Epidemiology

Introduction

Epidemiology is the study of the distribution and determinants of health-related states or events in specified populations and the application of that study to the control of health problems. It is a core science of public health, but its methods are also central to clinical research, environmental health, occupational health, social science, veterinary medicine, and health policy. Instead of asking only why one person became ill, epidemiology asks how health events are distributed across populations, which factors help explain those patterns, and what actions can improve population health.

At university level, you should learn to move between four connected activities: describing patterns, measuring disease frequency, comparing groups, and judging whether an observed association is likely to reflect a causal effect. This requires careful study design, valid measurement, appropriate statistical analysis, ethical reasoning, and clear communication of uncertainty.

The field includes both infectious and non-infectious outcomes. Epidemiologists may study influenza transmission, cancer risk, road injuries, mental health, medication safety, climate-related illness, maternal mortality, or the effects of public policies. The unifying feature is a population-based approach to health.

Learning Goals

By the end of this aiMOOC, you should be able to define core epidemiologic concepts, calculate and interpret common measures of disease frequency and association, distinguish major study designs, identify important threats to validity, explain principles of causal inference and surveillance, interpret outbreak data, evaluate screening tests, and communicate epidemiologic evidence responsibly.

You should also be able to connect epidemiologic reasoning with biostatistics, causal inference, health policy, evidence-based medicine, and data visualization.

Foundations of Epidemiologic Thinking

Epidemiology is both descriptive and analytic. Descriptive epidemiology characterizes health events by person, place, and time. Analytic epidemiology compares groups to evaluate hypotheses about exposures and outcomes. In practice, these activities are iterative: descriptive patterns generate hypotheses, analytic studies test them, and findings often lead to new surveillance questions or interventions.

Distribution, Determinants, and Populations

Distribution refers to the frequency and pattern of health events. Frequency concerns how often an event occurs. Pattern concerns how occurrence varies by time, place, and person. Determinants are factors that influence health events, including biological characteristics, behaviors, social conditions, environmental exposures, pathogens, healthcare access, policies, and combinations of these factors.

A population must be defined carefully. A rate is meaningful only when its numerator and denominator refer to compatible people, places, and time periods. An apparent increase in disease may reflect a true rise in risk, but it may also result from population growth, changes in diagnosis, changes in reporting, altered case definitions, or improved detection.

Person, Place, and Time

When you describe a health problem, ask who is affected, where events occur, and when they occur. Person variables can include age, occupation, socioeconomic position, behavior, or clinical characteristics. Place can range from households and workplaces to neighborhoods, countries, or ecological regions. Time can refer to hours, seasons, years, birth cohorts, or historical periods.

These dimensions are not explanations by themselves. A spatial cluster may suggest a shared exposure, but clustering can also arise through population density, healthcare access, reporting patterns, or chance. Descriptive epidemiology is therefore a foundation for hypothesis generation rather than automatic proof of causation.

A Historical Example: John Snow and Cholera

The investigation of cholera in nineteenth-century London is a classic example of epidemiologic reasoning. John Snow compared the geographic distribution of cholera deaths with water sources and argued that contaminated water played a major role in transmission. The historical importance of this work lies not in a single map alone, but in the combination of observation, comparison, natural experiment, and causal reasoning.

When you examine the map, identify the apparent concentration of deaths, consider the location of water pumps, and ask what alternative explanations would need to be evaluated. Modern epidemiology extends this logic with formal sampling, statistical modeling, laboratory evidence, causal diagrams, and explicit methods for quantifying uncertainty.

Measuring Disease Frequency

Measures of frequency translate counts into quantities that can be compared across populations. A count of 500 cases is difficult to interpret without knowing the population size, the period of observation, and who was actually at risk.

Incidence

Incidence concerns new events. An incidence proportion, sometimes called cumulative incidence or risk, is the proportion of an initially disease-free population that develops the outcome over a specified period.

Incidence proportion = new cases during a specified period / number of people at risk at the start of the period.

An incidence rate uses person-time in the denominator.

Incidence rate = new cases / total person-time at risk.

