Inequality in Infant Mortality: Cross-State Variation and Medical System Institutions

Inequality in Infant Mortality: Cross-State Variation and Medical System Institutions Abstract This article examines variation in the association between maternal education and infant mortality across the 50 U.S. states. An analysis of 22,967,018 Vital Statistics records from 1997-2002, reveals evidence of dramatic cross-state differences. In some states, infants born to mothers with less than 12 years of schooling are more than twice as likely to die as infants of mothers with 4 years of college or more. Other states see far more equality between these groups. I then evaluate two components of state medical systems that are predicted to be associated with this variation in the magnitude of inequalities in infant mortality: neonatal intensive care units and primary care physician supply. More widespread availability of neonatal intensive care is associated with reduced inequalities in infant mortality. In contrast, the supply of primary care is linked to slightly larger differences in infant mortality between mothers with low and high education. infant mortality, inequality, institutions, health disparities, United States The likelihood of infant mortality differs dramatically depending on a mother’s social position. Across sociodemographic indicators, including education, income, and race, infants born to mothers from more advantaged backgrounds experience consistently lower rates of mortality (Finch 2003; Gortmaker 1979; Hummer 1993; Hummer etal. 1999; Mathews, Menacker, and MacDorman 2004; Singh and Kogan 2007). These inequalities in infant mortality have inspired research on the role that socioeconomic resources play in promoting infant health (e.g., Blumenshine etal. 2010; Conley and Bennett 2000; Strully, Rehkopf, and Xuan 2010). This work is consistent with a broader literature on the SES-health gradient and the role of socioeconomic position as a “fundamental cause” of health outcomes (Elo 2009; Link and Phelan 1995). Research in this tradition has been extremely influential in efforts to explain the persistent association between socioeconomic position and health (e.g., Miech etal. 2011; Phelan, Link, and Tehranifar 2010). However, there is also evidence that the magnitude of this association varies substantially across different contexts (Beckfield, Olafsdottir, and Bakhtiari 2013; Chetty etal. 2016; Mackenbach etal. 2008). Moreover, there is growing interest in the institutional predictors of such variation (Beckfield etal. 2015). In the case of the United States, there is evidence of substantial cross-state differences in health disparities between socioeconomic groups (Subramanian, Kawachi, and Kennedy 2001; Xu 2006). So far, most of this work has focused on inequalities in adult health outcomes (cf. Wildeman 2012). Yet, there is reason to expect that socioeconomic inequalities in infant mortality may also vary across states and state institutions. The theory of socioeconomic position as a fundamental cause of disease highlights the fact that interventions designed to improve health outcomes often result in greater health inequalities (Phelan etal. 2010). When utilization of these interventions is not universal, those with more socioeconomic resources are typically better positioned to take advantage of them (Link etal. 1998). This principle has important implications for two state medical system institutions that have been shown to influence infant mortality: neonatal intensive care units (NICUs) and primary care physician supply (Gortmaker and Wise 1997; Shi etal. 2004; Starfield, Shi, and Macinko 2005; Wise 2003). This article contributes to the emerging literature on variation in the magnitude of health disparities and the institutional predictors of this variation by exploring two research questions: (1) To what extent do educational inequalities in infant mortality differ in magnitude across U.S. states? and (2) Is cross-state variation in inequality associated with medical system institutions? To answer these questions, I analyze Vital Statistics data on 22,967,018 births from 1997-2002. Focusing on inequalities in infant mortality between mothers with less than 12 years of education and those with 4 years of college or more,1 1 In addition to aligning with the timing of important educational credentials, these categories highlight mothers in unambiguously different social positions (Goesling 2007). I first assess the extent to which this disparity varies across states and find evidence of substantial differences. I then use random intercept logistic regression models to evaluate how state-level differences in the availability of neonatal intensive care and primary care are associated with variation in the magnitude of these inequalities. BACKGROUND Maternal Education as a Fundamental Cause of Infant Mortality Infant mortality refers to deaths that occur after a live birth and before a child reaches one year of age. Over the past three decades, rates of infant mortality in the United States fell from 10.9 deaths per 1,000 live births in 1983 to 6.05 deaths per 1,000 live births in 2011 (MacDorman, Hoyert, and Mathews 2013; National Center for Health Statistics 2012). However, even as overall mortality rates declined, infants born to mothers with less than 12 years of schooling have remained approximately twice as likely to die as infants of mothers with 16 years of education or more (Mathews etal. 2004; Singh and Kogan 2007). Research on the association between maternal education and infant health has focused on pathways from education to infant mortality. One pathway involves the economic benefits of educational attainment (Currie and Moretti 2003). Education enables individuals to qualify for high status jobs and earn higher incomes, and such material advantages provide access to a number of resources that matter for infant health (Cramer 1995; Finch 2003; Strully etal. 2010). For example, resources like proper nutrition, health insurance, prenatal care, and nontoxic environments are all linked to infant health and mortality (Abu-Saad and Fraser 2010; Currie, Greenstone, and Moretti 2011; Moss and Carver 1998; Vintzileos etal. 2002). In addition, education may provide mothers with knowledge and cognitive skills that are beneficial to infant health (Baker etal. 2011). There is evidence of a relationship between education and health-promoting behaviors such as exercise, responsible alcohol use, and not smoking (Currie and Moretti 2003; Cutler and Lleras-Muney 2010; Ross and Wu 1995; Salihu etal. 2003), and these factors are strongly associated with birth outcomes (Chen etal. 2009; Kleinman etal. 1988; Passaro etal. 1996; Ventura etal. 2003). Education may also enhance a mother’s ability to navigate the health care system and adhere to treatment regimens during her pregnancy (Goldman and Smith 2002; Hummer etal. 1999). The presence of multiple reinforcing pathways from maternal education to infant health is consistent with the theory of social conditions as a fundamental cause of disease (Link and Phelan 1995). This perspective suggests that socioeconomic position is fundamentally linked to health because it provides access to an extensive array of health-promoting resources such as money, information, social support, and network connections. When faced with health risks, individuals with these resources have more opportunities to protect themselves than those constrained by limited resources (Phelan and Link 2005). Thus, only addressing proximate risks of disease like malnutrition and obesity neglects the broader socioeconomic factors that may pattern the distribution of these risk factors in the first place (Lutfey and Freese 2005). A key implication of fundamental cause theory is that the implementation of health-promoting interventions will often result in larger health disparities between socioeconomic groups because those with more resources are better positioned to take advantage (Phelan etal. 2010). For example, highly educated individuals may have more exposure to information about medical innovations and treatments and be more able to afford the cost of such advances (Glied and Lleras-Muney 2008). This process has been shown to help explain disparities in a range of health outcomes, including inequalities in infant mortality (Chang and Lauderdale 2009; Link etal. 1998; Mechanic 2005; Song and Burgard 2011). For example, W. Parker Frisbie and colleagues (2004) examine differences in infant mortality among blacks and whites following the introduction of surfactant therapy to treat respiratory distress syndrome in 1990. Consistent with the idea that black infants would be less likely to receive surfactant therapy than white infants, they find that although the overall infant mortality rate declined after the introduction of surfactants, racial inequalities in infant mortality increased due to larger reductions among white infants. Variation in the Magnitude of Health Inequalities In recent years, scholars have documented substantial variation in the association between socioeconomic position and health across contexts. This includes cross-national differences (Beckfield etal. 2013; Mackenbach etal. 2008) and variation within the United States across states, counties, and commuting zones (Chetty etal. 2016; Singh and Siahphush 2006; Wilkinson and Pickett 2008; Xu 2006). Differences in the magnitude of health inequalities across populations highlight the need for theories that can help to explain this variation. These explanations can be divided into two general categories. Compositional explanations trace population-level variation in the extent to which socioeconomic position matters for health outcomes to demographic differences (McLeod, Nonnemaker, and Call 2004; Ross and Mirowsky 2008). For example, even after controlling for basic socioeconomic indicators, rates of infant mortality among individuals of Mexican origin living in the United States are much lower than rates for non-Hispanic blacks (Hummer etal. 2007; Hummer etal. 1999; Mathews and MacDorman 2012).2 2 Despite attempts to explain this phenomenon, it largely remains an epidemiologic paradox (Hummer etal. 2007). Thus, in states where a large proportion of those with low education are of Mexican origin, the association between maternal education and infant mortality will likely be smaller in magnitude than in states where few Mexican Americans but many African Americans are represented among those with low education. As this example demonstrates, attempts to understand state-level variation in health inequalities must account for compositional differences across states. Variation in the magnitude of health inequalities may also be a product of differences in institutional context across populations (Beckfield etal. 2015). Institutions represent the rules, policies, infrastructures, and organizations that are part of any society, and an extensive literature highlights the role of such institutions in shaping the distribution of economic resources (e.g., McCall and Percheski 2010; Moller, Nielsen, and Alderson 2009; Morris and Western 1999). Just as economic inequality can be traced to institutional structures, there is growing interest in the role of institutions in shaping health disparities (Beckfield and Krieger 2009; Olafsdottir 2007). Institutional arrangements are likely to play a key role in explaining variation in inequalities in infant mortality across U.S. states. There are notable differences in the scale and organization of state medical system institutions and states differ in the availability of medical personnel and facilities. This infrastructure can have consequences for infant health (Matteson, Burr, and Marshall 1998), and evidence suggests that characteristics of state medical systems, like the availability of health practitioners, contribute to cross-state variation in rates of infant mortality (Bird and Bauman 1995; Shi etal. 2004). So far, the relationship between state medical system institutions and health inequalities has only been subject to limited evaluation in research on variation in the magnitude of health disparities (e.g., Xu 2006). To generate hypotheses about the nature of this association, I turn to fundamental cause theory, specifically the notion that those in more advantaged social positions are better positioned to benefit from many health-promoting interventions (Glied and Lleras-Muney 2008; Goldman and Lakdwalla 2005). This represents an important process through which institutions may influence the association between socioeconomic position and health. When medical system institutions primarily benefit those with high levels of education and other socioeconomic resources, health differences between social groups can be expected to increase (Phelan etal. 2010). Moreover, a complementary (although less widely established) proposition is that institutions that broaden usage of health interventions are expected to reduce health inequalities because the advantages granted by socioeconomic resources in utilizing such interventions will be diminished (Gortmaker and Wise 1997). In this article, I build on these ideas to generate hypotheses about the relationship between state medical system institutions and inequalities in infant mortality. State Medical System Institutions and Inequalities in Infant Mortality Although research has established links between infant mortality and social institutions in U.S. states (Bird and Bauman 1995, 1998; Matteson etal. 1998),3 3 A related line of inquiry explores the role of institutions in explaining variation in infant mortality rates at the national level (Pampel and Pillai 1986). Conley and Springer (2001) provide evidence of an association between welfare state spending and infant mortality in 19 affluent nations. scholars have yet to explore the role of institutions in helping to explain state-level variation in inequalities in infant mortality. Building on fundamental cause theory and the literature on the institutional predictors of infant mortality, I identify two aspects of state medical systems that are hypothesized to influence the relationship between maternal education and infant mortality in U.S. states: neonatal intensive care facilities and primary care physician supply. Neonatal Intensive Care Advances in neonatal intensive care have been a driving force behind reductions in infant mortality in recent decades (Wise 2003). Hospital neonatal intensive care units are equipped with the technology and personnel to treat newborns whose lives are threatened by extreme prematurity, very low birth weight, illnesses, or other delivery complications. There is extensive evidence that the appropriate level of neonatal care is effective in reducing mortality among these high-risk infants (Horbar and Lucey 1995; Paneth etal. 1982; Phibbs etal. 1996; Richardson etal. 1998). Although this care is provided to all high-risk infants born in hospitals with NICUs, not all hospitals have these facilities, and there are considerable differences in NICU availability between states. In 2000, 76 percent of very low birth weight infants were delivered in the hospitals with the appropriate neonatal care facilities in Georgia compared to just 31 percent in Mississippi (Shanahan, Perry, and DeClerque 2012). Differential availability of neonatal intensive care could influence the association between socioeconomic position and infant health (Gortmaker and Wise 1997). Fundamental cause theory suggests that high levels of educational attainment will provide mothers with the resources and information to increase the likelihood that they give birth in hospitals with the facilities necessary for treating high-risk pregnancies, and available evidence supports this notion (Howell and Vert 1993; Samuelson etal. 2002). Thus, in states where NICUs are not widely available and NICU usage depends in part on educational attainment, inequality in infant mortality between education groups is expected to be larger. In contrast, in states where hospitals with NICU facilities are widespread, the disadvantage of low maternal education in securing neonatal intensive care is likely to be diminished (Eberstein, Nam, and Hummer 1990; Gortmaker and Wise 1997). Based on this theory, I predict that greater availability of neonatal intensive care units will be associated with smaller disparities in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more (H1 in Table 1). Table 1. Hypothesized Relationships Between State Medical System Institutions and Inequalities in Infant Mortality Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Table 1. Hypothesized Relationships Between State Medical System Institutions and Inequalities in Infant Mortality Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Primary Care Physician Supply Another aspect of state medical systems that has been linked to infant health is the availability of primary care.4 4 This class of physician includes family and general practitioners, general internists, and general pediatricians. Research on this issue suggests that primary care influences infant mortality in several important ways. For one, primary care is linked to improved infant care practices. Primary care physicians help mothers identify and treat infections and other illnesses common in newborns. They also teach mothers about healthy practices like safe sleeping positions and safety at home and in vehicles (Shi etal. 2004). These represent some of the principal risk factors for mortality (Starfield 1985) and highlight the potential for primary care to benefit infant health. In addition, primary care influences infant mortality through its effect on maternal health. Primary care promotes reduced smoking and alcohol use, healthier sexual practices, and improved nutrition (Shi etal. 2004), and there are clear links between these behaviors and birth weight and infant mortality (Chen etal. 2009; Kleinman etal. 1988; Passaro etal. 1996; Ventura etal. 2003). Consistent with these linkages between primary care and infant health, differences in the supply of state primary care physicians per capita are associated with state-level differences in infant mortality (Shi etal. 1999; Shi etal. 2004). In addition to influencing absolute levels of infant mortality, the availability of primary care physicians in a state may be associated with inequalities in infant mortality between socioeconomic groups. Barbara Starfield, Leiyu Shi, and James Macinko (2005) report that primary care provides more substantial reductions in infant mortality in states with high social inequality than in states with lower inequality (see also Shi etal. 2004). Based on this result, they suggest that the supply of primary care physicians in a state can reduce health inequality by increasing the availability of key health services (Starfield etal. 2005). However, Starfield and colleagues (2005) also acknowledge that an increased supply of primary care physicians per capita may not guarantee more universal usage of primary care. Consistent with fundamental cause theory, the supply of primary care physicians may actually be linked to greater inequalities in infant mortality if mothers with higher education are better positioned to make use of this care due to better health coverage and other socioeconomic advantages. To help adjudicate between these predictions, I evaluate the hypothesis that disparities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 years of college or more will be smaller in states where the supply of primary care physicians is greater (H2 in Table 1). DATA AND METHODS The primary data for this article are birth and infant death records from the National Vital Statistics System (NVSS). The NVSS, run by the National Center for Health Statistics, links birth and death certificates for all infants born in the United States (National Center for Health Statistics 2001-2006).5 5 In practice, the NVSS is able to successfully link almost 99 percent of infant deaths to a corresponding birth certificate. For example, in 2002, only 292 out of 27,527 infant deaths were unlinked. 