Investigating Age-Period-Cohort Effects Using Lexis Diagrams and Intrinsic Estimator Regression Modelling to Understanding Scotland’s ‘Excess Mortality’

Background: Mortality rates are higher in Scotland relative to England and Wales, even after accounting for deprivation. This ‘excess’ mortality is partly due to higher mortality from suicide, alcohol- (ARD) and drug-related deaths (DRD). We investigated whether age, period or cohort effects from exposure to adversity from the 1980s might explain the recent trends in these outcomes in Scotland. Methods: Data on ARD (36,635), suicide (31,061) and DRD (15,427) were obtained from the National Records of Scotland for 1974-2013 and analysed by sex and deprivation using line plots, shaded contour plots, commonly used in demography but limited in epidemiology and health sciences, and intrinsic estimator regression modelling to identify and quantify relative age-period-cohort effects. Findings: Cohort effects were identified for people born between 1960 and 1980 for both DRDs and suicide. The 1960-1980 birth cohort had approximately a 30% higher risk of suicide than other generations, whilst the increase in risk of DRD was at least twice as high. Both cohort effects are largely driven by males and those living in the most deprived areas but the cohort most affected by suicide occurs slightly earlier than for DRDs. The timing of the cohort at highest risk also differs by deprivation for both suicide and DRD, being earlier for males in the most deprived group. In contrast, an age-period effect for working-age adults, particularly in the most deprived areas, helped explain the trends in ARDs. The risk of ARD for these adults increased at least two fold from 1990 to the mid-2000s. Interpretation: These results for DRD and suicide are consistent with the hypothesis that exposure to the changing socioeconomic and political contexts of the 1980s created a delayed negative health impact because the populations most affected are the same as those subsequently having the higher risks of mortality. Limitations include defining death outcomes, changes to coding, the use of year of registration, difficulties in disentangling age-period-cohort effects and only having a partial view of each birth cohort. The use of several methods in triangulation added strength to the findings