Step 1 Public Health Sciences
Public Health Sciences on USMLE Step 1 covers epidemiology, biostatistics, ethics, healthcare delivery, and quality improvement. Questions span study design, statistical interpretation, bioethical reasoning, and healthcare systems — roughly 10–15% of the exam. Most vignettes integrate these concepts with clinical scenarios rather than testing definitions in isolation. If you are looking for high-yield biostatistics and epidemiology topics for Step 1, the look-alike concept pairs below decide more questions than any single formula.
The hardest part is that many concepts look similar until the details matter. Odds ratios versus relative risk, PPV versus sensitivity, capacity versus competence, lead-time bias versus length-time bias — the exam specifically probes whether you know the distinction. Students consistently confuse PPV with sensitivity: a test can be highly sensitive yet have terrible PPV in a low-prevalence population, because prevalence directly determines PPV. Getting comfortable with 2x2 tables, bias identification in study design vignettes, and the autonomy-first ethical framework will cover a large share of the questions.
Step 1 ethics vignettes are often more straightforward than students expect — the answer almost always defaults to respecting the competent patient's wishes — but the traps are specific. Another common misconception: students conflate clinical capacity with legal competence, when capacity is assessed at the bedside by any physician while competence is a court determination. Medicare versus Medicaid eligibility, NNT calculation, and the difference between RCA and FMEA are the kind of precise details that separate a good USMLE score from a great one. Do not treat this section as lower yield just because it is not organ-based.
Observational Study Designs
Ecological, cross-sectional, case report, and case series designs — units of analysis and why none establish causation.
- Confuses ecological study unit of analysis (population) with individual-level data
- Assumes cross-sectional design can establish temporal sequence between exposure and outcome
Case-Control Studies
Outcome-to-exposure directionality, odds ratio calculation, and the biases that differentially distort case-control results.
- Incorrectly calculates relative risk instead of odds ratio from case-control data
- Confuses case-control directionality (outcome → exposure) with cohort directionality (exposure → outcome)
Cohort Studies
Exposure-to-outcome directionality, relative risk versus attributable risk, and the Framingham prototype.
- Confuses retrospective cohort design with case-control design because both use past data
- Conflates relative risk (strength of association) with attributable risk (excess absolute risk) from cohort data
Randomized Controlled Trials
Randomization's role in distributing unknown confounders, trial phases, blinding levels, and intention-to-treat versus per-protocol analysis.
- Misattributes randomization's power to balancing measured variables rather than distributing unknown confounders
- Confuses Phase II (preliminary efficacy/dosing, small sample) with Phase III (large-scale efficacy)
Systematic Review and Meta-Analysis
Quantitative pooling versus qualitative synthesis, and why publication bias and heterogeneity can corrupt meta-analytic conclusions.
- Conflates systematic review (qualitative synthesis) with meta-analysis (quantitative pooling)
- Assumes combining many studies in a meta-analysis eliminates publication bias
Selection Bias
Berkson bias, healthy worker effect, and Neyman bias — each distorts which subjects enter or remain in a study.
- Misattributes Berkson bias to disease severity rather than differential hospitalization rates for exposure and disease
- Reverses the direction of healthy worker effect bias (underestimates rather than overestimates occupational risk)
Measurement and Information Bias
Recall bias, observer bias, and Hawthorne effect — how information is distorted after subjects are enrolled.
- Fails to recognize that recall bias is a particular threat to case-control studies, not prospective cohort studies
- Confuses Hawthorne effect (participant behavior change due to observation) with observer/interviewer bias (investigator expectation)
Confounding
Three required conditions for a confounder, design versus analysis control strategies, and how effect modification differs from bias.
- Treats effect modification as a bias to eliminate rather than a real finding to report separately by stratum
- Assumes matching alone fully controls confounding without recognizing the need for matched analytic methods
Lead-Time and Length-Time Bias
Lead-time and length-time bias make screened patients appear to survive longer without any true mortality benefit.
- Interprets longer post-diagnosis survival in screened patients as proof of benefit, missing lead-time bias
- Misunderstands length-time bias as relating to duration of screening rather than enrichment for indolent disease
Sensitivity, Specificity, and Cutoff / ROC
SNOUT and SPIN, the sensitivity-specificity tradeoff at different cutoffs, and ROC curve interpretation.
- Reverses SNOUT and SPIN mnemonics, applying specificity to rule-out and sensitivity to rule-in
- Confuses sensitivity/specificity (prevalence-independent) with PPV/NPV (prevalence-dependent)
Predictive Values (PPV and NPV)
Prevalence determines PPV and NPV — even a highly accurate test yields poor PPV in a low-prevalence population.
- Confuses PPV/NPV as test-intrinsic properties rather than prevalence-dependent values
- Fails to recognize the rare-disease screening paradox where low prevalence collapses PPV
Likelihood Ratios
Prevalence-independent LR+ and LR- update pretest odds to posttest odds using Bayesian reasoning.
- Confuses LRs as prevalence-dependent when they are actually test-intrinsic
- Misunderstands the direction of effect for LR- versus LR+
Risk and Effect Measures (OR, RR, ARR, NNT)
OR, RR, ARR, RRR, NNT, and NNH — when each applies and why RRR alone misleads without baseline risk context.
- Confuses when OR is a valid approximation of RR versus when it overestimates effect size
- Overinterprets RRR without considering baseline risk or ARR
Hypothesis Testing, Power, and Confidence Intervals
Type I and II errors, correct p-value interpretation, power determinants, and CI-crosses-null rules for ratios versus differences.