6 In 2003, a substantial revision of the birth certificate was introduced. The changes included a new measure of maternal education that was deemed incompatible with the previous standard (Mathews and MacDorman 2012). States adopted this revision gradually and the process was not completed until 2015. As a result, 2002 is the last available year that all states used the same standard measure of maternal education. 7 I measure infant’s sex with a dummy variable for male infants. Plural birth is measured with a dummy variable for plural infants. Mother’s age is measured with dummy variables for each age group < 20, 20-24, 25-29 (reference category), 30-34, 35-39, and 40+. Mother’s race is measured with dummy variables for white (reference category), black, Hispanic, and other race. Birth history is measured with dummy variables for first birth, second birth (reference category), third birth, fourth birth, and fifth or more births. Maternal health condition is measured with a dummy variable indicating the presence of one or more health problems reported on birth records (measured conditions include anemia, cardiac disease, acute or chronic lung disease, diabetes, hemoglobinopathy, chronic hypertension, and renal disease). 8 There is also evidence that prenatal care may influence infant health (Vintzileos etal. 2002; cf Fiscella 1995). In supplemental analyses, I control for the receipt and timing of prenatal care. The results are robust to the inclusion of these measures. Since prenatal care is likely to intervene on the pathway between maternal education and infant mortality, I exclude this factor from the analyses shown here. I use records from births occurring in 1997-2002.6 The linked data files include information on an infant’s birth and death as well as maternal educational attainment. I code infant mortality as a dichotomous variable indicating whether an infant died in the first year of life. The measure of maternal education includes four categories of educational attainment: less than 12 years of schooling, 12 years of schooling, less than 4 years of college, and 4 years of college or more. In my analyses, I employ dummy variables for each education category (with less than 12 years of schooling serving as the reference category). The linked birth-death records include information on a number of additional infant and maternal characteristics. Here, I use measures of infant’s sex, plural birth status, maternal age, maternal race, maternal birth history, and maternal health conditions in order to control for factors relevant to infant mortality risk (Mathews etal. 2004).7 For example, a mother’s age, race, and health status have the potential to influence both her educational attainment and birth outcomes. Controlling for these potential confounders helps reduce bias in estimates of the association between maternal education and infant mortality.8 Moreover, controlling for factors like race and maternal age helps account for the role of demographic composition in driving state-level variation in the extent to which maternal education matters for infant mortality. I combine the six years of linked birth-death records with state-level data on medical systems. I measure the availability of neonatal intensive care with a measure of NICUs per 10,000 state residents (American Hospital Association 1997-2004) I measure primary care supply as primary care physicians per 10,000 residents (American Medical Association 1997-2004). In addition, I control for state-level characteristics that have the potential to influence the relationship between educational inequalities in infant mortality and state medical systems. First, I control for the number of hospitals per 10,000 state residents to ensure that the measures of NICU availability and primary care supply capture the effects of these institutions beyond the effects of a state's hospital infrastructure. In addition, I control for the state infant mortality rate to account for the broader infant health context. Finally, I control for a set of variables that capture socioeconomic conditions at the state level. These include the log of per-capita GDP, income inequality (measured with the Gini coefficient), the unemployment rate, and the poverty rate.9 9 Data on state GDP come from the Bureau of Economic Analysis (2012). Data on state Gini coefficients come from the U.S. Census Bureau’s Annual Social and Economic Supplements (2012). State unemployment data come from the U.S. Bureau of Labor Statistics’ Local Area Unemployment Statistics program (2012). State poverty data come from the Census Bureau’s Small Area Income and Poverty Estimates program (2012b). Data on state hospitals come from the American Hospital Association (1997-2004). State infant mortality data come from the CDC’s WONDER online database (2012). Descriptive statistics for all variables are displayed in Table 2. All results are based on unweighted data, but the use of weights that account for unlinked death records does not substantively change the results presented here. Table 2. Descriptive Statistics for All Individual and State-Level Variables, 1997-2002 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Table 2. Descriptive Statistics for All Individual and State-Level Variables, 1997-2002 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 The analysis proceeds in three parts. I first investigate potential variation in the association between maternal education and infant mortality across U.S. states. I measure this association in each state with separate logistic regression models for all 50 states. Each model includes dummy variables for each category of maternal education and also controls for race, maternal age, sex, plural birth, maternal birth history, and maternal health conditions. The models pool data from 1997-2002 and include year fixed effects to account for time trends. Based on these models, I calculate the predicted probability of mortality for infants of mothers from two education groups in each state: those with less than 12 years of schooling and those with 4 years of college or more. When calculating predicted probabilities, I hold the values of control variables constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. Infants with these characteristics are the least likely to experience mortality (Mathews etal. 2004). Thus, this represents a conservative assessment of the level of infant mortality risk. In the second part of the analysis, I combine the individual infant birth-death records with the state-level institutional measures. This results in a multi-level data set with 22,967,018 individual records clustered in 50 states over six years. Using these data, I analyze the association between maternal education and infant mortality using logistic regression models with random intercepts for each of the 50 states. These models are well-suited for analysis of cross-state institutional differences because they account for variation both within and across states. Year fixed effects account for any national-level time trends. In order to assess whether the association between maternal education and infant mortality varies across state medical system institutions, I introduce cross-level interaction terms in which the indicator variables for maternal education are allowed to interact with the measures of NICU availability and primary care supply. I focus on the interaction between these medical system variables and the indicator for four plus years of college in order to explore inequalities in infant mortality between mothers with less than 12 years of schooling and 4 years of college or more. Finally, after evaluating whether institutional measures are significantly associated with educational inequalities in infant mortality, I present graphs showing how the predicted probability of infant mortality varies across the observed levels of neonatal intensive care and primary care for mothers with less than 12 years of schooling and those with 4 years of college or more. Graphs are generated based on the random effects logistic regression models of infant mortality on maternal education detailed above. For each graph, individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level controls and institutional variables are held constant at their mean values from 1997-2002. RESULTS I begin the analysis by exploring the possibility that educational inequalities in infant mortality vary in magnitude across U.S. states. Based on a series of state-specific logistic regression models of infant mortality on maternal education, I calculate the predicted probability of mortality for infants of mothers from two education groups in each state: those with less than 12 years of schooling and those with 4 years of college or more.10 10 Results of these analyses are displayed in Appendix Figure A1. In order to assess inequalities between these education groups, I calculate the relative risk ratio as the probability of infant mortality for mothers with low education over the probability for highly educated mothers.11 11 I also compare this measure of inequality to a measure based on absolute differences in the probability of infant mortality between mothers with low and high education. The correlation between the absolute and relative measures of inequality is .89 and the geographic patterning is similar across states.Figure 1 maps the relative risk for each state and displays considerable variation in the magnitude of inequality across states. In some states (including Alaska, North Dakota, and Kentucky), infants born to mothers with less than 12 years of schooling are more than twice as likely as those born to mothers with 4 years or more of college to die. In other states, the risk ratios are as low as 1.3 (and as low as 1.13 in Hawaii). This analysis controls for key compositional factors including race, suggesting that population demographics do not fully account for cross-state variation in the extent to which maternal education is associated with infant mortality. Figure 1. View largeDownload slide Relative Risk of Infant Mortality by Maternal Education—Less than 12 Years of Schooling vs. 4 Years of College or More, 1997-2002 (with controls) Note: Risk ratios of infant mortality by education for non-Hispanic white, non-plural, second-born daughters of mothers age 25–29 with no prior health conditions. Figure 1. View largeDownload slide Relative Risk of Infant Mortality by Maternal Education—Less than 12 Years of Schooling vs. 4 Years of College or More, 1997-2002 (with controls) Note: Risk ratios of infant mortality by education for non-Hispanic white, non-plural, second-born daughters of mothers age 25–29 with no prior health conditions. I then evaluate whether these cross-state differences in inequality in infant mortality are statistically significant. I estimate a model that combines observations from all 50 states and includes maternal education, sociodemographic controls, state and year dummy variables, and state * maternal education interactions (not shown). I assess whether the association between maternal education and infant mortality differs significantly across states with a joint F-test of the null hypothesis that the state*maternal education interaction coefficients are all equal to each other. This hypothesis can be rejected (p < .0001), demonstrating significant cross-state differences in the effect of education on infant mortality. In addition, I assess whether the effects of maternal education on infant mortality depend on the state with a joint F-test of the null hypothesis that the interaction coefficients are all equal to 0. This hypothesis can also be rejected (p < .0001), providing evidence of a significant role of state-level factors in the association between maternal education and infant mortality.12 12 Unlike the state-specific models, the effects of the sociodemographic control variables are necessarily assumed to be the same across states in this combined model. 13 All results are robust to the exclusion of all non-significant state-level control variables. After highlighting significant differences in the magnitude of inequalities in infant mortality across states, I evaluate hypotheses about the role of state medical system institutions that are predicted to be associated with this variation using a series of random effects logit models. Table 3 displays the results of this analysis using odds ratios. Model 1 of Table 3 presents a baseline analysis of maternal education and infant mortality from 1997-2002 that controls for key individual and state-level factors. This reveals a clear education-mortality gradient, with each increasing level of education reducing the odds of infant mortality relative to mothers with less than 12 years of schooling. Odds ratios for the individual-level control variables are signed in the expected direction with black, male, plural infants of mothers with existing health conditions having higher odds of mortality relative to each reference category. Of the state-level control variables, only the odds ratios for the infant mortality rate and hospitals per 10,000 are significant, though the lack of significant predictors at the state-level is not surprising given that the state random effects accounts for much of the state-level variation.13 Table 3. Logistic Regression of Infant Mortality on Maternal Education and State Medical System Institutions, 1997-2002 (state random intercepts and year fixed effects) Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Notes: Models also include dummy variables for birth history and mother’s age. Standard errors in parentheses. * p < .05 **p < .01 (two-tailed tests) Table 3. Logistic Regression of Infant Mortality on Maternal Education and State Medical System Institutions, 1997-2002 (state random intercepts and year fixed effects) Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Notes: Models also include dummy variables for birth history and mother’s age. Standard errors in parentheses. * p < .05 **p < .01 (two-tailed tests) Model 2 of Table 3 adds the two key measures of state medical systems: NICUs per 10,000 state residents and primary care physicians per 10,000 residents. Both measures are negatively associated with infant mortality, and while the odds ratio for NICUs per 10,000 is significant at the .05 level, the odds ratio for primary care physicians per 10,000 is not significant. In order to assess whether the association between maternal education and infant mortality varies across state medical system institutions, I introduce a series of cross-level interaction terms in which the indicator variables for maternal education are allowed to interact with the measures of NICU availability (Model 3) and primary care supply (Model 4). I focus on the interaction term for mothers with four plus years of college because this value reflects the extent to which the disparity in infant mortality between mothers with less than 12 years of schooling and those with 4 years of college or more varies with the state medical system variables. In Model 3, the odds ratio of 5.412 is equivalent to a logit coefficient of 1.689, which is positively signed and significant. This indicates that compared to mothers with 4 or more years of college, the NICU effect is greater among mothers with less than 12 years of schooling. Model 4 switches the focus to the primary care supply and includes cross-level interactions between primary care physicians per 10,000 residents and maternal education. The odds ratio of .987 for the interaction between primary care physicians per 10,000 and four plus years of college corresponds with a logit coefficient that is negatively signed and significant (-.013), indicating that the effect of primary care supply is greater