Disability Mediates the Impact of Common Conditions on Perceived Health
Figure 1
General mediation model used in analyses.
The figure displays the general mediation model that has been used to estimate effects according to path-diagrammatic conventions. Squares represent variables. Di is one of the p = 19 disorders under consideration, Mj is one of the k = 8 mediating variables (disability dimensions), and VAS is the final outcome. Arrows represent regression slope parameters from independent variables to outcomes. The δ parameters stand for the direct effect regression from disorders to the final outcome. The ι parameters indicate the two regression components of the disorder indirect effects as mediated by M: a) p x k regression parameters from D to M (ιDij) and b) k regression parameters from M to VAS (ιMj). For each disorder the model can be decomposed in two paths: 1) VAS regressed on disorders, and 2) a causal mediation chain of VAS regressed on mediators which in turn are regressed on disorders. The partial indirect effect of a certain disorder Di through a mediator Mj is ιDij x ιMj, whereas its total indirect effect is the sum of the k products across mediators (). Total effects for a disorder are the sum of direct and total indirect effect (δi+Ιi). Directionality cannot be assumed as a causal association in our study due to its cross-sectional, observational nature. Also notice that in the general model, the effect of each disorder on each mediator is adjusted by the direct effect of the remaining disorders (thus controlling for comorbidity), while the impact of a disorder on VAS is controlled by the total effects of the other disorders. Disability is thus fully taken into account, even though it is decomposed in subscales. The effects on VAS are also controlled for age, gender, employment status and country.