Comparison of direct and indirect standardisation

Direct and indirect standardisation usually give similar results in practice. In general, direct standardisation is preferred to the indirect method. This is because, in direct standardisation, the age-specific rates of the study populations are applied to just one standard population i.e. the weights applied to the age-specific rates are the same. In indirect standardisation, the weights applied to the standard age-specific rates depend on the age structure of the study populations. If there are large differences in age structure between the study populations the SMRs calculated would be based on different weightings and not comparable.

A further advantage of direct standardisation is that, because the denominator used is the same, comparisons across multiple sets of data are transitive. Thus, if direct standardisation is applied to three sets of data (C, D and E), if the CMF between C and D is 2, and that between D and E is 0.5, this would imply that the CMF between C and E is 2/0.5 =4.

The choice of method may be affected by several other considerations, including:

  1. Direct standardisation requires that we know the age-specific rates of mortality (or morbidity) in all the populations under study. Indirect standardisation only requires that we know the total number of deaths (or cases) and the age structure of the study population, and thus indirect standardisation may be the only feasible method if age-specific rates are not available.
  2. Indirect standardisation is preferable when there are small numbers in particular age groups.If we undertook direct standardisation under these circumstances, the estimated rates would be subject to substantial sampling variation.

    With indirect adjustments we can choose rates from a large population as standard, thus minimising the effects of sampling error.



If we were comparing the rates of a rare cancer (such as cancer of the lip) in different regions of a country, when subdivided by region, there might be only be a few cases of cancer in each age group.

Thus if we did a direct standardisation, there would be substantial variation due to random error. Under these circumstances, indirect standardisation would produce more reliable comparisons.

A common use of indirect standardisation is to compare mortality in sub-populations using the age-specific rates of the whole population, as in the example that follows. In such a situation, the number of events observed in the sub-population may be too few to derive robust estimates of mortality rates