Simultaneous modelling of time trends and regional variation in mortality rates.

Abstract

We seek to model the regional component of the variance in the mortality rates in the UK and to ascertain if there is evidence that this regional variance is increasing in recent periods. Age Period Cohort (APC) models, based on the local 'curvatures', are used in each region to describe the changes in the trends in the mortality rates. This is extended to a multilevel model to estimate the regional component of the variance in the rates and to estimate the effect of regional differences in the trends in the rates. We show how the use of a multilevel APC model can help to distinguish the cohort and period trends in the mortality rates from the cohort and period effects on the regional variance in these rates. For both sexes, but particularly for females, a reduction in the rate of decrease in mortality was found around 1960. In addition, particularly for females, cohorts born after 1930 appear to show reductions in mortality at an increased rate. It is demonstrated that there is evidence that the between-region variation in the rates has not remained constant and that it is much less now than it was at the beginning of the data series. Further, there is evidence that the trends in the rates are not the same in all regions and that while there is a convergence of the rates in many regions, Scotland, in particular, stands out as a region which contributes most to the regional variation in mortality rates. Evidence of regional variation in mortality rates has been found with a suggestion of a decrease over the period of the study though with some stability since 1951.