoholic filed under Nutrition/Community Medicine. This is a clinical neurology/medicine question.
10f16150 — Opioid dependence pharmacotherapy (disulfiram) filed under Environmental Health. Should be under Mental Health/Substance Use.

Biostatistics/demographic items filed under wrong sub-topics:

49551d41 — Effective literacy rate age group filed under Biostatistics. Should be under Demography or Health Statistics.
917a25cf — Literacy rate denominator filed under Non-Communicable Diseases. Should be under Health Statistics.
14227607 — Cost accounting features filed under Health Economics/Community Medicine. Tests administrative definitions with no epidemiological relevance.

Recommended disposition

Misclassified items that are otherwise of acceptable quality should be reclassified to the correct topic — this is a metadata fix, not a content rewrite. Items that are misclassified AND belong to a different subject entirely (2f42243b, a66b90e1) should be disabled from the Community Medicine pool and transferred to the appropriate subject. The frequency of misclassification in this sample suggests a topic taxonomy review is needed for the full subject pool, particularly at the Infectious Diseases / NCD / Epidemiology boundary and the Community Medicine / Clinical Medicine boundary.

5. Structural Item-Writing Flaws: "All of the Above" and "None of the Above" as Keyed Answers

Why this pattern is bad

"All of the above" as the correct answer is a well-documented item-writing flaw. A test-wise candidate who recognises that two options are correct can select "All of the above" without evaluating the remaining options, reducing the item to a partial-knowledge test. It also means that if any single option is incorrect, the entire item is broken. "None of the above" as a distractor (rather than the correct answer) is a weaker but related problem — it signals to candidates that the correct answer is among the named options and reduces the effective distractor count to three. Neither format appears in any benchmark or PYQ exemplar.

How it shows up

This pattern appears across multiple shards and topics:

4df5c54c — NBCC question with "All of the above" as the keyed correct answer.
3eea1af8 — Medical sociology question with "All of the above" as the keyed correct answer. Beyond the structural flaw, the question tests a definition so broad it has no discriminatory value.
b62bbb14 — "All of the following are greenhouse gases" with "All of the options" as the answer. Trivial and structurally broken.
8c400621 — "Secondary attack rate is a measure of?" with "None of the above" as option D.
d06950b1 — Uses "None of the above" as a distractor.
45a0ce58 — Ergonomics definition question with "None of the above" as a distractor.
ab7df93c — Filariasis elimination question with "None of the above" as option D.

Recommended disposition

Items where "All of the above" is the keyed correct answer should be fixed by selecting the single most important or discriminating correct statement as the answer and replacing the others with plausible distractors, or by converting to a "which of the following is NOT" format. Items where "None of the above" appears as a distractor should be fixed by replacing it with a substantive, plausible incorrect option. Items where the concept being tested is itself low-value (e.g., 3eea1af8, b62bbb14) should be disabled rather than fixed, since the structural flaw is compounded by low discriminatory value.

6. Near-Duplicate and Redundant Item Clusters

Why this pattern is bad

When multiple questions test the same narrow concept at the same cognitive level with the same format, they consume pool capacity without adding coverage breadth or cognitive diversity. In a subject with 10,989 total questions, redundancy at the low end of the quality spectrum is particularly wasteful — it means the pool is large in volume but narrow in effective coverage. Near-duplicates also create a risk of the same concept appearing multiple times in a single test assembly, which distorts difficulty calibration and frustrates candidates.

How it shows up

Several near-duplicate clusters were identified in the reviewed set:

Epidemiological time intervals:

5a94f971 (latent period definition) and 5c906cc2 (incubation period definition) are conceptually adjacent Bloom's 1–2 recall items. Neither provides a clinical scenario. Together they test the same narrow domain at the same level without the scenario that would make the distinction meaningful.

Vaccine efficacy study design:

98fdacd6 (RCT confirms vaccine efficacy) and e77235cb (field trial for vaccine in 1-year-old) ask essentially the same conceptual question — what study design confirms vaccine efficacy — with slightly different framings. Without a clear contextual distinction (phase of evaluation, resource setting), these create confusion rather than coverage.

Incubation period matching (PYQ-level):

The PYQ set itself contains two near-identical incubation period matching questions (18e8d4a1 and 1fb79883) — one matching syphilis/SARS/Hepatitis A/Chickenpox and one matching syphilis/Chickenpox/COVID-19/Hepatitis A. These are structurally the same question with one disease swapped. Both are flagged in the gold set as blooms-1.

Sampling method questions:

Multiple questions across shards test sampling method identification (stratified, cluster, systematic, simple random) at Bloom's 2 with scenario descriptions. While individually acceptable, the cluster is disproportionately large relative to the exam weight of this sub-topic.

