post by Paul Kelleher
Last week I concluded a discussion of an important recent article on health gradients with this claim:
Since not all [health] gradients are created in the same way, we cannot assume that they will be eliminated in the same way.
This conclusion stemmed from the article's finding that a social gradient in risky health behaviors largely accounts for the social gradient in health in the British Whitehall II study, but not in the French GAZEL study. Comparing these two studies shows that one should be sensitive to the different ways that different gradients may arise. And because of their different causes, what flattens one gradient may not flatten another.
It does not follow from this conclusion, however, that if two gradients have different causes, then they cannot be flattened in the same way. Unfortunately, the authors commit precisely this fallacy:
Our results suggest that policies that target unhealthy behaviours in the socially disadvantaged groups are likely to lead to decreases in social inequalities in health in the Whitehall II cohort but not in the GAZEL cohort, as health behaviours are not major mediators of the SES-health association in GAZEL.
Here the authors assume that a gradient in health can be flattened only by eliminating its mediating cause. Since only the Whitehall gradient has its origins in a health behavior gradient, only the Whitehall gradient can be addressed by addressing health behaviors. This way of thinking probably results from what we might call a "gradient-elimination fixation": if our end goal is to eliminate the gradient in overall health, then we need to find the gradient that accounts for it and flatten that. But there is more than one way to eliminate a gradient. A second way is to create a new, reverse gradient in one of the indicators that partially determines overall health. Following this path would achieve what the authors of the above quotation say can't be done. For example, in GAZEL, those at the bottom of the overall gradient could be helped to improve their health behaviors while those at the top are left alone. This in theory could lead to an improvement at the bottom that offsets whatever mechanism led to the original gradient in overall health outcomes.
Thus, while it is right to stress the importance of understanding the mechanisms behind gradients, there may be things we can do to flatten gradients even if we have no idea how they arose. In the case of GAZEL, we can create a new reverse gradient in health behaviors where no gradient in them had existed. Whether this is wise and/or permissible is a normative question that this post doesn't address.