Westerling R
Department of Social Medicine, Uppsala University, Sweden.
Int J Epidemiol. 1995 Jun;24(3):552-8. doi: 10.1093/ije/24.3.552.
The effect of using multiple causes of death, i.e. all causes of death mentioned on the death certificates, compared to using the underlying cause of death only for analyses of regional variation among small areas in mortality for asthma, diabetes, hypertensive disease and cerebrovascular disease was studied.
Standardized mortality ratios (SMR) were analysed for the different health administrative areas in Sweden using data from the Swedish cause of death register for the period 1987-1991. The SMR were calculated for each of the selected underlying causes of death as well as for multiple causes of death. The correlations between SMR for underlying and multiple causes were analysed.
The highest level of correlation of the SMR between multiple and underlying causes of death was found for cerebrovascular disease (0.96) and the lowest for hypertensive disease in the age group 0-64 years (0.51). For hypertensive disease, diabetes and asthma, when using multiple causes of death some further areas were found to have high SMR and the level of significance was higher. Significantly high SMR using underlying causes of death were, however, not shown to be false when multiple causes of death were used. In the case of cerebrovascular disease little additional information was gained.
By including multiple causes of death in small-area analysis more statistical outliers can be detected and the risk of false warning signals due to random effect can be limited. Analyses of underlying causes of death and multiple causes of death should preferably be combined.
