A cross-sectional study measures exposure and outcome in the same people at the same time.
A cross-sectional study measures exposure and outcome in the same people at the same time. It is fast and cheap and it produces a picture of a population at a moment, which is genuinely useful for some questions.
How common something is in a population right now is exactly what this design answers, and no other design answers it as directly. Surveys of this kind underpin most public health planning.
Because both are measured together, the design cannot say which came first. An association is compatible with exposure causing outcome, outcome causing exposure, or both following something else entirely.
People with a condition appear in the sample only if they still have it and are still available to be surveyed. Conditions that resolve quickly or end badly are under-represented, which distorts any comparison involving duration.
Who answers a survey differs from who does not, in ways usually related to the topic. A low response rate undermines a cross-sectional study more fundamentally than it undermines designs with internal comparisons.
Surveying a population repeatedly shows how the population changed. It does not show how individuals changed, and conclusions about individual trajectories from repeated snapshots are a known error.
For questions about current distribution, for generating hypotheses cheaply, and for situations where following people forward is impossible. In those cases the alternative is usually no evidence at all.
Take the prevalence figures seriously, treat any association as a hypothesis, and check the response rate before either.
Survey samples rarely match the population, so results are usually weighted to correct for who responded. The weights depend on assumptions about non-responders, and two defensible weighting schemes can produce noticeably different headline figures from identical responses.
Prevalence counts who has a condition now; incidence counts who develops it over a period. A cross-sectional design measures the first and cannot measure the second, and summaries routinely use the words interchangeably in ways that change the meaning substantially.
The claims on this page about a snapshot of a population, and why sequence is missing from it can be checked against sources outside this site: