Age, period and cohort effects; reading trends
- Concept
A rate changes because of people's age, because of something that happened in a given year, or because of the generation they were born in.
What it is
Trends in a rate break down into three effects: age (risk changes as people grow older), period (something affects all ages at once in a given year) and cohort (a generation born in certain years carries its own exposure throughout life). Before accepting that a trend is real, artefactual changes must be ruled out: new classification codes, better diagnostic methods, screening or errors in denominators.
Worked example
- Rate of a cancer per 100,000 in 2000, 2010 and 2020. Age 40–49: 10, 20, 20. Age 50–59: 20, 20, 40.
- Age 60–69: 40, 40, 40. In every year the rate rises with age: an age effect.
- The rise appears in 2010 only at 40–49 and in 2020 at 50–59: always in people born in 1961–1970.
- That generation has double the previous one's rate at the same age: 20 against 10, and 40 against 20.
- A period effect would have raised all ages in the same year; that does not happen here.
- Reading: a cohort effect; look for a new exposure in that generation. Around 2030 the rise would be expected at 60–69.
Three effects in a trend
| Effect | How it shows in the table | Example |
|---|---|---|
| Age | Rises or falls down each year's column | Cancer more common with age |
| Period | All ages change in the same year | A new test that finds more cases at all ages |
| Cohort | The change follows one generation diagonally | Generations that took up smoking in large numbers |
| Artefact | A sudden jump that coincides with a recording change | A change of ICD version |
Markers
- Age
- Period
- Cohort
- Trend
In practice
- Before announcing an epidemic, check whether the case definition, code, diagnostic test or screening changed.
- Rising incidence with stable mortality suggests more diagnosis (overdiagnosis) rather than more disease.
- Use age-adjusted rates to compare years: the population is ageing.
Common pitfalls
- Reading the age curve of a single year as the story of one person growing older.
- Comparing rates without checking whether the denominator changed (new census, migration).
- Blaming chance for a jump that coincides with an ICD change or a new surveillance system.
Test yourself
Think of the answer, then tap the question to see it.
What pattern in an age-by-year table shows a cohort effect?
A change that appears first in the youngest and moves diagonally to older ages with the same generation.
What pattern shows a period effect?
A simultaneous change at all ages in the same year.
A cancer's incidence doubles after screening starts and mortality does not change. What does it suggest?
A rise that is largely artefactual from detection (including overdiagnosis), not more disease.
Why can the three effects not be fully separated?
Because they are linearly dependent: year of birth is calendar year minus age.
Connections
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: Gen X Projected to Experience Greater Increases in Cancer Incidence than Previous Generations (National Cancer Institute, Division of Cancer Epidemiology and Genetics, 2024); SEER Cancer Statistics Review 1975-2016: Introduction (National Cancer Institute, 2019); Age adjustment (National Center for Health Statistics, Centers for Disease Control and Prevention, 2026).
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.