Crude rates provide a useful summary measure to compare similar populations of different sizes, but crude rates are sensitive to differences in age compositions.
For example, a county with an older population will have a higher crude death rate due to cancer, even though its risk exposure levels and age-specific cancer rates may the same as those in other counties. One might incorrectly attribute the high cancer rate to some characteristic of the county other than age. Age-adjustment may also be used to control for age effects when comparing across several years of data, as the age distribution of the population changes over time. Calculating age-adjusted rates may be accomplished using direct, or indirect age standardization.
Direct Age-adjustmentDirect age-adjustment (or age standardization) is the same as calculating a weighted average. It weights the age-specific rates observed in a population of interest by the proportion of each age group in a standard population (Lilienfeld & Stolley, 1994).
In 1999, the Centers for Disease Control and Prevention (CDC) replaced the 1940 U.S. standard population weights that had been used for the several previous decades with revised standard population weights for direct age-adjustment (Klein & Schoenborn). Table 1, below, contains the CDC standard population weights, which represent the proportion of the U.S. 2000 population in each age group, and sum to 1.0.
Compare only age-adjusted rates that have been adjusted to the
Unless otherwise noted, all age-adjusted rates in NJSHAD have been adjusted using the US 2000 population standard.
Age-adjusted rates should be viewed as
Table 1. U.S. 2000 Standard Population Weights for Age Standardization
Calculating Age-Adjusted Rates Using the Direct Method
To apply direct age-adjustment to a set of rates, the age-specific rate for each age group in the study population is multiplied by the appropriate weight in the standard population. The sum of these products is the directly age-adjusted, or age-standardized rate. The age-adjusted rate can be considered an average of each of the individual age-specific rates, but rather than being a simple average, it is a weighted average with each age-specific rate weighted by the proportion of people in the same age group in the standard population.
Tables 2a. and 2b. demonstrate the method used by NJSHAD in calculating age-adjusted rates. Notice that using crude death rates in Tables 2a. and 2b., one might conclude that whites have a higher underlying risk for heart disease death compared with blacks. How should the age-adjusted death rates be interpreted? You could use confidence intervals to assist in interpreting these data (NJSHAD automatically provides 95% confidence intervals for all rates).
Table 2a. Age-Adjusted Death Rate due to Heart Disease among Whites, New Jersey, 2004
Table 2b. Age-Adjusted Death Rate due to Heart Disease among Blacks, New Jersey, 2004
Age adjustment is not appropriate if the age-specific death rates in the population of interest do not have a consistent relationship. For example, if the death rate among younger persons is increasing over time, but the death rate among older persons is decreasing over time, one would not want to age-adjust rates across years. One's conclusion of the trend in this death rate would be different, depending on which standard population is used. A younger standard population (such as the US 1940) would show an increase, whereas an older standard population (such as the US 2000) would show a decrease, or no change at all. Care should be taken so that the selection of the standard population does not affect the comparisons. For more information, see Curtin & Klein.
When reporting age-adjusted rates, always report the standard population used, and when comparing results to other data, be sure to document that those data were also age-adjusted and report the standard population. The age-adjusted rate is hypothetical, and is useful only for
Although age-adjustment may be used with broad population age groups, such as adults (e.g., age 18+), it is not necessary (or meaningful) to age-adjust data for smaller age groups (e.g., age 18-24).
FAQs for Age-Adjustment:
Event Rates for a Subpopulation
When NOT to Age-Adjust
Age/Sex Adjusted Rates
Confidence Intervals for Age-Adjusted Rates
In some cases, such as when there are too few cases to stratify by age, "indirect age standardization" may be used. Indirect standardization is based on standard mortality and morbidity ratios (SMR), and adjusts the age-specific rates found in the standard population to the age distribution of the smaller area or sub-population. According to Curtin & Klein, "One of the problems with [direct age adjustment] is that rates based on small numbers of deaths will exhibit a large amount of random variation. A very rough guideline is that there should be at least 25 total deaths over all age groups." NJSHAD follows NCHS's guideline of 20 total deaths. When fewer than 20 health events occurred over a time period, you may consider combining years, or using indirect age-adjustment.
The direct method can present problems when population sizes are particularly small. Calculating directly standardized rates requires calculating age-specific rates, and for small areas these age-specific rates may be based on only one or two events. In such cases, indirect standardization of rates may be used.
Indirectly standardized rates are based on the standard mortality or morbidity ratio (SMR) and the crude rate for a standard population. Indirect standardization adjusts the overall standard population death rate to the age distribution of the small area (Lilienfeld & Stolley, 1994). It is technically appropriate to compare indirectly standardized rates only with the rate in the standard population, not with each other.
Deciding Which Measure to Use
The measure that best informs the question you are trying to answer is the one to use. This is a guideline, not a hard and fast rule, but generally:
1. Anderson RN, Rosenberg HM. Age Standardization of Death Rates: Implementation of the Year 2000 Standard. National vital statistics reports; vol 47 no.3. Hyattsville, Maryland: National center for Health Statistics. 1998.
2. Klein RJ, Schoenborn CA. Age-Adjustment Using the 2000 Projected U.S. Population. Statistical notes; no.20. Hyattsville, Maryland: National center for Health Statistics. January 2001.
3. Curtin, LR, Klein, RJ. Direct Standardization (Age-Adjusted Death Rates). Statistical notes; no.6. Hyattsville, Maryland: National center for Health Statistics. March 1995.
4. Fleis, JL. Statistical methods for rates and proportions. John Wiley and Sons, New York, 1973. As cited in Curtin and Klein, 1995.
5. Klein RJ, Schoenborn CA., 2001.
6. Lilienfeld, DE and Stolley, PD. Foundations of Epidemiology, 3rd Ed. Oxford University Press, 1994.