An interesting tension was raised several times at the first Drivers of Health meeting in Princeton. (You can watch the webcast of the meeting here.) On the one hand, there’s a temptation — even a policy need — to separate social determinants and health care.

On the other hand, the two seem tightly intertwined. For instance, access to and adoption of health care treatments varies by socioeconomic status.

“To what extent do we want Medicaid to address social determinants of health?” asked David Cutler, one of our speakers. This is an important policy question. A dollar spent on social needs is a dollar less for traditional health care. For which is the dollar best spent?

This line of thought leads us to a view that SDoH and the health system are at odds. It’s one or the other.

Of course it isn’t, and I don’t just mean that because we can do some of both (because it’s still true that more of one means less of another in a world of finite resources). I mean that in a different sense. They seem inextricably linked. At the meeting, Paula Braveman, our other speaker, expressed skepticism that decompositions of health into areas of SDoH and the health system are meaningful. We don’t know how to separate them empirically any more than we do practically.

She added that we don’t need to degrade medical care when we raise up the importance of social determinants to our health and wellbeing.

You get a different sense of the relative importance of SDoH and the health system depending on what kind of evidence you examine. David Cutler showed that at a point in time, the vast majority of variation in longevity across geographic regions of the U.S. is not due to health care. This suggests social determinants are much more important, and they’re clearly important for addressing inequities across populations at a point in time.

But over time, improvements in health care technology are responsible for about 40 percent of increases in longevity. This suggests that social determinants are still very important, but not as important as cross sectional studies suggest. However, not all populations may experience the same longevity gains. Those with better access to health care will do so first, others later. This is the interaction of SDoH with health care technology.

The gains in longevity are concentrated in a few conditions — cardiovascular disease has been one big winner in improvements due to the health system. Some conditions are more or less prevalent across different socioeconomic groups. Some conditions, again with varying degrees of prevalence across different groups, have benefited very little from the health system, including dementia and pain.

In work Cutler discussed, he said that differences in self-reported knee pain varied by education, but those variations were not associated with detectable physiological differences. So what is the source of pain differences? Do we say that education is the source of the difference? Or some other social factor correlated with education?

It’s just another way in which it seems hard to pull apart medical care and social factors. Epicycles within epicycles.