Date of Award

January 2014

Document Type


Degree Name

Medical Doctor (MD)



First Advisor

Harlan Krumholz

Subject Area(s)

Medicine, Public health, Epidemiology


Socioeconomic disparities in health are well-documented, but the precise reasons for these disparities are poorly understood. Traditional explanations for health disparities focus on the influence of person-level disadvantages, such as those related to income, education, and health insurance status. However, the contribution of place to these health disparities is increasingly appreciated. By "place", we refer to the sum of the environmental and community-level factors that may contribute to health, as distinguished from person-level disadvantages associated with low socioeconomic status (SES).

These contextual factors can be categorized as 1) characteristics of the built environment (e.g. access to healthy food, space for exercise); 2) characteristics of the social environment (e.g. community norms regarding smoking, obesity, or treatment-seeking); and 3) direct psychosocial and physical stressors (e.g. pollution, crime). Such contextual factors may act independently of SES, they may mediate the effects of SES, and they may exhibit other complex relationships with SES and related factors such as income and education.

Contextual factors may be modifiable, and are thus an important potential target for health policy and interventions aimed at reducing health disparities. To inform the design of such interventions, is it critical to demonstrate definitively that place is important, to identify which specific contextual factors matter, and to understand the mechanisms by which they affect health. There is currently little evidence in each of these areas.

Isolating the contribution of any single environmental factor to health disparities is challenging. We have begun with a more fundamental question - to what extent does place matter in life expectancy and cardiovascular outcomes? Specifically, we sought to 1) characterize the contribution of place to life expectancy, by identifying geographic disparities in life expectancy that persist after adjusting for individual SES and race; 2) measure the effect of place - using neighborhood-level SES as a proxy - on outcomes after acute myocardial infarction (AMI); and 3) measure the relative effects of individual SES and place on delays to seeking treatment for AMI.

A proposed conceptual model for these relationships is given below. We hypothesize that SES and place exhibit complex interactions. For example, the effect of a person's SES on health may be modified by the context in which they live. Conversely, a person's SES may modify any effects of place on health. There are a wide range of potential mechanisms for the effects of SES and place on health, which we summarize as being related to 1) direct environmental exposures (e.g. pollution, crime); 2) healthcare access and quality; and 3) health-related behaviors. Together, these factors would mediate the effect of SES and place on health outcomes and, subsequently, life expectancy.

To thoroughly test this model would require a complete characterization of contextual exposures, which is outside the scope of this study. We instead limit our focus to shaded items in the diagram above. We first establish the plausibility of such a model by characterizing the association of SES and other contextual factors (i.e. place) with life expectancy. We then use acute myocardial infarction (AMI) as a model condition for measuring the contribution of SES and place to health-related behaviors and health outcomes. Our health-related behavior of interest is delay to hospital presentation in the setting of AMI. Our outcomes of interest include mortality, rehospitalization, and angina symptoms after AMI.

In summary, our study aims include the following: 1) To characterize the contribution of place to life expectancy, by identifying geographic disparities in life expectancy that persist after adjusting for individual SES and race 2) To measure the effect of place - using neighborhood SES as a proxy - on outcomes after AMI 3) To measure the relative effects of individual SES and place on delays to seeking treatment for AMI.

Aim 1: Characterizing the contribution of place to life expectancy

We began at the macro-level, with the objective of understanding the reasons behind variation in life expectancy across the United States. This variation has been previously described, and previous studies have shown that SES is significantly associated with life expectancy at a regional level. Ours is the first study to quantify the degree to which differences in SES can account for geographic disparities in longevity nationwide. Using county as the unit of analysis, our results show that we find that SES does explain many of the striking geographic differences in life expectancy in the United States. This is consistent with traditional conceptions of health disparities as being primarily driven by socioeconomic factors.