Person-time is useful when individuals are followed for different lengths of time. Someone contributes time while they are under observation and at risk of the event. Incidence rates therefore describe the speed at which new events occur rather than the probability that an individual will experience an event.

Prevalence

Prevalence concerns existing cases. Point prevalence is the proportion of a population with a condition at a particular time. Period prevalence considers whether the condition was present at any time during a specified period.

Prevalence = existing cases / population assessed.

Prevalence depends on both the rate at which new cases occur and the duration of disease. In a relatively stable population with a relatively rare condition and stable incidence and duration, prevalence can sometimes be approximated from incidence and average duration. This approximation should not be treated as a universal identity.

Mortality and Case Fatality

A mortality rate relates deaths to the population and time at risk. A cause-specific mortality rate focuses on deaths from a particular cause. A case-fatality proportion describes the proportion of people with a specified disease who die from that disease during a defined period or episode. It is therefore a measure of severity among cases, not a population mortality rate.

Age, sex, and other population structures can strongly influence crude rates. When populations differ in composition, stratified or standardized rates may provide fairer comparisons.

Descriptive Epidemiology and Surveillance

Public health surveillance is the ongoing, systematic collection, analysis, interpretation, and dissemination of health-related data for action. Surveillance is not simply data storage. Its purpose is to produce timely information that supports prevention, control, planning, and evaluation.

Surveillance Systems

Surveillance can be passive, active, sentinel, syndromic, laboratory-based, registry-based, or increasingly genomic and digital. Each system trades off timeliness, completeness, cost, representativeness, and specificity. A passive system may cover a large population at modest cost but miss cases. An active system may achieve better ascertainment but require more resources.

A surveillance case definition specifies operational criteria for classifying cases consistently. It may include clinical, laboratory, epidemiologic, person, place, and time criteria. The purpose is standardization for surveillance or investigation, not necessarily individual clinical diagnosis. If case definitions or testing practices change over time, trend interpretation must account for that change.

Epidemic Curves

An epidemic curve plots cases by time of symptom onset or another relevant event. Its shape can suggest features of transmission, timing, incubation, and exposure patterns. A sharp rise and fall may be compatible with a point-source exposure, while successive waves can be compatible with propagated transmission. Real outbreaks can produce more complicated curves because of reporting delays, multiple exposures, interventions, and changing case ascertainment.

A historical epidemic curve should be interpreted together with information about case definitions, surveillance intensity, laboratory confirmation, geography, and interventions.

Epidemiologic Study Designs

Choosing a study design means matching a scientific question to a feasible and ethically acceptable method. The direction of sampling, timing of exposure and outcome measurement, comparison group, and assignment mechanism all shape what can be estimated.

No simple hierarchy can replace critical appraisal. A well-designed observational study may answer a question that an experiment cannot ethically or practically address. Likewise, randomization is powerful for causal inference but does not automatically solve problems such as nonadherence, attrition, poor outcome measurement, or limited generalizability.

Descriptive and Ecological Studies

Case reports and case series can identify unusual events and generate hypotheses, but they usually lack a comparison group. Ecological studies compare groups or populations using aggregate exposure and outcome data. They are useful for population-level questions, but individual-level conclusions can be invalid if group-level associations are assumed to apply to individuals. This error is called the ecological fallacy.

Cross-Sectional Studies

Cross-sectional studies measure exposure and outcome within the same general time window. They are efficient for estimating prevalence and exploring associations. Their main limitation for etiologic questions is often temporality: if exposure and outcome are measured at about the same time, it can be difficult to establish which came first.

Case-Control Studies

Case-control studies sample people according to outcome status. Researchers compare the exposure histories of cases with those of suitable controls. This design is efficient for rare outcomes or outcomes with long latency. Control selection is crucial because controls should represent the exposure distribution in the source population that produced the cases.

The odds ratio is the standard measure of association in a conventional case-control study. Case-control studies are often described as retrospective, but the defining feature is sampling by outcome status, not simply looking backward in calendar time. A case-control study can also be nested within a prospective cohort.

