Treat ABPM preventive medicine preparation as decision conversion rather than fact recall. The core difficulty is that a single vignette bundles an intervention, a study, and a claim together, and the correct answer depends on naming the point of intervention, the design's direction of inquiry, and the measure's exact meaning. Anchor each item to a named framework — prevention levels, explicit screening criteria, cohort versus case-control logic, and the major threats to validity — then practice converting stems into decisions with the worked scenarios and rubric below. One caution before any administrative step: the abbreviation ABPM is shared by a preventive medicine board and a pain medicine board, so confirm the exact issuing organization on its own official pages before relying on eligibility or logistics information.
How to classify primary, secondary, and tertiary prevention without swapping labels
Classify by the intervention's timing relative to disease: before onset is primary, during asymptomatic early disease is secondary, and after diagnosis to limit damage or disability is tertiary.
Primary prevention acts before disease exists: immunization, counseling to quit smoking, or removing a workplace noise hazard. Secondary prevention detects disease early in people without symptoms, such as blood pressure screening or organized cancer screening. Tertiary prevention manages diagnosed disease to prevent complications, rehabilitation after a stroke being the classic example. The boundary between primary and secondary is whether disease already exists, not whether the program sounds preventive.
Scenario: a worksite program offers on-site blood pressure checks and refers employees with elevated readings to treatment. A plausible mistake is labeling the whole program primary prevention because it happens at work and is framed as wellness. The better decision is to classify component by component: the blood pressure check is secondary prevention because it detects existing asymptomatic disease, while a smoking-cessation session in the same program is primary. Mixed-program stems test exactly this skill, and the same reasoning transfers to occupational and community health vignettes.
Judging a screening proposal with explicit criteria, not enthusiasm
Screening questions reward structured appraisal. Test the proposal against named criteria: disease burden, a detectable preclinical phase, an acceptable test, and treatment given early that improves outcomes.
Walk each criterion in order: is the condition an important problem, is its natural history understood, does an early detectable stage exist, is a suitable and acceptable test available, and does treatment at the detected stage outperform treatment after symptoms appear? The last criterion is the one stems hide. A test can be highly accurate yet useless if early treatment offers no advantage, so the disease-and-treatment pair matters more than test statistics alone.
Scenario: a memo reports that patients whose cancer was found by screening survived five years on average, versus one year for symptom-detected patients, and concludes that screening saves lives. The plausible mistake is accepting survival from diagnosis as evidence of benefit — lead-time bias means screening simply moved the diagnosis earlier without changing the date of death. The better decision is to ask whether cause-specific mortality fell in the whole screened population. The same reasoning underlies length-time bias and overdiagnosis, so separating survival from mortality is the repeatable skill worth drilling.
Matching the study design to the question the stem is asking
Design identification is a matching task: cohort for incidence and temporality, case-control for rare outcomes, cross-sectional for prevalence, randomized trial for treatment effect, ecologic for population comparisons.
Infer the design from the direction of sampling rather than vocabulary. If the study starts with an outcome and looks back at exposure, it is case-control; if it starts with exposure and follows forward to outcome, it is cohort; if exposure and disease are measured in one snapshot, it is cross-sectional; if participants are randomly assigned, it is a trial. A stem can describe a design that sounds rigorous while its actual direction answers a different question than the one asked.
Use the table below as a decision aid. Read the question column first, then check whether the described sample, timing, and comparison match the design you selected. When two rows seem plausible, identify which limitation the stem is hinting at: loss to follow-up points to a cohort, recall problems point to a case-control study, and an inability to establish temporality points to a cross-sectional survey. Let the limitation confirm the design rather than guessing from tone.
| Question the stem asks | Design that fits | Signature limitation |
|---|---|---|
| What is the incidence of disease in exposed versus unexposed groups? | Cohort | Loss to follow-up; long duration |
| Why do people who already have the disease differ in past exposure? | Case-control | Recall bias; difficult control selection |
| How common are exposure and disease right now? | Cross-sectional | Cannot establish temporality |
| Does the intervention itself cause the benefit? | Randomized controlled trial | Limited generalizability; feasibility of randomization |
| Do population-level exposure rates track population outcomes? | Ecologic | Ecologic fallacy: group patterns may not hold for individuals |
Computing measures of association from a two-by-two table
Risk, relative risk, odds ratio, attributable risk, and number needed to treat answer different questions. Build the two-by-two table first, then select the measure the question actually names.
