Estimation of Smoking Related Deaths: an Age-Period-Cohort Model based Approach
presentation
posted on 2023-06-07, 20:33authored byMike Murphy, Mariachiara Di Cesare
Tobacco is the second most important risk factor for mortality in Europe. The impact of smoking on the burden of disease varies among countries and sexes depending on differences among cohorts in lifetime exposure to smoking. Although it is widely recognised that many of the smoking related causes of death have a strong cohort effect, especially in the case of lung cancer, most of the estimates are period based. This paper presents cohort-based estimates for lung cancer mortality (using data from the WHO Mortality database) for a range of European countries that include a variety of levels and trends in smoking-related causes of death. We use an Age-Period-Cohort (APC) model-based approach. The APC model for the log-death rates m(apc), at age a, in period p for persons in birth cohort c=p-a, is: ln(m(apc)) = f(a) + g(p) +h(c). We argue that after attributing maximum variation to the cohort dimension, period patterns are relatively unimportant and therefore a simplified model may be used. We use the cohort function for lung cancer as an indicator of total smoking attributable mortality to detect cohort and gender differences across countries. The pace of increase and decease in lung cancer is related to the pace of cohorts uptake and reductions in smoking rather than levels per se, therefore we discuss estimation and use the first derivative of the h(c) cohort function to identify the most favoured cohorts.