Cohort Studies

Cohort studies classify participants by exposure and compare the occurrence of outcomes over follow-up. They can be prospective or retrospective with respect to data collection. Because the population at risk and follow-up are observed, investigators can directly estimate incidence, risks, and rates when the design and data support them.

Cohort studies are particularly useful for studying rare exposures and multiple outcomes, but they may be inefficient for very rare diseases or diseases with long latency.

Randomized and Quasi-Experimental Studies

In a randomized controlled trial, eligible participants or groups are assigned by a random mechanism to intervention conditions. Randomization aims to create exchangeable groups by balancing both measured and unmeasured baseline causes on average. Allocation concealment, adherence, masking where possible, complete follow-up, and appropriate analysis remain important.

When randomization is impossible, quasi-experimental designs such as interrupted time series, regression discontinuity, or difference-in-differences can strengthen causal inference if their assumptions are plausible.

Measures of Association and Effect

Epidemiology compares disease occurrence across groups. Different designs naturally support different measures.

The Two-by-Two Table

A conventional two-by-two table cross-classifies exposure and outcome. Let a represent exposed cases, b exposed non-cases, c unexposed cases, and d unexposed non-cases.

	Outcome present	Outcome absent
Exposed	a	b
Unexposed	c	d

Risk among exposed = a / (a + b).

Risk among unexposed = c / (c + d).

Risk ratio = risk among exposed / risk among unexposed.

Risk difference = risk among exposed - risk among unexposed.

Odds ratio = (a times d) / (b times c).

A risk ratio of 1 indicates equal risk in the two groups. A risk difference of 0 indicates no absolute risk difference. Measures on ratio and difference scales answer different questions and can lead to different judgments about public health importance.

Relative and Absolute Effects

Relative measures describe proportional contrasts, while absolute measures describe differences in event frequency. Suppose a treatment reduces risk from 2 percent to 1 percent. The risk ratio is 0.5, but the risk difference is 1 percentage point. If baseline risk were 20 percent and treated risk 10 percent, the same risk ratio would correspond to a 10 percentage-point risk difference.

Good communication often presents both relative and absolute effects. Policy decisions also require attention to uncertainty, harms, costs, feasibility, and equity.

Validity, Bias, Confounding, and Effect Modification

An estimate can be numerically precise but scientifically wrong. Epidemiologists therefore distinguish random error from systematic error and ask whether the study validly estimates the quantity of interest.

Selection Bias

Selection bias occurs when the relationship between exposure and outcome differs between those included in the analysis and the target relationship because of how participants enter, remain in, or are selected for the study. Examples include differential loss to follow-up, inappropriate control selection, and conditioning on participation affected by both exposure and outcome causes.

Selection mechanisms should be considered during design, not only after data collection.

Information Bias and Misclassification

Information bias arises from systematic error in measuring exposure, outcome, or covariates. Recall bias can occur when cases and controls remember past exposures differently. Interviewer bias can occur when data collection differs according to participant status. Misclassification can be differential or nondifferential, and its direction is not always predictable.

Reliability is not the same as validity. A measure can be highly reproducible yet consistently wrong.

Confounding

Confounding occurs when a mixing of effects produces a distorted exposure-outcome association. A confounder is related to exposure and independently causes or predicts the outcome, and it is not merely an intermediate on the causal pathway of interest. Design strategies include randomization, restriction, and matching. Analysis strategies include stratification, standardization, regression adjustment, weighting, and other causal methods.

Adjustment should be guided by substantive causal knowledge rather than by an automatic rule that every measured variable belongs in a model. Adjusting for a mediator can block part of the effect you intend to estimate, while adjusting for a collider can create bias.

Effect Modification

Effect modification means that the effect of an exposure differs across levels of another variable on a specified effect scale. Unlike confounding, effect modification is not a nuisance that must always be removed. It can reveal meaningful biological, social, or policy heterogeneity. You should therefore state the scale on which effect differences are being discussed and present stratum-specific effects when appropriate.

