Study guide · BEP
Biostatistics & Epidemiology/Population Health
4 to 6% of Step 1 · 14 of the 280 questions on our mock · 7 sections
This system covers biostatistics, epidemiology, population health, and how to read medical research. It is about 4% to 6% of Step 1, and its items test practice-based learning and improvement. Many questions give a 2 x 2 table, a results table, or a short study description. You must name the design, find the bias, or calculate a measure such as sensitivity, relative risk, or the number needed to treat. Most calculations are short if you set up the table carefully and check which row or column is the denominator.
A study feature written by ExamDragon, not part of the real exam. The headings follow the USMLE Content Outline for this system, in our words. Pharmacology here is about how drugs work, as on Step 1.
At a glance
BEPof Step 1, the published range
- On our mock
- 14
- Sections
- 7
- Key terms
- 10
What this guide covers
1Measures of disease frequency and health status
- Incidence = new cases / population at risk over a period of time. Prevalence = all existing cases / total population at one time.
- Prevalence is about incidence x average duration. A treatment that prolongs life without cure raises prevalence; a cure or a faster death lowers it.
- Crude rates use the whole population. Adjusted (standardized) rates remove differences in age structure so populations can be compared. The standardized mortality ratio is observed deaths / expected deaths.
- Infant mortality rate = deaths under 1 year of age / live births. Maternal mortality ratio = maternal deaths / live births, usually per 100,000.
- Years of potential life lost give more weight to early deaths. Quality-adjusted and disability-adjusted life years combine length and quality of life.
- Attack rate = ill people / people exposed during an outbreak. Herd immunity protects unvaccinated people when enough of the population is immune.
- Worked exampleA disease has an incidence of 2 per 1000 per year and lasts 5 years on average. Prevalence is about 2 x 5 = 10 per 1000. If a drug doubles survival to 10 years, prevalence rises to about 20 per 1000, while incidence stays at 2 per 1000.
- Where questions turnDo not let a rising prevalence convince you that more new cases are occurring. Ask whether the duration of disease changed or whether detection improved.
2Study designs
- Case report and case series: describe patients; no comparison group; they generate hypotheses.
- Cross-sectional study: exposure and outcome are measured at one time; it gives prevalence; it cannot show which came first.
- Ecological study: the unit is a population; beware the ecological fallacy.
- Case-control study: start with the outcome and look back for exposure; good for rare diseases; gives an odds ratio; prone to recall bias.
- Cohort study: start with exposure and follow for the outcome (prospective or retrospective); gives incidence and relative risk; good for rare exposures.
- Randomized controlled trial: randomization balances known and unknown confounders; blinding reduces information bias; analyze by intention to treat. Noninferiority trials test whether a new treatment is not worse by more than a set margin.
- Systematic review and meta-analysis: pool studies; forest plots show each study and the pooled result; funnel plot asymmetry suggests publication bias.
- Worked exampleResearchers pick 150 people with a rare cancer and 300 people without it, then ask about past pesticide use. They started from the outcome, so this is a case-control study. They report an odds ratio, not a relative risk.
- Where questions turnName the design by how participants were selected, not by when the data were collected. A cohort study can be retrospective if participants are grouped by past exposure.
3Measures of association and impact
- Relative risk (RR) = incidence in exposed / incidence in unexposed (cohort studies and trials). Odds ratio (OR) = ad / bc (case-control studies); it is close to the RR when the disease is rare.
- Attributable risk (risk difference) = incidence in exposed - incidence in unexposed. Attributable risk percent = (RR - 1) / RR.
- Absolute risk reduction (ARR) = control risk - treated risk. Relative risk reduction = ARR / control risk. Number needed to treat = 1 / ARR, rounded up.
- Number needed to harm = 1 / absolute risk increase, rounded up.
- Hazard ratio: the ratio of event rates over time from survival analysis; 1 means no difference.
- Population attributable risk shows how much disease in the whole population is due to an exposure; it depends on how common the exposure is.
- Worked exampleIn a trial, deaths are 10% with placebo and 8% with the drug. ARR = 2%. NNT = 1 / 0.02 = 50. Relative risk reduction = 2 / 10 = 20%. A 20% relative reduction sounds large, but 50 people must be treated to prevent one death.
- Where questions turnDo not confuse relative and absolute measures. The relative risk reduction often stays similar when baseline risk changes, but the NNT changes a lot.
4Testing and screening
- Sensitivity = TP / (TP + FN), the share of diseased people who test positive. A very sensitive test, when negative, helps rule a disease out.
- Specificity = TN / (TN + FP). A very specific test, when positive, helps rule a disease in.
- PPV = TP / (TP + FP). NPV = TN / (TN + FN). Predictive values change with prevalence; sensitivity and specificity do not.
- LR+ = sensitivity / (1 - specificity). LR- = (1 - sensitivity) / specificity. Posttest odds = pretest odds x LR.
- Lowering a cutoff raises sensitivity and lowers specificity. The ROC curve plots sensitivity against 1 - specificity; the area under the curve measures overall accuracy.
- Screening biases: lead-time bias (earlier diagnosis without longer life) and length-time bias (screening finds slow-growing disease). Compare mortality, not survival from diagnosis.