among mothers with at least 4 years of college than among mothers with less than 12 years of schooling. The analyses in Table 3 provide evidence that state medical system institutions moderate the association between maternal education and infant mortality. To offer a more detailed picture of this relationship, I present graphs showing how the predicted probability of under-five mortality varies with the level of neonatal intensive care and primary care for infants from these two groups when other key factors are held constant. Figure 2 illustrates how the predicted probability of infant mortality changes over the observed range of NICU facilities per 10,000 state residents for mothers with less than 12 years of schooling and mothers with 4 years of college or more. With more NICUs, inequality in the predicted probability of infant mortality is reduced. This reduction is driven by the negative relationship between NICU availability and infant mortality among mothers with low educational attainment. For example, at a very low level of NICU availability like .01 NICUs per 10,000, the predicted probability of mortality for infants of mothers with less than 12 years of schooling is .0053. At a high level of NICU availability like .09 NICUs per 10,000, this probability declines to .0048. In contrast, the predicted probability for infants of mothers with 4 years of college or more is .0028 with .01 NICUs per 10,000 and .0029 with .09 NICUs per 10,000. This means that for mothers with less than 12 years of schooling, there is one fewer death for every 2,000 live births in states with .09 NICUs per 10,000 compared to states with .01 NICUs per 10,000. For mothers with 4 years of college or more, there is one more death per 10,000 live births at greater levels of state NICU availability. While this difference in inequality is small in absolute terms, the relative risk of infant mortality between these education groups declines 12.5 percent, from 1.91 to 1.67 when comparing low versus high NICU availability. This supports H1 that state neonatal intensive care availability will be negatively associated with inequality in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more. Figure 2. View largeDownload slide Predicted Probability of Infant Mortality by NICUs per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 2. View largeDownload slide Predicted Probability of Infant Mortality by NICUs per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 3 shows how educational inequality in infant mortality varies over the observed range of primary care physicians per 10,000 state residents. A greater primary care physician supply is linked to reductions in the predicted probability of infant mortality for both mothers with less than 12 years of schooling and mothers with 4 years of college or more. However, this reduction is more pronounced for mothers with at least a college education. For example, expressed in absolute terms, mothers in the highly educated group experience 1.1 fewer deaths for every 3,000 live births among in states with 7 primary care physicians per 10,000 (a low level of primary care supply) compared to states with 13 primary care physicians per 10,000 (a high primary care supply). In contrast, mothers with low education experience .9 fewer deaths for every 3,000 births at these levels of primary care. This small absolute change in inequality represents a modest relative difference. At 7 primary care physicians per 10,000, the relative risk of infant mortality between these groups is 1.80. At 13 primary care physicians per 10,000, the relative risk increases 8 percent to 1.95. Thus, inequality in the predicted probability of infant mortality is slightly larger where there are more primary care physicians. This contradicts H2 that state primary care supply will be negatively associated with inequality in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more. Figure 3. View largeDownload slide Predicted Probability of Infant Mortality by Primary Care Physicians per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 3. View largeDownload slide Predicted Probability of Infant Mortality by Primary Care Physicians per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. DISCUSSION This article explores two research questions about the relationship between maternal education and infant mortality in U.S. states. I first evaluate the extent to which the magnitude of this association varies across states and find evidence of substantial differences. Infants born to mothers with less than 12 years of schooling are more than twice as likely to die as infants born to mothers with four or more years of college (even after accounting for factors like race, maternal age, and birth history). In contrast, other states see minimal differences in the risk of infant mortality by maternal education. This analysis contributes to the growing body of research that documents variation in the relationship between socioeconomic position and health across populations (e.g., Beckfield etal. 2013; Chetty etal. 2016; Mackenbach etal. 2008) with the first study focused on cross-state differences in educational inequalities in infant mortality. The presence of dramatic cross-state differences in inequality in infant mortality calls attention to the institutional predictors of this variation. In the second portion of the analysis, I explore whether variation in inequality is associated with state medical system institutions. I focus on two components of state medical systems that have proven relevant to infant mortality: neonatal intensive care availability and primary care supply. Greater NICU availability is associated with a reduction in the risk of infant mortality among mothers with less than 12 years of education. In contrast, the probability of infant mortality does not vary with neonatal intensive care for mothers with 4 years of college or more. Thus, with more NICUs per 10,000, inequalities in the risk of infant mortality between these groups are smaller in magnitude. A possible explanation for this negative association between the availability of state neonatal intensive care and educational inequalities in infant mortality is that the importance of maternal socioeconomic resources in accessing this care varies based on the availability of NICUs. Existing evidence suggests that individuals with more resources are advantaged in making use of health interventions (Chang and Lauderdale 2009; Link and Phelan 1995). Thus, in states where hospitals with NICUs are not widely available, education, health insurance, and other resources are likely to play an important role in helping mothers locate and travel to the hospitals equipped with these facilities (Howell and Vert 1993; Samuelson etal. 2002). Mothers without these resources face more obstacles to receiving this care, which likely contributes to the observed disparity in infant mortality in states with low NICU availability. However, I also find that inequality in infant mortality is reduced in states with more NICUs per 10,000 residents, which suggests that the advantages provided by education in securing neonatal intensive care are diminished in contexts where NICU facilities are widely available and there are fewer barriers to accessing NICU care. This may reflect the nature of NICUs—unlike many other health interventions, hospitals equipped with NICUs provide care to all infants in need of these facilities, regardless of their families’ socioeconomic resources. In contrast, greater primary care physician supply is associated with slightly larger disparities in infant mortality between maternal education groups. Mothers with less than 12 years of schooling and those with 4 years of college or more both face lower absolute risk of infant mortality in states with more primary care physicians per 10,000 residents, but this reduction is more pronounced among mothers with a college education. While this finding does not support the hypothesis that primary care leads to reductions in health inequality (Starfield etal. 2005), it is consistent with the notion that health-promoting interventions can increase health disparities between social groups (Phelan and Link 2005). Unlike NICUs, which provide care to all at-risk infants who are born in hospitals with these facilities, an extensive supply of primary care physicians does not necessarily broaden usage (Matteson etal. 1998). Instead, making use of primary care requires health insurance coverage, the ability to attend regular medical appointments, and information on the benefits of this care. Thus, mothers with higher levels of education are likely to be better positioned to benefit from a greater supply of primary care physicians. This calls attention to a key difference between these two components of state medical systems, and this difference has implications for both theory and public policy. The finding that primary care is associated with greater health inequalities is consistent with the fundamental cause perspective because it highlights how interventions that improve health outcomes can also exacerbate health inequalities (Phelan etal. 2010). However, the link between NICU availability and smaller disparities in infant mortality by maternal education highlights an important implication of fundamental cause theory that has received insufficient attention. While evaluations of this theory have focused on the role of health interventions in increasing inequality (Chang and Lauderdale 2009; Glied and Lleras-Muney 2008; Song and Burgard 2011), the results presented here suggest that interventions like NICUs that broaden usage of medical services may be able to reduce health inequalities by minimizing the health risks of a disadvantaged socioeconomic position (Gortmaker and Wise 1997). This is a notable extension of research on social position as a fundamental cause of health outcomes because it suggests that the salience of socioeconomic resources for an individual’s health varies based on the institutional context. When health-promoting institutions are not widely accessible, socioeconomic resources appear to play a more central role in determining health outcomes. Yet, when institutions broaden usage of health services and interventions, the importance of socioeconomic resources is diminished. This is a key pathway through which social institutions can shape health disparities and represents a promising starting point for future efforts to understand differences in the nature of health disparities across contexts. This analysis also highlights the power of social policy to shape health inequalities (Beckfield and Krieger 2009). Even beyond the potential value of investments in medical systems that further expand the availability of NICUs, the results suggest that policies that expand usage of health interventions and services among socioeconomically disadvantaged individuals can be effective in reducing health disparities. For example, policies that broaden usage of primary care might enable mothers with low education to receive the same infant health benefits of this service as highly educated mothers. If so, then policy reforms like the 2010 Affordable Care Act (ACA), which expanded health insurance coverage to many individuals who could not previously afford care, have the potential to play a key role in efforts to reduce health differences between social groups. Moreover, the differential adoption of the ACA’s Medicaid expansion across states suggests that there will likely be considerable differences in the law’s impact at the state level (Blumenthal and Collins 2014). This variation provides an opportunity to evaluate the effects of large-scale institutional change on the relationship between socioeconomic position and health and represents a promising opportunity for future research. As scholars continue to study the relationship between state institutions and inequalities in infant mortality, several additional issues stand out as particularly important for further exploration. A limitation of this article is that the state-level analysis prevents an examination of the geographic distribution of medical system institutions within states. While the number of NICUs per 10,000 residents provides a broad measure of the availability of neonatal intensive care, it does not account for the fact that NICUs may not be proportionally distributed within states. As a result, future research should seek to incorporate more detailed geographic data on the location of NICUs in order to evaluate whether within-state differences in NICU availability influence inequalities in infant mortality. Another useful way to extend the analysis would be to account for the process by which infants are selected into neonatal intensive care. Pregnant mothers in disadvantaged socioeconomic positions face a multitude of health risks (e.g., malnutrition, unsafe housing, lack of prenatal care), and their infants are more likely to be born prematurely and at low birth weight (Blumenshine etal. 2010). If the greater likelihood of being born in these vulnerable conditions results in a greater likelihood of treatment in NICU facilities, then the observed association between state NICU availability and inequalities in infant mortality could partially reflect this selection effect. In supplemental analyses, I control for birth weight in order to account for a measure of health status at birth that predicts NICU entry. Although the results are robust to this alternative specification,14 14 I control for birth weight with dummy variables for infants who are born at low (2,500-1,500 grams) or very low (< 2,500 grams) weight. Results of this analysis are available upon request. birth weight is an imperfect measure of an infant’s health status at birth (Schempf etal. 2007). This highlights the value of collecting more detailed data on infant health at birth and the care infants receive after being born in order to analyze the extent to which infants are selected into NICUs. In addition, while existing research has focused on state medical systems as key institutional predictors of infant mortality, there may also be institutions that operate outside the medical system that influence the relationship between socioeconomic position and health (Beckfield and Krieger 2009). For example, institutions like labor unions and welfare policies have been shown to influence a multitude of health outcomes (e.g., Cho 2011; Reynolds and Brady 2012). Research that explores the extent to which these and other political institutions have distinct effects on individuals in different social positions represents an important next step in the study of health inequalities. The presence of state-level variation in the association between maternal education and infant mortality shows that socioeconomic position varies in its importance as a predictor of health outcomes across contexts. This draws attention to the role of institutions in the production of health inequalities, and the results presented here suggest that efforts to reduce inequality should focus on institutions that broaden the usage of health interventions. The author wishes to thank Jason Beckfield, Bruce Western, Sandy Jencks, Alexandra Killewald, Marie McCormick, Jack Shonkoff, Benjamin Sommers, Rourke O’Brien, and Chong-Min Fu for help and suggestions. Direct correspondence to: Benjamin Sosnaud, Department of Sociology and Anthropology, Trinity University, San Antonio, TX 78212. E-mail: bsosnaud@trinity.edu. APPENDIX Figure A1. View largeDownload slide Predicted Probability of Infant Mortality and 95 Percent Confidence Intervals for Mothers with Low and High Education by State, 1997-2002 (with controls) Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. Figure A1. View largeDownload slide Predicted Probability of Infant Mortality and 95 Percent Confidence Intervals for Mothers with Low and High Education by State, 1997-2002 (with controls) Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. REFERENCES Abu-Saad Kathleen , Fraser Drora . 2010 . “Maternal Nutrition and Birth Outcomes.” Epidemiologic Reviews 32 : 5 - 25 . Google Scholar CrossRef Search ADS PubMed American Hospital Association . 1997-2004 . Hospital Statistics . Chicago : Health Forum LLC . American Medical Association . 1997-2004 . Physician Characteristics and Distribution in the United States . Chicago : American Medical Association . Baker David P. , Leon Juan , Smith Greenaway Emily G. , Collins John , Movit Marcela . 2011 . “The Education Effect on Population Health: A Reassessment.” Population and Development Review 37 : 307 - 32 . Google Scholar CrossRef Search ADS PubMed Beckfield Jason , Bambra Clare , Eikemo Terje A. , Huijts Tim , McNamara Courtney , Wendt Claus . 2015 . “An Institutional Theory of Welfare State Effects on the Distribution of Population Health.” Social Theory & Health 13 : 227 - 44 . Google Scholar CrossRef Search ADS Beckfield Jason , Krieger Nancy . 2009 . “Epi + demos + cracy: Linking Political Systems and Priorities to the Magnitude of Health Inequities—Evidence, Gaps, and a Research Agenda.” Epidemiologic Reviews 31 : 152 - 77 . Google Scholar CrossRef Search ADS PubMed Beckfield Jason , Olafsdottir Sigrun , Bakhtiari Elyas . 2013 . “Health Inequalities in Global Context.” American Behavioral Scientist 57 : 1014 - 39 . Google Scholar CrossRef Search ADS Bird Sheryl Thorburn , Bauman Karl E . 1995 . “The Relationship Between Structural and Health Services Variables and State-Level Infant Mortality in the United States.” American Journal of Public Health 85 : 26 - 29 . Google Scholar CrossRef Search ADS PubMed Bird Sheryl Thorburn , Bauman Karl E. 1998 . “State-Level Infant, Neonatal, and Postneonatal Mortality: The Contribution of Selected Structural Socioeconomic Variables.” International Journal of Health Services 28 : 13 - 27 . Google Scholar CrossRef Search ADS PubMed Blumenthal David , Collins Sara R . 