Socioeconomic Disparities and Health Outcomes

To the Editor:

The ability to link data from sources such as the U.S. Census is now enabling researchers to direct their focus toward reporting neighborhood and contextual characteristics that increase the risk for adverse health outcomes and are independent of patient-level attributes. This is all the more important because disparities in health outcomes likely arise as a result of both individual exposures and contextual factors (1). Research regarding disparities have until recently been challenging because of the high response bias associated with collecting individual-level socioeconomic measures (2). However, area-based measures from the U.S. Census’s American Community Survey and the National Center for Health Statistics Urban-Rural Classification Scheme can be used to gain insight into the role of area-based measures as independent risk factors for diseases, as demonstrated in the work by Raju and colleagues (3). Understanding area-based risk factors could help researchers design, target, monitor, and assess public health programs, including prevention interventions.

First, some limitations of Raju and colleagues’ analysis need to be emphasized. Although the authors used census tract–based determinants as area-based measures, it is important to acknowledge the possibility of ecological fallacy, and that these determinants provide information regarding the neighborhood that is not reducible to the individual level (4). Although the authors have defined neighborhoods as census tracts, nearby neighborhoods may also influence health outcomes and disparities.

Second, data structures arising from both individual and neighborhood levels are inherently hierarchical and correlated. To account for geographical correlation, it is important to analyze such data using multilevel models. Multilevel models can account for a lack of independence, evaluate multivariate associations, incorporate covariates at both individual and geographic levels, and model interactions between variables (5). Multilevel models have been used to evaluate health disparities and to describe the relationship between geographic exposures for a wide variety of health outcomes. They can also help researchers quantify the proportion of variability associated with being in a specific neighborhood.

The authors are to be applauded for taking the research on chronic obstructive pulmonary disease risk factors a step further by investigating area-based measures. Although the availability of public datasets such as that provided by the U.S. Census make such investigations possible, it must be emphasized that their implementation is challenging. Several resources, such as the Public Health Disparities Geocoding Project (6), are available to provide guidance on techniques to conduct research on socioeconomic gradients in health.