[Frequency analysis for achieving health goals--I: Standardization]

The formal mathematical steps of direct and indirect standardisation are deduced and it is pointed out that standardisation is an analytical concept for processing data within the framework of structurised populations to enable correct interpretation of these data in a manner modified according to specific problems. In particular, this concept is not confined to transforming age-structured mortality data of a population to the age distribution of the members of a larger population by means of a rule of three. Standardisation is in fact a model of the macrolevel (population). It makes no demands on the modalities of structuring. The structure selected as standard is arbitrarily set. With time series, standardisation usually modifies only the levels of the figures and not the frequencies. The results of standardisation are in every case valid only if the standard is also stated. A rule of thumb is: In retrospective analysis it is meaningful to standardise on the structure at the beginning or end of the time series; in extrapolations standardisation should be performed on the last structure that had been determined, whereas for comparisons a median structure should be standardised. The standardisation approach can be used in case of weighted arithmetic, geometric or harmonic means. Before effecting standardisation, the type of connection between the employed data should be examined. The examples of concepts of direct standardisation refer to cardiovascular and accident mortality in West Berlin from 1963 to 1991, whereas those of indirect standardisation are based on the accident mortality in West Berlin in 1987 structured according to city districts.