Disparities

My professional work involves analysis of health disparities, so I tend to consider myself at least somewhat knowledgeable on the literature on disparities, but I was blown away by the news that

Black women who feel they've been victims of racial discrimination are more likely than their peers to develop breast cancer, a large study suggests.

The study, which followed 59,000 African-American women for six years, found that those who reported more incidents of racial bias had a higher risk of breast cancer.

This finding gives a new dimension to the notion of health disparities, which traditionally revolve around differences in health care, access, and outcomes experienced by minorities and underserved populations even after controlling for virtually all imaginable confounders (income, insurance, education, etc.)  Disparities in clinical trials revolve around the notion that minorities and underserved populations tend to be underrepresented in clinical trials even while many of these communities bear disproportionate disease burdens.

The above finding adds a new dimension: women who experience racial discrimination are at higher risk of developing breast cancer.  These findings to me seem to echo Daniels, Kawachi, and Kennedy's argument that justice is good for our health, and more broadly resonate in the subdiscipline of social epidemiology, which posits that social structures have a pronounced effect* on patterns of disease and illness.

The Abstract of the study, which was published in the July 1 version (vol. 166, no. 1) of the American Journal of Epidemiology, is as follows:

Perceived discrimination may contribute to somatic disease. The association between perceived discrimination and breast incidence was assessed in the Black Women's Health Study. In 1997, participants completed questions on perceived discrimination in two domains: "everyday" discrimination (e.g., being treated as dishonest) and major experiences of unfair treatment due to race (job, housing, and police). Cox proportional hazards models were used to estimate incidence rate ratios, controlling for breast cancer risk factors. From 1997 to 2003, 593 incident cases of breast cancer were ascertained. In the total sample, there were weak positive associations between cancer incidence and everyday and major discrimination. These associations were stronger among the younger women. Among women aged less than 50 years, those who reported frequent everyday discrimination were at higher risk than were women who reported infrequent experiences. In addition, the incidence rate ratio was 1.32 (95% confidence interval: 1.03, 1.70) for those who reported discrimination on the job and 1.48 (95% confidence interval: 1.01, 2.16) for those who reported discrimination in all three situations—housing, job, and police—relative to those who reported none. These findings suggest that perceived experiences of racism are associated with increased incidence of breast cancer among US Black women, particularly younger women. 

*Caution: Correlation is not causation.  Nevertheless, the findings of this study are certainly consistent with a vastly more complicated, multifactorial view of disease causality.  More on this in subsequent posts, as it is a current interest of mine (though I have been interested in philosophical, scientific, and medical causation for years).

Comments

[kevin whited]

The above finding adds a new dimension: women who experience racial discrimination are at higher risk of developing breast cancer.

Women who experience racial discrimination, or who feel they are discriminated against?

They can be two different things, yet you seem to use the terms interchangeably.

So many of these sorts of studies find correlation, yet never seem to construct an explanatory model that describes causal linkages satisfactorily (in a way that produces any decent sort of R-squared). Academics have been known to spend lots of time/money finding all sorts of interesting correlations in natural and social phenomena, but it's much more interesting when there is a robust explanatory model involved.

[Daniel Goldberg]

Hey Kevin,

Women who experience racial discrimination, or who feel they are discriminated against?

The latter.

They can be two different things, yet you seem to use the terms interchangeably.

Hmmm. Well, whether they are two different things is certainly a debatable metaphysical proposition. Assuming that there is a meaningful distinction to be drawn between instances of racial discrimination and merely perceived instances of racial distinction -- an assumption I would very much wish to problematize -- I'm not 100% sure how much impact that would have on the study itself.

This is because, as medical anthropologists indicate, belief is one of the most powerful forces in illness experiences. Belief in spiritual or religious precepts, for example, is strongly correlated with all sorts of different positive health outcomes. Belief in therapeutic efficacy -- what is commonly termed the placebo effect -- also shows strong correlations with certain kinds of positive health care experiences (analgesia, for one).

So, let's say that the subjects of the study have not in fact been discriminated against, but they fervently believe that they have. If that fervent belief is strongly correlated with subsequent disease, I'd still suggest that finding is at least worth further investigation (if nothing else to determine whether we can get at a more "robust explanatory model," as you suggest).

So many of these sorts of studies find correlation, yet never seem to construct an explanatory model that describes causal linkages satisfactorily (in a way that produces any decent sort of R-squared).

Actually, I agree completely. I mentioned that I am quite interested in disease causality, and the correlation-causation problem is, I think, extremely important when talking about disease. This is why I noted exactly this point in the Caution above.

I also think -- and others have made this point as well -- that we tend to lionize reductionist, linear models of disease causality in lieu of more complicated, multifactorial causal attributions, the latter of which are typical of nonlinear dynamical systems like, say, population health.