Epidemiology is the study of how diseases and health conditions spread through populations, who gets sick, and why. A review of definitions published between 1978 and 2017 found that five terms appeared in more than half of all formal definitions: “population,” “study,” “disease,” “health,” and “distribution.”1PLoS ONE. What is epidemiology? Changing definitions of epidemiology 1978-2017 Those five words capture the essence of the field, but the practice is far broader and more varied than any one-line definition can convey.
Where the Field Came From
People have noticed patterns in disease for thousands of years, but epidemiology as a recognizable discipline traces back to the mid-1800s. The figure most often credited as the founder is John Snow, a London physician who investigated cholera outbreaks at a time when most doctors still believed disease spread through foul air. Snow painstakingly mapped cholera cases, used spatial analysis and statistics to link them to a contaminated water pump on Broad Street, and convinced local officials to remove the pump handle. His work challenged the dominant theory of his era and laid the groundwork for modern public health.2PubMed Central. John Snow: The Pioneer of Modern Epidemiology and Anesthesia Snow’s approach, gathering data from communities, looking for patterns, and testing hypotheses about causes, remains the template for how epidemiologists work today.
How Epidemiologists Count Disease
Before you can figure out what makes people sick, you need to know how much sickness there is and where it is happening. Two of the most basic measurements in the field are prevalence and incidence. Prevalence captures the total number of existing cases of a disease in a population at a given moment. Incidence captures the number of new cases appearing over a period of time.3PubMed. Measures of disease frequency: prevalence and incidence The distinction matters because a disease with high prevalence but low incidence is one that persists for a long time in those who have it, while a disease with low prevalence but high incidence is appearing rapidly but being resolved quickly, either through recovery or death.
Prevalence itself comes in a few flavors. Point prevalence is a snapshot at a single moment, period prevalence covers a defined time window, and lifetime prevalence asks whether someone has ever had the condition.4Concepts of Epidemiology. The concept of risk and fundamental measures of disease frequency: Incidence and prevalence Each version answers a slightly different question, and picking the right one depends on what you are trying to learn. Tracking the incidence of flu in a community tells you whether a new wave is building. Measuring the prevalence of diabetes tells you how many people need ongoing care right now.
Measuring Whether an Exposure Actually Matters
Counting cases is just the starting point. Epidemiologists also need to quantify how much a particular exposure, whether that is a chemical, a behavior, a drug, or a living condition, changes a person’s risk. They do this using both relative and absolute measures of effect.5PubMed. Measures of effect: relative risks, odds ratios, risk difference, and ‘number needed to treat’
Relative measures compare two groups. A relative risk of 2 means the exposed group gets sick twice as often as the unexposed group. An odds ratio works similarly but compares odds rather than straightforward probabilities, which is useful in certain study designs. Absolute measures, like risk difference, tell you the actual gap in disease rates between the two groups.6PubMed Central. Key Measures in Epidemiology: Risk Difference, Relative Risk and Odds Ratio Both perspectives are needed. A relative risk of 3 sounds alarming, but if the baseline risk is one in a million, the absolute increase is tiny. Conversely, a modest relative risk applied to a very common exposure can translate into a large number of affected people.
Study Designs and How They Differ
Not all epidemiological studies are built the same way, and the design determines what questions you can answer and how confident you can be in the results. Two workhorses of the field are the cohort study and the case-control study.
A cohort study starts with a group of people defined by their exposure, say, smokers and nonsmokers, and follows them forward in time to see who develops a disease. A case-control study flips the logic: it starts with people who already have the disease (cases) and people who do not (controls), then looks backward to see what exposures each group had. Both designs can produce similarly robust estimates of association when done well.7PubMed Central. Core concepts in pharmacoepidemiology: Fundamentals of the cohort and case–control study designs Which one researchers choose often comes down to practicalities. Case-control studies are more efficient when the disease is rare, because you do not have to follow thousands of healthy people for years waiting for a handful of cases. Cohort studies are better when you want to look at many different outcomes from a single exposure, or when you need to measure absolute risk rather than just a ratio.
Beyond these two, epidemiologists use cross-sectional studies (a snapshot of a population at one time), ecological studies (comparing disease rates across whole regions or countries), and randomized controlled trials (where the researcher assigns the exposure, as in a clinical drug trial). Each step up in design complexity tends to give stronger evidence about causation, but also costs more money and time.
The Hard Part: Proving Cause and Effect
Showing that two things are associated is not the same as proving one causes the other. People who carry lighters are more likely to develop lung cancer, but the lighter is not the cause; smoking is. Sorting out genuine causes from mere associations is one of the central challenges of the field.
