Table of Contents
Introduction to Epidemiology for USMLE Beginners
Epidemiology is the science that studies how disease is distributed in populations and the factors that influence or determine this distribution. For the USMLE, epidemiology questions usually test your ability to interpret data, understand study designs, and apply measures of disease frequency and association to clinical scenarios. You are not expected to be a statistician, but you must be fluent with the core concepts and formulas that appear again and again in exam questions.
This chapter focuses on what is unique to epidemiology: how we measure disease occurrence, how we compare groups, how different study designs work, and what can go wrong through bias and confounding. The detailed mathematical treatment is covered under biostatistics, so here the emphasis is on concepts and interpretation.
Measures of Disease Frequency
A central task in epidemiology is to quantify how often health events occur. The three foundational measures are prevalence, incidence, and mortality. They all look similar but answer different questions and are used in different contexts on the exam.
Prevalence
Prevalence describes how many people in a population have a disease at a given time. It is like a snapshot. USMLE questions often use prevalence when they are asking about burden of disease or when they provide cross sectional data.
The basic formula is:
$$\text{Prevalence} = \frac{\text{Number of existing cases}}{\text{Total population at that time}}$$
Point prevalence refers to one specific point in time, for example: “What proportion of adults have depression today?” Period prevalence refers to a time interval, such as “in the past year.”
Prevalence is heavily influenced by both incidence and disease duration. A disease that is long lasting and rarely fatal will have a high prevalence even if few new cases occur each year. Conversely, a disease that is rapidly fatal or quickly cured can have low prevalence even with high incidence.
Rule: Prevalence increases with increased incidence or increased duration of disease.
Rule: For chronic diseases with stable incidence and population, $ \text{Prevalence} \approx \text{Incidence} \times \text{Duration} $.
USMLE questions frequently require you to reason about how improved treatment affects prevalence. Treatments that prolong life without curing disease usually increase prevalence, because more people are living with the condition.
Incidence
Incidence measures the occurrence of new cases over a specified time period. It is about risk of developing the disease, so it is central when the question mentions “future risk,” “follow up,” or “newly diagnosed.”
There are two related concepts: incidence proportion and incidence rate.
Incidence proportion (also called cumulative incidence) is:
$$\text{Incidence proportion} = \frac{\text{New cases during a period}}{\text{Population at risk at start of period}}$$
Incidence rate uses person time in the denominator and captures changing populations and different lengths of follow up:
$$\text{Incidence rate} = \frac{\text{New cases}}{\text{Total person time at risk}}$$
Incidence is used to estimate risk and is essential in cohort studies and clinical trials. When you read “participants were followed for 5 years” you should think of incidence.
Mortality and Case Fatality
Mortality focuses on deaths rather than disease occurrence. The simplest measure is the crude mortality rate, which is:
$$\text{Mortality rate} = \frac{\text{Number of deaths in a time period}}{\text{Total population}}$$
Mortality can also be cause specific, for example deaths due to lung cancer. Remember that mortality rates are not the same as incidence of disease. A disease can have low incidence but high mortality, or high incidence but low mortality.
Case fatality is the proportion of people with a disease who die from it, usually in a given time frame:
$$\text{Case fatality} = \frac{\text{Deaths from the disease}}{\text{Individuals with the disease}}$$
Case fatality illustrates how lethal a condition is among those affected, while mortality rate reflects the burden of deaths in the entire population.
Measures of Association and Effect
Once we can measure disease frequency, we want to compare groups. For example, what is the effect of smoking on lung cancer, or of a drug on heart attacks? Measures of association quantify the relationship between an exposure and an outcome and are core for interpreting study results on the USMLE.
Relative Risk
Relative risk (RR) compares the risk of an outcome in an exposed group with the risk in an unexposed group. It is used mainly in cohort studies and clinical trials, where you can measure incidence.
If you have a standard $2 \times 2$ table:
| Outcome present | Outcome absent | Total | |
|---|---|---|---|
| Exposed | a | b | a + b |
| Not exposed | c | d | c + d |
Then:
$$\text{Risk in exposed} = \frac{a}{a + b}, \quad \text{Risk in unexposed} = \frac{c}{c + d}$$
and
$$\text{Relative risk} = \frac{\text{Risk in exposed}}{\text{Risk in unexposed}} = \frac{a / (a + b)}{c / (c + d)}$$
Interpretation is central on the exam. If RR = 1 there is no association. If RR > 1 the exposure is associated with increased risk. If RR < 1 the exposure is associated with decreased risk, which suggests a protective factor.
Students often confuse RR with risk difference. RR is a ratio that answers “how many times higher or lower is the risk?” Risk difference is a subtraction that answers “how much higher or lower is the risk in absolute terms?” Both can appear in questions about benefit or harm of interventions.
Odds Ratio
The odds ratio (OR) compares the odds of exposure or outcome between groups. It is the primary measure of association in case control studies, and on USMLE it often approximates relative risk when the outcome is rare.
