Mental health spillovers from serious family illness: Doubly robust estimation using EQ-5D-5L population normative data

Abstract

People are interconnected and ill-health is rarely experienced in isolation. However, while there has been extensive research on health spillovers related to informal caregiving, there is comparatively little evidence on how ill-health may impact upon non-caregiving family members. This paper analyses EQ-5D-5L normative data from a nationally representative sample of adult residents of Ireland to estimate the independent relationship between serious family illness and five distinct dimensions of health. The empirical strategy combines inverse probability weighting and multivariate ordered probit regression in a doubly robust estimation. We find that experience of serious family illness is associated with large mental health decrements that are independent of caring responsibilities, while similar results are not evident for the four other health dimensions. Furthermore, stratified sub-sample analyses indicate considerable heterogeneity by sex and by income. In particular, we find evidence consistent with larger mental health spillovers for females than for males, as well as for low- and medium-income households relative to high-income households. The latter suggests that such spillovers may be substantially worse for those with fewer resources. Overall, the findings have a range of potential implications, including for the provision of mental health supports and services, for equity of health outcomes, as well as for health economic evaluation. For example, we calculate that our estimates of health spillovers are consistent with a 1.3% reduction in health utility for non-caregiving family members.

Keywords: Doubly robust estimation; EQ-5D-5L; Health spillovers; Inverse probability weighting; Mental health; Multivariate ordered probit model.