Indirect Standardization

Rates (incidence, mortality, etc.) calculated from different study groups or populations are often not directly comparable if the groups differ with regard to the distribution of some characteristic associated with [Page 530]the outcome of interest. Adjustment of rates to facilitate comparison across populations or for the same population across time is called standardization and involves adjusting the crude rates ‘as if’ they were calculated based on the same underlying population. The crude rates are ‘adjusted’ for the characteristic on which the groups differ. For example, if the mortality rates of two different communities are to be compared, but the age distributions of the two communities are sufficiently dissimilar, comparison of the crude rates is inappropriate and will likely yield inaccurate conclusions.

There are two methods of standardization, direct (discussed elsewhere) and indirect. Indirect standardization is most often applied when the specific rates for the study population are not calculable, as is required for direct standardization, or when some strata of the study group are so sparsely populated that stable rates are not determinable.

Indirect standardization accomplishes adjustment by calculating the rates for a reference population and then applying weights (as person-years of follow-up) derived from the study population to the rates calculated from the reference population. Thus, the expected number of events in the study group is estimated from the observed number of events in the reference population. The resulting ratio of observed events to expected events is the adjusted rate ratio that compares the study group with the reference group. The standard mortality ratio (SMR) is an example of the use of indirect standardization to compare the mortality rate of a study group, an occupational cohort, for example, with that of a reference population.

Ideally, the chosen reference population would be as closely representative of the study population as is feasible. When two different study groups are indirectly standardized to the same reference population, the adjusted rates are directly comparable. When a single study group is compared with a single reference group, the direct and indirect methods of standardization are equivalent.

In general, presenting standardized rates instead of rates specific to a particular study group is advantageous in several ways. First, and perhaps most important, as discussed in the foregoing, the standardizing of rates in effect adjusts for potential confounding by those factors on which the rates are standardized, such as age. A standardized rate is effectively a summary measure that is then easier to compare with other similar summary measures than are unstandardized rates.

Statistically speaking, standardized rates have a smaller standard error than do unstandardized rates. Standardized estimates are more stable and less influenced by small cell sizes or sparsely populated strata. Finally, specific rates may be unavailable for certain groups of interest, so indirect standardization is the method for making estimates about such groups.

Standardization is not without its disadvantages, however. In instances where there is effect modification, a standardized rate, because it is in essence an averaging of the specific rates, will tend to obscure any differences between strata. In addition, the magnitude of the standardized rate depends on the chosen standard population and is thus arbitrary. The standardized rate itself doesn't necessarily yield useful information, but the difference or ratio between rates is important.

...