Public Health, Epidemiology, and Prevention
What this domain tests
This domain asks you to read population data critically: measure disease frequency, choose and interpret study designs, quantify association, and detect bias. Prevention questions test the levels of prevention and the criteria for a worthwhile screening program.
Core rules
- Incidence counts new cases per population at risk per unit time; prevalence counts existing cases and rises with longer survival and lower mortality. Prevalence ≈ incidence × duration.
- Study designs: randomized trials for causation, cohort studies for incidence and multiple outcomes, case-control studies for rare outcomes (odds ratio only), cross-sectional for prevalence, ecological for group-level correlation that cannot be applied to individuals.
- Measures: relative risk = risk in exposed / risk in unexposed; attributable risk = difference; odds ratio approximates relative risk only when the outcome is rare (roughly under 10%).
- Absolute measures drive decisions: ARR = control event rate − treatment event rate; NNT = 1/ARR; NNH = 1/absolute risk increase.
- Bias: selection, information and recall, confounding (a third factor linked to both exposure and outcome), lead-time bias, length-time bias, and immortal time bias.
- Prevention levels: primary (prevent onset, e.g., vaccination), secondary (early detection, e.g., screening), tertiary (limit disability, e.g., rehabilitation), quaternary (avoid unnecessary intervention).
- Screening criteria include an important condition, accepted treatment, adequate test performance, and net benefit; surveillance systems need completeness, timeliness, and case definitions.
Worked example
A trial reports death in 20% of controls and 15% of treated patients. Relative risk is 0.75, a 25% relative reduction. Absolute risk reduction is 5 percentage points, so NNT = 1/0.05 = 20 patients treated to prevent one death. The relative figure sounds dramatic; the absolute figure tells the patient what it means. If the baseline risk were 2% rather than 20%, the same relative risk would give an NNT of 200.
Traps that cost marks
- Reporting relative risk reduction alone as if it were patient-level benefit.
- Using an odds ratio as a risk ratio for a common outcome.
- Confusing incidence with prevalence when survival changes.
- Attributing improved screening survival to earlier detection when lead-time bias explains it.
- Generalizing an ecological correlation to individuals.