Emelie Thern

Introduction The social gradient in consumption behaviours has been suggested to partly explain health inequalities. The majority of previous studies have only included baseline measurements and not considered potential changes in behaviours over time. The study aimed to investigate the contribution of alcohol consumption and smoking to the social gradient in mortality and to assess whether the use of repeated measurements results in larger attenuations of the main association compared to using single baseline assessments. Methods Longitudinal survey data from the population‐based Stockholm Public Health Cohort from 2006 to 2014 was linked to register data on mortality until 2018 for 13,688 individuals and analysed through Cox regression. Results Low socioeconomic position (SEP) was associated with increased mortality compared with high SEP; hazard ratios 1.56 (95% CI 1.30–1.88) for occupational status and 1.77 (95% CI 1.49–2.11) for education, after adjustment for demographic characteristics. Using repeated measurements, alcohol consumption and smoking explained 44% of the association between occupational status and all‐cause mortality. Comparing repeated and baseline measures, the percentage attenuation due to alcohol consumption increased from 11% to 18%, whereas it remained similar for smoking (25–23%). Discussion and Conclusions Smoking and alcohol consumption explained a large part of the association between SEP and mortality. Comparing results from time‐fixed and time‐varying models, there was an increase in overall percentage attenuation that was mainly due to the increased proportion explained by alcohol consumption. Repeated measurements provide a better estimation of the contribution of alcohol consumption, but not smoking, for the association between SEP and mortality.