Socioeconomic inequalities in smoking-attributable mortality in Europe; understanding trends 2000-2020 23 2 also known as temporary life expectancies or restricted mean survival times,22, 23 were estimated from all-cause mortality life tables for each year. Second, to estimate the smoking-attributable fraction (SAF) we used the PrestonGlei-Wilmoth (PGW) indirect estimation method.24 In short, the PGW method uses a regression model, run on 21 high-income countries over 1950-2003, to estimate the statistical relationship between lung cancer mortality and mortality from other causes of death, accounting for age, calendar year, country, and interactions between mortality and calendar year and mortality and age. The coefficients from this model,24 when combined with the assumption that mortality from lung cancer among non-smokers resembles age and sex estimates from the Cancer Prevention Study II (CPS-II),18 are used to estimate the overall smoking-attributable fraction (SAF) of deaths. SAF was estimated for each year, sex, age and educational group in each country. Age-specific SAF was then combined with all-cause mortality to estimate its impact on life expectancy. Finally, SAF was standardized using the European Standard Population (2013) to allow for comparison between countries.25 Third, cause-deleted life tables that use death rates where the risk of dying from a specified cause is eliminated, in this case smoking-related causes, were calculated to estimate the partial life expectancy loss attributable to smoking between age 50 and 80.26, 27 Finally, for each country and sex, the changes in partial life expectancy (age 50 to 80) by education between 2000 and 2020, and the absolute educational differences in life expectancy between low and high educated in each year, were decomposed into the age-specific contributions in years attributable and nonattributable to smoking, using the continuous change model developed by Horiuchi et al.28 All analyses were done using R (version 4.2.2) and the DemoDecomp package for the decompositions.29 An extensive guide on how to use the Horiuchi model in R is available.30 Given the differences in mortality patterns between men and women, results are presented separately by sex. Results Over time large differences between countries and educational groups were observed in smoking attributable mortality, as seen in Figure 1. For women, SAF was higher for lower educated in all countries, and stable or increasing in many countries regardless of educational level. The exceptions were Denmark and Sweden, who have both experienced a decrease in SAF since mid-2010. Among men SAF declined in all countries and among all educational groups, and SAF was highest in the lower educational group for all countries (Appendix II).
RkJQdWJsaXNoZXIy MTk4NDMw