Using the same $2 \times 2$ table:
| Outcome present | Outcome absent | Total | |
|---|---|---|---|
| Exposed | a | b | a + b |
| Not exposed | c | d | c + d |
Odds in the exposed group is $a / b$. Odds in the unexposed group is $c / d$. The odds ratio is:
$$\text{Odds ratio} = \frac{a/b}{c/d} = \frac{ad}{bc}$$
In a case control study you typically start by selecting cases and controls, then you look back to see exposure. You cannot directly measure incidence, so you cannot directly calculate relative risk. Instead, you use OR.
When the disease is rare the odds ratio approximates the relative risk, and exam questions may ask you to recognize or state that this is valid.
Rule: In rare diseases (low prevalence), $ \text{OR} \approx \text{RR} $.
Rule: Use RR in cohort and clinical trial data, use OR in case control data.
Attributable Risk and Attributable Risk Percent
Attributable risk (AR) answers how much of the risk in the exposed group is due to the exposure itself. It is the difference in risk between exposed and unexposed:
$$\text{Attributable risk} = \text{Risk in exposed} - \text{Risk in unexposed}$$
Attributable risk percent expresses this as a proportion of the total risk in the exposed group:
$$\text{AR\%} = \frac{\text{Risk in exposed} - \text{Risk in unexposed}}{\text{Risk in exposed}} \times 100\%$$
For the USMLE, AR and AR% are often used to interpret the impact of a harmful exposure, such as smoking, and to think about how much disease could be prevented by removing that exposure from the exposed group.
Population attributable risk (PAR) extends this concept to the entire population, not just those exposed, by taking into account the prevalence of exposure. It is conceptually important for public health decisions, though you are less often required to calculate it explicitly.
Number Needed to Treat and Number Needed to Harm
Number needed to treat (NNT) and number needed to harm (NNH) are ways of expressing risk differences in clinically meaningful terms. They are highly testable because they connect epidemiology with patient care.
NNT is the number of patients who need to receive a treatment for one additional patient to benefit compared with a control. When an intervention reduces a bad outcome, NNT is based on the absolute risk reduction (ARR). NNH is the analogous concept for causing harm.
If $ \text{ARR} $ is the absolute risk reduction, then
$$\text{NNT} = \frac{1}{\text{ARR}}$$
For harm, if $ \text{AR}_{\text{harm}} $ is the absolute increase in risk due to the exposure, then
$$\text{NNH} = \frac{1}{\text{AR}_{\text{harm}}}$$
On the exam, remember that lower NNT means a more effective treatment, and lower NNH means a more dangerous exposure or side effect profile.
Rule: Always use absolute risk reduction (not relative risk reduction) when calculating NNT or NNH.
Rule: Round NNT and NNH up to the next whole patient when interpreting.
Study Designs in Epidemiology
Epidemiology relies on different types of studies to answer different questions. For USMLE purposes, you must be able to identify the design from a brief description, know its main direction of inquiry, and recognize its major strengths and limitations.
Cross Sectional Studies
Cross sectional studies examine a population at one point in time. They measure exposure and outcome simultaneously and usually estimate prevalence. Many survey based studies fall into this category.
In question stems, words like “snapshot,” “single time point,” “prevalence of disease and associated factors” often indicate cross sectional design.
This design is useful for describing disease burden and for generating hypotheses, but it cannot reliably establish temporal sequence, so you cannot say whether exposure preceded outcome. That limits its ability to infer causality.
Case Control Studies
Case control studies start with people who already have the outcome (cases) and compare them with people who do not have the outcome (controls). Investigators then look backward to assess prior exposure.
Timeline is a key clue: first select by outcome status, then look back in time for exposure. Because you choose the number of cases and controls, you usually cannot calculate incidence or prevalence directly. That is why odds ratios are used.
Case control studies are especially useful for rare diseases and diseases with long latency periods, for example rare cancers. They are relatively quick and inexpensive, but they are more vulnerable to certain biases, especially recall bias, where cases remember exposures differently than controls.
Cohort Studies
Cohort studies begin with participants who are free of the outcome and classify them according to exposure status. They are then followed forward in time to see who develops the outcome.
The key pattern is: select based on exposure, then measure incidence of outcome. You can directly calculate incidence, relative risk, and attributable risk.
Cohort studies can be prospective, in which time moves forward as the study progresses, or retrospective, in which past records are used to reconstruct exposure and follow up. Both are cohort designs because they start with exposure and move toward outcome.
They are ideal for studying multiple outcomes of a single exposure and for establishing temporal relationships, but they can be expensive and time consuming, and they may suffer from loss to follow up.
Clinical Trials
Clinical trials are experimental cohort studies that test interventions, commonly drugs or procedures. Participants are assigned to intervention or control groups, often by randomization, then followed for outcomes.
Randomization aims to balance known and unknown confounders between groups. Blinding seeks to minimize various biases by keeping participants, investigators, or outcome assessors unaware of group assignments.
On USMLE, key identifiers include random assignment, control/placebo groups, and explicit interventions. Clinical trials provide high quality evidence about causality but are subject to ethical constraints, cost, and issues like nonadherence.