Odds ratio / relative risk calculation:

The pool contains multiple OR and RR calculation questions at varying quality levels. The benchmark set already provides high-quality exemplars for both. Lower-quality versions of the same calculation (e.g., dee507e7 is good; dad4a1d2 is a pure definition item for the same concept) should not coexist.

Recommended disposition

Within near-duplicate clusters, retain the highest-quality item (highest Bloom's level, best scenario embedding, cleanest distractors) and disable the lower-quality siblings. For the epidemiological time interval cluster, the recommended fix is to merge the two recall items into a single scenario-based question that requires the candidate to distinguish latent period from incubation period in a described clinical case. For the vaccine study design cluster, retain e77235cb (field trial, Bloom's 3) and disable 98fdacd6 (more generic, lower discriminatory value).

Prioritization

The six issue categories are not equally urgent. The following prioritization reflects both the severity of the problem and the operational effort required to address it.

Immediate action (before any deployment from this pool):

Factual errors and unsafe answer keys — Items with demonstrably wrong answer keys (71053b22, e70433b8, b3808c8a, c0e71363, ab74392c) must be disabled before any test assembly draws from this pool. These are the highest-risk items because they actively harm candidates who know the correct answer.
Broken image-dependent items — Items with missing images (bfe60bcb, 964d6a66, 411cb845, 2a7bf4b7, 817cc642) are non-functional and must be disabled immediately. A full image-integrity audit of the subject pool is warranted.

High priority (address in the next content sprint):

Bare-recall overload — The dominant volume problem. The recommended approach is a systematic sweep of all Bloom's 1 items in the pool, disabling those that test concepts already covered at Bloom's 3–4 by benchmark or PYQ items, and flagging the remainder for scenario conversion. This is the largest single remediation task but also the one with the clearest decision rule.
Outdated programme content — Items referencing superseded DOTS categories, old IGNDPS amounts, and historical MCH targets should be audited against current NTP/NHM guidelines and either updated or disabled. This requires a subject-matter expert pass, not just editorial review.

Medium priority (address in the following sprint):

Topic misclassification — Reclassification is a metadata fix that does not require content rewriting. It should be done systematically across the full pool, particularly at the Infectious Diseases / NCD / Epidemiology boundary. Clinical medicine questions that have been misrouted into Community Medicine should be transferred or disabled.
Structural item-writing flaws — "All of the above" and "None of the above" items should be swept and fixed. This is a mechanical fix for most items, though some will need to be disabled if the underlying concept is also low-value.

Ongoing:

Near-duplicate consolidation — Deduplication within clusters (time intervals, sampling methods, OR/RR calculations) should be done as part of the broader Bloom's audit rather than as a separate pass.
New item development — The most important long-term action is not fixing weak items but commissioning new Bloom's 3–4 scenario-based items in the categories where the benchmark standard is not met: applied epidemiology calculations (attributable risk, NNT, PPV/NPV in defined populations), national programme application scenarios, and environmental/occupational health exposure-consequence reasoning. The current pool cannot reach benchmark quality through remediation alone.

Example Keep / Fix / Disable Calls

The following examples are drawn from across the reviewed shards and illustrate the application of the issue categories above to specific items.

KEEP — dee507e7 Odds ratio calculation from a case-control study with real numbers (cases = 90, smokers = 75; controls = 100, smokers = 40). Bloom's 3, requires correct 2×2 table construction, answer (7.5) is arithmetically verifiable. Structurally aligned with benchmark 9cd5b0d1. No structural flaws. Deploy as-is.

KEEP — d99d1107 Specific death rate for TB calculation: clean data, single correct calculation (1500/250,000 × 1,000 = 6), good distractor spread. Bloom's 3, matches benchmark style for applied epidemiology. Deploy as-is.

KEEP — 9101fb3d PPV calculation with explicit prevalence, sensitivity, and specificity values in a defined population of 10,000. Bloom's 3, requires 2×2 table construction. Numerically sound. Minor improvement: add a clinical scenario wrapper (e.g., a screening programme context) to reach full benchmark standard, but deployable as-is.

KEEP — a55b9e53 Bleaching powder calculation for well chlorination using Horrocks' test. Multi-step calculation involving well volume, chlorine demand, and bleaching powder purity. Genuine Bloom's 3 applied numerics. Uncommon in question banks. Keep with arithmetic verification of the correct answer.

KEEP — 668d99ae Dengue virus multi-statement "which are correct" format: Bloom's 4, integrates serotypes, reservoir, age-severity pattern, and DHF pathogenesis. Well-constructed, UPSC-CMS PYQ, plausible distractors. Deploy as-is.

