Fig. 2
Contribution of health behaviours and body mass index, used as time-dependent covariates, in explaining the association between socioeconomic status and all-cause and cause-specific mortality, the CHAMP study. Abbreviations: HR, hazard ratio; SHR, sub-hazard ratio; PA, physical activity; BMI, body mass index. N = 1527. For all-cause mortality, repeated assessment of health-related behaviours, and body mass index at baseline, first, second, and third follow-up were entered into the model as time-varying covariates; for cause-specific mortality repeated assessment at baseline, first, and second follow-up were entered into the model. The study population was divided into three groups by teritles (tertile1: lowest disadvantages-tertile3: highest disadvantages). a We used calendar year as the time scale, with survivors having a censoring date of 31 December 2017 (person-years follow-up = 13,814). b We used calendar year as the time scale, with survivors having a censoring date of 31 December 2015 (person-years follow-up = 12,180). c Hazard ratio and sub-hazard ratio for lowest versus highest tertile (least disadvantages; reference group) of cumulative socioeconomic status. d Adjusted for age, country of birth, and living arrangement. Percent attenuation =100 × (βModel1 − βModel1 + health behaviour(s))/ (βModel1), where β = log(Hazard ratio)