2014 . “Health Care Coverage under the Affordable Care Act—A Progress Report.” New England Journal of Medicine 371 : 275 - 81 . Google Scholar CrossRef Search ADS PubMed Blumenshine Philip , Egerter Susan , Barclay Colleen J. , Cubbin Catherine , Braveman Paula A . 2010 . “Socioeconomic Disparities in Adverse Birth Outcomes: A Systematic Review.” American Journal of Preventive Medicine 39 : 263 - 72 . Google Scholar CrossRef Search ADS PubMed Bureau of Economic Analysis . 2012 . Regional Economic Accounts . Retrieved September 13, 2017 (www.bea.gov/regional/). Chang Virginia W. , Lauderdale Diane S . 2009 . “Fundamental Cause Theory, Technological Innovation, and Health Disparities: The Case of Cholesterol in the Era of Statins.” Journal of Health and Social Behavior 50 : 245 - 60 . Google Scholar CrossRef Search ADS PubMed Chen Aimin , Feresu Shingairai A. , Fernandez Cristina , Rogan Walter J . 2009 . “Maternal Obesity and the Risk of Infant Death in the United States.” Epidemiology 20 : 74 - 81 . Google Scholar CrossRef Search ADS PubMed Chetty Raj , Stepner Michael , Abraham Sarah , Lin Shelby , Scuderi Benjamin , Turner Nicholas , Bergeron Augustin , Cutler David . 2016 . “The Association Between Income and Life Expectancy in the United States, 2001–2014.” Journal of the American Medical Association 315 : 1750 - 66 . Google Scholar CrossRef Search ADS PubMed Cho Rosa M. 2011 . “Effects of Welfare Reform Policies on Mexican Immigrants’ Infant Mortality Rates.” Social Science Research 40 : 641 - 53 . Google Scholar CrossRef Search ADS Conley Dalton , Springer Kristen W . 2001 . “Welfare State and Infant Mortality.” American Journal of Sociology 107 : 768 - 807 . Google Scholar CrossRef Search ADS Conley Dalton , Bennett Neil G . 2000 . “Is Biology Destiny? Birth Weight and Life Chances.” American Sociological Review 65 : 458 - 67 . Google Scholar CrossRef Search ADS Cramer James C. 1995 . “Racial and Ethnic Differences in Birthweight: The Role of Income and Financial Assistance.” Demography 32 : 231 - 47 . Google Scholar CrossRef Search ADS PubMed Currie Janet , Moretti Enrico . 2003 . “Mother’s Education and the Intergenerational Transmission of Human Capital: Evidence from College Openings.” Quarterly Journal of Economics 118 : 1495 - 1532 . Google Scholar CrossRef Search ADS Currie Janet , Greenstone Michael , Moretti Enrico . 2011 . “Superfund Cleanups and Infant Health.” American Economic Review 101 : 435 - 41 . Google Scholar CrossRef Search ADS PubMed Cutler David M. , Lleras-Muney Adriana . 2010 . “Understanding Differences in Health Behaviors by Education.” Journal of Health Economics 29 : 1 - 28 . Google Scholar CrossRef Search ADS PubMed Eberstein Isaac , Nam Charles , Hummer Robert . 1990 . “Infant Mortality by Cause of Death: Main and Interaction Effects.” Demography 27 : 413 - 30 . Google Scholar CrossRef Search ADS PubMed Elo Irma T. 2009 . “Social Class Differentials in Health and Mortality: Patterns and Explanations in Comparative Perspective.” Annual Review of Sociology 35 : 553 - 72 . Google Scholar CrossRef Search ADS Finch Brian Karl. 2003 . “Early Origins of the Gradient: The Relationship Between Socioeconomic Status and Infant Mortality in the United States.” Demography 40 : 675 - 99 . Google Scholar CrossRef Search ADS PubMed Fiscella Kevin. 1995 . “Does Prenatal Care Improve Birth Outcomes? A Critical Review.” Obstetrics & Gynecology 85 : 468 - 79 . Google Scholar CrossRef Search ADS Frisbie W. Parker , Song Seung-eun , Powers Daniel A. , Street Julie A . 2004 . “The Increasing Racial Disparity in Infant Mortality: Respiratory Distress Syndrome and Other Causes.” Demography 41 : 773 - 800 . Google Scholar CrossRef Search ADS PubMed Glied Sherry , Lleras-Muney Adriana . 2008 . “Technological Innovation and Inequality in Health.” Demography 45 : 741 - 61 . Google Scholar CrossRef Search ADS PubMed Goesling Brian. 2007 . “The Rising Significance of Education for Health?” Social Forces 85 : 1621 - 44 . Google Scholar CrossRef Search ADS Goldman Dana P. , Lakdawalla Darius N . 2005 . “A Theory of Health Disparities and Medical Technology.” Contributions in Economic Analysis & Policy 4 : 1 - 30 . Google Scholar CrossRef Search ADS Goldman Dana P. , Smith James P . 2002 . “Can Patient Self-Management Help Explain the SES Health Gradient?” Proceedings of the National Academy of Sciences 99 : 10929 - 34 . Google Scholar CrossRef Search ADS Gortmaker Steven L. 1979 . “Poverty and Infant Mortality in the United States.” American Sociological Review 44 : 280 - 97 . Google Scholar CrossRef Search ADS PubMed Gortmaker Steven L. , Wise Paul H . 1997 . “The First Injustice: Socioeconomic Disparities, Health Services Technology, and Infant Mortality.” Annual Review of Sociology 23 : 147 - 70 . Google Scholar CrossRef Search ADS PubMed Horbar Jeffrey D. , Lucey Jerold F . 1995 . “Evaluation of Neonatal Intensive Care Technologies.” The Future of Children 5 : 139 - 61 . Google Scholar CrossRef Search ADS PubMed Howell Embry Martin , Vert Paul . 1993 . “Neonatal Intensive Care and Birth Weight-Specific Perinatal Mortality in Michigan and Lorraine.” Pediatrics 91 : 464 - 69 . Google Scholar PubMed Hummer Robert A. 1993 . “Racial Differentials in Infant Mortality in the U.S.: An Examination of Social and Health Determinants.” Social Forces 72 : 529 - 54 . Google Scholar CrossRef Search ADS Hummer Robert A. , Powers Daniel A. , Pullum Starling G. , Gossman Ginger L. , Frisbie W. Parker . 2007 . “Paradox Found (Again): Infant Mortality Among the Mexican-Origin Population in the United States.” Demography 44 : 441 - 57 . Google Scholar CrossRef Search ADS PubMed Hummer Robert A. , Biegler Monique , De Turk Peter B. , Forbes Douglas , Frisbie W. Parker , Hong Ying , Pullum Starling G . 1999 . “Race/Ethnicity, Nativity, and Infant Mortality in the United States.” Social Forces 77 : 1083 - 1117 . Google Scholar CrossRef Search ADS Kleinman Joel C. , Pierre Mitchell B. , Madans Jennifer H. , Land Garland H. , Schramm Wayne F . 1988 . “The Effects of Maternal Smoking on Fetal and Infant Mortality.” American Journal of Epidemiology 127 : 274 - 82 . Google Scholar CrossRef Search ADS PubMed Link Bruce G. , Phelan Jo . 1995 . “Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior 35 : 80 - 94 . Google Scholar CrossRef Search ADS Link Bruce G. , Northridge Mary E. , Phelan Jo C. , Ganz Michael L . 1998 . “Social Epidemiology and the Fundamental Cause Concept: On the Structuring of Effective Cancer Screens by Socioeconomic Status.” Milbank Quarterly 76 : 375 - 402 . Google Scholar CrossRef Search ADS PubMed Lutfey Karen , Freese Jeremy . 2005 . “Toward Some Fundamentals of Fundamental Causality: Socioeconomic Status and Health in the Routine Clinic Visit for Diabetes.” American Journal of Sociology 110 : 1326 - 72 . Google Scholar CrossRef Search ADS MacDorman Marian F. , Hoyert Donna L. , Mathews T. J . 2013 . Recent Declines in Infant Mortality in the United States 2005-2011. Hyattsville, MD : National Center for Health Statistics . Mackenbach Johan P. , Stirbu Irina , Roskam Albert-Jan , Schaap Maartje M. , Menvielle Gwenn , Leinsalu Mall , Kunst Anton E . 2008 . “Socioeconomic Inequalities in Health in 22 European Countries.” The New England Journal of Medicine 358 : 2468 - 81 . Google Scholar CrossRef Search ADS PubMed Mathews T. J. , Menacker Fay , MacDorman Marian F . 2004 . Infant Mortality Statistics from the 2002 Period: Linked Birth/Infant Death Data Set. Hyattsville, MD : National Center for Health Statistics . Mathews T. J. , MacDorman Marian F . 2012 . Infant Mortality Statistics from the 2008 Period: Linked Birth/Infant Death Data Set. Hyattsville, MD : National Center for Health Statistics . Matteson Donald W. , Burr Jeffrey A. , Marshall James R . 1998 . “Infant Mortality: A Multi-Level Analysis of Individual and Community Risk Factors.” Social Science & Medicine 47 : 1841 - 54 . Google Scholar CrossRef Search ADS McCall Leslie , Percheski Christine . 2010 . “Income Inequality: New Trends and Research Directions.” Annual Review of Sociology 36 : 329 - 47 . Google Scholar CrossRef Search ADS McLeod Jane D. , Nonnemaker James M. , Call Kathleen Thiede . 2004 . “Income Inequality, Race, and Child Well-Being: An Aggregate Analysis in the 50 United States.” Journal of Health and Social Behavior 45 : 249 - 64 . Google Scholar CrossRef Search ADS PubMed Mechanic David. 2005 . “Policy Challenges in Addressing Racial Disparities and Improving Population Health.” Health Affairs 24 : 335 - 38 . Google Scholar CrossRef Search ADS PubMed Miech Richard , Pampel Fred , Kim Jinyoung , Rogers Richard G . 2011 . “The Enduring Association Between Education and Mortality The Role of Widening and Narrowing Disparities.” American Sociological Review 76 : 913 - 34 . Google Scholar CrossRef Search ADS PubMed Moller Stephanie , Nielsen François , Alderson Arthur S . 2009 . “Changing Patterns of Income Inequality in U.S. Counties, 1970-2000.” American Journal of Sociology 114 : 1037 - 1101 . Google Scholar CrossRef Search ADS Morris Martina , Western Bruce . 1999 . “Inequality in Earnings at the Close of the Twentieth Century.” Annual Review of Sociology 25 : 623 - 57 . Google Scholar CrossRef Search ADS Moss Nancy E. , Carver Karen . 1998 . “The Effect of WIC and Medicaid on Infant Mortality in the United States.” American Journal of Public Health 88 : 1354 - 61 . Google Scholar CrossRef Search ADS PubMed National Center for Health Statistics . 2001-2006 . “Linked Birth/Infant Death Data Set. Public Use Data files.” National Center for Health Statistics, Hyattsville, MD. National Center for Health Statistics . 2012 . Health, United States, 2011: With Special Feature on Socioeconomic Status and Health. Hyattsville, MD : National Center for Health Statistics . Olafsdottir Sigrun. 2007 . “Fundamental Causes of Health Disparities: Stratification, the Welfare State, and Health in the United States and Iceland.” Journal of Health and Social Behavior 48 : 239 - 53 . Google Scholar CrossRef Search ADS PubMed Pampel Fred C. , Pillai Vijayan K . 1986 . “Patterns and Determinants of Infant Mortality in Developed Nations, 1950- 1975.” Demography 23 : 525 - 42 . Google Scholar CrossRef Search ADS PubMed Paneth Nigel , Kiely John L. , Wallenstein Sylvan , Marcus Michele , Pakter Jean , Susser Mervyn . 1982 . “Newborn Intensive Care and Neonatal Mortality in Low-Birth-Weight Infants: A Population Study.” The New England Journal of Medicine 307 : 149 - 55 . Google Scholar CrossRef Search ADS PubMed Passaro Kristi Tolo , Little Ruth E. , Savitz David A. , Noss John , and The ALSPAC Study Team . 1996 . “The Effect of Maternal Drinking Before Conception and in Early Pregnancy on Infant Birthweight.” Epidemiology 7 : 377 - 83 . Google Scholar CrossRef Search ADS PubMed Phelan Jo C. , Link Bruce G . 2005 . “Controlling Disease and Creating Disparities: A Fundamental Cause Perspective.” The Journals of Gerontology 60B : S27 – S33 . Google Scholar CrossRef Search ADS Phelan Jo C. , Link Bruce G. , Tehranifar Parisa . 2010 . “Social Conditions as Fundamental Causes of Health Inequalities Theory, Evidence, and Policy Implications.” Journal of Health and Social Behavior 51 : S28 - S40 . Google Scholar CrossRef Search ADS PubMed Phibbs Ciaran S. , Bronstein Janet M. , Buxton Eric , Phibbs Roderic H . 1996 . “The Effects of Patient Volume and Level of Care at the Hospital of Birth on Neonatal Mortality.” Journal of the American Medical Association 276 : 1054 - 59 . Google Scholar CrossRef Search ADS PubMed Reynolds Megan M. , Brady David . 2012 . “Bringing You More than the Weekend: Union Membership and Self-Rated Health in the United States.” Social Forces 90 : 1023 - 49 . Google Scholar CrossRef Search ADS Richardson Douglas K. , Gray James E. , Gortmaker Steven L. , Goldman Donald A. , Pursley DeWayne M. , McCormick Marie C . 1998 . “Declining Severity Adjusted Mortality: Evidence of Improving Neonatal Intensive Care.” Pediatrics 102 : 893 - 99 . Google Scholar CrossRef Search ADS PubMed Ross Catherine E. , Wu Chia-ling . 1995 . “The Links Between Education and Health.” American Sociological Review 60 : 719 - 45 . Google Scholar CrossRef Search ADS Ross Catherine E. , Mirowsky John . 2008 . “Neighborhood Socioeconomic Status and Health: Context or Composition?” City & Community 7 : 163 - 79 . Google Scholar CrossRef Search ADS Salihu Hamisu M. , Aliyu Muktar H. , Pierre-Louis Bosny J. , Alexander Greg R . 2003 . “Levels of Excess Infant Deaths Attributable to Maternal Smoking During Pregnancy in the United States.” Maternal and Child Health Journal 7 : 219 - 27 . Google Scholar CrossRef Search ADS PubMed Samuelson Julia L. , Buehler James W. , Norris Dianne , Sadek Ramses . 2002 . “Maternal Characteristics Associated with Place of Delivery and Neonatal Mortality Rates Among Very-Low-Birthweight Infants, Georgia.” Paediatric and Perinatal Epidemiology 16 : 305 – 13 . Google Scholar CrossRef Search ADS PubMed Schempf Ashley H. , Branum Amy M. , Lukacs Susan L. , Schoendorf Kenneth C . 2007 . “The Contribution of Preterm Birth to the Black–White Infant Mortality Gap, 1990 and 2000.” American Journal of Public Health 97 : 1255 - 60 . Google Scholar CrossRef Search ADS PubMed Shanahan Ellen , Perry Julie R. , DeClerque Julia L . 2012 . “Consensus in Region IV: Women and Infant Health Indicators for Planning and Assessment.” Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, Chapel Hill, NC. Shi Leiyu , Starfield Barbara , Kennedy Bruce , Kawachi Ichiro . 1999 . “Income Inequality, Primary Care, and Health Indicators.” Journal of Family Practice 48 ( 4 ): 275 - 84 . Google Scholar PubMed Shi Leiyu , Macinko James , Starfield Barbara , Xu Jiahong , Regan Jerrilynn , Politzer Robert , Wulu John . 2004 . “Primary Care, Infant Mortality, and Low Birth Weight in the States of the USA.” Journal of Epidemiology and Community Health 58 : 374 - 80 . Google Scholar CrossRef Search ADS PubMed Singh Gopal K. , Kogan Michael D . 2007 . “Persistent Socioeconomic Disparities in Infant, Neonatal, and Postneonatal Mortality Rates in the United States, 1969-2001.” Pediatrics 119 : e928 - e39 . Google Scholar CrossRef Search ADS PubMed Singh Gopal K. , Siahpush Mohammad . 2006 . “Widening Socioeconomic Inequalities in U.S. Life Expectancy, 1980–2000.” International Journal of Epidemiology 35 : 969 - 79 . Google Scholar CrossRef Search ADS PubMed Song Shige , Burgard Sarah A . 2011 . “Dynamics of Inequality Mother’s Education and Infant Mortality in China, 1970-2001.” Journal of Health and Social Behavior 52 : 349 - 64 . Google Scholar CrossRef Search ADS PubMed Starfield Barbara. 1985 . “Postneonatal Mortality.” Annual Review of Public Health 6 : 21 - 40 . Google Scholar CrossRef Search ADS PubMed Starfield Barbara , Shi Leiyu , Macinko James . 2005 . “Contribution of Primary Care to Health Systems and Health.” Milbank Quarterly 83 : 457 - 502 . Google Scholar CrossRef Search ADS PubMed Strully Kate W. , Rehkopf David H. , Xuan Ziming . 2010 . “Effects of Prenatal Poverty on Infant Health State Earned Income Tax Credits and Birth Weight.” American Sociological Review 75 : 534 - 62 . Google Scholar CrossRef Search ADS PubMed Subramanian S. V. , Kawachi Ichiro , Kennedy Bruce P . 2001 . “Does the State You Live in Make a Difference? Multilevel Analysis of Self-rated Health in the U . S.” Social Science & Medicine 53 : 9 - 19 . Google Scholar CrossRef Search ADS U.S. Bureau of Labor Statistics . 2012 . Local Area Unemployment Statistics . Retrieved September 13, 2017 (www.bls.gov/lau/home.htm). U.S. Census Bureau . 2012a . Annual Social and Economic Supplements. Retrieved September 28, 2017 (https://www.census.gov/did/www/saipe/data/model/info/cpsasec.html). U.S. Census Bureau . 2012b . “Small Area Income and Poverty Estimates—State and County Data.” Retrieved September 13, 2017 (www.census.gov/did/www/saipe/index.html). U.S. Centers for Disease Control and Prevention (CDC) . 2012 . “Linked Birth / Infant Death Records.” Retrieved September 13, 2017 (http://wonder.cdc.gov/lbd.html). Ventura Stephanie J. , Hamilton Brady E. , Mathews T. J. , Chandra Anjani . 2003 . “Trends and Variations in Smoking During Pregnancy and Low Birth Weight: Evidence From the Birth Certificate, 1990-2000.” Pediatrics 111 : 1176 . Google Scholar PubMed Vintzileos Anthony M. , Ananth Cande V. , Smulian John C. , Scorza William E. , Knuppel Robert A . 2002 . “The Impact of Prenatal Care on Neonatal Deaths in the Presence and Absence of Antenatal High-Risk Conditions.” American Journal of Obstetrics and Gynecology 186 : 1011 - 16 . Google Scholar CrossRef Search ADS PubMed Wildeman Christopher. 2012 . “Imprisonment and Infant Mortality.” Social Problems 59 : 228 - 57 . Google Scholar CrossRef Search ADS Wilkinson Richard G. , Pickett Kate E . 2008 . “Income Inequality and Socioeconomic Gradients in Mortality.” American Journal of Public Health 98 : 699 - 704 . Google Scholar CrossRef Search ADS PubMed Wise Paul H. 2003 . “The Anatomy of a Disparity in Infant Mortality.” Annual Review of Public Health 24 : 341 - 62 . Google Scholar CrossRef Search ADS PubMed Xu Ke Tom. 2006 . “State-Level Variations in Income-Related Inequality in Health and Health Achievement in the U.S.” Social Science & Medicine 63 : 457 - 64 . Google Scholar CrossRef Search ADS © The Author 2017. Published by Oxford University Press on behalf of the Society for the Study of Social Problems.All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Social Problems Oxford University Press

Inequality in Infant Mortality: Cross-State Variation and Medical System Institutions

Social Problems , Volume Advance Article – Oct 12, 2017

Loading next page...
 
/lp/ou_press/inequality-in-infant-mortality-cross-state-variation-and-medical-gt3Pi0rmyI
Publisher
University of California Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of the Society for the Study of Social Problems.All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0037-7791
eISSN
1533-8533
D.O.I.