- Reliability is repeatability (precision); validity is accuracy against a reference standard.
- Worked exampleA test has a sensitivity of 90% and a specificity of 80%. LR+ = 0.9 / 0.2 = 4.5. With a pretest probability of 10%, pretest odds are 0.1 / 0.9 = 0.11. Posttest odds = 0.11 x 4.5 = 0.5. Posttest probability = 0.5 / 1.5 = 33%.
- Where questions turnRead the orientation of the table before you calculate. Sensitivity and specificity use the disease columns; predictive values use the test rows.
5Statistics: distributions, hypothesis testing, and errors
- Normal distribution: about 68%, 95%, and 99.7% of values fall within 1, 2, and 3 standard deviations of the mean.
- Right (positive) skew: mean > median > mode. Left skew: mean < median < mode. The median resists outliers.
- Type I error (alpha): rejecting a true null hypothesis, a false positive. Type II error (beta): missing a true difference. Power = 1 - beta.
- Power rises with a larger sample, a larger effect, less variability, and a higher alpha. Many tests at once raise the chance of a false positive.
- A 95% CI for a difference that includes 0, or for a ratio that includes 1, is not significant at p < 0.05.
- Tests: the t-test compares two means; ANOVA compares three or more means; the chi-square test compares proportions.
- Correlation r runs from -1 to +1 and shows the strength of a linear relationship; r squared is the share of variation explained. Correlation does not prove causation.
- Worked exampleA trial of 30 patients finds p = 0.30 for a drug effect. It may have too little power. The result does not prove that the drug has no effect; a type II error is possible.
- Where questions turnA p value is not the probability that the null hypothesis is true. It is the probability of results at least this extreme if the null hypothesis were true.
6Bias, confounding, and validity
- Selection bias: the study groups do not represent the target population (for example, Berkson bias, the healthy worker effect, loss to follow-up).
- Information bias: errors in measuring exposure or outcome (recall bias; observer bias when assessors are not blinded).
- Confounding: a third factor linked to both exposure and outcome, and not on the causal path. Control it with randomization, matching, restriction, stratification, or multivariable analysis.
- Hawthorne effect: people change behavior when observed. Placebo effect: improvement from belief in a treatment.
- Regression to the mean: extreme values tend to move toward the average when measured again.
- Internal validity: the result is true for the study sample. External validity (generalizability): the result applies to other populations. Efficacy is benefit under ideal conditions; effectiveness is benefit in real practice.
- Effect modification is different from confounding: the effect truly differs across subgroups.
- Worked exampleA study finds that carrying matches is linked to lung cancer. Smoking explains both. When the analysis is done separately for smokers and nonsmokers, the link disappears. Smoking was a confounder.
- Where questions turnMatch the bias to the design. Recall bias belongs to case-control studies, loss to follow-up to cohorts and trials, and lead-time bias to screening.
7Prevention, public health, and research ethics
- Primary prevention stops disease before it starts (vaccines, quitting smoking). Secondary prevention finds disease early (screening). Tertiary prevention limits harm from established disease (rehabilitation).
- Community-level interventions include tobacco and soda taxes, smoke-free laws, school policies, and access to healthy food and safe environments.
- Outbreak investigation: confirm the outbreak, define a case, describe cases by time, place, and person, form and test hypotheses, control the source, and report.
- Vaccine effectiveness = (attack rate in unvaccinated - attack rate in vaccinated) / attack rate in unvaccinated.
- Drug trial phases: I (safety in a small group, often healthy people), II (efficacy and dose), III (large comparison for approval), IV (postmarketing surveillance).
- Research needs informed consent, institutional review board (IRB) review, and protection of health data (HIPAA). A data safety monitoring board may stop a trial early for clear benefit, harm, or futility.
- Worked exampleAnnual low-dose CT for a former heavy smoker aims to find lung cancer before symptoms. That is secondary prevention. Helping him stay off cigarettes is primary prevention.
- Where questions turnScreening is secondary prevention, even though it is done in people who look healthy. Vaccination is primary prevention.
Key terms
- Incidence
- The number of new cases in a population at risk over a period of time.
- Prevalence
- The proportion of a population that has a disease at one time.
- Confounder
- A third factor linked to both exposure and outcome that distorts their apparent relationship.
- Odds ratio
- The odds of exposure in cases divided by the odds of exposure in controls.
- Likelihood ratio
- How much a test result changes the odds of disease.
- Power
- The probability that a study finds a true difference when one exists.
- Confidence interval
- A range of values that likely contains the true value.
- Intention-to-treat analysis
- Analyzing every participant in the group to which they were randomized.
- Lead-time bias
- Apparent longer survival caused only by earlier diagnosis.
- Number needed to treat
- How many patients must be treated to prevent one bad outcome.
After this guide: After this guide, a learner should be able to name a study design, spot its main bias, and calculate and interpret the common measures from a 2 x 2 table.
Practise Biostatistics & Epidemiology/Population HealthSources: The USMLE Content Outline and the Step 1 content specifications (usmle.org); the explanations are ours. Every source this path is built on.
