2.4 Calculating standardized rates

It is particularly important when working with geographical health event data that prevalence rates take into account the nature of the population at risk. For example, an area with many elderly residents is likely to have many more cases of coronary heart disease than a city centre area with younger residents. This is because elderly residents are much more likely to have heart disease than the young. When faced with this problem, health geographers use a technique called standardisation to ‘factor in’ the nature of the population into disease rates. Standardisation produces a figure known as a standardised morbidity (or mortality) rate, which shows the variation in disease rates after taking into account the effect of the age of the population living in different areas. We shall consider the standardization of rates here in relation to mortality data, although the principles apply equally to morbidity data. We will also focus in on a particular form of standardisation (indirect standardisation), which is commonly used by geographers.

Mortality rates are one the most frequently used health indicators and are derived from information about death registrations. Death certificates vary in detail between countries but will generally include a record of the age, sex, place of residence of the person who has died and cause of death. In many countries additional information such as employment allow social class and other analyses to be conducted on death registration records. A standard encoding system known as the International Classification of Diseases (ICD) is generally used in order to classify the cause of death recorded on the certificate for statistical purposes. Crude rates – expressed as the number of occurrences in relation to a population denominator – use denominator populations of different sizes according to the rarity of the event. The paper by Bunting and Kelly (1997) that examines geographical pattern in suicide rates uses crude rates per million population, for example, while the World Health Organization data used here are reported per 1,000 population.

The table below shows main causes of death by age-group in Scotland in 1999 (death rates per 1,000 population) from the World Health Organization Global Health Observatory http://www.who.int/gho/en/.

 

The table reveals very clearly how different causes of death are dominant at different stages of life, and how the rates of death differ greatly with age. They are initially high in relation to perinatal and congenital conditions then decline dramatically during childhood and adolescence when accidents are the main cause of death. During young adulthood suicide briefly takes the first position, while at all ages over 35 malignant neoplasms (cancers) and circulatory diseases are the main causes of death – cancers dominant in the middle age groups and circulatory diseases among the elderly. This pattern is very typical of a developed western economy although we would find extremely different patterns in other societies. It is these characteristic differences in the mortality rates between population groups that help us to understand why standardization is necessary. The rates displayed in the table below are taken from the World Health Organization Global Health Observatory http://www.who.int/gho/en/. The table above shows deaths per 1,000 population, known as crude rates. Using these figures, the regions with the highest death rates will simply be those with the greatest concentrations of the elderly. It is thus clear that we need some way to standardize these rates in order to take into account the population structure of a region if we are to compare its rates with those of other regions (or countries). standardization is thus an important stage in geographical analysis.

We will illustrate the calculation of a standardized mortality rate by using some more data for Scotland in 1999 (technically, the particular technique we will use is known as indirect standardisation. There is also a direct method of standardising rates too, though this is more commonly used for very large areas such as entire countries). The following table shows male deaths in Scotland in 1999 standardized to male death rates in the whole of the UK during the same period.

 

The first column of the table (a) shows the total male population (in 000s) in Scotland in 1999, by age group. The second column (b) shows the number of male death rates in that age group for the whole of the UK . The third column calculated as (a x b) thus shows the number of male deaths in each age group we would expect in Scotland if the pattern conformed to the rates for the whole of the UK, and can be summed to give the total number of deaths expected. The final column shows the actual deaths, which are higher than those expected for every age group from 5-14 upwards.

In order to calculate the SMR we take the total actual deaths, divide by the expected deaths and multiply by 100:

SMR = Total actual male deaths / total expected male deaths * 100

SMR = 28605 / 24227.71 * 100 = 118.067

The interpretation of SMR is that a value of 100 would represent exactly the number of deaths that would be expected if our region conformed to the overall UK levels, adjusted for its age structure. In this case, it is apparent that the male SMR for Scotland is some 18% higher than that for the UK as a whole. If our intention is to map death rates, then we should begin by calculating SMRs. It is conventional to standardize by age and sex, although it is entirely possible to standardize for any variable (e.g. social class) for which data are available at the level of the regions to be mapped and the reference area as a whole.

This is an example of indirect standardization, where disease rates from the reference population are applied to the local population. The other version of standization, direct standardization, involves doing the calculation the other way around and applying the rates from the local population to the reference population. This indirect standardization technique is frequently used in a GIS context, with disease rates for many local populations being standardized. There is typically one local population with disease data to be standardized for each polygon on a map layer. Sometimes in a GIS context, the reference population is created by summing together all of the population cohort data for the sub-regions across an area. There are also standard reference population data sets that are used for standardising rates. The World Health Organization use a standard global population data set, for example, when they calculate standardised mortality rates for particular diseases for the different countries of the world.

---

---