EMREE public health practice questions

Correct answer: B. Lower sensitivity and higher specificity

Raising the threshold for a positive result classifies fewer people as positive. Fewer diseased people are detected, so sensitivity falls, while fewer healthy people are misclassified, so specificity rises. Lowering the cut-off produces the reverse. Only a better test, not a changed threshold, improves both.

• Why A is wrong. Fewer positives mean fewer false positives, so specificity cannot fall.
• Why C is wrong. A threshold change cannot raise both measures; it trades one against the other.
• Why D is wrong. Fewer positives mean fewer true positives, so sensitivity cannot rise.

Topic: Sensitivity: effect of raising a cut-off. Reference: Gordis Epidemiology: Assessing the validity and reliability of diagnostic and screening tests; Fletcher RH. Clinical Epidemiology: The Essentials; Last JM. A Dictionary of Epidemiology

Correct answer: D. Detection (ascertainment) bias

When outcome assessors know the treatment allocation and the outcome is subjective, they may rate the intervention group more favourably, creating detection (ascertainment) bias. Blinding of assessors, or use of objective outcomes, reduces this. Selection bias concerns how groups were formed, and recall bias concerns retrospective reporting.

• Why A is wrong. Lead-time bias applies to screening survival comparisons.
• Why B is wrong. Recall bias affects retrospective reporting, not clinician scoring.
• Why C is wrong. Selection bias concerns group formation, which randomisation addresses.

Topic: Bias: unblinded outcome assessment. Reference: Cochrane Handbook for Systematic Reviews of Interventions: Risk of bias; CONSORT 2010 Statement; Gordis Epidemiology: Randomized trials

Correct answer: B. Selection bias from broken allocation concealment; conceal the sequence until enrolment

Allocation concealment means the person enrolling a participant cannot know or predict the next assignment, for example by using central telephone or web randomisation or sequentially numbered opaque sealed envelopes. When the sequence is visible, recruiters can steer particular patients to a chosen arm, which is selection bias and undoes the benefit of random sequence generation. Concealment is distinct from blinding, which operates after allocation.

• Why A is wrong. Performance bias arises after allocation from differences in care and is addressed by blinding, not by hiding the sequence.
• Why C is wrong. Attrition bias arises from losses after randomisation, whereas the problem here occurs at enrolment.
• Why D is wrong. Detection bias concerns outcome measurement after allocation, which is not the problem described.

Topic: Trial design: allocation concealment. Reference: CONSORT 2010 Statement; Schulz KF, Grimes DA. Allocation concealment in randomised trials: defending against deciphering. Lancet 2002; Cochrane Handbook for Systematic Reviews of Interventions: Risk of bias

Correct answer: A. Selection (attrition) bias from differential loss to follow-up

When loss to follow-up differs by both exposure and outcome, the remaining participants are no longer representative, introducing selection (attrition) bias. Here the sickest exposed workers disappear, which biases the association towards the null. Recall bias concerns inaccurate reporting of past exposure, and lead-time bias applies to screening.

• Why B is wrong. Recall bias affects how participants report past exposures, not who remains in the study.
• Why C is wrong. Lead-time bias applies to survival comparisons in screening, not to cohort attrition.
• Why D is wrong. Confounding by indication arises in treatment studies where prognosis influences prescribing.

Topic: Selection bias: differential loss to follow-up. Reference: Gordis Epidemiology: Observational studies and sources of bias; Szklo M, Nieto FJ. Epidemiology: Beyond the Basics; Last JM. A Dictionary of Epidemiology

Correct answer: A. 5.0

The positive likelihood ratio is sensitivity divided by (1 - specificity): 0.75 / (1 - 0.85) = 0.75 / 0.15 = 5.0. A positive result is therefore 5 times as likely in a diseased person as in a healthy person. The negative likelihood ratio would be (1 - 0.75) / 0.85 = 0.29.

• Why B is wrong. This is the negative likelihood ratio, which describes a negative test.
• Why C is wrong. This results from dividing specificity by (1 - sensitivity), the wrong way round.
• Why D is wrong. No combination of the stated values gives 10; it is a miscalculation.

Topic: Likelihood ratios: from sensitivity and specificity. Reference: Gordis Epidemiology: Assessing the validity and reliability of diagnostic and screening tests; Jaeschke R et al., JAMA 1994: Users guides to the medical literature: how to use an article about a diagnostic test; Fletcher RH. Clinical Epidemiology: The Essentials

Correct answer: C. About 8 percent

Of 10,000 people, 100 have the condition and 9,900 do not. True positives are 90 percent of 100, which is 90. False positives are 10 percent of 9,900, which is 990. Total positives are 90 + 990 = 1,080, so the positive predictive value is 90 / 1,080, or about 8 percent. When prevalence is low, most positives are false even with a good test, which is why screening needs a confirmatory test.

