socioeconomic inequalities, diet cost and diet quality

The current study aims to investigate the mediating role of dietary costs in the association between educational level and diet quality in a Dutch study population. This will be investigated using the macro Process by Hayes.

We hypothesize that socioeconomic inequalities in diet quality are partly explained by differential dietary costs.

Age, sex, study center and educational level were measured at baseline (between 1993 and 1997). In 2015, respondents to the 2011 questionnaire on electromagnetic radiation who were still alive, , living in the Netherlands and who gave informed consent (n=13,421) were invited to fill out a second food-frequency questionnaire.

The mediating role of dietary cost in the association between (household) educational level and adherence to the DHD15-index and DASH diet was assessed using multiple linear regression analyses. This was done using the macro PROCESS developed by Hayes. Firstly, the total effect was estimated by assessing the associations between the independent variables educational level and household educational level and the dependent variables adherence to the DHD15-index and the DASH diet (c-path) separately. Secondly, the association between the independent variables and the potential mediator dietary cost (a-path) was assessed. Thirdly, the associations between the potential mediator dietary cost and the dependent variables, adjusted for the independent variables (b-path), were assessed separately. Lastly, the direct effect was estimated by assessing the associations between the independent variables and the dependent variables, adjusted for the mediator dietary cost (c’-path), separately for the two dependent variables. In order to assess whether dietary cost indeed mediated the association between SES and adherence to the DHD15-index and the DASH diet, the indirect effect and the corresponding 95% confidence interval were assessed. The indirect effects were calculated by multiplying the regression coefficient of the a-path with the regression coefficient of the b-path (a-path x b-path). The bootstrapped 95% confidence intervals around the indirect effects were based on 1234 seeds and 5,000 bootstrap resamples. Statistical significance of the indirect effects were determined if the upper and lower bound of the bias corrected 95% bootstrap confidence intervals did not contain zero.