Other Descriptive and Analytic Designs
Several other designs appear in exam questions, mostly in descriptive contexts.
Case reports and case series describe one or a few patients with unusual presentations. They cannot establish association but can generate hypotheses. Ecological studies use group level data, such as comparing countries or regions, and may demonstrate correlations at the population level, but they are vulnerable to the ecological fallacy, where group level associations are incorrectly applied to individuals.
Randomized controlled trials, cohort, case control, and cross sectional studies are the core analytic designs that the exam tests for cause effect questions.
Validity, Bias, and Confounding
An epidemiologic study is only as useful as its validity. Internal validity refers to whether the observed association within the study is true for the study participants. External validity asks whether the results can be generalized to other populations. The main threats to validity are bias and confounding.
Types of Bias
Bias is a systematic error in the design, conduct, or analysis of a study that leads to an incorrect estimate of the association between exposure and outcome. On the USMLE, you are expected to recognize common biases and understand how to reduce them.
Selection bias occurs when the way participants are selected or retained in the study results in non comparable groups. Examples include loss to follow up in cohort studies where dropouts differ systematically from those who remain, or non response bias in surveys. Hospital based cases and controls may differ from those in the community, which can distort the exposure outcome association.
Information bias arises from incorrect measurement or classification of exposure or outcome. Recall bias is common in case control studies when cases remember exposures more accurately than controls. Observer bias happens when the investigators’ knowledge of exposure status influences their assessment of outcome. Surveillance bias occurs when one group is monitored more closely than another, leading to higher detection rates that may mimic increased incidence.
To reduce information bias, studies use standardized data collection, objective measurements, and blinding of assessors.
Confounding
Confounding is a distortion of the apparent effect of an exposure on an outcome caused by a third variable that is associated with both the exposure and the outcome but is not on the causal pathway between them.
A confounder must satisfy three conditions. It is associated with the exposure. It is an independent risk factor for the outcome. It is not a step in the causal chain between exposure and outcome. Classic examples involve age, smoking, and socioeconomic status.
Confounding can create a spurious association or mask a real association. When you control for the confounder, for example by stratification or regression, the association between exposure and outcome should change.
Distinguishing confounding from effect modification is important on the exam. In effect modification, the association between exposure and outcome genuinely differs across levels of a third variable. For instance, a drug may be effective overall but much more effective in younger patients. In effect modification the difference between strata is real and informative. In confounding it represents distortion that you want to reduce.
Controlling Bias and Confounding
Various strategies help minimize bias and confounding in study design and analysis.
Randomization in clinical trials spreads both known and unknown confounders evenly between groups. Restriction limits the study to participants with certain characteristics and can reduce confounding but may reduce generalizability. Matching in case control studies selects controls that are similar to cases on potential confounders, such as age or sex.
At the analysis stage, stratified analysis and multivariable regression can adjust for confounders. Blinding reduces observer and performance biases by preventing knowledge of group assignment from influencing behavior or assessment. Intention to treat analysis preserves the benefits of randomization by analyzing participants in their assigned groups, regardless of adherence.
Rule: Confounding distorts the true association. Effect modification reveals a real, different effect in subgroups.
Rule: Randomization prevents confounding by distributing confounders. Blinding prevents information and observer bias.
Screening and Epidemic Concepts in Epidemiology
Epidemiology also provides the framework for screening programs and for understanding how diseases spread in populations.
Screening and Epidemiologic Measures
Screening uses tests to identify disease in asymptomatic people. The epidemiologic perspective focuses on how screening affects disease statistics.
Screening can increase the apparent incidence of disease because more cases are detected earlier. It also increases prevalence in the short term, because previously undiagnosed cases are now counted. Lead time bias occurs when earlier detection makes survival time from diagnosis appear longer, even though the time of death does not change. Length time bias occurs when screening preferentially detects slower growing, less aggressive forms of disease, making screened patients seem to have better outcomes.
These biases are central in epidemiology questions that ask whether screening actually improves mortality or only appears to improve survival statistics. The key outcome that screening should improve is disease specific or all cause mortality, not just time since diagnosis.
Basic Concepts of Disease Spread
Epidemiology also studies how infectious diseases spread in populations, using terms like endemic, epidemic, and pandemic.
An endemic disease is consistently present at a baseline level in a population. An epidemic is a sudden increase in disease occurrence above the expected level. A pandemic is an epidemic that has spread over multiple countries or continents.
The reproductive number $R_0$ represents how many secondary cases one case generates in a fully susceptible population. Control measures, such as vaccination or behavior changes, aim to reduce the effective reproductive number below 1 to stop an epidemic.
These concepts provide context for understanding outbreak investigations and population level interventions, which often appear in scenario based questions.
Epidemiology, with its focus on populations, measures, and study designs, provides the language through which medical evidence is created and interpreted. Mastery of these concepts allows you to navigate USMLE questions that tie clinical decisions to underlying data and to judge whether a reported association is likely to be real, biased, or confounded.