KEEP — 89aa1e8d Scenario about a household with a visiting brother's family; tests the definition of nuclear vs. extended/joint family with a realistic ambiguity (temporary visitors). Bloom's 3, well-constructed, PYQ-tagged. Deploy as-is.

KEEP — c52b2dea Malaria surveillance scenario requiring integration of API, ABER, and SPR to classify surveillance status. Bloom's 4, applied reasoning, good discriminatory potential. Meets benchmark standard for applied epidemiology.

FIX — 28baf9fc Specificity formula presented as four algebraic expressions. The concept is appropriate but the format is pure formula-recall. Fix: convert to a 2×2 table vignette with actual numbers and ask the candidate to calculate specificity, matching benchmark dadd54bf. Estimated effort: moderate (requires new scenario construction).

FIX — a32a2e0b "Which is NOT a method of survival analysis?" — Kruskal-Wallis is listed as a distractor but is itself not a survival analysis method, creating two potentially correct answers. Fix: replace Kruskal-Wallis with Cox proportional hazards regression (a genuine survival analysis method), making "Survival rate" the unambiguous odd-one-out. Estimated effort: low (single distractor replacement).

FIX — 7eff478e Kyasanur forest disease NOT true item. The distractor "transmitted by soft ticks" is itself incorrect (KFD is transmitted by hard ticks, Haemaphysalis). Two options are arguably wrong. Fix: correct the tick-type distractor to "hard ticks" and replace with a genuinely plausible wrong statement. Estimated effort: low.

FIX — 8cc29534 Measles outbreak vaccination at 6 months is only correct in outbreak/high-risk settings, not the routine schedule. Fix: add the qualifier "during an outbreak" to the correct option, or replace with an unambiguous fact about outbreak thresholds. Estimated effort: low.

FIX — 2a7bf4b7 Box plot interpretation question (AIIMS 2018 PYQ) with missing image. The question is worth retaining. Fix: restore the box plot image showing symmetric distribution and verify rendering before re-enabling. Estimated effort: low if image asset is available.

FIX — 76b8d7e1 IMNCI pneumonia classification: "fast breathing" only applies to non-severe pneumonia; chest indrawing classifies severe pneumonia. Fix: specify the classification level in the stem (e.g., "non-severe pneumonia in a child aged 2–59 months") to remove ambiguity. Estimated effort: low.

FIX — 566f8e6c "B-X duration" in natural history of disease is meaningless without the Leavell & Clark diagram. Fix: embed the diagram or rewrite the stem to define the reference points explicitly (e.g., "the interval between biological onset and the point detectable by screening"). Estimated effort: moderate.

FIX — 4df5c54c NBCC question with "All of the above" as the keyed correct answer. Fix: select the single most discriminating correct statement as the answer and replace the others with plausible distractors. Estimated effort: moderate (requires content knowledge to select the best single answer).

DISABLE — 71053b22 Factual error: marks "Life expectancy at birth" as NOT included in the Human Development Index. Life expectancy is a foundational HDI dimension. Disable immediately; do not deploy under any circumstances until corrected.

DISABLE — e70433b8 Marks "Oral polio vaccine" as NOT a live attenuated vaccine. OPV is definitionally live attenuated. Answer key is wrong. Disable immediately.

DISABLE — bfe60bcb "The formula shown is:" with no formula present. Completely non-functional as a standalone text item. Disable until image is embedded and verified.

DISABLE — 45a0ce58 "What is the definition of ergonomics?" with "None of the above" as a distractor. Bloom's 1, no clinical context, no discriminatory value. Disable.

DISABLE — 87e9b0bd IGNDPS monthly pension amount (Rs. 200). Administrative trivia with an outdated figure. No clinical or epidemiological reasoning. Disable.

DISABLE — 78ac8f20 Council of ministers' accountability to Lok Sabha. Civics content with no community medicine relevance. Disable.

DISABLE — 2f42243b Red eye after malaria filed under Non-Communicable Diseases. Wrong subject, questionable answer key (post-malarial red eye is more commonly anterior uveitis or subconjunctival hemorrhage, not viral keratitis). Disable from Community Medicine pool.

DISABLE — a66b90e1 Wernicke's encephalopathy in a chronic alcoholic. Clinical neurology question misrouted into Community Medicine. Disable from this subject pool.

DISABLE — 21505e0a Endometrial TB DOTS category question using the obsolete Category I/II/III classification. Outdated and potentially misleading under current NTP framework. Disable unless updated.

DISABLE — b62bbb14 "All of the following are greenhouse gases" with "All of the options" as the answer. Trivial, Bloom's 1, structurally broken. Disable.