10.1093/socpro/spx034
Publisher site
See Article on Publisher Site

Abstract

Abstract This article examines variation in the association between maternal education and infant mortality across the 50 U.S. states. An analysis of 22,967,018 Vital Statistics records from 1997-2002, reveals evidence of dramatic cross-state differences. In some states, infants born to mothers with less than 12 years of schooling are more than twice as likely to die as infants of mothers with 4 years of college or more. Other states see far more equality between these groups. I then evaluate two components of state medical systems that are predicted to be associated with this variation in the magnitude of inequalities in infant mortality: neonatal intensive care units and primary care physician supply. More widespread availability of neonatal intensive care is associated with reduced inequalities in infant mortality. In contrast, the supply of primary care is linked to slightly larger differences in infant mortality between mothers with low and high education. infant mortality, inequality, institutions, health disparities, United States The likelihood of infant mortality differs dramatically depending on a mother’s social position. Across sociodemographic indicators, including education, income, and race, infants born to mothers from more advantaged backgrounds experience consistently lower rates of mortality (Finch 2003; Gortmaker 1979; Hummer 1993; Hummer etal. 1999; Mathews, Menacker, and MacDorman 2004; Singh and Kogan 2007). These inequalities in infant mortality have inspired research on the role that socioeconomic resources play in promoting infant health (e.g., Blumenshine etal. 2010; Conley and Bennett 2000; Strully, Rehkopf, and Xuan 2010). This work is consistent with a broader literature on the SES-health gradient and the role of socioeconomic position as a “fundamental cause” of health outcomes (Elo 2009; Link and Phelan 1995). Research in this tradition has been extremely influential in efforts to explain the persistent association between socioeconomic position and health (e.g., Miech etal. 2011; Phelan, Link, and Tehranifar 2010). However, there is also evidence that the magnitude of this association varies substantially across different contexts (Beckfield, Olafsdottir, and Bakhtiari 2013; Chetty etal. 2016; Mackenbach etal. 2008). Moreover, there is growing interest in the institutional predictors of such variation (Beckfield etal. 2015). In the case of the United States, there is evidence of substantial cross-state differences in health disparities between socioeconomic groups (Subramanian, Kawachi, and Kennedy 2001; Xu 2006). So far, most of this work has focused on inequalities in adult health outcomes (cf. Wildeman 2012). Yet, there is reason to expect that socioeconomic inequalities in infant mortality may also vary across states and state institutions. The theory of socioeconomic position as a fundamental cause of disease highlights the fact that interventions designed to improve health outcomes often result in greater health inequalities (Phelan etal. 2010). When utilization of these interventions is not universal, those with more socioeconomic resources are typically better positioned to take advantage of them (Link etal. 1998). This principle has important implications for two state medical system institutions that have been shown to influence infant mortality: neonatal intensive care units (NICUs) and primary care physician supply (Gortmaker and Wise 1997; Shi etal. 2004; Starfield, Shi, and Macinko 2005; Wise 2003). This article contributes to the emerging literature on variation in the magnitude of health disparities and the institutional predictors of this variation by exploring two research questions: (1) To what extent do educational inequalities in infant mortality differ in magnitude across U.S. states? and (2) Is cross-state variation in inequality associated with medical system institutions? To answer these questions, I analyze Vital Statistics data on 22,967,018 births from 1997-2002. Focusing on inequalities in infant mortality between mothers with less than 12 years of education and those with 4 years of college or more,1 1 In addition to aligning with the timing of important educational credentials, these categories highlight mothers in unambiguously different social positions (Goesling 2007). I first assess the extent to which this disparity varies across states and find evidence of substantial differences. I then use random intercept logistic regression models to evaluate how state-level differences in the availability of neonatal intensive care and primary care are associated with variation in the magnitude of these inequalities. BACKGROUND Maternal Education as a Fundamental Cause of Infant Mortality Infant mortality refers to deaths that occur after a live birth and before a child reaches one year of age. Over the past three decades, rates of infant mortality in the United States fell from 10.9 deaths per 1,000 live births in 1983 to 6.05 deaths per 1,000 live births in 2011 (MacDorman, Hoyert, and Mathews 2013; National Center for Health Statistics 2012). However, even as overall mortality rates declined, infants born to mothers with less than 12 years of schooling have remained approximately twice as likely to die as infants of mothers with 16 years of education or more (Mathews etal. 2004; Singh and Kogan 2007). Research on the association between maternal education and infant health has focused on pathways from education to infant mortality. One pathway involves the economic benefits of educational attainment (Currie and Moretti 2003). Education enables individuals to qualify for high status jobs and earn higher incomes, and such material advantages provide access to a number of resources that matter for infant health (Cramer 1995; Finch 2003; Strully etal. 2010). For example, resources like proper nutrition, health insurance, prenatal care, and nontoxic environments are all linked to infant health and mortality (Abu-Saad and Fraser 2010; Currie, Greenstone, and Moretti 2011; Moss and Carver 1998; Vintzileos etal. 2002). In addition, education may provide mothers with knowledge and cognitive skills that are beneficial to infant health (Baker etal. 2011). There is evidence of a relationship between education and health-promoting behaviors such as exercise, responsible alcohol use, and not smoking (Currie and Moretti 2003; Cutler and Lleras-Muney 2010; Ross and Wu 1995; Salihu etal. 2003), and these factors are strongly associated with birth outcomes (Chen etal. 2009; Kleinman etal. 1988; Passaro etal. 1996; Ventura etal. 2003). Education may also enhance a mother’s ability to navigate the health care system and adhere to treatment regimens during her pregnancy (Goldman and Smith 2002; Hummer etal. 1999). The presence of multiple reinforcing pathways from maternal education to infant health is consistent with the theory of social conditions as a fundamental cause of disease (Link and Phelan 1995). This perspective suggests that socioeconomic position is fundamentally linked to health because it provides access to an extensive array of health-promoting resources such as money, information, social support, and network connections. When faced with health risks, individuals with these resources have more opportunities to protect themselves than those constrained by limited resources (Phelan and Link 2005). Thus, only addressing proximate risks of disease like malnutrition and obesity neglects the broader socioeconomic factors that may pattern the distribution of these risk factors in the first place (Lutfey and Freese 2005). A key implication of fundamental cause theory is that the implementation of health-promoting interventions will often result in larger health disparities between socioeconomic groups because those with more resources are better positioned to take advantage (Phelan etal. 2010). For example, highly educated individuals may have more exposure to information about medical innovations and treatments and be more able to afford the cost of such advances (Glied and Lleras-Muney 2008). This process has been shown to help explain disparities in a range of health outcomes, including inequalities in infant mortality (Chang and Lauderdale 2009; Link etal. 1998; Mechanic 2005; Song and Burgard 2011). For example, W. Parker Frisbie and colleagues (2004) examine differences in infant mortality among blacks and whites following the introduction of surfactant therapy to treat respiratory distress syndrome in 1990. Consistent with the idea that black infants would be less likely to receive surfactant therapy than white infants, they find that although the overall infant mortality rate declined after the introduction of surfactants, racial inequalities in infant mortality increased due to larger reductions among white infants. Variation in the Magnitude of Health Inequalities In recent years, scholars have documented substantial variation in the association between socioeconomic position and health across contexts. This includes cross-national differences (Beckfield etal. 2013; Mackenbach etal. 2008) and variation within the United States across states, counties, and commuting zones (Chetty etal. 2016; Singh and Siahphush 2006; Wilkinson and Pickett 2008; Xu 2006). Differences in the magnitude of health inequalities across populations highlight the need for theories that can help to explain this variation. These explanations can be divided into two general categories. Compositional explanations trace population-level variation in the extent to which socioeconomic position matters for health outcomes to demographic differences (McLeod, Nonnemaker, and Call 2004; Ross and Mirowsky 2008). For example, even after controlling for basic socioeconomic indicators, rates of infant mortality among individuals of Mexican origin living in the United States are much lower than rates for non-Hispanic blacks (Hummer etal. 2007; Hummer etal. 1999; Mathews and MacDorman 2012).2 2 Despite attempts to explain this phenomenon, it largely remains an epidemiologic paradox (Hummer etal. 2007). Thus, in states where a large proportion of those with low education are of Mexican origin, the association between maternal education and infant mortality will likely be smaller in magnitude than in states where few Mexican Americans but many African Americans are represented among those with low education. As this example demonstrates, attempts to understand state-level variation in health inequalities must account for compositional differences across states. Variation in the magnitude of health inequalities may also be a product of differences in institutional context across populations (Beckfield etal. 2015). Institutions represent the rules, policies, infrastructures, and organizations that are part of any society, and an extensive literature highlights the role of such institutions in shaping the distribution of economic resources (e.g., McCall and Percheski 2010; Moller, Nielsen, and Alderson 2009; Morris and Western 1999). Just as economic inequality can be traced to institutional structures, there is growing interest in the role of institutions in shaping health disparities (Beckfield and Krieger 2009; Olafsdottir 2007). Institutional arrangements are likely to play a key role in explaining variation in inequalities in infant mortality across U.S. states. There are notable differences in the scale and organization of state medical system institutions and states differ in the availability of medical personnel and facilities. This infrastructure can have consequences for infant health (Matteson, Burr, and Marshall 1998), and evidence suggests that characteristics of state medical systems, like the availability of health practitioners, contribute to cross-state variation in rates of infant mortality (Bird and Bauman 1995; Shi etal. 2004). So far, the relationship between state medical system institutions and health inequalities has only been subject to limited evaluation in research on variation in the magnitude of health disparities (e.g., Xu 2006). To generate hypotheses about the nature of this association, I turn to fundamental cause theory, specifically the notion that those in more advantaged social positions are better positioned to benefit from many health-promoting interventions (Glied and Lleras-Muney 2008; Goldman and Lakdwalla 2005). This represents an important process through which institutions may influence the association between socioeconomic position and health. When medical system institutions primarily benefit those with high levels of education and other socioeconomic resources, health differences between social groups can be expected to increase (Phelan etal. 2010). Moreover, a complementary (although less widely established) proposition is that institutions that broaden usage of health interventions are expected to reduce health inequalities because the advantages granted by socioeconomic resources in utilizing such interventions will be diminished (Gortmaker and Wise 1997). In this article, I build on these ideas to generate hypotheses about the relationship between state medical system institutions and inequalities in infant mortality. State Medical System Institutions and Inequalities in Infant Mortality Although research has established links between infant mortality and social institutions in U.S. states (Bird and Bauman 1995, 1998; Matteson etal. 1998),3 3 A related line of inquiry explores the role of institutions in explaining variation in infant mortality rates at the national level (Pampel and Pillai 1986). Conley and Springer (2001) provide evidence of an association between welfare state spending and infant mortality in 19 affluent nations. scholars have yet to explore the role of institutions in helping to explain state-level variation in inequalities in infant mortality. Building on fundamental cause theory and the literature on the institutional predictors of infant mortality, I identify two aspects of state medical systems that are hypothesized to influence the relationship between maternal education and infant mortality in U.S. states: neonatal intensive care facilities and primary care physician supply. Neonatal Intensive Care Advances in neonatal intensive care have been a driving force behind reductions in infant mortality in recent decades (Wise 2003). Hospital neonatal intensive care units are equipped with the technology and personnel to treat newborns whose lives are threatened by extreme prematurity, very low birth weight, illnesses, or other delivery complications. There is extensive evidence that the appropriate level of neonatal care is effective in reducing mortality among these high-risk infants (Horbar and Lucey 1995; Paneth etal. 1982; Phibbs etal. 1996; Richardson etal. 1998). Although this care is provided to all high-risk infants born in hospitals with NICUs, not all hospitals have these facilities, and there are considerable differences in NICU availability between states. In 2000, 76 percent of very low birth weight infants were delivered in the hospitals with the appropriate neonatal care facilities in Georgia compared to just 31 percent in Mississippi (Shanahan, Perry, and DeClerque 2012). Differential availability of neonatal intensive care could influence the association between socioeconomic position and infant health (Gortmaker and Wise 1997). Fundamental cause theory suggests that high levels of educational attainment will provide mothers with the resources and information to increase the likelihood that they give birth in hospitals with the facilities necessary for treating high-risk pregnancies, and available evidence supports this notion (Howell and Vert 1993; Samuelson etal. 2002). Thus, in states where NICUs are not widely available and NICU usage depends in part on educational attainment, inequality in infant mortality between education groups is expected to be larger. In contrast, in states where hospitals with NICU facilities are widespread, the disadvantage of low maternal education in securing neonatal intensive care is likely to be diminished (Eberstein, Nam, and Hummer 1990; Gortmaker and Wise 1997). Based on this theory, I predict that greater availability of neonatal intensive care units will be associated with smaller disparities in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more (H1 in Table 1). Table 1. Hypothesized Relationships Between State Medical System Institutions and Inequalities in Infant Mortality Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Table 1. Hypothesized Relationships Between State Medical System Institutions and Inequalities in Infant Mortality Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Hypothesis Predicted Relationship Measure H1 Greater state NICU availability will be associated with smaller inequalities in infant mortality between mothers with less than 12 years of education and mothers with 4 plus years of college. NICUs per 10,000 state residents H2 Greater state supply of primary care physicians will be associated with smaller inequalities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 plus years of college. Primary care physicians per 10,000 state residents Primary Care Physician Supply Another aspect of state medical systems that has been linked to infant health is the availability of primary care.4 4 This class of physician includes family and general practitioners, general internists, and general pediatricians. Research on this issue suggests that primary care influences infant mortality in several important ways. For one, primary care is linked to improved infant care practices. Primary care physicians help mothers identify and treat infections and other illnesses common in newborns. They also teach mothers about healthy practices like safe sleeping positions and safety at home and in vehicles (Shi etal. 2004). These represent some of the principal risk factors for mortality (Starfield 1985) and highlight the potential for primary care to benefit infant health. In addition, primary care influences infant mortality through its effect on maternal health. Primary care promotes reduced smoking and alcohol use, healthier sexual practices, and improved nutrition (Shi etal. 2004), and there are clear links between these behaviors and birth weight and infant mortality (Chen etal. 2009; Kleinman etal. 1988; Passaro etal. 1996; Ventura etal. 2003). Consistent with these linkages between primary care and infant health, differences in the supply of state primary care physicians per capita are associated with state-level differences in infant mortality (Shi etal. 1999; Shi etal. 2004). In addition to influencing absolute levels of infant mortality, the availability of primary care physicians in a state may be associated with inequalities in infant mortality between socioeconomic groups. Barbara Starfield, Leiyu Shi, and James Macinko (2005) report that primary care provides more substantial reductions in infant mortality in states with high social inequality than in states with lower inequality (see also Shi etal. 2004). Based on this result, they suggest that the supply of primary care physicians in a state can reduce health inequality by increasing the availability of key health services (Starfield etal. 2005). However, Starfield and colleagues (2005) also acknowledge that an increased supply of primary care physicians per capita may not guarantee more universal usage of primary care. Consistent with fundamental cause theory, the supply of primary care physicians may actually be linked to greater inequalities in infant mortality if mothers with higher education are better positioned to make use of this care due to better health coverage and other socioeconomic advantages. To help adjudicate between these predictions, I evaluate the hypothesis that disparities in the likelihood of infant mortality between mothers with less than 12 years of schooling and those with 4 years of college or more will be smaller in states where the supply of primary care physicians is greater (H2 in Table 1). DATA AND METHODS The primary data for this article are birth and infant death records from the National Vital Statistics System (NVSS). The NVSS, run by the National Center for Health Statistics, links birth and death certificates for all infants born in the United States (National Center for Health Statistics 2001-2006).5 5 In practice, the NVSS is able to successfully link almost 99 percent of infant deaths to a corresponding birth certificate. For example, in 2002, only 292 out of 27,527 infant deaths were unlinked. 