This glossary collects the key concepts, people, and ideas you will meet in this lesson. Use it as a reference while you work through the material, or as a review before assessments. Type in the search box to filter entries.

Introduction and Overview

Lessons 1 and 2 worked at the level of the discipline: where epidemiology came from, what counts as knowledge, and how the published record can fail and be appraised. Lesson 3 takes the first step from those framing concerns into the working tools of the field. Almost every bias the manufactured-doubt strategy weaponizes, every analytic error the integrity-and-reform material in Lesson 1 catalogued, and every pre-analytic discipline EGAP demands — all of them play out concretely once you have a study design in front of you. So this lesson is about the design.

The four content sections proceed from general to specific. Section 1 sets up a classification scheme for every study you will encounter for the rest of HSCI 230. Section 2 covers the simplest of those study types — descriptive studies, where a comparison group is absent. Section 3 introduces the first true analytic observational design: the cross-sectional study. Section 4 pushes on its limitations and shows what reporting it well requires. By the end of the lesson, the language you'll need for case-control studies (Lesson 4) and cohort studies (Lesson 5) will already be in place.

Descriptive vs. Explanatory Studies

Epidemiologic studies can be classified into two major categories: descriptive and explanatory (analytic). This classification reflects both the study’s objectives and its ability to support causal inference. The two tabs below define each category in turn; click between them and notice that the difference is not about technique but about purpose — whether the study is built to characterize disease occurrence or to compare groups.

The descriptive–explanatory split asks what a study is trying to do. The next split — experimental versus observational — asks how the explanatory studies do it, and is the distinction this entire lesson hinges on.

Experimental vs. Observational Studies

In experimental studies, the investigator controls (usually through randomisation) the allocation of study subjects to exposure groups. In contrast, in observational studies, the investigators try not to influence the natural course of events for the study subjects.

The price paid through the use of observational studies is that considerable efforts are required to prevent confounding (bias) of the exposure–disease association. Experiments are the preferred choice when the treatment is straightforward and easily manipulated, such as a vaccine or a specific therapeutic agent. The major advantage of the experimental approach is the ability to control potential confounders through the process of randomisation.

If observational studies have to fight confounding the way experiments avoid it, the natural question is how to design them so that the fight is winnable. Two well-known methodologists — Miguel Hernan and Donald Rubin — offer the same answer in slightly different language: borrow the discipline of the experiment, even when you cannot run one. That answer has three parts, summarized in the accordion below.

A Unified Approach to Study Design

Hernan (2005) stressed that when considering an observational study design, we should think about the design of a field experiment to accomplish the same objective. This approach, reinforced by Rubin (2007), emphasises that ‘design trumps analysis’ and that all elements of the study design should be completed before seeing any outcome data. Expand each of the three steps below in order — the order matters, because each step constrains the next.

The unified approach gives you the discipline that any single study should have. Stepping back further, we can also rank study types by how much causal weight any one well-designed instance of them can carry. That ranking — the so-called hierarchy of evidence — is the last piece of the classification map for this section.

Hierarchy of Evidence for Causal Inference

From the perspective of drawing causal inferences, experimental studies are generally referred to as the gold standard (hierarchy of evidence). The hierarchy of causal evidence (from strongest to weakest) is typically as shown in the table below. The Lesson 1 caveat is worth holding onto here: the hierarchy describes causal weight, not absolute quality. A well-conducted cross-sectional study can be far more useful than a poorly-conducted RCT for many real-world questions.

The reflection below asks you to use the classification scheme we just built to evaluate a research question of your own. The knowledge check that follows tests the conceptual content directly. Once you have worked through both, Section 2 takes the bottom of the hierarchy — the descriptive designs — and asks what useful work they can do despite their lack of a comparison group.

Introduction and Overview

The hierarchy at the end of Section 1 placed descriptive studies at the bottom for causal inference, and that ranking is fair — but it can give the wrong impression about their usefulness. Descriptive studies are where many real epidemiologic investigations start — from James Lind's 1753 treatise on scurvy to John Snow's 1854 mapping of the Broad Street cholera outbreak. They generate the hypotheses that the analytic designs in later sections then test.

Descriptive studies are used to describe the main features of a disease or health-related outcome. Although they are not designed to evaluate associations between exposures and outcomes, the observations made in a descriptive study can form the basis of hypotheses which are then further investigated in analytic studies. Three forms of descriptive studies are case reports, case-series reports, and surveys. The three flip cards below introduce each in turn; click through them and notice how each one expands the scope of observation — from a single unusual case, to a documented group of cases, to a full population sample.

