How to be smart about socioeconomic status in studies

About Joe Rojas-Burke

Joe Rojas-Burke is AHCJ’s core topic leader on the social determinants of health, working to help journalists broaden the frame of health coverage to include factors such as education, income, neighborhood and social network. Send questions or suggestions to joe@healthjournalism.org or @rojasburke.

Image by Jay Reimer via flickr.

Image by Jay Reimer via flickr.

Medical study authors routinely claim to have “controlled” for socioeconomic status.

That kind of sweeping assertion should set off alarm bells. The authors probably haven’t come close to fully accounting for something as difficult to measure as a person’s place in the hierarchy of self-determination and power, neighborhood quality, working conditions, job security, income and wealth.

To assume otherwise is a mistake that can lead to misleading conclusions.

Consider, for example, a recent study in the journal Nature Medicine describing a genetic variation that might account for lower heart disease survival among African Americans. News coverage of the study caught my attention because whatever role genetics plays in the black/white disparity in heart disease, it’s probably small.

Some researchers have concluded that socioeconomic disadvantage is the most significant root of the problem, not genetic differences. And there is pretty good evidence that the traditional risk factors (diabetes, high blood pressure, lack of physical activity, obesity, smoking) account for all of the difference in heart disease mortality between black and white men in the United States, and most of the difference between black and white women.

The authors of the Nature Medicine paper offered a different view that played up the importance of a single gene involved in blood clotting. “Race is an independent predictor of survival in coronary heart disease even when demographic, socioeconomic and clinical factors are considered,” they asserted, before concluding that there must be other factors, such as undiscovered genetic variations, driving the racial disparity.

Here’s the problem. When attempting to adjust for socioeconomic status, researchers typically rely on proxies such as years of education or household income. It’s convenient to do so, and sometimes no better data are available. But it’s not the same as “controlling” for all socioeconomic differences.

The Nature authors relied on a study that ranked people’s socioeconomic position by their neighborhood’s home values and median household income. It ignored a lot of potentially very important factors driving worse outcomes among African Americans, such as rates of medication use and access to medical providers and heart procedures. There is evidence that African Americans don’t have equal access to medical and surgical care, and are less likely to start or stay on prescribed cardiac medications. It’s not clear that income and home prices capture the impact of segregation, that is, the concentration of poverty, social adversity and lower quality schools where many African Americans live.

There are probably quite a few unmeasured socioeconomic differences driving racial and ethnic health disparities, according to Paula Braveman, M.D., M.P.H., a professor at the University of California, San Francisco School of Medicine. In an informative review article, Braveman and colleagues make a trenchant point:

“Rather than claiming to have ‘controlled for SES,’ researchers should acknowledge the potentially relevant aspects of SES that could not be measured and explicitly consider the implications of unmeasured socioeconomic influences when interpreting findings.”

3 thoughts on “How to be smart about socioeconomic status in studies

  1. Robert C. Bowman, M.D.

    This is incredibly on target. An example are studies finding adverse outcomes in African Americans regarding cancer detection or treatment. The cancer databases have race, gender, and location but not specific socioeconomic markers. So the geographic location is turned into county income or other markers. The only truly research subject specific data is race/ethnicity. Differences in income are not really being considered. The marker “African American” takes on the missing social determinant factors that were not collected. This does not mean that the studies are incorrect or correct. The methods used cloud any differences or lack of differences.

    Studies that attempt to claim better outcomes from primary care, from teaching hours restrictions upon residents in training, by pay for performance, by hospital type, or by type of clinician are seriously flawed due to missing socioeconomic variables – not to mention apples to oranges comparison errors. Rural Critical Access hospitals with entirely different patient populations to serve, entirely different patients retained to be at the hospital (compared to those transported), entirely different personnel, and entirely different and inadequate funding are compared to hospitals that are different in these ways and more.

    The NP vs MD study in JAMA widely cited by many used different sites, the NP site was changed during the data collection, the patients were different, and the study showed no difference – as you would expect from care of the same types of patients.

    Hong et al in a JAMA article attempted to find a way to adjust SES so that clinicians serving the underserved were not punished by Pay for Performance. They failed to do so. This was one of the best studies attempting sociodemographic markers and illustrated the expected failure. Most try with more limited resources and data. JAMA has had 1 or 2 of the best and some of the worst studies – entirely due to failure to understand psycho social and situational determinants of health.

    As Deming noted, quality lies in the matrix of relationships. Attempting to do studies in rural health, inner city, or other places with lower concentrations of clinicians are fraught with error. Generally the same determinants that shape lower concentrations of clinicians shape lesser health outcomes. This is why 3% of urban hospitals have lost 1 – 2 percent of Medicare revenues this year or the top penalties compared to 9% of rural hospitals and over 14% of rural hospitals where workforce is more limited.

    Experimentation upon vulnerable populations has actually increased – with continued poor results involving innovative payment systems that generally result in even less payment where payment is least. Increasing cost of delivery forced upon vulnerable providers is another problem and as Commonwealth noted, the funding is limited to offset the extra costs. In general those less organized do not know how to manipulate the system to offset costs or to design more dollars their way.

    This overall result in turn damages local social determinants and speculatively would result in lesser health status and outcomes. When your track Medicare payment and consider centralization of workforce, training, ancillary services, and more – it is possible to understand how the health care designs for payment and for training result in continued and widening inequities.

  2. Joe Rojas-Burke

    Hear, Hear! You’ve raised tons of ideas for health reporters to pursue, Robert. For instance, how hospitals serving disadvantaged populations may be harmed by new payment models.

  3. Pingback: Weekend reads: How to rescue science, what “censorship” really means, worst paper of the year? | Retraction Watch

Leave a Reply