Causal Inference

Causal inference asks what would happen to outcomes under different exposure or intervention conditions. Because the same person cannot simultaneously experience both exposure states at the same time, causal effects are inherently counterfactual and require assumptions.

Core Assumptions

Three ideas frequently appear in modern causal inference. Exchangeability means that the groups being compared are sufficiently comparable with respect to causes of the outcome. Positivity means that relevant exposure alternatives are possible for the covariate patterns under study. Consistency links observed outcomes to well-defined exposure or intervention states.

These assumptions are not guaranteed by sophisticated statistical software. They depend on study design, subject-matter knowledge, measurement, and the specific causal question.

Directed Acyclic Graphs

A directed acyclic graph, often called a DAG, represents assumptions about causal relationships. DAGs help you distinguish confounders, mediators, and colliders and think about which variables should or should not be conditioned on.

A DAG is not proof that the assumed arrows are correct. Its value lies in making assumptions explicit so that they can be criticized, revised, and linked to analytic decisions.

Causal Criteria and Triangulation

Historical considerations such as temporality, strength, consistency, dose-response patterns, biological plausibility, and experimental evidence can support causal reasoning. They should be treated as considerations rather than a mechanical checklist. Modern causal assessment often combines evidence from different designs whose biases are unlikely to be identical. This strategy is sometimes called triangulation.

Screening and Diagnostic Test Evaluation

Epidemiologists often evaluate tests used for screening, diagnosis, case finding, or surveillance. A test result must be interpreted in relation to a reference standard, the tested population, and the intended use.

Sensitivity and Specificity

Sensitivity is the probability that a test is positive among people who truly have the condition. Specificity is the probability that a test is negative among people who truly do not have the condition.

Positive predictive value is the probability that a person has the condition given a positive test. Negative predictive value is the probability that a person does not have the condition given a negative test. Predictive values depend strongly on prevalence or pretest probability, so a test can have the same sensitivity and specificity in two settings but different predictive values.

Screening programs should not be judged by test accuracy alone. Benefits, harms, overdiagnosis, false positives, false negatives, follow-up capacity, acceptability, cost, and equity all matter.

Infectious Disease Epidemiology

Infectious disease epidemiology adds features such as transmission, contact structure, immunity, pathogen evolution, generation intervals, and feedback between individual behavior and population dynamics.

Reproduction Numbers and Transmission

The basic reproduction number, usually written R0, is the expected number of secondary infections generated by a typical infectious individual in a fully susceptible population under specified conditions. It is not a fixed biological constant independent of context. Contact patterns, environment, behavior, and the infectiousness profile all matter.

The effective reproduction number, often written Rt or Re, changes as susceptibility, behavior, interventions, and immunity change. Values above 1 are compatible with growing transmission under the conditions represented by the estimate, while values below 1 are compatible with declining transmission. Estimates are uncertain and sensitive to assumptions about generation intervals and reporting.

Compartmental Models

The SIR model divides a population into susceptible, infectious, and recovered compartments. Individuals move between compartments according to modeled transition rates. Such models can clarify mechanisms and explore scenarios, but they simplify reality.

More detailed models can incorporate exposed states, age structure, spatial movement, vaccination, waning immunity, or heterogeneity in contact patterns. A model should be judged by whether its assumptions are appropriate for the decision problem, not merely by mathematical complexity.

Outbreak Investigation

Outbreak investigations combine surveillance, laboratory science, environmental assessment, descriptive epidemiology, analytic studies, and communication. The order of activities may overlap because investigators often need to act before every uncertainty is resolved.

A Practical Investigation Sequence

A typical investigation includes preparing for field work, establishing whether an outbreak exists, verifying the diagnosis, constructing a working case definition, finding and line-listing cases, describing cases by time, place, and person, developing hypotheses, evaluating hypotheses, integrating epidemiologic findings with laboratory and environmental evidence, implementing control measures, maintaining surveillance, and communicating findings.

Control measures should not always wait for a completed analytic study. If a plausible exposure presents an urgent risk and a low-harm intervention is available, public health teams may act while continuing to investigate.

