Written and maintained by CASRAI Editorial Board
Last updated
Age standardization (also called age adjustment) converts the rates of two or more populations into rates that would be observed if each had the same age structure. It exists because age is tied to most health outcomes, so a population with more older people will show a higher crude rate even when its age-specific risks are identical to another population’s.
Definition in one paragraph
An age-standardized rate is a weighted average of age-specific rates, where the weights come from a single standard population rather than from the population being studied. Because every population is weighted by the same age distribution, differences between the resulting rates cannot be caused by differences in age structure. The US National Center for Health Statistics (NCHS) cautions that age-adjusted rates “should be viewed as relative indexes rather than actual measures of risk”: they are for comparison, not for stating how many people are affected. For the underlying concept of a rate, see the guide to mortality rate.
Age standardization vs age adjustment: are they the same?
In practice, yes. NCHS and the Pennsylvania Department of Health use “age adjustment” and “age-adjusted rate”; WHO’s discussion paper on the topic is titled “Age Standardization of Rates”, and the NCI SEER program lists “standard populations”. Both terms describe the same operation. The two main methods are the direct method (described next) and the indirect method, which produces a ratio such as the standardized mortality ratio instead of a rate.
Formula for the direct method, with every symbol explained
NCHS (Health, United States, “Age adjustment”) gives the direct method as:
Age-adjusted rate = Σ rᵢ × (pᵢ / P), summed over i = 1 to n
- rᵢ: the rate in age group i of the population of interest (for example, deaths per 100,000 people aged 45-54).
- pᵢ: the number of people in age group i in the standard population.
- P: the sum of all pᵢ, the total standard population.
- pᵢ / P: the weight for age group i; the weights add up to 1.
- n: the number of age groups.
The Pennsylvania Department of Health describes the same calculation as steps: compute each age-specific rate (deaths divided by population), multiply it by the standard population in that age group to get expected deaths, sum the expected deaths, and divide by the total standard population (times 100,000 for a rate per 100,000).
Which standard population to use
NCHS uses the projected year 2000 US standard population for its age-adjusted death rates. The NCI SEER program lists the standards in common use:
| Standard | Where it is documented |
|---|---|
| US standard populations (1940, 1950, 1960, 1970, 1980, 1990, 2000) | NCI SEER Standard Populations |
| World (WHO 2000-2025) | NCI SEER Standard Populations, citing WHO Discussion Paper 31, “Age Standardization of Rates: A New WHO Standard” |
| World (Segi 1960) | NCI SEER Standard Populations, citing WHO Discussion Paper 31 |
| European (Scandinavian 1960) | NCI SEER Standard Populations, citing WHO Discussion Paper 31 |
| European (EU-27 plus EFTA, 2011-2030) | NCI SEER Standard Populations |
The choice changes the answer. NCHS reports that the mortality ratio for the Black to White populations falls from 1.6 using the 1940 standard to 1.4 using the 2000 standard, because the 2000 standard puts more weight on older ages. The practical rule from the Pennsylvania Department of Health is to use the same standard population in every adjusted rate you compare.
Worked example with real numbers
The two populations below are hypothetical, invented for this illustration. Population A and Population B have exactly the same age-specific death rates, but B is much younger. The weights are the NCHS year 2000 standard million weights for 11 age groups (the 75-84 weight is rounded up so the total is 1.000000, as NCHS notes).
| Age group | Rate per 100,000 (both populations) | Standard weight | Rate × weight |
|---|---|---|---|
| Under 1 | 200 | 0.013818 | 2.764 |
| 1-4 | 20 | 0.055317 | 1.106 |
| 5-14 | 10 | 0.145565 | 1.456 |
| 15-24 | 40 | 0.138646 | 5.546 |
| 25-34 | 70 | 0.135573 | 9.490 |
| 35-44 | 150 | 0.162613 | 24.392 |
| 45-54 | 400 | 0.134834 | 53.934 |
| 55-64 | 1,100 | 0.087247 | 95.972 |
| 65-74 | 2,800 | 0.066037 | 184.904 |
| 75-84 | 6,500 | 0.044842 | 291.473 |
| 85 and over | 15,000 | 0.015508 | 232.620 |
| Total | 1.000000 | 903.655 |
Age-adjusted rate = 903.7 per 100,000 for both populations.
