3.4 Predictors of caregiver quality of life
As shown in Table 3, a three-step hierarchical linear regression model was conducted to assess the ability of AD severity, illness perceptions and caregiver SE to predict caregiver QoL. Only IPQ-R subscales that significantly correlated with DFIQ scores were included in the model. No demographic variables significantly correlated with QoL so were excluded.
Model 1, which included AD severity as the sole predictor, explained 34.6% of the variance in caregiver QoL. Model 2 explained 71.4% of the variance in caregiver QoL and included AD severity which accounted for 34.6% of the variance, and illness perceptions which accounted for 36.8%. Significant illness perception predictors were emotional representations, consequences, and identity. Model 3 explained 72.2% of variance of caregiver QoL. Adding caregiver SE to model 3 contributed 0.9% to the variance explained and did not result in a significant difference. Additionally, caregiver SE was not a significant predictor. The strongest predictors were consequences (β=0.62) emotional representations (β=0.49), identity (β=0.31) and AD severity (β=0.29) meaning that for every 1-unit increase in consequences, emotional representations, identity and AD severity scores there were 0.62, 0.49, 0.31 and 0.29 increases in family impact scores respectively.
3.5 The mediating role of caregiver self-efficacy between illness perceptions and quality of life
To examine whether caregiver SE mediates the relationship between illness perceptions and caregiver QoL, 7 separate models using different illness perceptions as a predictor in each model were produced. For each model, simple linear regressions were run for pathways a, b and c (see supplement table d). Path a is the direct effect of the predictor (illness perception) on the mediator (caregiver SE). Path b is the direct effect of the mediator (caregiver SE) on the outcome (caregiver QoL). Path c is the direct effect of the predictor (illness perception) on the outcome (caregiver QoL). Path c’ is the direct effect between the illness perception and caregiver QoL when the mediator is taken into account. This pathway and the indirect/mediating effect were produced in the PROCESS macro in SPSS.
Mediation analysis confirmed the direction of the regression paths and found that caregiver SE had a statistically significant mediating effect between all illness perceptions and QoL, except emotional representations (see Table 4). The relationships between timeline acute/chronic, illness coherence and identity and QoL were no longer significant when the indirect effect via SE was accounted for, indicating a full mediation effect. This implies that the effect of these illness perceptions on QoL within those models was transmitted entirely through SE. SE partially mediated between consequences, personal control and treatment control and QoL. SE did not mediate between emotional representations and QoL.