Attack Rates in an Outbreak

Suppose 80 people attended a meal. Twenty became ill. The overall attack proportion is 20 / 80 = 25 percent. If 30 people ate a particular food and 18 of them became ill, the attack proportion among exposed people is 60 percent. If 50 did not eat the food and 2 became ill, the attack proportion among unexposed people is 4 percent.

The risk ratio is 0.60 / 0.04 = 15. This strong association would support further investigation of that food, but it would not by itself prove causation. You would still evaluate the timing of exposure and illness, food-handling practices, laboratory findings, co-exposures, selection of attendees, and data quality.

Epidemiologic Data Analysis and Interpretation

Good epidemiologic analysis begins before statistical modeling. Define the target population, exposure, outcome, time zero, follow-up, estimand, and comparison strategy. Then examine data quality, missingness, distributions, and whether the analysis matches the design.

Confidence Intervals and P Values

A confidence interval communicates sampling uncertainty under a statistical model. A conventional frequentist 95 percent confidence interval should not be interpreted as a 95 percent probability that the fixed true parameter lies inside the interval after the interval has been computed. Rather, it comes from a procedure that would cover the target parameter in 95 percent of repeated samples under its assumptions.

A p value measures how incompatible the observed data or more extreme data are with a specified statistical model that includes a null hypothesis. It is not the probability that the null hypothesis is true, and statistical significance does not measure effect size, importance, or absence of bias.

Missing Data and Reproducibility

Missing data can produce bias when the reasons for missingness are related to variables in the analysis. Complete-case analysis is not automatically valid. Depending on assumptions and the research question, strategies can include multiple imputation, inverse-probability weighting, sensitivity analysis, and explicit modeling of missingness.

Reproducible epidemiology documents data provenance, preprocessing, code, model specifications, and analytic decisions. Preregistration or protocol registration can reduce selective reporting, while open code and clear reporting make analyses easier to audit and extend.

Evidence Synthesis and Population Decision-Making

Single studies rarely settle complex health questions. Systematic reviews identify and appraise relevant studies using explicit methods. Meta-analysis can statistically combine comparable estimates, but a precise pooled estimate is not automatically trustworthy if the underlying studies share bias or address meaningfully different questions.

A forest plot commonly displays study-specific effect estimates and confidence intervals together with a pooled estimate. Heterogeneity can reflect real differences in populations, interventions, exposure definitions, outcome definitions, follow-up, or bias. Publication bias and selective outcome reporting can distort the available evidence.

From Evidence to Action

Epidemiologic evidence contributes to decisions but does not make value judgments by itself. Public health choices can depend on disease burden, effect size, uncertainty, cost, feasibility, distribution of benefits and harms, legal authority, community priorities, and equity.

Ethics, Equity, and Communication

Epidemiology uses information about people and populations, so ethical practice is central. Ethical concerns include privacy, confidentiality, consent, proportionality, data governance, community engagement, stigma, fairness, and the distribution of benefits and burdens.

Public health surveillance and research can be governed by different legal and ethical frameworks. You should not assume that a public health purpose eliminates ethical responsibilities. Trust is essential because surveillance and outbreak response depend on people and institutions sharing accurate information.

Equity in Epidemiologic Reasoning

Health differences between groups can reflect unequal exposure to hazards, unequal access to resources, discrimination, structural conditions, measurement practices, or differential healthcare. Epidemiologists should avoid treating socially constructed categories as simple biological explanations.

When reporting disparities, describe the relevant context and mechanisms, consider whether data collection itself is unequal, and avoid stigmatizing communities. Equity also matters in intervention evaluation: an intervention can improve an overall average while leaving a disadvantaged group behind.

Communicating Uncertainty

Communicate what is known, what remains uncertain, how estimates were produced, and what could change the conclusion. Avoid presenting association as causation when the design does not justify that inference. Report absolute as well as relative effects when possible, explain denominators, and use graphics that do not exaggerate differences.