The most widely cited framework for thinking about causation in epidemiology comes from Sir Austin Bradford Hill, who in 1965 proposed nine viewpoints to help researchers evaluate whether an observed association is causal.8PubMed Central. Applying the Bradford Hill criteria in the 21st century: how data integration has changed causal inference in molecular epidemiology These include things like the strength of the association, whether it is consistent across different studies, whether a plausible biological mechanism exists, and whether the exposure comes before the disease in time. They are sometimes called “criteria,” but Hill himself was careful to call them viewpoints. No single one is required, and no checklist can guarantee a correct answer. Researchers still debate how best to apply them, and modern work has pushed to update the framework to incorporate newer methods of causal thinking.9PubMed Central. Assessing causality in epidemiology: revisiting Bradford Hill to incorporate developments in causal thinking
Confounding, where a hidden third variable explains the association between an exposure and an outcome, is the constant nemesis. Epidemiologists have developed an extensive toolkit for dealing with it, from statistical adjustment to study designs that minimize its influence. Even so, residual confounding is always a concern in observational research, and entire methodological papers are devoted to assessing whether uncontrolled confounders could have shifted a study’s results.10JNCI Monographs. Evaluation of Confounding and Selection Bias in Epidemiological Studies of Populations Exposed to Low-Dose, High-Energy Photon Radiation
Disease Detectives in the Field
Not all epidemiology happens at a desk with datasets. Field epidemiologists are the people who show up when a mysterious outbreak occurs, whether it is a cluster of food poisoning at a wedding, a new respiratory virus in a city, or a spike in birth defects in a farming region. Their job is to figure out what is happening, where it is coming from, and how to stop it.
Outbreak investigation follows a series of steps, though in practice these overlap and repeat. Investigators confirm the outbreak is real, define what counts as a case, count and describe the cases, generate hypotheses about the source, test those hypotheses with data, and recommend control measures.11Infectious Disease Emergencies: Preparedness and Response. 10. Field epidemiology and outbreak investigation The process is messy and iterative. Hypotheses get revised as new cases appear. Control measures sometimes start before the cause is confirmed, because waiting for certainty can cost lives. This is the branch of the field most visible during pandemics, when epidemiologists become public figures whether they want to or not.
The Many Branches of Epidemiology
The field has branched into numerous specialties, each applying epidemiological methods to different domains. A few of the most active ones are worth knowing about.
Social epidemiology examines how social conditions, poverty, racism, education, housing, neighborhood environments, produce patterns of disease. Research in this area consistently finds that people in poorer social circumstances fare worse across nearly every health measure, from cause-specific mortality to cognitive function to behavioral risk factors.12International Journal of Epidemiology. Commentary: Social inequalities in health, social epidemiology and social value The field has pushed back against a purely biomedical view of disease by arguing that social systems are fundamental causes of health disparities, not just background noise.13PubMed. Social Epidemiology: Past, Present, and Future
Molecular epidemiology integrates biological measurements, like genetic markers, biomarkers of exposure, or viral genome sequences, into traditional epidemiological study designs. The goal is to refine risk estimates and trace disease pathways at a molecular level. The rise of genome-wide association studies has given this branch enormous datasets to work with.14PubMed Central. Molecular epidemiology, candidate genes versus genome-wide scans
Pharmacoepidemiology focuses on drugs and their effects in real-world populations. Clinical trials before a drug reaches the market are carefully controlled, but they typically involve limited sample sizes and short follow-up periods, which means rare or delayed side effects often go undetected.15PubMed Central. Methods for drug safety signal detection using routinely collected observational electronic health care data: A systematic review Pharmacoepidemiologists monitor drugs after they are approved, using insurance claims, electronic health records, and prescription databases to spot safety signals that only emerge once millions of people are taking a medication.16Bulletin of Pharmacy Practice. APPLICATIONS OF PHARMACOEPIDEMIOLOGY IN DRUG SAFETY SURVEILLANCE AND PUBLIC HEALTH DECISION-MAKING: A NARRATIVE REVIEW
Climate epidemiology is a newer specialty that brings together epidemiologists and climate scientists to project how changing weather patterns will affect human health. By combining climate model outputs with health data, researchers can estimate how different greenhouse gas emission scenarios might alter heat-related deaths, vector-borne disease ranges, and other climate-sensitive outcomes.17American Journal of Epidemiology. The Future of Climate Epidemiology: Opportunities for Advancing Health Research in the Context of Climate Change
Herd Immunity and the Math Behind Vaccination
One of the most publicly visible applications of epidemiology is estimating how many people need to be vaccinated to protect a community. This depends on the basic reproduction number, often written as Râ‚€, which represents how many people one infected person will typically infect in a fully susceptible population. The higher the Râ‚€, the more people need to be immune to break the chain of transmission.