Key Characteristics of Study Types

Descriptive studies differ from analytic observational studies in important ways. The following comparison highlights these differences:

The third descriptive type — the survey — is also the bridge into the rest of the lesson. Once a survey starts collecting risk-factor information alongside disease information, it stops being purely descriptive and becomes the design Section 3 is about.

From Survey to Analytic Study

Kalsbeek and Heiss (2000), and Speybroeck et al (2003) have described the appropriate analysis of surveys bearing in mind the study design. If the survey is designed to collect information about both an outcome of interest and potential exposures (risk factors) beyond the categories of people, place, and time, it then becomes a cross-sectional analytic study and as such, can be used to evaluate associations between exposures and outcomes. The Ontario Hypertension Survey scenario below is a good example of this transition in action; we will return to its design choices in Section 3.

The reflection below puts the descriptive–analytic boundary to work. Once you have answered it and the knowledge check, Section 3 turns the question around: given that we now have a true analytic observational design (the cross-sectional study), how should we actually build one?

Introduction and Overview

Section 1 placed observational analytic studies in the larger classification map, and Section 2 walked through the descriptive designs that lead up to them. This section is the first time we sit inside an analytic observational design and look at how it actually works. We start with two classification distinctions that apply to every observational study you will encounter (prospective vs retrospective; the three sampling approaches), and then narrow in on the design that sits at the centre of this lesson: the cross-sectional study.

Observational Studies Overview

Observational studies (a subgroup of analytic or explanatory studies) have an explicit formal contrast as part of their design: the prevalence of the outcome by exposure category groups is the central foundation. They differ from descriptive studies in that the comparison of two or more groups is central, and from experiments in that the researcher has no control over the allocation of study subjects to the exposure groups.

Prospective vs. Retrospective Designs

Observational studies can also be classified as prospective or retrospective. In prospective studies, the disease or outcome has not occurred at the time the study starts. In retrospective studies, both the exposure and the outcome have occurred when the study begins—hence cross-sectional studies are inherently retrospective in nature.

Sampling Drives the Design

Of the three sampling approaches above, the cross-sectional study is the one this lesson will fully dissect. Lessons 5 and 6 will pick up case-control and cohort designs in the same way. So the rest of this section is the architectural tour of one design, but the questions it asks — how to obtain study subjects, how to assess exposure, how to define the outcome, how to control confounding — are the same questions you will be asking of every observational design from now on.

Cross-Sectional Study Design

The defining feature of a cross-sectional study is that it is an observational study whose outcome frequency measure is prevalence. The basis of the cross-sectional design is that a sample, or census, of subjects is obtained from the source population and the presence or absence of the outcome is ascertained at that point. The accordion below walks through the four design steps in the order a researcher would take them: define the study group, measure exposure, ascertain outcome, and design in protections against confounding from the start.

The scenario below is a real Canadian example that uses every step of the accordion above. As you read it, ask yourself what the researchers did at each step — how the source population was defined, how exposure was measured, how the outcome was ascertained — and where you can already see the design being constrained by the cross-sectional choice. The reflection that follows is built around exactly that question.

Introduction and Overview

Section 3 introduced the cross-sectional design and walked through how to build one. This section is its honest counterweight. Once a study is in the field, three practical questions tend to arrive in sequence: What can this design not tell us? Can we squeeze incidence information out of it? And how should we report the whole thing so the next reader can judge it? Those three questions are the three subsections that follow.

Inferential Limitations of Cross-Sectional Studies

By its nature, a cross-sectional study design measures prevalence, which is a function of both incidence and duration of the disease. Consequently, it is often difficult to disentangle factors associated with persistence of the outcome from factors associated with developing the outcome in the first instance (i.e., becoming a new incident case).

Cross-sectional studies are best suited for time-invariant exposures such as race or sex. In these instances, the investigator can be certain that the exposure preceded, or at least was not caused by, the outcome.

Reverse causation and the prevalence/duration confound are the conceptual limits of the design. The next thing to learn is what the design does let you do, computationally — namely, the cross-tabulation that almost every analytic move in this course is built on.