Worked example: in a hypothetical cohort of 1,000 exposed and 1,000 unexposed workers, disease occurs in 100 exposed and 50 unexposed. The risks are 10 percent and 5 percent, so the relative risk is 2.0 and the risk difference is 5 excess cases per 100. The attributable fraction among the exposed is 50 percent. If the same numbers came from a randomized trial, the 5 percent absolute risk reduction yields a number needed to treat of 20. Each number answers a distinct question: strength of association, excess burden, fraction preventable, or effort per benefit.
Scenario: a colleague reads a case-control study reporting an odds ratio of 3.0 and tells a patient that the exposure triples disease risk. The plausible mistake is treating an odds ratio as an incidence-based relative risk without checking whether the outcome is rare, which is the condition under which the two approximately coincide. The better decision is to report it as an odds ratio, an estimate of relative risk under a stated rare-disease assumption. Precision about what a measure can say also changes population-level claims, such as how much disease in the whole population is attributable to an exposure.
Sensitivity, specificity, PPV, and why prevalence moves the answers
Sensitivity and specificity are properties of the test and stay stable across settings; predictive values are properties of the patient and shift with disease prevalence. Stems exploit that split.
Sensitivity is the proportion of diseased people who test positive; specificity is the proportion of disease-free people who test negative. Lowering the cutoff to raise sensitivity lowers specificity, so screening and diagnostic uses often sit at different points on the same tradeoff. The positive predictive value, the proportion of positive results that are true disease, depends on how many diseased people are in the tested group in the first place.
Worked example: a test that is 90 percent sensitive and 95 percent specific is used in a specialty clinic with 10 percent prevalence and in a general population with 1 percent prevalence. Among 1,000 specialty patients, 90 true positives meet 45 false positives, giving a positive predictive value near 67 percent; among 1,000 general-population patients, 9 true positives meet roughly 50 false positives, giving a value near 15 percent. The plausible mistake is quoting the clinic value as a property of the test. The better decision is to state that predictive values travel with prevalence, which is why a positive result carries different meaning before and after a screening referral.
Naming the threat to validity: confounding, bias, or effect modification
Validity stems ask you to name the flaw. Confounding mixes in a third variable's effect; selection and information bias distort who enters or what is recorded; effect modification means the association genuinely differs by subgroup.
Distinguish them with concrete occupational examples. The healthy worker effect, in which employed populations are systematically healthier than the general population, is selection bias. Recall bias, in which cases remember past exposures differently from controls, is information bias. Smoking distorting an observed alcohol-lung cancer association is confounding. If the association is strong in men and absent in women, that is effect modification, and reporting a single pooled estimate hides real, opposite-direction or magnitude-varying effects rather than distorting them.
Use a fixed decision procedure. First ask who entered and stayed in the study, which surfaces selection problems. Second ask how exposure and outcome were measured, which surfaces information problems. Third ask whether a third variable plausibly travels with both exposure and outcome, which surfaces confounding. Fourth ask whether stratified estimates differ meaningfully, which surfaces effect modification. Note that randomization controls confounding by design, but a trial can still suffer attrition or measurement problems, so a randomized label does not immunize a study against every threat.
A headline-critique exercise, self-check rubric, and study sequence
Close each session by critiquing a real health news item: identify its design, measure, and one validity threat, then score yourself. Readiness means repeatable correct classifications, not finished flashcard decks.
Exercise: collect three health news items that cite a study. For each, write the likely study design, the association or test measure used, one bias or confounder the design cannot escape, and whether the headline's claim concerns individuals or populations. Expected observations: relative risks converted into absolute claims, ecologic comparisons presented as individual advice, and the survival-versus-mortality confusion from the screening section. If you complete all four steps for three items in one sitting without notes, the classification skills are consolidating.
Self-check rubric: score each of the four steps 0 for a guess, 1 for correct with hesitation, and 2 for correct and justified; a useful milestone before mixed review is 5 of 8 on two consecutive items. These milestones are learning signals, not predictions of any exam outcome. For an adaptable sequence, work in the order of this guide: frameworks first, then calculations, then validity threats and ethics reasoning, then mixed timed sets. One administrative caution: the ABPM abbreviation belongs to more than one organization. The site at the supplied abpm.org URL is the American Board of Pain Medicine, a different credential and a different board. To confirm eligibility, applications, or testing logistics for preventive medicine certification, locate the preventive medicine board's own official site independently and verify you are on the correct organization's pages before relying on any detail.
- Weeks 1-2: rebuild the prevention-level and screening-criteria frameworks; classify twenty short vignettes.
- Weeks 3-4: drill study designs and two-by-two calculations using tables you construct yourself.
- Weeks 5-6: add validity-threat reasoning and public health ethics to the same vignettes.
- Final stretch: run mixed timed sets plus the headline-critique exercise until you reach the rubric milestone twice.
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.