Appendices 1-5

Abstract

If one needs to compare morbidity or mortality rates of two or more populations, a variable (for example, age or gender) which is related to the outcome of interest can distort the comparison if the populations differ with respect to the distribution of that variable. One therefore needs to standardise or adjust rates before they can be compared. This appendix outlines standardisation with respect to age distribution-a similar approach would be used to standardise for gender or for age and gender. Standardisation is also dealt with in , as it is often used in mortality studies. Chapter 15 Standardisation can be direct or indirect, depending on the type of information that is available for the populations (see). Table A1.1 To perform direct age-standardisation, the researcher needs age-specific rates for each population, as well as a standard or reference age distribution. A step-by-step example is provided in. For Table A1.2 each population, the standardised rate is calculated by (a) multiplying each age-specific rate by the number of persons in that age category in the standard population, (b) adding these numbers, and (c) dividing by the number of people in the standard population. (The same results can be derived by multiplying each age-specific rate by the proportion of the standard population which falls into that age category, and then adding these weighted age-specific rates.) Directly standardised rates are therefore obtained by mathematical manipulation of category-specific rates (in this case age-specific rates). These standardised rates are hypothetical summary rates for purposes of comparison. The researcher can choose as standard distribution the age distribution of either of the populations of interest, their joint distribution, or some other standard such as the World Standard (Ahmad et al, 2001). As standardised rates depend on which standard is chosen, various standards can lead to varying conclusions. Researchers often pay too little attention to the informative category-specific rates by focusing only on the standardised rates. No single summary rate can reflect the richness of category-specific rates, but if there are many categories it may be difficult to make useful deductions based just on category-specific. rates Indirect age standardisation is used when age-specific rates are not available for one or more of the populations of interest. For this standardisation method, the researcher uses the crude number of cases in each of the populations (the 'observed' number), the age distribution of each of the populations, and the age-specific rates of a standard population (which could be one of the populations being compared, if the information is available). See for a step-by-step example. By (a) multiplying each age-specific Table A1.2 rate of the standard population by the number of people in the matching age category of the population of interest and then (b) adding these, one obtains the 'expected' number of cases in that population. By (c) dividing the observed by the expected number of cases, a standardised morbidity (or mortality) ratio (SMR) can be calculated for the population. The various standardised rates for the data provided in are summarised in. Table A1.2 Table A1.3 A word of caution about indirect standardisation: we use the category distribution (in our example, age distribution) of each population to calculate each population's expected number of cases. These distributions may differ vastly, and this method has therefore also been called the 'changing base' method. When one compares SMRs, one is actually comparing measures based on differing standards. By contrast, direct standardisation (or the 'fixed base' method) involves the comparison of measures based on the same standard, and therefore represents the essence of standardisation. It can be seen from that the value of a relative index depends on the method used to calculate Table A1.3 it. While the direction of the relationship will remain the same, the magnitude depends on which population was selected as standard. It is crucial to report details of the standard population and the method of standardisation when reporting standardised rates. Standardisation can also be used if the researcher has done a study on a sample in which the proportion