6 In 2003, a substantial revision of the birth certificate was introduced. The changes included a new measure of maternal education that was deemed incompatible with the previous standard (Mathews and MacDorman 2012). States adopted this revision gradually and the process was not completed until 2015. As a result, 2002 is the last available year that all states used the same standard measure of maternal education. 7 I measure infant’s sex with a dummy variable for male infants. Plural birth is measured with a dummy variable for plural infants. Mother’s age is measured with dummy variables for each age group < 20, 20-24, 25-29 (reference category), 30-34, 35-39, and 40+. Mother’s race is measured with dummy variables for white (reference category), black, Hispanic, and other race. Birth history is measured with dummy variables for first birth, second birth (reference category), third birth, fourth birth, and fifth or more births. Maternal health condition is measured with a dummy variable indicating the presence of one or more health problems reported on birth records (measured conditions include anemia, cardiac disease, acute or chronic lung disease, diabetes, hemoglobinopathy, chronic hypertension, and renal disease). 8 There is also evidence that prenatal care may influence infant health (Vintzileos etal. 2002; cf Fiscella 1995). In supplemental analyses, I control for the receipt and timing of prenatal care. The results are robust to the inclusion of these measures. Since prenatal care is likely to intervene on the pathway between maternal education and infant mortality, I exclude this factor from the analyses shown here. I use records from births occurring in 1997-2002.6 The linked data files include information on an infant’s birth and death as well as maternal educational attainment. I code infant mortality as a dichotomous variable indicating whether an infant died in the first year of life. The measure of maternal education includes four categories of educational attainment: less than 12 years of schooling, 12 years of schooling, less than 4 years of college, and 4 years of college or more. In my analyses, I employ dummy variables for each education category (with less than 12 years of schooling serving as the reference category). The linked birth-death records include information on a number of additional infant and maternal characteristics. Here, I use measures of infant’s sex, plural birth status, maternal age, maternal race, maternal birth history, and maternal health conditions in order to control for factors relevant to infant mortality risk (Mathews etal. 2004).7 For example, a mother’s age, race, and health status have the potential to influence both her educational attainment and birth outcomes. Controlling for these potential confounders helps reduce bias in estimates of the association between maternal education and infant mortality.8 Moreover, controlling for factors like race and maternal age helps account for the role of demographic composition in driving state-level variation in the extent to which maternal education matters for infant mortality. I combine the six years of linked birth-death records with state-level data on medical systems. I measure the availability of neonatal intensive care with a measure of NICUs per 10,000 state residents (American Hospital Association 1997-2004) I measure primary care supply as primary care physicians per 10,000 residents (American Medical Association 1997-2004). In addition, I control for state-level characteristics that have the potential to influence the relationship between educational inequalities in infant mortality and state medical systems. First, I control for the number of hospitals per 10,000 state residents to ensure that the measures of NICU availability and primary care supply capture the effects of these institutions beyond the effects of a state's hospital infrastructure. In addition, I control for the state infant mortality rate to account for the broader infant health context. Finally, I control for a set of variables that capture socioeconomic conditions at the state level. These include the log of per-capita GDP, income inequality (measured with the Gini coefficient), the unemployment rate, and the poverty rate.9 9 Data on state GDP come from the Bureau of Economic Analysis (2012). Data on state Gini coefficients come from the U.S. Census Bureau’s Annual Social and Economic Supplements (2012). State unemployment data come from the U.S. Bureau of Labor Statistics’ Local Area Unemployment Statistics program (2012). State poverty data come from the Census Bureau’s Small Area Income and Poverty Estimates program (2012b). Data on state hospitals come from the American Hospital Association (1997-2004). State infant mortality data come from the CDC’s WONDER online database (2012). Descriptive statistics for all variables are displayed in Table 2. All results are based on unweighted data, but the use of weights that account for unlinked death records does not substantively change the results presented here. Table 2. Descriptive Statistics for All Individual and State-Level Variables, 1997-2002 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Table 2. Descriptive Statistics for All Individual and State-Level Variables, 1997-2002 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 Variable Mean SD Individual-level variables  Infant died .007 .081  Less than 12 years of schooling .218 .413  12 years of schooling .320 .467  Less than 4 years college .219 .413  4+ years of college .243 .429  Male .512 .500  Plural birth .031 .173  Maternal health conditions, 1+ .074 .262  White .596 .491  Black .148 .355  Hispanic .200 .400  Other race .055 .228  Maternal age < 20 .118 .323  Maternal age 20-24 .250 .433  Maternal age 25-29 .270 .444  Maternal age 30-34 .230 .421  Maternal age 35-39 .110 .312  Maternal age 40+ .023 .145  1st birth .402 .490  2nd birth .326 .469  3rd birth .166 .373  4th birth .064 .245  5th birth+ .041 .199 State-level variables  GDP per capita (logged) 10.433 .134  Gini coefficient .456 .024  Unemployment rate 4.741 1.027  Poverty rate 12.270 2.703  Infant mortality rate (per 1,000 live births) 7.016 1.289  Hospitals per 10,000 .178 .084 State medical system institutions  NICUs per 10,000 .030 .010  Primary care physicians per 10,000 9.009 1.880 N = 22,967,018 The analysis proceeds in three parts. I first investigate potential variation in the association between maternal education and infant mortality across U.S. states. I measure this association in each state with separate logistic regression models for all 50 states. Each model includes dummy variables for each category of maternal education and also controls for race, maternal age, sex, plural birth, maternal birth history, and maternal health conditions. The models pool data from 1997-2002 and include year fixed effects to account for time trends. Based on these models, I calculate the predicted probability of mortality for infants of mothers from two education groups in each state: those with less than 12 years of schooling and those with 4 years of college or more. When calculating predicted probabilities, I hold the values of control variables constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. Infants with these characteristics are the least likely to experience mortality (Mathews etal. 2004). Thus, this represents a conservative assessment of the level of infant mortality risk. In the second part of the analysis, I combine the individual infant birth-death records with the state-level institutional measures. This results in a multi-level data set with 22,967,018 individual records clustered in 50 states over six years. Using these data, I analyze the association between maternal education and infant mortality using logistic regression models with random intercepts for each of the 50 states. These models are well-suited for analysis of cross-state institutional differences because they account for variation both within and across states. Year fixed effects account for any national-level time trends. In order to assess whether the association between maternal education and infant mortality varies across state medical system institutions, I introduce cross-level interaction terms in which the indicator variables for maternal education are allowed to interact with the measures of NICU availability and primary care supply. I focus on the interaction between these medical system variables and the indicator for four plus years of college in order to explore inequalities in infant mortality between mothers with less than 12 years of schooling and 4 years of college or more. Finally, after evaluating whether institutional measures are significantly associated with educational inequalities in infant mortality, I present graphs showing how the predicted probability of infant mortality varies across the observed levels of neonatal intensive care and primary care for mothers with less than 12 years of schooling and those with 4 years of college or more. Graphs are generated based on the random effects logistic regression models of infant mortality on maternal education detailed above. For each graph, individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level controls and institutional variables are held constant at their mean values from 1997-2002. RESULTS I begin the analysis by exploring the possibility that educational inequalities in infant mortality vary in magnitude across U.S. states. Based on a series of state-specific logistic regression models of infant mortality on maternal education, I calculate the predicted probability of mortality for infants of mothers from two education groups in each state: those with less than 12 years of schooling and those with 4 years of college or more.10 10 Results of these analyses are displayed in Appendix Figure A1. In order to assess inequalities between these education groups, I calculate the relative risk ratio as the probability of infant mortality for mothers with low education over the probability for highly educated mothers.11 11 I also compare this measure of inequality to a measure based on absolute differences in the probability of infant mortality between mothers with low and high education. The correlation between the absolute and relative measures of inequality is .89 and the geographic patterning is similar across states.Figure 1 maps the relative risk for each state and displays considerable variation in the magnitude of inequality across states. In some states (including Alaska, North Dakota, and Kentucky), infants born to mothers with less than 12 years of schooling are more than twice as likely as those born to mothers with 4 years or more of college to die. In other states, the risk ratios are as low as 1.3 (and as low as 1.13 in Hawaii). This analysis controls for key compositional factors including race, suggesting that population demographics do not fully account for cross-state variation in the extent to which maternal education is associated with infant mortality. Figure 1. View largeDownload slide Relative Risk of Infant Mortality by Maternal Education—Less than 12 Years of Schooling vs. 4 Years of College or More, 1997-2002 (with controls) Note: Risk ratios of infant mortality by education for non-Hispanic white, non-plural, second-born daughters of mothers age 25–29 with no prior health conditions. Figure 1. View largeDownload slide Relative Risk of Infant Mortality by Maternal Education—Less than 12 Years of Schooling vs. 4 Years of College or More, 1997-2002 (with controls) Note: Risk ratios of infant mortality by education for non-Hispanic white, non-plural, second-born daughters of mothers age 25–29 with no prior health conditions. I then evaluate whether these cross-state differences in inequality in infant mortality are statistically significant. I estimate a model that combines observations from all 50 states and includes maternal education, sociodemographic controls, state and year dummy variables, and state * maternal education interactions (not shown). I assess whether the association between maternal education and infant mortality differs significantly across states with a joint F-test of the null hypothesis that the state*maternal education interaction coefficients are all equal to each other. This hypothesis can be rejected (p < .0001), demonstrating significant cross-state differences in the effect of education on infant mortality. In addition, I assess whether the effects of maternal education on infant mortality depend on the state with a joint F-test of the null hypothesis that the interaction coefficients are all equal to 0. This hypothesis can also be rejected (p < .0001), providing evidence of a significant role of state-level factors in the association between maternal education and infant mortality.12 12 Unlike the state-specific models, the effects of the sociodemographic control variables are necessarily assumed to be the same across states in this combined model. 13 All results are robust to the exclusion of all non-significant state-level control variables. After highlighting significant differences in the magnitude of inequalities in infant mortality across states, I evaluate hypotheses about the role of state medical system institutions that are predicted to be associated with this variation using a series of random effects logit models. Table 3 displays the results of this analysis using odds ratios. Model 1 of Table 3 presents a baseline analysis of maternal education and infant mortality from 1997-2002 that controls for key individual and state-level factors. This reveals a clear education-mortality gradient, with each increasing level of education reducing the odds of infant mortality relative to mothers with less than 12 years of schooling. Odds ratios for the individual-level control variables are signed in the expected direction with black, male, plural infants of mothers with existing health conditions having higher odds of mortality relative to each reference category. Of the state-level control variables, only the odds ratios for the infant mortality rate and hospitals per 10,000 are significant, though the lack of significant predictors at the state-level is not surprising given that the state random effects accounts for much of the state-level variation.13 Table 3. Logistic Regression of Infant Mortality on Maternal Education and State Medical System Institutions, 1997-2002 (state random intercepts and year fixed effects) Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Notes: Models also include dummy variables for birth history and mother’s age. Standard errors in parentheses. * p < .05 **p < .01 (two-tailed tests) Table 3. Logistic Regression of Infant Mortality on Maternal Education and State Medical System Institutions, 1997-2002 (state random intercepts and year fixed effects) Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Model 1 Model 2 Model 3 Model 4 Maternal education  Less than 12 years of schooling (reference)  12 years of schooling .852** .852** .874** .811** (.006) (.006) (.017) (.027)  Less than 4 years college .697** .697** .706** .727** (.006) (.006) (.016) (.028)  4+ years of college .539** .539** .513** .610** (.005) (.005) (.013) (.025) Infant/mother characteristics  Male (reference female) 1.229** 1.229** 1.229** 1.229** (.006) (.006) (.006) (.006)  Plural birth (reference singleton birth) 5.713** 5.713** 5.713** 5.714** (.044) (.044) (.044) (.044)  Maternal health conditions (1+) 1.099** 1.099** 1.099** 1.099** (.010) (.010) (.010) (.010)  Black (reference white) 1.992** 1.992** 1.992** 1.991** (.013) (.013) (.013) (.013)  Hispanic .902** .902** .902** .902** (.008) (.008) (.008) (.008)  Other race 1.050** 1.050** 1.050** 1.050** (.014) (.014) (.014) (.014) State-level controls  GDP per capita (logged) .949 .984 .982 .984 (.075) (.077) (.075) (.075)  Gini coefficient .757 .801 .802 .800 (.155) (.167) (.167) (.166)  Unemployment rate .988 .989 .989 .988 (.007) (.007) (.007) (.007)  Poverty rate .992 .992* .992* .992* (.004) (.004) (.004) (.004)  Infant mortality rate 1.096** 1.097** 1.097** 1.096** (.007) (.007) (.007) (.007)  Hospitals per 10,000 1.488** 1.589** 1.582** 1.586** (.005) (.139) (.137) (.137) Medical system institutions  NICUs per 10,000 .263* .297 .261* (.145) (.201) (.143)  Primary care physicians per 10,000 .994 .992 .994 (.006) (.006) (.003) Medical system *maternal education interactions  NICUs * 12 years of schooling .419 (.249)  NICUs * < 4 years college .679 (.465)  NICUs * 4+ years of college 5.412* (4.129)  Primary care physicians * 12 years of schooling 1.006 (.004)  Primary care physicians * < 4 years of college .995 (.004)  Primary care physicians * 4+ years of college .987** (.004) Constant .006** .004** .004** .004** (.005) (.003) (.003) (.003) ρ .001 .001 .001 .001 States 50 50 50 50 Infants 22,967,018 22,967,018 22,967,018 22,967,018 Notes: Models also include dummy variables for birth history and mother’s age. Standard errors