ABPM Public Health common mistakes
ABPM Public Health · Understand the traps and how to avoid them.
Computing a screening test's positive predictive value from sensitivity and specificity alone, without using the stated prevalence, and reporting the sensitivity as if it were the PPV.
Candidates memorize the 2x2 table formulas but treat PPV as an intrinsic test property; they forget that PPV depends on prevalence while sensitivity and specificity do not.
What to do instead
Before answering any screening question, write PPV = (sens x prev) / [(sens x prev) + ((1 - spec) x (1 - prev))]. If prevalence is not given, it cannot be computed. Check whether the question asks about the test's discrimination (sens/spec) or its clinical usefulness in a population (PPV/NPV).
Reading a relative risk or odds ratio of 2.0 as 'twice as many cases' or as a large absolute burden, ignoring the baseline risk and confidence interval.
Relative measures feel intuitive and are reported prominently, so candidates skip the absolute risk difference and fail to notice when the CI crosses 1.0.
What to do instead
For every effect measure, compute the absolute difference (risk in exposed minus risk in unexposed) and state whether the 95% CI excludes the null. If the CI includes 1.0 (or 0 for a difference), the association is not statistically significant at alpha = 0.05.
Confusing incidence with prevalence when interpreting a study, e.g., calling a cross-sectional proportion an incidence rate or using prevalence as the denominator for a risk calculation.
Both are expressed as proportions or per-capita rates and the words are used loosely in stems; candidates pattern-match on 'rate' without checking the design.
What to do instead
Ask two questions: does the measure have a time-at-risk denominator (incidence) or a point-in-time count (prevalence)? And does prevalence approximately equal incidence x duration for stable diseases? Use that identity to detect swapped measures.
Selecting a study design that cannot answer the stated question, such as choosing a case-control design to estimate incidence, or a cross-sectional survey to establish temporality.
Design names are memorized as a list of pros and cons rather than mapped to the specific inferential goal (association vs. incidence vs. causality).
What to do instead
Match the question to the design's output: prevalence to cross-sectional, incidence and temporality to cohort, rare outcomes to case-control, causal effect to RCT. If the stem asks for incidence or a temporal sequence, eliminate case-control and cross-sectional options first.
Applying a one-tailed test, or a paired test to unpaired data, and then reporting a p-value without checking the assumptions of the chosen procedure.
Candidates default to the test they practiced most (often the two-sample t-test) and skip the assumption checklist under time pressure.
What to do instead
Run a fixed decision sequence: what is the outcome type, how many groups, are observations paired, are variances equal, are expected cell counts adequate? State the assumption aloud, then choose the test. Default to two-tailed unless the stem explicitly specifies a directional hypothesis.
Interpreting a p-value as the probability that the null hypothesis is true, or as the probability the result will replicate.
The definition is counterintuitive and often stated loosely in study reports, so the misreading feels correct.
What to do instead
Use the fixed sentence: a p-value is the probability of data at least this extreme if the null were true. It says nothing about the probability the null is true. For that reasoning, look for a confidence interval or a Bayesian statement, not a p-value.
Choosing a health services administration answer that maximizes clinical effectiveness while ignoring cost, access, or feasibility constraints stated in the scenario.
Clinicians are trained to prioritize the best clinical outcome and underweight the resource, reimbursement, and equity dimensions that administration questions test.
What to do instead
Before selecting, list the constraints in the stem (budget, staffing, payer, population served) and eliminate any option that violates one. In administration items, the best answer is usually the one that is effective and feasible, not the one that is merely most effective.
Confusing Medicare and Medicaid eligibility, financing, and administration when answering coverage questions.
Both are public programs with overlapping-sounding names, and candidates recall fragments without a clean structural contrast.
What to do instead
Anchor on a single contrast: Medicare is federal, age- or disability-based, and not income-tested; Medicaid is joint federal-state, income-based, and administered by states under federal rules. Classify the person in the stem by age, disability, and income before choosing.
Treating an environmental exposure as a single acute event and ignoring latency, dose-response, and route of exposure when interpreting risk.
Environmental questions often describe a release or hazard, and candidates reason about immediate harm rather than the chronic, delayed outcomes that drive most environmental disease.
What to do instead
For each exposure item, name the route (inhalation, ingestion, dermal), the dose and duration, and the expected latency. If the outcome is chronic (e.g., cancer, neurodevelopmental), look for a latency period and a dose-response relationship in the correct answer.
Answering social and behavioral items with an individual-level explanation when the stem describes a population-level pattern, or vice versa.
Both levels use familiar language, and candidates pick the plausible-sounding individual behavior (e.g., poor diet) over the structural determinant (e.g., food access) without checking the level of the question.
What to do instead
Identify the unit of analysis in the stem first: individual, community, or policy. Then choose the explanation at that level. If the stem shows a group difference across neighborhoods or income levels, prefer the structural or upstream determinant over an individual behavior.