Interactive Tasks

Quiz: Test Your Knowledge

Which statement best defines epidemiology? (The study of health event distribution and determinants in populations and its application to control) (!The laboratory study of pathogens without population data) (!The treatment of individual patients using clinical examination) (!The use of statistics without a defined health question)

Which measure focuses on new cases occurring during follow-up? (Incidence) (!Prevalence) (!Specificity) (!Case fatality)

Which measure of association is standard in a conventional case-control study? (Odds ratio) (!Risk difference from direct incidence) (!Incidence rate from full cohort follow-up) (!Population prevalence ratio only)

What primarily distinguishes a cohort study? (Participants are classified by exposure and outcomes are compared over follow-up) (!Participants are selected only because they already have the outcome) (!Exposure and outcome must be measured at exactly the same instant) (!Researchers must randomly assign every exposure)

Which statement best describes confounding? (A third factor distorts an exposure outcome association because it is related to exposure and outcome) (!Random sampling error always changes the effect estimate in one direction) (!A mediator must always be adjusted for) (!Effect modification is another name for measurement error)

What does test sensitivity describe? (The probability of a positive result among people who truly have the condition) (!The probability of disease among people with a positive result) (!The probability of a negative result among people who have the condition) (!The proportion of the total population that has the condition)

What is an epidemic curve mainly used to display? (Cases according to time of onset or another relevant event) (!Individual genome sequences arranged by chromosome) (!Only the geographic distance between households) (!The cost of an intervention over a lifetime)

What is the main causal advantage of randomization in a well-conducted trial? (It tends to balance baseline causes of the outcome between groups on average) (!It guarantees that no participant will be lost to follow-up) (!It removes the need to measure outcomes accurately) (!It guarantees that results generalize to every population)

What does the basic reproduction number represent under specified conditions? (The expected secondary infections from a typical infectious individual in a fully susceptible population) (!The percentage of infected people who require hospitalization) (!The number of laboratory tests needed to confirm one case) (!The exact biological constant for a pathogen in every setting)

What makes public health surveillance more than a one-time survey? (It involves ongoing systematic data collection analysis dissemination and use for action) (!It always requires random assignment) (!It includes only data from hospitals) (!It excludes laboratory and registry information)

Memory Game

Incidence	New health events arising in a population at risk over time
Prevalence	Existing health conditions present in a population at a specified time or period
Confounder	A factor that can distort an exposure outcome association when causal structure permits
Case definition	Operational criteria used to classify cases consistently
Person-time	The accumulated time individuals contribute while observed and at risk
Odds ratio	A comparison of exposure or outcome odds commonly used in case-control studies

Drag and Drop

Match the correct terms.	Topic
Measures exposure and outcome in the same general time window	Cross-sectional study
Samples participants according to outcome status and compares prior exposure	Case-control study
Classifies participants by exposure and observes outcome occurrence	Cohort study
Uses a random mechanism to assign an intervention	Randomized trial
Compares aggregate exposure and outcome measures across groups	Ecological study

...

Crossword Puzzle

Incidence	What term describes the occurrence of new cases in a population at risk?
Prevalence	What term describes the proportion of a population with an existing condition?
Cohort	Which study design commonly follows exposure groups to compare outcome occurrence?
Confounding	What term describes distortion of an association by a third causal factor?
Surveillance	What ongoing public health process collects analyzes and disseminates health data for action?
Randomization	What allocation process uses chance to assign intervention groups?

LearningApps

Cloze Text

Epidemiology studies the distribution and

of health-related states in populations. New cases are summarized by measures of

. Existing disease in a population is summarized by

. A case-control study usually estimates an

. Systematic error created by a third causal factor can be called

. Test

is the probability of a positive result among people who truly have the condition. An epidemic curve displays cases in relation to

. Ongoing systematic collection and use of health data is called

. In causal inference, randomization aims to improve baseline

. Ethical epidemiology requires attention to privacy, fairness, and public

.

