Overview of standardisation

Introduction to age standardisation

Disease and mortality rates often vary widely by age. Such variation complicates any comparisons made between two populations that have different age structures. For example, consider two areas A and B with equal-sized populations and identical crude all-age death rates. At first glance they appear to have a similar mortality experience. Suppose, however, that area A has a younger age structure than area B. Given that mortality rates increase with age, one would expect the older population in area B to experience more deaths. The fact that the two have identical rates means that the younger population in area A must have a relatively worse mortality experience.

The most comprehensive way of comparing the disease experience of two populations is to present and compare their age-specific rates. However, when the number of populations being compared increases, the volume of data to consider quickly becomes unmanageable. What is needed is a single, easily interpreted, summary figure for each population that is adjusted to account for its age structure. Such summary figures are calculated using age standardisation methods. It may also be desirable to standardise for other variables, such as sex or level of deprivation, that may also potentially confound any comparisons. The two most common methods of age standardisation are:

  • indirect: the age-specific rates of a chosen standard population (usually the relevant national or regional population) are applied to the age structure of the subject population to give an expected number of events. The observed number of events is then compared with that expected. This is usually expressed as a ratio (observed divided by expected). A common example is the standardised mortality ratio (SMR).

  • direct: the age-specific rates of the subject population are applied to the age structure of the standard population. This gives the overall rate that would have occurred in the subject population if it had the standard age profile and the observed age-specific rates.


Indirect standardisation

Advantages

  • The indirect method requires only the total number of observed events in the subject population and may therefore be used in some situations where the direct method cannot.

  • Indirect standardisation is more stable as it minimises the variance, giving a smaller standard error and narrower confidence intervals than the direct method. It can therefore be more appropriate when dealing with the statistical significance of small populations (Breslow and Day, 1987).

Disadvantages

  • Indirect standardisation does not necessarily preserve consistency between the populations being compared. In extreme situations this may give misleading results (see Figure 1).

  • Indirectly standardised ratios for areas A and B may be compared with the standard but should only be directly compared to each other if the age structures of areas A and B are similar to the standard, or if the ratio of their age-specific mortality rates to that of the standard is consistent across the age groups (Tsai and Wen, 1986).

  • Indirect standardisation can be used for comparisons over time, but only by freezing the standard rates at a fixed point in time. This can seem less intuitive. The validity of the comparisons is subject to the same conditions as described above for comparison between areas.


In Figure 1, even though Population A has double the rate of Population B in every age group, the SMR for Population A is lower than that for Population B. This arises because the population distribution is very different, and the ratios of the local rates to the reference rates are very uneven. Although it is not unusual for different populations to have very different age distributions, it is unusual for the ratios to be so different. In general rates will tend to vary with age in a consistent way across different populations.


(a) Population A has an SMR of 89
(b) Population B has an SMR of 99
(c) Reference rates
Figure 1: Example of misleading indirectly age-standardised ratios

Direct standardisation

Advantages

  • The direct method preserves consistency between the populations being compared. This means that if each age-specific rate in area A is greater than each of the corresponding age-specific rates in area B, then the directly standardised rate for area A will always be higher than that of area B irrespective of the standard population used (Inskip, 1983; Breslow and Day, 1987). As a result, this is the preferred method for comparing a number of different populations with each other.

  • Directly standardised rates can readily be compared over time, provided the same standard population is used.

Disadvantages

  • The direct method requires that the observed events in the subject population are available broken down by age. If this information is not available, the directly standardised rate cannot be calculated.

  • Where the subject population is small, age-specific rates are based on small numbers and consequently are unstable. This means that small changes in the number of deaths in a particular age band may result in large changes in the directly standardised rate. However, as long as the total number of deaths across all age groups is at least 10, the calculation has been shown to be valid. (Morris et al., 2018)

The European Standard Population (ESP) is often used for direct standardisation. This is a hypothetical population structure which does not change and is the same for both sexes. The advantage of using such a hypothetical population is one of greater comparability, for example, between different countries, across time periods, and between sexes. To allow such comparisons, the method applies a weighting to age groups. This means it gives different weights to each event, such as deaths or admissions, according to age. When used in general populations, such as whole local authorities, there is no significant problem. It is, however, not usually appropriate to use the ESP where the denominator is a subset of the population with a very different age distribution, for example those with a certain condition or occupation or patients admitted to hospital. Ideally, the standard population should be reasonably similar to the populations being standardised. In summary, the ESP must be used when comparisons are to be made with other rates which have been standardised to the ESP, but for self- contained comparative analyses, a more closely matched standard population may be preferred.

Summary

In practice, direct and indirect standardisation generally give comparable results. Moreover, it has been demonstrated by Breslow and Day (Breslow and Day, 1987) that when the two do differ, the direct method is not necessarily the more ‘correct’. Where the ratio between the age-specific rates of areas A and B varies by age group the choice of standard population and rates becomes crucially important for both indirect and direct methods. This choice is often as, or more, important than the choice of method (Goldman and Brender, 2000).


Page last updated: August 2024