During the COVID-19 pandemic, estimated Râ‚€ values ranged widely, from about 1.4 to nearly 7, depending on the variant and the study. That wide range translated into herd immunity thresholds anywhere from roughly 29% to 85%.18PubMed Central. R0 of COVID-19 and its impact on vaccination coverage: compared with previous outbreaks The actual vaccination coverage needed is higher still, because no vaccine is 100% effective. Research on more transmissible variants like Omicron confirmed that both high vaccination coverage and high vaccine effectiveness are necessary to block transmission when Râ‚€ climbs toward the upper end of that range.19PubMed Central. Percentages of Vaccination Coverage Required to Establish Herd Immunity against SARS-CoV-2 These calculations are never static. A new variant, a waning immune response, or a change in population behavior can shift the numbers. Epidemiologists constantly update these models, which is why public health recommendations changed repeatedly during the pandemic.
Newer Frontiers: Sewage, Search Data, and One Health
Some of the most interesting developments in epidemiology involve data sources that would have seemed strange a generation ago. Wastewater-based surveillance, for instance, monitors sewage for traces of pathogens. Because infected people shed viral genetic material before they feel sick or get tested, sampling wastewater can detect a surge in infections days before hospital admissions rise.20PubMed Central. Water and wastewater digital surveillance for monitoring and early detection of the COVID-19 hotspot: industry 4.0 The approach gained enormous traction during COVID-19 and is now being expanded to track other pathogens. Machine learning is increasingly being applied to wastewater data to extract meaningful public health signals from what is inherently a noisy data source.21PubMed. Machine learning-enabled wastewater-based surveillance for emerging pathogen detection and monitoring: current applications, challenges, and future prospects
Researchers have also experimented with combining wastewater data with internet search behavior, since people tend to search for symptoms online around the same time viral loads appear in sewage. One framework integrated wastewater surveillance with a major online search platform to improve early warning for COVID-19.22Engineering. Integrating Urban Wastewater Surveillance and Internet Search Behavior to Strengthen Early Warning for Infectious Diseases The idea is that no single data stream gives you the full picture, but layering them can catch signals earlier and more reliably than any one source alone.
Another growing area is the One Health approach, which recognizes that human health, animal health, and environmental conditions are deeply intertwined. Most emerging infectious diseases in humans originate in animals, a process called zoonotic spillover. Network analyses have confirmed that the probability of zoonotic spillover increases at certain human-animal interfaces, particularly around livestock and food production.23Nature Communications. A One Health framework for exploring zoonotic interactions demonstrated through a case study Researchers have proposed interdisciplinary frameworks that bring together ecologists, veterinarians, social scientists, and epidemiologists to investigate the drivers of spillover events, from deforestation to wildlife trade to changes in farming practices.24PubMed Central. A framework for the study of zoonotic disease emergence and its drivers: spillover of bat pathogens as a case study
When Models Meet the Public
The COVID-19 pandemic put epidemiological models in front of the general public in an unprecedented way. Projections of case counts, hospitalizations, and deaths appeared in news briefings and shaped government decisions on lockdowns, school closures, and vaccine rollout. But the uncertainty built into those models was often poorly communicated, both by modelers and by the media that reported on their work.25PubMed Central. Communicating uncertainty in epidemic models
Every model rests on assumptions, about how a pathogen spreads, how people behave, how quickly interventions take effect. When those assumptions are wrong, the projections can be far off. That is not a failure of the model; it is how models work. They are tools for exploring scenarios, not crystal balls. The problem arises when the public and policymakers interpret a model’s output as a forecast rather than a conditional projection. When reality diverges from the projection, trust erodes. Researchers working in this area have argued that better visual formats, clearer labeling of scenarios versus predictions, and more transparent communication about what models can and cannot do are essential for maintaining public confidence in epidemiological science.26PubMed Central. Infectious disease modeling for public health practice: projections, scenarios, and uncertainty in three phases of outbreak response
The Ethics of Watching Populations
Public health surveillance, the backbone of epidemiology, creates an inherent tension between collective benefit and individual rights. Disease registries, mandatory reporting of certain infections, and population-level monitoring all depend on collecting personal health information, and this data is often gathered without individual informed consent. A systematic review of the ethical literature found that this tension is one of the most frequently debated issues in the field. Some argue that collecting surveillance data without consent is justified because it produces more complete and reliable datasets, which in turn protects communities. Others see it as an infringement on privacy and autonomy that requires stronger safeguards.27PubMed Central. Ethical issues in public health surveillance: a systematic qualitative review
Wastewater surveillance adds a new wrinkle to this debate. Monitoring sewage for pathogens is anonymous by design, since you are sampling the output of an entire sewer system rather than testing individuals. But in small communities, that anonymity can thin out. If a building or a neighborhood has its own sewer line, wastewater data can point to a very small group of people. How much geographic precision is acceptable, who gets access to the results, and what actions can be taken on the basis of community-level pathogen detection are questions the field is still working through. The ethical frameworks that governed traditional epidemiological surveillance were built for a world of paper forms and hospital reports. Adapting them to digital data streams and real-time molecular monitoring is an ongoing process.