Open-Ended Tasks

Easy

Epidemic curve sketch: Use a small hypothetical outbreak dataset to draw an epidemic curve, label the axes, identify the apparent peak, and write a short interpretation that includes at least one alternative explanation for the pattern.
Health headline audit: Choose a recent health news headline, identify the study design behind the claim, and write a 300-word critique distinguishing association, causation, relative effects, and absolute effects.
Case definition design: Create a hypothetical surveillance case definition for a campus respiratory illness and explain how making it more sensitive or more specific could change case counts.
Epidemiology infographic: Produce a one-page image that explains incidence, prevalence, risk ratio, and odds ratio to students outside the health sciences without using misleading visual scales.

Standard

Mini cohort protocol: Draft a two-page protocol for a cohort study on a university-relevant exposure and outcome, including target population, eligibility, exposure measurement, outcome measurement, follow-up, likely confounders, and a primary effect measure.
Case-control interview: Design an interview guide for a hypothetical foodborne outbreak, conduct a role-play interview with a classmate, and reflect on how recall and interviewer bias could affect the data.
Diagnostic test simulation: Create a spreadsheet or simple simulation showing how positive predictive value changes when disease prevalence changes while sensitivity and specificity remain fixed, then interpret the result.
Risk communication video: Produce a three-minute video explaining an uncertain epidemiologic finding to the public, including the denominator, absolute risk, main limitation, and what evidence would increase confidence.

Advanced

Causal diagram project: Draw a directed acyclic graph for a real epidemiologic question, justify each major arrow, identify a minimally sufficient adjustment strategy, and explain which variables should not be adjusted for.
Outbreak investigation simulation: In a team, conduct a tabletop outbreak investigation using a synthetic line list, construct an epidemic curve, generate hypotheses, calculate an appropriate measure of association, propose control measures, and present findings in a briefing.
Reproducible epidemiology analysis: Analyze an open public health dataset with documented code, a data dictionary, a prespecified primary question, descriptive tables, an effect estimate with uncertainty, and a sensitivity analysis.
Public health field interview: Visit or virtually meet a local public health department, epidemiology unit, occupational health service, or research center, interview a professional about surveillance and evidence-to-action decisions, and create a reflective report comparing practice with methods learned in this course.

Learning Assessment

Study design transfer: Given three new public health questions, choose an appropriate design for each, defend the choice, identify the main estimand, and explain one important limitation.
Bias diagnosis: Analyze a short observational study scenario, construct a causal diagram, distinguish selection bias, information bias, confounding, and effect modification, and propose design or analysis remedies.
Outbreak reasoning: Interpret a synthetic line list and epidemic curve, calculate attack proportions and a risk ratio, generate a biologically plausible hypothesis, and explain what evidence would justify immediate control measures.
Screening policy analysis: Compare the expected consequences of introducing the same screening test in low-prevalence and high-prevalence populations, including predictive values, false positives, false negatives, and follow-up burden.
Causal claim critique: Select a published epidemiologic claim and evaluate temporality, comparison groups, measurement, exchangeability, positivity, consistency, precision, and generalizability before deciding how strong the causal interpretation should be.
Evidence-to-policy brief: Synthesize evidence from several study designs into a two-page policy brief that communicates absolute effects, uncertainty, equity implications, implementation constraints, and a justified recommendation.

Evidence of Learning

Knowledge: You can accurately explain disease frequency, measures of association, study designs, surveillance, outbreak investigation, screening, bias, confounding, effect modification, and core causal concepts.

Analytic skills: You can calculate and interpret incidence proportions, rates, prevalence, risk ratios, risk differences, odds ratios, attack proportions, sensitivity, specificity, and predictive values when the necessary data are available.

Design skills: You can translate a population health question into a target population, exposure, outcome, time frame, comparison group, study design, and estimand, while identifying plausible threats to validity.

Causal reasoning: You can use subject-matter knowledge and causal diagrams to distinguish confounders, mediators, and colliders and to justify an adjustment strategy rather than selecting covariates mechanically.

Products: Strong evidence may include a protocol, epidemic curve, reproducible analysis, causal diagram, data visualization, interview report, risk communication video, or policy brief.

Transfer: You can apply epidemiologic reasoning to unfamiliar health questions, detect overclaiming in public communication, distinguish population association from individual causation, and connect statistical results with ethical and policy decisions.