We used population-based data from 4 state-based birth defect surveillance programs (Arizona, New York, New Jersey, and Texas) to conduct a retrospective cohort study. The study population included live-born infants delivered from 1999 to 2007 with a diagnosis of 1 of the following 7 CHDs: common truncus arteriosus, transposition of the great vessels, tetralogy of Fallot, atrioventricular septal defect, aortic valve stenosis, hypoplastic left heart syndrome, and coarctation of the aorta. We selected these defects for inclusion in the analysis because of the high reliability with which they are ascertained by public health birth defect surveillance programs and because of the relatively high mortality associated with each defect. We classified infants as having one of the included CHDs by using a modified British Pediatrics Association (BPA) coding system³⁴ for births in Arizona, New York, and Texas, and the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)³⁵ codes for births in New Jersey.

Overall, we identified 10 578 infants with at least 1 of the 7 selected CHDs from the 4 population-based birth defect surveillance programs. Of those, 9853 infants (93%) had census tract information that could be used for linkage with the 2000 US Census. Maternal and infant characteristics of the final cohort by participating state are provided in Table A and infant mortality by CHD type are provided in Figure A (both available as supplements to the online version of this article at http://www.ajph.org).

Socioeconomic disadvantage was adversely associated with the survival of infants born with CHDs. Survival varied by most but not all community indicators that we examined, and the statistical significance of the survival difference was stronger in the postneonatal period than in the neonatal period. The community factors most predictive of infant death were related to income, poverty, education, and occupation. Socioeconomic disadvantage related to these factors increased the infant mortality risk by up to 47%, and the associated mortality risk increased significantly in some subpopulations, such as those of Hispanic ethnicity. Among infants born to Hispanic mothers, those who lived in communities with high residential crowding had more than a fourfold increased infant mortality risk compared with those living in communities with the least residential crowding.

Because infants with severe CHDs require early and continued surgical and medical intervention, increased access to and use of specialized health care resources would be expected to improve the likelihood of survival; however, measuring access or barriers to care in population-based studies is challenging.^5,29 Individual-level information on health insurance and SES is not frequently available in population-based data in the United States; therefore, a growing body of research has focused on determining appropriate community measures of SES that might serve as a proxy for unavailable individual-level information.⁴⁰ Previous studies observed an association between individual- and area-level SES and the etiology of birth defects,^{25,27,28,41,42} but population-based survival studies in the United States have yet to incorporate community measures of SES to better understand factors influencing survival among infants born with birth defects. Limited studies in Canada, which has universal health care coverage, found inconsistent evidence that neighborhood income was associated with survival of medically needy infants,^43,44 but this association has not been thoroughly examined in the United States. On the basis of findings from previous work, our study included a range of measures to better understand the type of SES factors that might be associated with mortality in our study population.^28,30,45

CHD subtypes can range in severity. Those included in this study are considered severe and were selected because of the reliability with which they are clinically diagnosed and accurately detected by birth defect surveillance programs. Because of the severe nature of these conditions, infants require surgical intervention and appropriate follow-up care, with additional surgeries frequently required. The complexity of continued care might explain why the impact of SES was greater on postneonatal survival. Higher levels of education might be needed to understand and process sophisticated medical information and to research and make selection decisions about the physicians and institutions that might provide higher-quality care. Family income is associated with gradients in both children’s health and access to health care, and it might influence the ability to seek out and use high-quality care, especially to overcome barriers presented by significant travel distances.⁴⁶ Teaching hospitals and institutions that conduct a high volume of pediatric cardiac surgeries have been shown to have better outcomes than low-volume hospitals, and the ability to control for differential patterns of referrals and for the risk-adjusted mortality of hospitals might elucidate the association between SES factors and survival.^47–51

The finding that occupation influenced survival was interesting and potentially more difficult to explain. This variable might serve as a proxy for health literacy or family resources, although it was not highly correlated with the measures for income and education. It is more likely that occupation type is an indicator of the level of medical insurance coverage, which is often employer based. Unemployment was not associated with mortality; however, Medicaid eligibility might be higher in areas of high unemployment, with covered individuals having sufficient insurance coverage and access to care. Those more likely to be in operator or laborer occupations might have incomes too high to qualify for Medicaid but have no employer-provided coverage or have plan options that provide inadequate coverage. The impact of medical insurance on infant survival among individuals born with CHDs has not been well examined, although several hospital-based studies found that infants with public insurance had higher postoperative mortality than infants covered by private insurance.^18,19