in parentheses. * p < .05 **p < .01 (two-tailed tests) Model 2 of Table 3 adds the two key measures of state medical systems: NICUs per 10,000 state residents and primary care physicians per 10,000 residents. Both measures are negatively associated with infant mortality, and while the odds ratio for NICUs per 10,000 is significant at the .05 level, the odds ratio for primary care physicians per 10,000 is not significant. In order to assess whether the association between maternal education and infant mortality varies across state medical system institutions, I introduce a series of cross-level interaction terms in which the indicator variables for maternal education are allowed to interact with the measures of NICU availability (Model 3) and primary care supply (Model 4). I focus on the interaction term for mothers with four plus years of college because this value reflects the extent to which the disparity in infant mortality between mothers with less than 12 years of schooling and those with 4 years of college or more varies with the state medical system variables. In Model 3, the odds ratio of 5.412 is equivalent to a logit coefficient of 1.689, which is positively signed and significant. This indicates that compared to mothers with 4 or more years of college, the NICU effect is greater among mothers with less than 12 years of schooling. Model 4 switches the focus to the primary care supply and includes cross-level interactions between primary care physicians per 10,000 residents and maternal education. The odds ratio of .987 for the interaction between primary care physicians per 10,000 and four plus years of college corresponds with a logit coefficient that is negatively signed and significant (-.013), indicating that the effect of primary care supply is greater among mothers with at least 4 years of college than among mothers with less than 12 years of schooling. The analyses in Table 3 provide evidence that state medical system institutions moderate the association between maternal education and infant mortality. To offer a more detailed picture of this relationship, I present graphs showing how the predicted probability of under-five mortality varies with the level of neonatal intensive care and primary care for infants from these two groups when other key factors are held constant. Figure 2 illustrates how the predicted probability of infant mortality changes over the observed range of NICU facilities per 10,000 state residents for mothers with less than 12 years of schooling and mothers with 4 years of college or more. With more NICUs, inequality in the predicted probability of infant mortality is reduced. This reduction is driven by the negative relationship between NICU availability and infant mortality among mothers with low educational attainment. For example, at a very low level of NICU availability like .01 NICUs per 10,000, the predicted probability of mortality for infants of mothers with less than 12 years of schooling is .0053. At a high level of NICU availability like .09 NICUs per 10,000, this probability declines to .0048. In contrast, the predicted probability for infants of mothers with 4 years of college or more is .0028 with .01 NICUs per 10,000 and .0029 with .09 NICUs per 10,000. This means that for mothers with less than 12 years of schooling, there is one fewer death for every 2,000 live births in states with .09 NICUs per 10,000 compared to states with .01 NICUs per 10,000. For mothers with 4 years of college or more, there is one more death per 10,000 live births at greater levels of state NICU availability. While this difference in inequality is small in absolute terms, the relative risk of infant mortality between these education groups declines 12.5 percent, from 1.91 to 1.67 when comparing low versus high NICU availability. This supports H1 that state neonatal intensive care availability will be negatively associated with inequality in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more. Figure 2. View largeDownload slide Predicted Probability of Infant Mortality by NICUs per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 2. View largeDownload slide Predicted Probability of Infant Mortality by NICUs per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 3 shows how educational inequality in infant mortality varies over the observed range of primary care physicians per 10,000 state residents. A greater primary care physician supply is linked to reductions in the predicted probability of infant mortality for both mothers with less than 12 years of schooling and mothers with 4 years of college or more. However, this reduction is more pronounced for mothers with at least a college education. For example, expressed in absolute terms, mothers in the highly educated group experience 1.1 fewer deaths for every 3,000 live births among in states with 7 primary care physicians per 10,000 (a low level of primary care supply) compared to states with 13 primary care physicians per 10,000 (a high primary care supply). In contrast, mothers with low education experience .9 fewer deaths for every 3,000 births at these levels of primary care. This small absolute change in inequality represents a modest relative difference. At 7 primary care physicians per 10,000, the relative risk of infant mortality between these groups is 1.80. At 13 primary care physicians per 10,000, the relative risk increases 8 percent to 1.95. Thus, inequality in the predicted probability of infant mortality is slightly larger where there are more primary care physicians. This contradicts H2 that state primary care supply will be negatively associated with inequality in infant mortality between mothers with less than 12 years of education and mothers with 4 years of college or more. Figure 3. View largeDownload slide Predicted Probability of Infant Mortality by Primary Care Physicians per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. Figure 3. View largeDownload slide Predicted Probability of Infant Mortality by Primary Care Physicians per 10,000, 1997-2002 Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. State-level variables are held constant at their mean values from 1997-2002. DISCUSSION This article explores two research questions about the relationship between maternal education and infant mortality in U.S. states. I first evaluate the extent to which the magnitude of this association varies across states and find evidence of substantial differences. Infants born to mothers with less than 12 years of schooling are more than twice as likely to die as infants born to mothers with four or more years of college (even after accounting for factors like race, maternal age, and birth history). In contrast, other states see minimal differences in the risk of infant mortality by maternal education. This analysis contributes to the growing body of research that documents variation in the relationship between socioeconomic position and health across populations (e.g., Beckfield etal. 2013; Chetty etal. 2016; Mackenbach etal. 2008) with the first study focused on cross-state differences in educational inequalities in infant mortality. The presence of dramatic cross-state differences in inequality in infant mortality calls attention to the institutional predictors of this variation. In the second portion of the analysis, I explore whether variation in inequality is associated with state medical system institutions. I focus on two components of state medical systems that have proven relevant to infant mortality: neonatal intensive care availability and primary care supply. Greater NICU availability is associated with a reduction in the risk of infant mortality among mothers with less than 12 years of education. In contrast, the probability of infant mortality does not vary with neonatal intensive care for mothers with 4 years of college or more. Thus, with more NICUs per 10,000, inequalities in the risk of infant mortality between these groups are smaller in magnitude. A possible explanation for this negative association between the availability of state neonatal intensive care and educational inequalities in infant mortality is that the importance of maternal socioeconomic resources in accessing this care varies based on the availability of NICUs. Existing evidence suggests that individuals with more resources are advantaged in making use of health interventions (Chang and Lauderdale 2009; Link and Phelan 1995). Thus, in states where hospitals with NICUs are not widely available, education, health insurance, and other resources are likely to play an important role in helping mothers locate and travel to the hospitals equipped with these facilities (Howell and Vert 1993; Samuelson etal. 2002). Mothers without these resources face more obstacles to receiving this care, which likely contributes to the observed disparity in infant mortality in states with low NICU availability. However, I also find that inequality in infant mortality is reduced in states with more NICUs per 10,000 residents, which suggests that the advantages provided by education in securing neonatal intensive care are diminished in contexts where NICU facilities are widely available and there are fewer barriers to accessing NICU care. This may reflect the nature of NICUs—unlike many other health interventions, hospitals equipped with NICUs provide care to all infants in need of these facilities, regardless of their families’ socioeconomic resources. In contrast, greater primary care physician supply is associated with slightly larger disparities in infant mortality between maternal education groups. Mothers with less than 12 years of schooling and those with 4 years of college or more both face lower absolute risk of infant mortality in states with more primary care physicians per 10,000 residents, but this reduction is more pronounced among mothers with a college education. While this finding does not support the hypothesis that primary care leads to reductions in health inequality (Starfield etal. 2005), it is consistent with the notion that health-promoting interventions can increase health disparities between social groups (Phelan and Link 2005). Unlike NICUs, which provide care to all at-risk infants who are born in hospitals with these facilities, an extensive supply of primary care physicians does not necessarily broaden usage (Matteson etal. 1998). Instead, making use of primary care requires health insurance coverage, the ability to attend regular medical appointments, and information on the benefits of this care. Thus, mothers with higher levels of education are likely to be better positioned to benefit from a greater supply of primary care physicians. This calls attention to a key difference between these two components of state medical systems, and this difference has implications for both theory and public policy. The finding that primary care is associated with greater health inequalities is consistent with the fundamental cause perspective because it highlights how interventions that improve health outcomes can also exacerbate health inequalities (Phelan etal. 2010). However, the link between NICU availability and smaller disparities in infant mortality by maternal education highlights an important implication of fundamental cause theory that has received insufficient attention. While evaluations of this theory have focused on the role of health interventions in increasing inequality (Chang and Lauderdale 2009; Glied and Lleras-Muney 2008; Song and Burgard 2011), the results presented here suggest that interventions like NICUs that broaden usage of medical services may be able to reduce health inequalities by minimizing the health risks of a disadvantaged socioeconomic position (Gortmaker and Wise 1997). This is a notable extension of research on social position as a fundamental cause of health outcomes because it suggests that the salience of socioeconomic resources for an individual’s health varies based on the institutional context. When health-promoting institutions are not widely accessible, socioeconomic resources appear to play a more central role in determining health outcomes. Yet, when institutions broaden usage of health services and interventions, the importance of socioeconomic resources is diminished. This is a key pathway through which social institutions can shape health disparities and represents a promising starting point for future efforts to understand differences in the nature of health disparities across contexts. This analysis also highlights the power of social policy to shape health inequalities (Beckfield and Krieger 2009). Even beyond the potential value of investments in medical systems that further expand the availability of NICUs, the results suggest that policies that expand usage of health interventions and services among socioeconomically disadvantaged individuals can be effective in reducing health disparities. For example, policies that broaden usage of primary care might enable mothers with low education to receive the same infant health benefits of this service as highly educated mothers. If so, then policy reforms like the 2010 Affordable Care Act (ACA), which expanded health insurance coverage to many individuals who could not previously afford care, have the potential to play a key role in efforts to reduce health differences between social groups. Moreover, the differential adoption of the ACA’s Medicaid expansion across states suggests that there will likely be considerable differences in the law’s impact at the state level (Blumenthal and Collins 2014). This variation provides an opportunity to evaluate the effects of large-scale institutional change on the relationship between socioeconomic position and health and represents a promising opportunity for future research. As scholars continue to study the relationship between state institutions and inequalities in infant mortality, several additional issues stand out as particularly important for further exploration. A limitation of this article is that the state-level analysis prevents an examination of the geographic distribution of medical system institutions within states. While the number of NICUs per 10,000 residents provides a broad measure of the availability of neonatal intensive care, it does not account for the fact that NICUs may not be proportionally distributed within states. As a result, future research should seek to incorporate more detailed geographic data on the location of NICUs in order to evaluate whether within-state differences in NICU availability influence inequalities in infant mortality. Another useful way to extend the analysis would be to account for the process by which infants are selected into neonatal intensive care. Pregnant mothers in disadvantaged socioeconomic positions face a multitude of health risks (e.g., malnutrition, unsafe housing, lack of prenatal care), and their infants are more likely to be born prematurely and at low birth weight (Blumenshine etal. 2010). If the greater likelihood of being born in these vulnerable conditions results in a greater likelihood of treatment in NICU facilities, then the observed association between state NICU availability and inequalities in infant mortality could partially reflect this selection effect. In supplemental analyses, I control for birth weight in order to account for a measure of health status at birth that predicts NICU entry. Although the results are robust to this alternative specification,14 14 I control for birth weight with dummy variables for infants who are born at low (2,500-1,500 grams) or very low (< 2,500 grams) weight. Results of this analysis are available upon request. birth weight is an imperfect measure of an infant’s health status at birth (Schempf etal. 2007). This highlights the value of collecting more detailed data on infant health at birth and the care infants receive after being born in order to analyze the extent to which infants are selected into NICUs. In addition, while existing research has focused on state medical systems as key institutional predictors of infant mortality, there may also be institutions that operate outside the medical system that influence the relationship between socioeconomic position and health (Beckfield and Krieger 2009). For example, institutions like labor unions and welfare policies have been shown to influence a multitude of health outcomes (e.g., Cho 2011; Reynolds and Brady 2012). Research that explores the extent to which these and other political institutions have distinct effects on individuals in different social positions represents an important next step in the study of health inequalities. The presence of state-level variation in the association between maternal education and infant mortality shows that socioeconomic position varies in its importance as a predictor of health outcomes across contexts. This draws attention to the role of institutions in the production of health inequalities, and the results presented here suggest that efforts to reduce inequality should focus on institutions that broaden the usage of health interventions. The author wishes to thank Jason Beckfield, Bruce Western, Sandy Jencks, Alexandra Killewald, Marie McCormick, Jack Shonkoff, Benjamin Sommers, Rourke O’Brien, and Chong-Min Fu for help and suggestions. Direct correspondence to: Benjamin Sosnaud, Department of Sociology and Anthropology, Trinity University, San Antonio, TX 78212. E-mail: bsosnaud@trinity.edu. APPENDIX Figure A1. View largeDownload slide Predicted Probability of Infant Mortality and 95 Percent Confidence Intervals for Mothers with Low and High Education by State, 1997-2002 (with controls) Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. Figure A1. View largeDownload slide Predicted Probability of Infant Mortality and 95 Percent Confidence Intervals for Mothers with Low and High Education by State, 1997-2002 (with controls) Notes: Individual-level characteristics are held constant as non-Hispanic white, non-plural, second-born daughters of mothers age 25-29 with no prior health conditions. REFERENCES Abu-Saad Kathleen , Fraser Drora . 2010 . “Maternal Nutrition and Birth Outcomes.” Epidemiologic Reviews 32 : 5 - 25 . Google Scholar CrossRef Search ADS PubMed American Hospital Association . 1997-2004 . Hospital Statistics . Chicago : Health Forum LLC . American Medical Association . 1997-2004 . Physician Characteristics and Distribution in the United States . Chicago : American Medical Association . Baker David P. , Leon Juan , Smith Greenaway Emily G. , Collins John , Movit Marcela . 2011 . “The Education Effect on Population Health: A Reassessment.” Population and Development Review 37 : 307 - 32 . Google Scholar CrossRef Search ADS PubMed Beckfield Jason , Bambra Clare , Eikemo Terje A. , Huijts Tim , McNamara Courtney , Wendt Claus . 