Now compare the crude rates. Population A has 850,000 people (mostly middle-aged and older) and 12,798 deaths: 12,798 / 850,000 × 100,000 = 1,505.6 per 100,000. Population B has 840,000 people (mostly young) and 2,494 deaths: 2,494 / 840,000 × 100,000 = 296.9 per 100,000. The crude rates differ about fivefold, yet the age-specific risks are identical; the adjusted rates correctly show no difference. As a check, weighting the same rates by the raw NCHS standard counts (total 274,633,642) gives 903.65 per 100,000, matching the table.
Interpretation: reading the result
| Situation | How to read it |
|---|---|
| Crude rates differ, adjusted rates are similar | The crude difference came from age structure, as in the example above. |
| Adjusted rates differ | The difference remains after the age structures are made equal, so something other than age is involved (an adjusted rate alone does not say what). |
| A single adjusted rate | Not an absolute level of risk. The Pennsylvania Department of Health notes it is not an indicator of the absolute level of mortality and is useful only for comparison. |
| Same data, different standard | The numbers and sometimes the size of a ratio change (NCHS: 1.6 vs 1.4). Compare only rates built on the same standard. |
Adjusting for age does not remove other confounding. The same steps can be repeated to adjust for sex or race, according to the Pennsylvania Department of Health.
Assumptions and common mistakes
- Using different standards across rates. Rates built on different standard populations cannot be compared with each other.
- Using different age groups. NCHS notes that estimates can differ between reports when different age groups are used; state the groups.
- Treating the adjusted rate as the real rate. It is a relative index. Where precision matters, age-specific rates can be more informative, as the Pennsylvania Department of Health points out.
- Comparing indirectly standardized ratios with each other. A ratio such as the SMR is weighted by each study population’s own age structure; for head-to-head comparisons use the direct method on a common standard (see also observed-to-expected ratios).
- Confusing prevalence and incidence inputs. The rates you standardize must be defined consistently across groups; see prevalence vs incidence.
How to report an age-standardized rate (APA example)
Name the method, the standard population, the age groups and the unit. For example:
Age-standardized mortality rates were calculated by the direct method using the year 2000 US standard population in 11 age groups (National Center for Health Statistics, n.d.). The adjusted rate was 903.7 deaths per 100,000 in both populations.
Reference entry: National Center for Health Statistics. (n.d.). Age adjustment. In Health, United States. Centers for Disease Control and Prevention. https://www.cdc.gov/nchs/hus/sources-definitions/age-adjustment.htm
How to calculate it in Excel, R or Python
Each snippet computes the direct age-adjusted rate from age-specific rates and standard population counts (here, the 11 NCHS year 2000 counts); the example values reproduce 903.65 per 100,000.
Excel
Put age-specific rates in B2:B12 and standard population counts in C2:C12, then:
=SUMPRODUCT(B2:B12, C2:C12) / SUM(C2:C12)R
rates <- c(200, 20, 10, 40, 70, 150, 400, 1100, 2800, 6500, 15000)
std <- c(3794901, 15191619, 39976619, 38076743, 37233437, 44659185,
37030152, 23961506, 18135514, 12314793, 4259173)
weighted.mean(rates, std) # 903.654Python
import numpy as np
rates = [200, 20, 10, 40, 70, 150, 400, 1100, 2800, 6500, 15000]
std = [3794901, 15191619, 39976619, 38076743, 37233437, 44659185,
37030152, 23961506, 18135514, 12314793, 4259173]
print(np.average(rates, weights=std)) # 903.654FAQ
What is age standardization in simple terms?
It is a way of comparing rates between populations after giving each the same age structure, so that age differences do not distort the comparison.
Is an age-adjusted rate the same as a crude rate?
No. A crude rate is total events divided by total population and reflects the population’s own age mix. An age-adjusted rate applies age-specific rates to a standard population.
Which standard population should I use?
Use the one your audience or comparison data use, and name it. US reports from NCHS use the year 2000 standard; NCI SEER also lists WHO world and European standards. Never mix standards in one comparison.
What is the difference between the direct and indirect method?
The direct method needs an age-specific rate for every age group, as in the formula above. The indirect method gives a ratio rather than a rate; see the standardized mortality ratio guide.
Sources
- National Center for Health Statistics, Health, United States: Age adjustment. https://www.cdc.gov/nchs/hus/sources-definitions/age-adjustment.htm (checked 2026-10-29)
- Pennsylvania Department of Health, Age-Adjusted Rates. https://www.pa.gov/en/agencies/health/health-statistics/statistical-resources/understanding-health-statistics/tools-of-the-trade/age-adjusted-rates.html (checked 2026-10-29)
- National Cancer Institute SEER Program, Standard Populations. https://seer.cancer.gov/stdpopulations/ (checked 2026-10-29)