• Why A is wrong. 50 percent would need a much higher prevalence, and it ignores the 990 false positives.
• Why B is wrong. 90 percent is the sensitivity, which is a different measure from the PPV.
• Why D is wrong. 10 percent is the false-positive rate (1 minus specificity), and it is not the PPV.

Topic: Predictive value: effect of low prevalence in a screened population. Reference: Bonita R, Beaglehole R, Kjellstrom T. Basic Epidemiology, 2nd edition (WHO), screening chapter

Correct answer: B. 2 per 100 person-years

An incidence rate uses the time at risk as the denominator, so it is the number of new cases divided by the total person-years: 12 / 600 = 0.02 per person-year, or 2 per 100 person-years. Dividing by the starting number of people (12 / 400 = 3 percent) gives the proportion who developed the disease over follow-up, not a rate. Inverting the fraction (600 / 12 = 50) gives the person-years per case.

• Why A is wrong. 12 per 100 person-years would require only 100 person-years of follow-up.
• Why C is wrong. 12 / 400 gives the proportion who developed asthma, and it ignores time at risk.
• Why D is wrong. 600 / 12 = 50 is the person-years per case, and it inverts the rate.

Topic: Cumulative incidence versus incidence rate: person-years at risk. Reference: Bonita R, Beaglehole R, Kjellstrom T. Basic Epidemiology, 2nd edition (WHO), measuring disease chapter

Correct answer: B. Tell her honestly and promptly, apologise, explain the effects and report the incident

The professional duty of candour (GMC and NMC) requires openness and honesty when something goes wrong that has caused, or could have caused, harm or distress, even when the harm is minor. The patient should be told promptly, receive an apology, an explanation of what happened and what will be done to monitor her, and the incident should be reported so that the system can learn. Concealment, delayed disclosure or telling relatives first damages trust and breaches her right to information.

• Why A is wrong. Waiting for questions makes disclosure depend on her noticing, which breaches candour.
• Why C is wrong. Recording alone is not disclosure, and the patient has a right to know about errors in her care.
• Why D is wrong. The patient has the capacity to be told, and relatives are not the first to be informed.

Topic: Ethics: candour after a medication error. Reference: General Medical Council and Nursing and Midwifery Council: Openness and honesty when things go wrong, the professional duty of candour

Correct answer: A. Arrange a qualified medical interpreter and record this in the notes

Valid consent requires that she understands the information about the procedure, its risks and alternatives. GMC guidance expects doctors to use an interpreter or translation service when a patient has difficulty understanding spoken English. A qualified medical interpreter gives accurate, impartial translation and protects confidentiality, and the interpreter used is recorded in the notes. A child may lack the vocabulary, may distort information to protect or influence a parent, and should not carry the burden of such a discussion. A colleague with conversational skills is not trained for medical terms, and an English form does not allow understanding.

• Why B is wrong. A form she cannot read does not show that she has understood, so the consent would not be valid.
• Why C is wrong. Conversational skill does not cover medical terms or risks, and the colleague has no training.
• Why D is wrong. A 16-year-old is not an appropriate interpreter for risks and complications and may filter information.

Topic: Consent: communication through an interpreter. Reference: General Medical Council Decision making and consent (2020), guidance on communication and language

Correct answer: A. Length-time bias

Screening is more likely to detect slow-growing, less aggressive tumours, which have a long preclinical phase, and it often misses fast-growing tumours that present as interval cancers. This makes screen-detected cases appear to survive longer, even if screening does not change the outcome, and is known as length-time bias. Lead-time bias is the apparent extra survival from earlier diagnosis without a later death, and the stem describes tumour behaviour instead. Mortality in the whole population is the fair outcome to judge screening.

• Why B is wrong. Healthy volunteer bias describes who takes part, and the stem describes the type of tumour detected.
• Why C is wrong. Lead-time bias concerns earlier diagnosis in the same disease course, and the stem describes tumour speed.
• Why D is wrong. Recall bias affects the reporting of past exposures in case-control studies, and it does not apply here.

Topic: Bias: length-time bias in a screening programme. Reference: Bonita R, Beaglehole R, Kjellstrom T. Basic Epidemiology, 2nd edition (WHO), screening chapter