The consistency with which we observed lower infant survival for the most disadvantaged decile across all the SES indicators raised the issue of whether each indicator was indicative of some common underlying factor associated with economic disadvantage. If that were the case, having multiple indicators in the lowest decile would not be expected to increase the risk above that associated with a single indicator; however, we found a decrease in infant survival with an increasing number of indicators that were in the most disadvantaged decile. This suggests that different community indicators of SES might be separable measures of differing underlying risks, and analyses limited to 1 community indicator might underestimate the impact of disadvantaged communities on health outcomes. Another unknown was whether the community indicators were simply reflecting an individual-level risk or if the community risk was independent of the individual risk. For the most direct test, we examined constructs for which both the individual and community variables were associated with survival. Low maternal education was associated with lower survival, but even infants born to mothers with higher education had significantly lower survival if they lived in a community with a high proportion of residents with a low level of education. In fact, the community effect associated with low education appeared to be larger than the individual-level effect. Similarly, even though non-Hispanic White infants had a survival advantage compared with Hispanic infants, non-Hispanic White infants in communities with a high proportion of Hispanic residents experienced lower survival than non-Hispanic White infants living in communities with a low proportion of Hispanic residents, which is in notable contrast to the “Hispanic paradox” observed in the general population.⁵²

The observed racial/ethnic disparities in mortality among infants with CHDs corroborate earlier findings that have not been well explained.^{2,9,14,41,53–55} Racial and ethnic disparities in survival were highest in the postneonatal period, and adjustment for community factors did not appreciably reduce racial and ethnic disparities, a finding contrary to previous ones based on other health outcomes.⁵⁶ After adjustment for all individual- and community-level confounders, non-Hispanic Black and Hispanic infants still had a higher mortality risk than non-Hispanic White infants. Although we did not have information on the medical insurance status of infants, previous studies found that medical insurance explained some of the observed racial/ethnic disparities.^16,57 Therefore, it is plausible that referral patterns and access to and use of advances in pediatric cardiac care may be important predictors of disparity in survival among infants with a CHD, and that addressing these issues might be an important step in reducing these disparities. Because survival estimates vary across CHD subtypes, there might be a concern that unequal distribution of subtypes by race/ethnicity could explain differences in survival by race and ethnicity; however, a previous study showed modest variation in birth prevalence by race/ethnicity.²⁵ Data from that study showed that the CHD type with the highest mortality, hypoplastic left heart syndrome, was not higher among non-Hispanic Black infants (prevalence ratio [PR]=1.01) or Hispanic infants (PR=1.02) than among non-Hispanic White infants.

A strength of this study was that it combined data from several states to provide a larger study population size that was diverse racially and ethnically, regionally, and socioeconomically; however, a disproportionate number of infants were from urban communities. This large, diverse population reduced the level of random variability in the survival estimates and allowed evaluation of differences in these estimates across an array of factors that might influence survival of infants with CHDs. Birth defect surveillance programs use varying methodologies to identify and confirm cases of birth defects, and the programs from which these data were drawn are those that employ active or partially active case ascertainment, which provide more accurate clinical diagnoses and the most complete prevalence estimates.^34,58 Despite the increased accuracy gained by use of medical records, some misclassification of CHD types may occur; however, our approach of grouping the 7 severe CHDs reduces the impact of misclassification.

A potential limitation of this study was the lack of information on pregnancies affected by a CHD that resulted in fetal deaths, especially those that were elective terminations. Socioeconomic and other barriers to prenatal care can result in differential prenatal diagnosis by race and SES.^15,59 Consequently, more severe defects could be detected and terminated at higher rates among those with greater access to care, which would artificially inflate the true mortality risk of disadvantaged populations.^60–62 Another limitation was that infants who moved out of state and died might be assumed to be alive; however, many states have interstate agreements to share information on infant deaths.

Although birth defects are a leading cause of infant death, public health strategies to address the overall burden of infant deaths have largely been concentrated on other causes such as preterm delivery, sudden infant death syndrome, and injuries.⁶³ This lack of attention might be in part the result of the inaccessibility of quality data or because mortality associated with birth defects is perceived as dependent primarily on surgical intervention. Yet a growing number of population-based studies of birth defects have identified unexplained racial and socioeconomic disparities, which represent opportunities for strategic and targeted interventions. Although previous survival studies of infants with birth defects focused primarily on clinical factors associated with early mortality, this study provides evidence that community measures of SES are potentially useful predictors of survival patterns in infants born with CHDs and are more strongly associated with survival than other established individual-level sociodemographic risk factors. This population-based study provides evidence suggesting that public health interventions such as home visitation programs and early intervention services might more effectively identify infants with the highest risk of infant death if integrated with information on residential communities.