CPH (Other exams) common mistakes
CPH · Understand the traps and how to avoid them.
Treating 'incidence' and 'prevalence' as interchangeable when a question asks which measure to report, then choosing prevalence for a study of new cases over a defined period.
Both are described as 'how common a disease is', so candidates pattern-match on the word 'common' instead of on whether the question involves new events (incidence) or existing cases at a point in time (prevalence).
What to do instead
Before answering, underline the time cue: 'new cases'/'over one year'/'risk of developing' points to incidence; 'existing cases'/'at a given time'/'burden' points to prevalence. Write the definition of each next to the question stem.
Calculating a screening test's positive predictive value from sensitivity and specificity alone, without using prevalence.
Candidates memorize Se/Sp as the 'important' test properties and assume predictive values follow directly, forgetting that PPV/NPV depend on the population's disease prevalence.
What to do instead
Build a 2x2 table with a hypothetical 1,000-person population and insert the given prevalence first, then apply Se and Sp. Derive PPV/NPV from the table rather than from formulas in the abstract.
Confusing the direction of a protective association when interpreting an odds ratio or relative risk below 1.0, labeling OR = 0.4 as 'increased risk'.
Candidates anchor on the magnitude (0.4 looks like a number) rather than on whether the estimate is above or below the null value of 1.0.
What to do instead
State the null value explicitly (1.0 for ratio measures, 0 for difference measures) and say 'less than 1 = lower odds/risk in the exposed' before reading the number. Practice converting OR = 0.4 to '60% lower odds'.
Selecting a study design by the disease rather than by the question's timing and sampling structure-e.g., choosing a cohort design when the question describes sampling by outcome status and looking back at exposures.
Candidates associate certain diseases with certain designs (e.g., rare disease = case-control) but do not extract the sampling logic from the vignette, so they guess from the disease name.
What to do instead
For every design question, write two lines: 'Sampled by: exposure or outcome?' and 'Timing: forward or backward?'. Use that grid to map to cohort (exposure-sampled, forward), case-control (outcome-sampled, backward), or cross-sectional (both at once).
Treating a p-value as the probability that the null hypothesis is true or as a measure of effect size when judging whether a program result is meaningful.
Statistical significance language is familiar and feels authoritative, so candidates default to 'p < 0.05 means the intervention works' without separating statistical from practical significance.
What to do instead
For each result, answer three separate questions: What is the effect size and its confidence interval? Is the CI compatible with no effect? Does the effect size matter for the population? Reject any option that equates a small p-value with a large or important effect.
Applying a tertiary-prevention label to a primary-prevention activity (or vice versa) because the activity 'seems clinical', such as calling smoking-cessation counseling after a myocardial infarction primary prevention.
Candidates classify by the setting (clinic vs community) rather than by the point in the disease process at which the activity acts.
What to do instead
Ask 'Has disease occurred yet, and if so, are we preventing the event or the complication?'. Pre-event in a well person = primary; early detection in an asymptomatic person = secondary; limiting disability after disease = tertiary. Classify by disease stage, not location.
Confusing the three core public health functions or the essential services when a question asks which function a described agency activity represents (assessment vs policy development vs assurance).
The categories overlap in plain language-gathering data, making plans, and delivering services all sound like 'public health work'-so candidates choose by intuition rather than by the definitional anchor.
What to do instead
Memorize one anchor verb per function: Assess = monitor/diagnose/investigate; Policy development = inform/plan/develop policy; Assurance = enforce/link/assure a competent workforce and access. Match the activity's verb to the anchor.
Answering an ethics item by choosing the principle that 'sounds most benevolent' rather than the principle the scenario actually tests, e.g., selecting beneficence when the issue is respect for autonomy or justice.
All four principles are morally positive, so candidates pick the most emotionally appealing one instead of identifying which duty is in tension in the scenario.
What to do instead
Identify the conflict first: 'Whose choice is being overridden?' (autonomy), 'Who bears burden vs benefit?' (justice), 'Is harm being avoided vs good promoted?' (nonmaleficence vs beneficence). Name the tension, then pick the principle that resolves it.
Treating a crude rate as comparable across populations without recognizing confounding by age or another structure variable, and failing to select a standardized or stratified result.
Crude rates are the headline number and look authoritative; candidates forget that populations differ in composition and that standardization exists to remove that difference.
What to do instead
Whenever two populations are compared, ask 'Do they differ in age (or another strong risk factor)?'. If yes, look for an age-adjusted/standardized rate or a stratified analysis among the answer choices before selecting the crude comparison.
Choosing an intervention or policy option based on its effectiveness alone while ignoring feasibility, acceptability, equity, and cost-or the reverse, rejecting the most effective option for a minor logistical reason.
Candidates treat effectiveness as the single 'correct' criterion and overlook that practice and policy questions test whether a recommendation is implementable and equitable in the described context.
What to do instead
Score each candidate option against four criteria-effectiveness, feasibility, acceptability/equity, cost-and pick the option that wins on balance for the stated population and resources. Flag any option that is effective but infeasible in the scenario.