Abstract
Background
This thesis investigates the relationship between income and health in New Zealand adults, using longitudinal data from the household panel Survey of Family, Income and Employment, 2002 to 2005. Evidence for a cross-sectional relationship between income and health is strong, but may be affected by confounding. Longitudinal data, with repeated measures on the same individuals, can be analysed to control for much of the confounding that biases cross-sectional analyses, giving a more accurate estimate of the causal impact of income on health.
Methods and Results
The outcome used was self-rated health (SRH), a five-level ordinal variable. The primary analytical method used was fixed effects regression modelling, which controls for time-invariant confounders, but applying this method to an ordinal outcome was challenging. Two fixed effects ordinal models were compared. These found that an increase in income of $10,000 increased the odds of better SRH by 0.6% (95% confidence interval -0.3% to 1.5%; ‘hybrid’ proportional odds model) to 0.9% (95% confidence interval -0.4% to 2.3%; amalgamated conditional logistic regression model). Random effects proportional odds models, which did not adequately control for bias from unmeasured time-invariant confounders, gave larger and statistically significant income estimates.
Conclusions
A modest (and statistically insignificant) association of changing income with changing SRH in New Zealand adults was found over the short-term, having controlled rigorously for confounding. The small size of the income estimate was consistent with results from other longitudinal studies. However, this thesis (and similar research) may have underestimated the putative causal association between income and health due to measurement error in income and changes in income over time. Likewise, SRH, as a general health measure, prone to ceiling effects and measurement error, may not be an ideal repeated measures outcome. Changes in SRH over time occur in response to changes in both mental and physical health and short term income changes may have different effects on mental and physical health outcomes. Long term income rather than short term income changes may be more relevant for health and time lags longer than one year between changes in income and effects on health were not accounted for.