[Durgical](https://durgical.com/en/) › [DurgiPedia](https://durgical.com/en/durgipedia/) › [Epidemiology](https://durgical.com/en/epidemiology/) › [Study designs](https://durgical.com/en/epidemiology/areas/designs/) › Case-crossover design

# Case-crossover design

- Study design

Only cases are studied: each person is compared with themselves, just before the event and at other times in their life.

Updated Nov 13, 2026

## Key points

- Cases only; the control is the same person at another time.
- Hazard window (just before the event) versus control windows (other days or hours).
- Transient exposure + acute, sudden-onset event (myocardial infarction, crash, asthma attack).
- Controls each person's fixed confounders, even unmeasured ones.
- Analysed like a matched study: discordant pairs or conditional logistic regression.

On this page

- The basics Key points
- What it is
- How it is built
- Worked example
- Markers
- When to use it
- Common pitfalls
- Memory aid

- Clinical Case-crossover versus neighbouring designs

- Study Test yourself

- Connections Connections

## What it is

In a case-crossover design only cases are recruited, and each one serves as their own control. Exposure in the hazard window (the period just before the event) is compared with the same person's exposure in earlier control windows. It suits transient exposures that might trigger acute events, and by design it removes confounding by traits that do not change over time (genetics, sex, socioeconomic level).

## How it is built

- Cases of an acute event with a precise date and time are gathered.
- The hazard window is set by biology (for example, the previous hour).
- One or more control windows of the same length are chosen in the same person (the same weekday a week earlier, for example).
- Exposure in each window is measured from the same source (interview, records, environmental data).
- The matched OR is computed by comparing each case's windows.

## Worked example

- 200 adults with acute myocardial infarction are studied. Exposure: heavy exertion. Hazard window: the hour before; control: the same hour the day before.
- Exposed in both windows: 6; in neither: 154.
- Exposed only in the hazard window: 30; only in the control window: 10. Check: 6 + 154 + 30 + 10 = 200.
- Matched OR = 30 / 10 = 3.0 (the 160 concordant cases add nothing).
- Reading: in these made-up data, the odds of infarction in the hour after heavy exertion are about 3 times those of an hour without it.

## Case-crossover versus neighbouring designs

| Aspect | Case-crossover | Matched case-control | Crossover trial |
| --- | --- | --- | --- |
| Who the control is | The same case at another time | Another similar person | The same patient on another treatment |
| Who assigns exposure | Nobody: it is observed | Nobody: it is observed | The investigator, at random |
| Suitable exposure | Transient | Transient or long-lasting | Reversible treatment |
| Fixed confounders | Controlled by design | Only those matched or measured | Controlled by design |
| Main threat | Time trends and unequal recall between windows | Control selection | Carry-over |

## Markers

- Hazard window
- Control window
- Matched OR

## When to use it

- Triggers of myocardial infarction or stroke (exertion, anger, heavy meals).
- Air pollution on given days and asthma visits or cardiovascular deaths.
- Phone use while driving and crashes; acute effects of a newly started medicine.

## Common pitfalls

- Using it for chronic or slowly changing exposures: almost every case will be concordant.
- Choosing control windows during a time trend (exposure rising over time): trend bias.
- Relying on sharper recall of the hazard window than of the control window (recall bias).

## Memory aid

Case-crossover is ‘me today versus me yesterday’. Fleeting trigger, sudden event: compare the person with themselves.

## Test yourself

Think of the answer, then tap the question to see it.

## What kind of exposure and outcome suits a case-crossover design?

Transient exposures that might trigger an acute, sudden-onset event.

## Which confounders does this design control automatically?

The person's traits that do not change over the study period, such as genetics, sex or education, even if unmeasured.

## Why does it not work for long-term smoking and lung cancer?

Because the exposure is constant and the outcome is not acute: the hazard and control windows would have the same exposure.

[Add to my cards (3)](https://durgical.app/estudiar/?add=/en/epidemiology/case-crossover-design/)

## Connections

## Related 5

- [Case-control study](https://durgical.com/en/epidemiology/case-control-study/)
- [Matching](https://durgical.com/en/epidemiology/matching/)
- [Crossover trial](https://durgical.com/en/epidemiology/crossover-trial/)
- [Confounding](https://durgical.com/en/epidemiology/confounding/)
- [Odds ratio](https://durgical.com/en/epidemiology/odds-ratio/)

[← All of epidemiology](https://durgical.com/en/epidemiology/)

Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.

Sources: [Case-Crossover Study Design](https://www.publichealth.columbia.edu/research/population-health-methods/case-crossover-design) (Columbia University Mailman School of Public Health, 2026); [Principles of Epidemiology | Lesson 1 - Section 7](https://archive.cdc.gov/www_cdc_gov/csels/dsepd/ss1978/lesson1/section7.html) (Centers for Disease Control and Prevention, 2012); [Chapter 25: Assessing risk of bias in a non-randomized study](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-25) (Cochrane, 2024).

## **Spotted a mistake?** We are always improving for the community. Tell us what is wrong and our team will review it. Report a mistake

Representative model, for educational purposes only.

---
Source: https://durgical.com/en/epidemiology/case-crossover-design/
License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Cite as: DurgiPedia (durgical.com), https://durgical.com/en/epidemiology/case-crossover-design/