Yet despite the prevailing influence of SES, our results also reveal significant exceptions in which regional variation persists, or increases, after adjusting for SES. In particular, we identify several comparisons of areas which are virtually identical in terms of racial and SES composition, yet differ dramatically in terms of life expectancy. Based on such comparisons, along with the observation that disparities in life expectancy persist after controlling for SES, we conclude that contextual factors are important contributors to health disparities.

In a secondary analysis, we applied the concept of "deviance" to identify places in which life expectancy is significantly higher or lower than what would be expected based on the race and SES composition of the population. Our results identified counties of significant positive and negative deviance. We conclude that the existence of these positive deviance areas - many of which have low SES, high minority populations - demonstrates that the disadvantages of SES are not insurmountable with respect to health outcomes. We further conclude that in-depth investigation of these positive deviance areas - and comparison with negative deviance areas of similar (race and SES) composition - may reveal characteristics of their environments that drive health disparities.

Aim 2: Measuring the effect of place on outcomes after AMI

Our second objective was to measure the independent contribution of place to outcomes after acute myocardial infarction (AMI). We employed neighborhood SES in an individual's area of residence as a proxy for place. Neighborhood SES was measured as a composite of median household income and five other factors related to wealth, education, and occupation. Neighborhood SES has been previously linked with a range of cardiovascular risk factors and outcomes, including incidence of coronary heart disease. Importantly, these associations have been shown to persist after simultaneously adjusting for person-level SES (e.g. personal income, education, insurance status, and/or occupation). This suggests that neighborhood SES is not merely a proxy for individual-level SES, and that where one lives has an effect on cardiovascular health beyond that of one's own resources.

Using this framework, we performed an analysis among patients in the nationwide PREMIER registry, which includes 2321 patients with AMI from 19 US hospitals. Our results show that neighborhood SES is independently associated with the prevalence of angina and risk of rehospitalization in the 12 months after AMI. This association persists after accounting for individual SES variables, again demonstrating that context matters independent of a patient's personal socioeconomic circumstances. The magnitude of this association is comparable to that of individual SES with outcomes. From this we conclude that context may be as important as personal resources in driving health disparities.

Aim 3: Measuring the relative effects of individual SES and place on prehospital delays in AMI

Having demonstrated an influence of neighborhood context on outcomes, there is a need for further studies to identify mechanisms underlying the effect of neighborhood on health. Such mechanisms may represent targets for public policy and interventions to reduce health disparities. These are summarized in the conceptual model above as involving 1) poorer health-related behaviors; 2) poorer healthcare access; or 3) the direct influence of psychosocial and environmental stressors on health. We focus on the first category, noting that features of both the physical environment (e.g. proximity to healthy food sources and space for exercise) and social environment (e.g. local norms and attitudes toward healthcare) may have a significant impact on health-related behaviors such as smoking, obesity, physical activity, and treatment-seeking, all of which could explain our above findings related to AMI outcomes.

Specifically, our objective is to investigate whether place - again using neighborhood SES as a proxy - is related to delays in seeking treatment ("prehospital delays") for AMI among patients in VIRGO, a nationwide AMI registry. Longer prehospital delays are associated with delayed revascularization, and thus can contribute to worse outcomes after AMI. Moreover, prehospital delays in AMI can be viewed as a marker for an individual's propensity to seek medical care when in need, and as such serve as a proxy for overall healthcare-seeking behavior.

VIRGO includes a high proportion of young and female patients with significant racial diversity. This diversity allows us to further investigate possible differences in the importance of neighborhood and person-level SES by race and sex. Our results show that both low neighborhood SES and low individual income are independently associated with delays of greater than 2 hours. Based on this observed relationship between neighborhood SES and delays, we conclude that context affects treatment-seeking in the setting of AMI. This association could in part mediate our observed effect of neighborhood on outcomes after AMI. Moreover, we find differential effects of SES variables according to race. Specifically, for black patients only individual-level SES, and not neighborhood SES, is a significant predictor of delays. Conversely, only neighborhood SES is a significant predictor of delays among non- blacks. From these observations, we conclude that different demographics have varying sensitivity to the influence of place on health-related behaviors