- Misdefines p-value as the probability the null hypothesis is true rather than a conditional probability
- Swaps definitions of Type I (false positive) and Type II (false negative) errors
Statistical Tests (t-test, ANOVA, Chi-Square, Correlation)
Matching outcome type and group count to the correct test — t-test, ANOVA, chi-square, or regression.
- Applies chi-square to continuous outcomes instead of t-test or ANOVA
- Defaults to parametric tests without checking the normality assumption
Incidence, Prevalence, and Mortality
New cases versus existing cases, case fatality versus mortality rate, and how treatment shifts prevalence without changing incidence.
- Confuses incidence (new cases) with prevalence (all existing cases)
- Confuses population-based mortality rate with case fatality rate among diagnosed patients
Distributions — Normal, Skewed, Mean vs Median
Skewed distributions shift the mean toward the tail — median better represents skewed data than mean.
- Incorrectly places the mean toward the mode rather than toward the tail in skewed distributions
- Defaults to mean as a summary statistic even when data are skewed, where median is more appropriate
Prevention Levels and Screening Criteria
Primary through quaternary prevention, Wilson-Jungner screening criteria, and why sensitivity trumps specificity for screening tests.
- Confuses tertiary prevention (managing established disease) with secondary prevention (screening/early detection)
- Prioritizes specificity over sensitivity for screening tests, reversing the correct requirement
Four Core Principles (Autonomy, Beneficence, Non-Maleficence, Justice)
Autonomy, beneficence, non-maleficence, and justice — autonomy is the default priority for competent adult patients.
- Overrides patient autonomy with physician beneficence in vignettes involving competent patient refusal
- Conflates non-maleficence (avoid harm) with beneficence (promote good) as a single principle
Capacity vs Competence
Appelbaum's four criteria assess clinical capacity; legal competence is a court determination, not a physician call.
- Conflates clinical capacity (physician-assessed, time-specific) with legal competence (court-determined)
- Assumes psychiatric illness automatically negates decisional capacity without individual assessment
Informed Consent (Elements and Exceptions)
Valid consent requires disclosure, understanding, voluntariness, and alternatives — emergency and waiver exceptions are narrow.
- Confuses any emergency with the specific conditions required for the emergency exception to informed consent
- Incorrectly applies therapeutic privilege as a legitimate reason to withhold information during consent
Confidentiality and HIPAA
Tarasoff, mandatory reporting, and HIPAA exceptions define when confidentiality must or may be breached.
- Overapplies Tarasoff duty to warn to non-specific or non-identifiable threats
- Assumes family members have automatic rights to patient information regardless of patient consent
Minors — Consent, Confidentiality, Mature and Emancipated Minor
Emancipation, the mature minor doctrine, and which services — STIs, contraception, substance use — minors access without parental consent.
- Confuses financial independence or living away from home with legal emancipation
- Incorrectly requires parental consent for confidential services minors can access independently
End-of-Life — Advance Directives, DNR, Palliative/Hospice, Double Effect
Healthcare proxy trumps next of kin; DNR restricts only resuscitation; double effect justifies opioid analgesia with foreseen but unintended risk.
- Misinterprets DNR as a global 'do not treat' order rather than a specific resuscitation restriction
- Prioritizes next of kin over a formally designated healthcare proxy
Error Taxonomy, Swiss Cheese, and Just Culture
Slips versus mistakes, active versus latent errors, and near-miss versus sentinel event distinctions drive just culture responses.
- Confuses slips (execution errors) with mistakes (planning/knowledge errors)
- Confuses near-miss (no patient contact) with an adverse event that caused minor harm
QI Frameworks and Communication — RCA, FMEA, PDSA, SBAR, Checklists
RCA investigates past events retrospectively; FMEA prevents future failures prospectively; PDSA cycles iterative small-scale improvement.
- Confuses RCA (retrospective, post-event) with FMEA (prospective, pre-event) in timing and purpose
- Conflates Six Sigma (reduce variability/defects) with Lean (eliminate waste/improve flow)
Medicare and Medicaid
Medicare covers age and disability regardless of income; Medicaid covers low-income populations through joint federal-state funding.
- Confuses Medicare (age/disability-based) with Medicaid (income-based) eligibility criteria
- Misattributes outpatient prescription drug coverage to Part B instead of Part D
Managed Care — HMO, PPO, POS, ACO
HMO gatekeeping versus PPO self-referral, capitation versus fee-for-service incentives, and ACO shared savings structure.
- Confuses HMO gatekeeper model with the more flexible PPO structure that allows self-referral
- Reverses the incentive structures of fee-for-service versus capitation payment models
Social Determinants of Health
Five SDOH domains — economic, education, neighborhood, social, health care access — directly shape clinical outcomes and require community-level responses.
- Responds to SDOH concerns with medical specialist referral rather than community resource connection
- Dismisses SDOH as non-clinical background rather than recognizing their direct impact on patient outcomes
Evidence-Based Medicine
EBM integrates research evidence, clinician expertise, and patient values; USPSTF Grade C means selective use, not avoidance.
- Reduces EBM to research evidence alone, omitting clinician expertise and patient values
- Misinterprets USPSTF Grade C as a recommendation against the service rather than selective use with shared decision-making
See how your Anki deck covers Public Health Sciences.
Upload your deck for a free audit →