2 Epidemiology: Concepts and Methods

Learn how epidemiology describes health patterns in populations, measures disease frequency, uses study designs to investigate health questions, and interprets evidence for public health decisions.

Epidemiologic thinking

examines the distribution and determinants of health-related conditions in specified populations and uses that knowledge to prevent and control health problems. It asks who is affected, when and where events occur, what factors may explain the pattern, and what action could improve health. Its central focus is the population, not only the individual.

A common starting point is to describe health events by person, place, and time. For example, mapping foodborne illness cases by neighborhood and date may reveal a shared location or exposure. Descriptive patterns can generate hypotheses, which analytic studies then test by comparing groups.

Counts alone rarely allow fair comparisons. Twenty cases in a town of $1{,}000$ people represent a different burden from $20$ cases in a city of $100{,}000$ . Relate counts to an appropriate population denominator, and clearly define the condition, population, and time period.

and new cases

describes new cases arising in a population at risk during a specified period. Its denominator should include only people who could become cases.

The , also called risk, is the proportion of people initially free of the condition who develop it during a defined interval:

\text{Incidence proportion} = \frac{\text{new cases during the interval}}{\text{people at risk at the start of the interval}}

For example, if $20$ of $200$ disease-free people develop an illness during one year, the one-year is $\frac{20}{200}=10\%$ .

The relates new cases to the total time people were observed and at risk:

\text{Incidence rate} = \frac{\text{new cases}}{\text{total person-time at risk}}

When participants are followed for different lengths of time, a rate using person-time may be more appropriate than an .

and other frequency measures

is the proportion of a population that has a condition, whether newly diagnosed or already present, at a specified time or over a defined period:

\text{Prevalence} = \frac{\text{existing cases}}{\text{population assessed}}

Point is a snapshot. Period counts people who had the condition at any time during a specified interval. helps estimate how widespread a condition is and plan services.

is influenced by both the occurrence of new cases and how long people live with the condition, so it is not interchangeable with . Other frequency measures include mortality rates, which describe deaths in a population over time. For any measure, check the event being counted, who is included in the denominator, and the period and population to which it applies.

Study designs and what they can show

Study design determines how evidence is gathered and which conclusions the data can support.

Descriptive study: Summarizes health events by person, place, and time, and may use surveillance or case reports. It identifies patterns and generates hypotheses, but usually cannot establish why a pattern occurred.
Cross-sectional study: Measures exposure and health status at about the same time. It can estimate , such as the proportion of adults with hypertension, but unclear timing can limit conclusions about cause and effect.
: Groups people by exposure and follows them to compare new outcomes. It can be prospective or use existing records retrospectively. It can measure and establish whether exposure preceded outcome, but may require substantial time or a large population.
: Starts with people who have an outcome (cases) and a suitable comparison group without it (controls), then compares prior exposures. It is often efficient for rare outcomes; selecting controls and accurately recalling past exposures are important challenges.
Randomized trial: Investigators assign an intervention, often randomly, and compare outcomes between groups. When feasible and ethical, it can provide strong evidence about intervention effects. Assigning harmful exposures is not appropriate.

A risk ratio compares the in exposed people with that in unexposed people. An odds ratio in a compares the odds of prior exposure in cases and controls. Because people in a are selected by outcome, that design generally cannot directly calculate risk. Cross-sectional studies often compare .

Interpreting associations and uncertainty

An observed association is a clue, not automatic proof of causation. Ask whether the exposure occurred before the outcome, and consider other explanations that could affect the result.

Chance: A result may reflect random variation, particularly in a small study.
Bias: Systematic errors in who is included, what is measured, or how information is collected can distort findings. For example, cases may remember past exposures differently from controls.
: A third factor related to both exposure and outcome may partly or wholly explain their association.
Measurement and selection problems: Inaccurate case definitions, missing data, or an unrepresentative sample can weaken conclusions.

Consider the size and direction of an association and its , which shows the range of effect estimates compatible with the data and conveys precision. Statistical significance alone does not establish causality or practical importance. A small effect can be statistically significant in a very large study, while an important effect may remain uncertain in a small one.

Interpret findings in light of the study design, possible bias and , consistency with other evidence, and applicability to the population of interest.

Using evidence in public health decisions

For public health decisions, translate relative effects into likely absolute health impact: how many cases might be prevented, and among whom? Weigh the strength and relevance of the evidence alongside feasibility, potential benefits and harms, resources, and community context.

Evidence-informed decisions use the best available research while recognizing its limitations and the needs of the people affected. Epidemiologic measures and study findings support decisions, but their interpretation depends on the question, population, uncertainty, and likely impact.