2015 . “An Institutional Theory of Welfare State Effects on the Distribution of Population Health.” Social Theory & Health 13 : 227 - 44 . Google Scholar CrossRef Search ADS Beckfield Jason , Krieger Nancy . 2009 . “Epi + demos + cracy: Linking Political Systems and Priorities to the Magnitude of Health Inequities—Evidence, Gaps, and a Research Agenda.” Epidemiologic Reviews 31 : 152 - 77 . Google Scholar CrossRef Search ADS PubMed Beckfield Jason , Olafsdottir Sigrun , Bakhtiari Elyas . 2013 . “Health Inequalities in Global Context.” American Behavioral Scientist 57 : 1014 - 39 . Google Scholar CrossRef Search ADS Bird Sheryl Thorburn , Bauman Karl E . 1995 . “The Relationship Between Structural and Health Services Variables and State-Level Infant Mortality in the United States.” American Journal of Public Health 85 : 26 - 29 . Google Scholar CrossRef Search ADS PubMed Bird Sheryl Thorburn , Bauman Karl E. 1998 . “State-Level Infant, Neonatal, and Postneonatal Mortality: The Contribution of Selected Structural Socioeconomic Variables.” International Journal of Health Services 28 : 13 - 27 . Google Scholar CrossRef Search ADS PubMed Blumenthal David , Collins Sara R . 2014 . “Health Care Coverage under the Affordable Care Act—A Progress Report.” New England Journal of Medicine 371 : 275 - 81 . Google Scholar CrossRef Search ADS PubMed Blumenshine Philip , Egerter Susan , Barclay Colleen J. , Cubbin Catherine , Braveman Paula A . 2010 . “Socioeconomic Disparities in Adverse Birth Outcomes: A Systematic Review.” American Journal of Preventive Medicine 39 : 263 - 72 . Google Scholar CrossRef Search ADS PubMed Bureau of Economic Analysis . 2012 . Regional Economic Accounts . Retrieved September 13, 2017 (www.bea.gov/regional/). Chang Virginia W. , Lauderdale Diane S . 2009 . “Fundamental Cause Theory, Technological Innovation, and Health Disparities: The Case of Cholesterol in the Era of Statins.” Journal of Health and Social Behavior 50 : 245 - 60 . Google Scholar CrossRef Search ADS PubMed Chen Aimin , Feresu Shingairai A. , Fernandez Cristina , Rogan Walter J . 2009 . “Maternal Obesity and the Risk of Infant Death in the United States.” Epidemiology 20 : 74 - 81 . Google Scholar CrossRef Search ADS PubMed Chetty Raj , Stepner Michael , Abraham Sarah , Lin Shelby , Scuderi Benjamin , Turner Nicholas , Bergeron Augustin , Cutler David . 2016 . “The Association Between Income and Life Expectancy in the United States, 2001–2014.” Journal of the American Medical Association 315 : 1750 - 66 . Google Scholar CrossRef Search ADS PubMed Cho Rosa M. 2011 . “Effects of Welfare Reform Policies on Mexican Immigrants’ Infant Mortality Rates.” Social Science Research 40 : 641 - 53 . Google Scholar CrossRef Search ADS Conley Dalton , Springer Kristen W . 2001 . “Welfare State and Infant Mortality.” American Journal of Sociology 107 : 768 - 807 . Google Scholar CrossRef Search ADS Conley Dalton , Bennett Neil G . 2000 . “Is Biology Destiny? Birth Weight and Life Chances.” American Sociological Review 65 : 458 - 67 . Google Scholar CrossRef Search ADS Cramer James C. 1995 . “Racial and Ethnic Differences in Birthweight: The Role of Income and Financial Assistance.” Demography 32 : 231 - 47 . Google Scholar CrossRef Search ADS PubMed Currie Janet , Moretti Enrico . 2003 . “Mother’s Education and the Intergenerational Transmission of Human Capital: Evidence from College Openings.” Quarterly Journal of Economics 118 : 1495 - 1532 . Google Scholar CrossRef Search ADS Currie Janet , Greenstone Michael , Moretti Enrico . 2011 . “Superfund Cleanups and Infant Health.” American Economic Review 101 : 435 - 41 . Google Scholar CrossRef Search ADS PubMed Cutler David M. , Lleras-Muney Adriana . 2010 . “Understanding Differences in Health Behaviors by Education.” Journal of Health Economics 29 : 1 - 28 . Google Scholar CrossRef Search ADS PubMed Eberstein Isaac , Nam Charles , Hummer Robert . 1990 . “Infant Mortality by Cause of Death: Main and Interaction Effects.” Demography 27 : 413 - 30 . Google Scholar CrossRef Search ADS PubMed Elo Irma T. 2009 . “Social Class Differentials in Health and Mortality: Patterns and Explanations in Comparative Perspective.” Annual Review of Sociology 35 : 553 - 72 . Google Scholar CrossRef Search ADS Finch Brian Karl. 2003 . “Early Origins of the Gradient: The Relationship Between Socioeconomic Status and Infant Mortality in the United States.” Demography 40 : 675 - 99 . Google Scholar CrossRef Search ADS PubMed Fiscella Kevin. 1995 . “Does Prenatal Care Improve Birth Outcomes? A Critical Review.” Obstetrics & Gynecology 85 : 468 - 79 . Google Scholar CrossRef Search ADS Frisbie W. Parker , Song Seung-eun , Powers Daniel A. , Street Julie A . 2004 . “The Increasing Racial Disparity in Infant Mortality: Respiratory Distress Syndrome and Other Causes.” Demography 41 : 773 - 800 . Google Scholar CrossRef Search ADS PubMed Glied Sherry , Lleras-Muney Adriana . 2008 . “Technological Innovation and Inequality in Health.” Demography 45 : 741 - 61 . Google Scholar CrossRef Search ADS PubMed Goesling Brian. 2007 . “The Rising Significance of Education for Health?” Social Forces 85 : 1621 - 44 . Google Scholar CrossRef Search ADS Goldman Dana P. , Lakdawalla Darius N . 2005 . “A Theory of Health Disparities and Medical Technology.” Contributions in Economic Analysis & Policy 4 : 1 - 30 . Google Scholar CrossRef Search ADS Goldman Dana P. , Smith James P . 2002 . “Can Patient Self-Management Help Explain the SES Health Gradient?” Proceedings of the National Academy of Sciences 99 : 10929 - 34 . Google Scholar CrossRef Search ADS Gortmaker Steven L. 1979 . “Poverty and Infant Mortality in the United States.” American Sociological Review 44 : 280 - 97 . Google Scholar CrossRef Search ADS PubMed Gortmaker Steven L. , Wise Paul H . 1997 . “The First Injustice: Socioeconomic Disparities, Health Services Technology, and Infant Mortality.” Annual Review of Sociology 23 : 147 - 70 . Google Scholar CrossRef Search ADS PubMed Horbar Jeffrey D. , Lucey Jerold F . 1995 . “Evaluation of Neonatal Intensive Care Technologies.” The Future of Children 5 : 139 - 61 . Google Scholar CrossRef Search ADS PubMed Howell Embry Martin , Vert Paul . 1993 . “Neonatal Intensive Care and Birth Weight-Specific Perinatal Mortality in Michigan and Lorraine.” Pediatrics 91 : 464 - 69 . Google Scholar PubMed Hummer Robert A. 1993 . “Racial Differentials in Infant Mortality in the U.S.: An Examination of Social and Health Determinants.” Social Forces 72 : 529 - 54 . Google Scholar CrossRef Search ADS Hummer Robert A. , Powers Daniel A. , Pullum Starling G. , Gossman Ginger L. , Frisbie W. Parker . 2007 . “Paradox Found (Again): Infant Mortality Among the Mexican-Origin Population in the United States.” Demography 44 : 441 - 57 . Google Scholar CrossRef Search ADS PubMed Hummer Robert A. , Biegler Monique , De Turk Peter B. , Forbes Douglas , Frisbie W. Parker , Hong Ying , Pullum Starling G . 1999 . “Race/Ethnicity, Nativity, and Infant Mortality in the United States.” Social Forces 77 : 1083 - 1117 . Google Scholar CrossRef Search ADS Kleinman Joel C. , Pierre Mitchell B. , Madans Jennifer H. , Land Garland H. , Schramm Wayne F . 1988 . “The Effects of Maternal Smoking on Fetal and Infant Mortality.” American Journal of Epidemiology 127 : 274 - 82 . Google Scholar CrossRef Search ADS PubMed Link Bruce G. , Phelan Jo . 1995 . “Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior 35 : 80 - 94 . Google Scholar CrossRef Search ADS Link Bruce G. , Northridge Mary E. , Phelan Jo C. , Ganz Michael L . 1998 . “Social Epidemiology and the Fundamental Cause Concept: On the Structuring of Effective Cancer Screens by Socioeconomic Status.” Milbank Quarterly 76 : 375 - 402 . Google Scholar CrossRef Search ADS PubMed Lutfey Karen , Freese Jeremy . 2005 . “Toward Some Fundamentals of Fundamental Causality: Socioeconomic Status and Health in the Routine Clinic Visit for Diabetes.” American Journal of Sociology 110 : 1326 - 72 . Google Scholar CrossRef Search ADS MacDorman Marian F. , Hoyert Donna L. , Mathews T. J . 2013 . Recent Declines in Infant Mortality in the United States 2005-2011. Hyattsville, MD : National Center for Health Statistics . Mackenbach Johan P. , Stirbu Irina , Roskam Albert-Jan , Schaap Maartje M. , Menvielle Gwenn , Leinsalu Mall , Kunst Anton E . 2008 . “Socioeconomic Inequalities in Health in 22 European Countries.” The New England Journal of Medicine 358 : 2468 - 81 . Google Scholar CrossRef Search ADS PubMed Mathews T. J. , Menacker Fay , MacDorman Marian F . 2004 . Infant Mortality Statistics from the 2002 Period: Linked Birth/Infant Death Data Set. Hyattsville, MD : National Center for Health Statistics . Mathews T. J. , MacDorman Marian F . 2012 . Infant Mortality Statistics from the 2008 Period: Linked Birth/Infant Death Data Set. Hyattsville, MD : National Center for Health Statistics . Matteson Donald W. , Burr Jeffrey A. , Marshall James R . 1998 . “Infant Mortality: A Multi-Level Analysis of Individual and Community Risk Factors.” Social Science & Medicine 47 : 1841 - 54 . Google Scholar CrossRef Search ADS McCall Leslie , Percheski Christine . 2010 . “Income Inequality: New Trends and Research Directions.” Annual Review of Sociology 36 : 329 - 47 . Google Scholar CrossRef Search ADS McLeod Jane D. , Nonnemaker James M. , Call Kathleen Thiede . 2004 . “Income Inequality, Race, and Child Well-Being: An Aggregate Analysis in the 50 United States.” Journal of Health and Social Behavior 45 : 249 - 64 . Google Scholar CrossRef Search ADS PubMed Mechanic David. 2005 . “Policy Challenges in Addressing Racial Disparities and Improving Population Health.” Health Affairs 24 : 335 - 38 . Google Scholar CrossRef Search ADS PubMed Miech Richard , Pampel Fred , Kim Jinyoung , Rogers Richard G . 2011 . “The Enduring Association Between Education and Mortality The Role of Widening and Narrowing Disparities.” American Sociological Review 76 : 913 - 34 . Google Scholar CrossRef Search ADS PubMed Moller Stephanie , Nielsen François , Alderson Arthur S . 2009 . “Changing Patterns of Income Inequality in U.S. Counties, 1970-2000.” American Journal of Sociology 114 : 1037 - 1101 . Google Scholar CrossRef Search ADS Morris Martina , Western Bruce . 1999 . “Inequality in Earnings at the Close of the Twentieth Century.” Annual Review of Sociology 25 : 623 - 57 . Google Scholar CrossRef Search ADS Moss Nancy E. , Carver Karen . 1998 . “The Effect of WIC and Medicaid on Infant Mortality in the United States.” American Journal of Public Health 88 : 1354 - 61 . Google Scholar CrossRef Search ADS PubMed National Center for Health Statistics . 2001-2006 . “Linked Birth/Infant Death Data Set. Public Use Data files.” National Center for Health Statistics, Hyattsville, MD. National Center for Health Statistics . 2012 . Health, United States, 2011: With Special Feature on Socioeconomic Status and Health. Hyattsville, MD : National Center for Health Statistics . Olafsdottir Sigrun. 2007 . “Fundamental Causes of Health Disparities: Stratification, the Welfare State, and Health in the United States and Iceland.” Journal of Health and Social Behavior 48 : 239 - 53 . Google Scholar CrossRef Search ADS PubMed Pampel Fred C. , Pillai Vijayan K . 1986 . “Patterns and Determinants of Infant Mortality in Developed Nations, 1950- 1975.” Demography 23 : 525 - 42 . Google Scholar CrossRef Search ADS PubMed Paneth Nigel , Kiely John L. , Wallenstein Sylvan , Marcus Michele , Pakter Jean , Susser Mervyn . 1982 . “Newborn Intensive Care and Neonatal Mortality in Low-Birth-Weight Infants: A Population Study.” The New England Journal of Medicine 307 : 149 - 55 . Google Scholar CrossRef Search ADS PubMed Passaro Kristi Tolo , Little Ruth E. , Savitz David A. , Noss John , and The ALSPAC Study Team . 1996 . “The Effect of Maternal Drinking Before Conception and in Early Pregnancy on Infant Birthweight.” Epidemiology 7 : 377 - 83 . Google Scholar CrossRef Search ADS PubMed Phelan Jo C. , Link Bruce G . 2005 . “Controlling Disease and Creating Disparities: A Fundamental Cause Perspective.” The Journals of Gerontology 60B : S27 – S33 . Google Scholar CrossRef Search ADS Phelan Jo C. , Link Bruce G. , Tehranifar Parisa . 2010 . “Social Conditions as Fundamental Causes of Health Inequalities Theory, Evidence, and Policy Implications.” Journal of Health and Social Behavior 51 : S28 - S40 . Google Scholar CrossRef Search ADS PubMed Phibbs Ciaran S. , Bronstein Janet M. , Buxton Eric , Phibbs Roderic H . 1996 . “The Effects of Patient Volume and Level of Care at the Hospital of Birth on Neonatal Mortality.” Journal of the American Medical Association 276 : 1054 - 59 . Google Scholar CrossRef Search ADS PubMed Reynolds Megan M. , Brady David . 2012 . “Bringing You More than the Weekend: Union Membership and Self-Rated Health in the United States.” Social Forces 90 : 1023 - 49 . Google Scholar CrossRef Search ADS Richardson Douglas K. , Gray James E. , Gortmaker Steven L. , Goldman Donald A. , Pursley DeWayne M. , McCormick Marie C . 1998 . “Declining Severity Adjusted Mortality: Evidence of Improving Neonatal Intensive Care.” Pediatrics 102 : 893 - 99 . Google Scholar CrossRef Search ADS PubMed Ross Catherine E. , Wu Chia-ling . 1995 . “The Links Between Education and Health.” American Sociological Review 60 : 719 - 45 . Google Scholar CrossRef Search ADS Ross Catherine E. , Mirowsky John . 2008 . “Neighborhood Socioeconomic Status and Health: Context or Composition?” City & Community 7 : 163 - 79 . Google Scholar CrossRef Search ADS Salihu Hamisu M. , Aliyu Muktar H. , Pierre-Louis Bosny J. , Alexander Greg R . 2003 . “Levels of Excess Infant Deaths Attributable to Maternal Smoking During Pregnancy in the United States.” Maternal and Child Health Journal 7 : 219 - 27 . Google Scholar CrossRef Search ADS PubMed Samuelson Julia L. , Buehler James W. , Norris Dianne , Sadek Ramses . 2002 . “Maternal Characteristics Associated with Place of Delivery and Neonatal Mortality Rates Among Very-Low-Birthweight Infants, Georgia.” Paediatric and Perinatal Epidemiology 16 : 305 – 13 . Google Scholar CrossRef Search ADS PubMed Schempf Ashley H. , Branum Amy M. , Lukacs Susan L. , Schoendorf Kenneth C . 2007 . “The Contribution of Preterm Birth to the Black–White Infant Mortality Gap, 1990 and 2000.” American Journal of Public Health 97 : 1255 - 60 . Google Scholar CrossRef Search ADS PubMed Shanahan Ellen , Perry Julie R. , DeClerque Julia L . 2012 . “Consensus in Region IV: Women and Infant Health Indicators for Planning and Assessment.” Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, Chapel Hill, NC. Shi Leiyu , Starfield Barbara , Kennedy Bruce , Kawachi Ichiro . 1999 . “Income Inequality, Primary Care, and Health Indicators.” Journal of Family Practice 48 ( 4 ): 275 - 84 . Google Scholar PubMed Shi Leiyu , Macinko James , Starfield Barbara , Xu Jiahong , Regan Jerrilynn , Politzer Robert , Wulu John . 2004 . “Primary Care, Infant Mortality, and Low Birth Weight in the States of the USA.” Journal of Epidemiology and Community Health 58 : 374 - 80 . Google Scholar CrossRef Search ADS PubMed Singh Gopal K. , Kogan Michael D . 2007 . “Persistent Socioeconomic Disparities in Infant, Neonatal, and Postneonatal Mortality Rates in the United States, 1969-2001.” Pediatrics 119 : e928 - e39 . Google Scholar CrossRef Search ADS PubMed Singh Gopal K. , Siahpush Mohammad . 2006 . “Widening Socioeconomic Inequalities in U.S. Life Expectancy, 1980–2000.” International Journal of Epidemiology 35 : 969 - 79 . Google Scholar CrossRef Search ADS PubMed Song Shige , Burgard Sarah A . 2011 . “Dynamics of Inequality Mother’s Education and Infant Mortality in China, 1970-2001.” Journal of Health and Social Behavior 52 : 349 - 64 . Google Scholar CrossRef Search ADS PubMed Starfield Barbara. 1985 . “Postneonatal Mortality.” Annual Review of Public Health 6 : 21 - 40 . Google Scholar CrossRef Search ADS PubMed Starfield Barbara , Shi Leiyu , Macinko James . 2005 . “Contribution of Primary Care to Health Systems and Health.” Milbank Quarterly 83 : 457 - 502 . Google Scholar CrossRef Search ADS PubMed Strully Kate W. , Rehkopf David H. , Xuan Ziming . 2010 . “Effects of Prenatal Poverty on Infant Health State Earned Income Tax Credits and Birth Weight.” American Sociological Review 75 : 534 - 62 . Google Scholar CrossRef Search ADS PubMed Subramanian S. V. , Kawachi Ichiro , Kennedy Bruce P . 2001 . “Does the State You Live in Make a Difference? Multilevel Analysis of Self-rated Health in the U . S.” Social Science & Medicine 53 : 9 - 19 . Google Scholar CrossRef Search ADS U.S. Bureau of Labor Statistics . 2012 . Local Area Unemployment Statistics . Retrieved September 13, 2017 (www.bls.gov/lau/home.htm). U.S. Census Bureau . 2012a . Annual Social and Economic Supplements. Retrieved September 28, 2017 (https://www.census.gov/did/www/saipe/data/model/info/cpsasec.html). U.S. Census Bureau . 2012b . “Small Area Income and Poverty Estimates—State and County Data.” Retrieved September 13, 2017 (www.census.gov/did/www/saipe/index.html). U.S. Centers for Disease Control and Prevention (CDC) . 2012 . “Linked Birth / Infant Death Records.” Retrieved September 13, 2017 (http://wonder.cdc.gov/lbd.html). Ventura Stephanie J. , Hamilton Brady E. , Mathews T. J. , Chandra Anjani . 2003 . “Trends and Variations in Smoking During Pregnancy and Low Birth Weight: Evidence From the Birth Certificate, 1990-2000.” Pediatrics 111 : 1176 . Google Scholar PubMed Vintzileos Anthony M. , Ananth Cande V. , Smulian John C. , Scorza William E. , Knuppel Robert A . 2002 . “The Impact of Prenatal Care on Neonatal Deaths in the Presence and Absence of Antenatal High-Risk Conditions.” American Journal of Obstetrics and Gynecology 186 : 1011 - 16 . Google Scholar CrossRef Search ADS PubMed Wildeman Christopher. 2012 . “Imprisonment and Infant Mortality.” Social Problems 59 : 228 - 57 . Google Scholar CrossRef Search ADS Wilkinson Richard G. , Pickett Kate E . 2008 . “Income Inequality and Socioeconomic Gradients in Mortality.” American Journal of Public Health 98 : 699 - 704 . Google Scholar CrossRef Search ADS PubMed Wise Paul H. 2003 . “The Anatomy of a Disparity in Infant Mortality.” Annual Review of Public Health 24 : 341 - 62 . Google Scholar CrossRef Search ADS PubMed Xu Ke Tom. 2006 . “State-Level Variations in Income-Related Inequality in Health and Health Achievement in the U.S.” Social Science & Medicine 63 : 457 - 64 . Google Scholar CrossRef Search ADS © The Author 2017. Published by Oxford University Press on behalf of the Society for the Study of Social Problems.All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Social ProblemsOxford University Press

Published: Oct 12, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off