# Simulate 200 people: half exposed, with higher disease prevalence among exposed.
set.seed(230)
n <- 200
exposure <- rep(c("exposed", "unexposed"), each = n/2)
disease  <- c(rbinom(100, 1, 0.30), rbinom(100, 1, 0.10))

dat <- data.frame(exposure = exposure, disease = disease)

# 2x2 cross-tabulation -- the workhorse of descriptive epi.
tab <- table(dat$exposure, dat$disease,
             dnn = c("Exposure", "Disease"))
tab

# Prevalence (proportion with disease=1) within each exposure group.
prop.table(tab, margin = 1)   # margin=1 -> divide each row by its row total

Estimating Incidence from Cross-Sectional Studies

Although cross-sectional studies directly measure prevalence, there are approaches for estimating incidence from prevalence data. This is often desirable because incidence data are more useful for causal inference. The three tabs below introduce the most common approaches in increasing order of analytical sophistication. Each one trades a different practical cost (run two surveys, develop two assays, build a mathematical model) for a different epistemic gain.

The first incidence-estimation approach — running two cross-sections at different time points — raises a question of its own. If we are willing to track the same population twice, why not just track the same individuals through time? That is the difference between a repeated cross-section and a cohort.

Repeated Cross-Sectional vs. Cohort Studies

Sometimes it is desirable to follow a population over time. Two options exist: repeated cross-sectional samplings of the population, or a longitudinal study of the initial study subjects (a cohort approach). Each has distinct advantages, and the choice between them is a real one. The two cards below put their strengths and limitations side by side — cohorts return in detail in Lesson 5.

The previous subsections covered what cross-sectional studies cannot do, what they can do, and how they relate to longitudinal designs. The last piece of the puzzle is reporting — how to write up an observational study so the next reader can judge it without having to reverse-engineer the methods.

Reporting Observational Studies: The STROBE Statement

In 2004, a network of methodologists, researchers, and journal editors established what we now know as the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (von Elm et al., 2007; explained in detail by Vandenbroucke et al., 2007) — one of the EQUATOR-network reporting guidelines we previewed in Lesson 1. It provides a checklist of 22 items considered essential for good reporting of observational studies.

The reflection below is the practical exit ticket for this section — a textbook reverse-causation case dressed up in plausible language. After working through it and the knowledge check, the lesson moves to its comprehensive final assessment, which integrates the classification map (Section 1), the descriptive designs (Section 2), the cross-sectional design itself (Section 3), and the limitations and reporting standards covered here.

# Simulate 200 people: half exposed, with higher disease prevalence among exposed
set.seed(230)
n        <- 200
exposure <- rep(c("exposed", "unexposed"), each = n/2)
disease  <- c(rbinom(100, 1, 0.30), rbinom(100, 1, 0.10))

dat <- data.frame(exposure = exposure, disease = disease)

# 2x2 cross-tabulation -- the workhorse of descriptive epi
tab <- table(dat$exposure, dat$disease,
             dnn = c("Exposure", "Disease"))
tab

# Prevalence (proportion with disease=1) within each exposure group
prop.table(tab, margin = 1)   # margin=1 -> divide each row by its row total

## -----------------------------------------------------------------------------
## Stretch: prevalence ratio (a quick preview of measures-of-association)
## -----------------------------------------------------------------------------
prev_exposed   <- prop.table(tab, margin = 1)["exposed",   "1"]
prev_unexposed <- prop.table(tab, margin = 1)["unexposed", "1"]
prev_ratio     <- prev_exposed / prev_unexposed
cat("Prevalence ratio:", round(prev_ratio, 2), "\n")

Bringing It All Together

This lesson started with a classification map (Section 1), walked down it to the descriptive designs (Section 2), then up into the first true analytic observational design (Section 3), and finished by interrogating that design’s honest limits and reporting standards (Section 4). The arc was deliberate. The vocabulary you now have — descriptive vs. explanatory, experimental vs. observational, prospective vs. retrospective, the three sampling approaches, prevalence vs. incidence, the unified approach to design, the STROBE reporting framework — is the working vocabulary of every observational study you will read for the rest of HSCI 230.

The deeper move was the one Hernan and Rubin pushed: design trumps analysis. An observational study cannot randomise its way out of confounding, but it can be built so that the analytic decisions are committed in advance, the source population is clearly defined, and the confounding strategy is not an afterthought. That is what the unified approach (the thought experiment, design before seeing data, forward projection) makes operational. Applied seriously, it converts an observational study from something the manufactured-doubt strategy can attack into something a critical reader can actually evaluate.

The two designs that come next — the case-control study (Lesson 4) and the cohort study (Lesson 5) — do not introduce a new framework. They are two more sampling approaches inside the one you already have, each making different trade-offs around how subjects are selected, how outcomes and exposures are ascertained, and what measure of association the design naturally produces. Read those lessons through the lens of this one and the through-line will be clear.

Lesson 3 Comprehensive Assessment

This assessment covers all sections of Lesson 3. You must answer all 15 questions correctly to complete the lesson. Review the feedback after each attempt.