On Insurance and Health

We have noted on this blog the relatively weak evidence for the proposition that lack of insurance coverage and access to care has a pronounced effect on health.  For understandable reasons, many assume that a strong relationship exists between the two variables, but the evidence for this assumption is lacking.  What's more, given the Whitehall studies, there is good evidence that pronounced differences in health outcomes persist even when the population in question enjoys coverage to basic care. 

I have argued that the case for increasing access is therefore on sounder footing if made on ethical grounds than on policy grounds that doing so will substantially improve population health.  In addition, I continue to believe that we would do well to direct some of the energy and focus we spend on the access debate to agitating for a reallocation of health care dollars to public health and primary prevention, both of which are on much firmer empirical footing in terms of their likely effects on improving population health.

All of this is context for a study released yesterday in CA: A Cancer Journal of Clinicians (an American Cancer Society journal) entitled Association of Insurance with Cancer Care Utilization and Outcomes.  The study produced statistically significant survival differences based on insurance.  According to the ACS press release, the study found that

About 89% of privately insured white women with breast cancer survived at least 5 years, compared to 76% of white women with Medicaid or no insurance. Among African-American women, 81% of breast cancer patients with private insurance survived 5 years, compared to 65% of those on Medicaid and 63% of those without insurance. A similar pattern emerged in colorectal cancer. Among white patients with private insurance, 66% survived 5 years, compared to 50% of those with no insurance and 46% of those on Medicaid. Among African Americans, 60% with private insurance survived 5 years compared to 41% of the uninsured and Medicaid patients.

This study suggests, of course, that lack of coverage is relevant to health outcomes as to cancer.  Several points are worth noting here: first, given that all manner of primary prevention interventions are in fact contingent on coverage, one should be careful about drawing conclusions from this study as to the overall effect of acute care on population health.  The study notes this, of course, reporting that the uninsured subjects had significantly lower access to cancer prevention programs and education, as well as lower access to early detection and screening programs.

Second, only about 4% of U.S. cancer patients are uninsured.  Accordingly, it is fair to wonder whether, even if insurance status is strongly linked with cancer outcomes, increasing access will have a large impact across the entire cancer population, let alone the U.S. population.  Third, finding a significant positive correlation between insurance status and health outcomes should not be taken as proof that insurance status is a primary factor.  The evidence has generally been lacking that it is a significant factor at all, and this study does not evaluate the relative contribution insurance status makes to outcomes as compared with other factors (i.e., health literacy, relative socioeconomic disparities, early child development, etc.)

None of these cautions detract from the study, which seems to me to have been thorough and well-done, and which certainly provides evidence that there may be a stronger link between insurance and health outcomes than prior work has demonstrated.

Comments

[Frank]

I'm glad you're keeping track of these studies. I'm reminded of a recent study I saw that said that "inflammation" is a leading (or ultimate?) cause of death. But when pressed, the doctors behind this idea said, well, it's both cause and effect of disease.

Here, Margalit and Wolff's book "Disadvantage" may be of interest. They talk about the clustering of disadvantage. So it may well be very hard to figure out if, say, low SES is causing a) bad health or b) lack of insurance or c) both. . . . and if c), then if a) is largely due to b) or due to some independent force of low SES (say, driving up cortisol levels).

[Daniel Goldberg]

Hey Frank,

All good points, and many of the chapters in Marmot & Wilkinson's Social Determinants of Health (2d ed.) adduce similar points. Some of the sociologists' work on social conditions as fundamental causes of disease may be relevant here (the seminal article is Link & Phelan's 1995 piece, and Lutfey & Freese followed up in the last few years).

I'm actually only currently making my way through this literature. Regardless, I've never objected per se to the claim that lack of access is correlated with poor health. What bothers me is (a) that this relationship is widely taken for granted instead of established by solid evidence; and (b) that there is already excellent evidence that certain social, public health, and primary prevention interventions have a robust relationship with population health. As such, in thinking of the notions of relative priority that policy analysis demands, IMO, there is a case to be made for moving the latter much higher up the scale than it currently is.