OBJECTIVES. Black women in the United States are more likely to give birth to preterm and low birth-weight infants than their white counterparts, but little is known about variation in birth outcomes within the black population. This study aimed to test the hypothesis that the risk of low birth weight and preterm birth within the black population varies by maternal ancestry and nativity.
POPULATION AND METHODS. We conducted a retrospective cohort study using New York City birth records. All of the recorded live births to black women occurring in New York City between January 1, 1998, and December 31, 2002 (N = 168039), were divided into the following self-reported ancestry groups: African, American, Asian, Cuban, European, Puerto Rican, South and Central American (excluding Brazilian), and West Indian and Brazilian. To estimate adjusted risk ratios for low birth weight (weight at birth <2500 g) and preterm birth (gestational age at delivery <37 weeks, based on clinical estimate), we ran 3 models for each outcome, using negative binomial regression and Poisson regression with robust SE estimation. All of the models used blacks reporting American ancestry as the reference group. The first model included ancestry as the primary exposure variable along with covariates that included maternal age, parity, smoking, and education, as well as paternal education and race. Nativity (US- or foreign-born) was included in the second model, and terms representing interaction effects between ancestry and nativity were included in the third model.
RESULTS. There was substantial variation in risks of preterm birth and low birth weight among the black subgroups, with all of the groups having lower risks than the American black reference group, even after adjusting for maternal risk factors and other covariates. Risk ratios for low birth weight ranged from 0.55 among South/Central Americans to 0.91 among Cubans; risk ratios for preterm birth showed a similar pattern. Nativity was also associated with low birth weight and preterm birth; births to foreign-born women were less likely to be preterm or low birth weight than births to US-born women. Furthermore, nativity effects varied by ancestry group, with foreign-born status inversely associated with poor birth outcomes among South/Central Americans but not among West Indians/Brazilians.
CONCLUSIONS. Important health differences may be masked in studies that treat black women in America as a homogeneous group and do not take ethnic variation and nativity into account.
A substantial body of research has demonstrated that infants born to black women in the United States have higher rates of infant mortality and are at higher risk for low birth weight (LBW) and preterm delivery (PTD) than infants born to white women.1–9 The black population in the United States is not a homogeneous group, however, and birth outcomes seem to vary by nativity within the black population.10 For example, Caribbean-born blacks and African-born blacks display risks for LBW delivery and PTD closer to those of US-born non-Hispanic whites than to those of US born American blacks.4,5 This heterogeneity within racial groups has led researchers to hypothesize that cultural, social, and environmental contexts associated with nativity and ethnicity may be important as risk factors and/or as mediators of the impact of individual-level characteristics, such as educational attainment.11 However, the degree of heterogeneity within the black population with regard to LBW and PTD remains unknown, because much of the research to date has compared blacks as a whole to other ethnic groups or has stratified the black population by foreign-born and US-born status without regard to country of origin.4–6,12–14 This study makes use of the size and diversity of the New York City population to investigate the variation in birth outcomes within the black population associated with maternal ancestry and nativity.
Our target population consisted of all of the live infants born in New York City between January 1, 1998, and December 31, 2002, to women self-reporting their race as “black.” Our data were obtained from computerized public-use birth certificate records for calendar years 1998 to 2002 provided by the New York City Department of Health (NYC DOH). The NYC DOH coding scheme included the following 14 possible values for self-reported race: “Asian Indian,” “Korean,” “Samoan,” “Vietnamese,” “Guamanian,” “white,” “black,” “American Indian,” “Chinese,” “Japanese,” “Hawaiian,” “Filipino,” “Other Asian or Pacific Islander,” and “other.” Our sample was limited to those records in which maternal race was coded as “black” as well as those in which maternal race was coded as “white” and whose ancestry was coded as “American.” We then stratified the births to black women by maternal ancestry.
The self-reported ancestry variable had 25 possible values. These were recoded to create 8 geographically distinct ancestry categories, with the following limitation: the NYC DOH coding scheme grouped black women reporting Brazilian ancestry with those reporting West Indian and Caribbean ancestry (rather than with those reporting Central and South American ancestry) and eliminated the country-specific ancestry information. Therefore, we were unable to identify women with Brazilian ancestry and group them with those of South /Central American ancestry. The recoding resulted in the following ancestry categories: Mexican, Puerto Rican, Cuban, European, African, Asian (including South Central Asia, Southeast Asia, and the Pacific Islands), South or Central American (excluding Brazil), and West Indian, Brazilian, or Caribbean. However, we were skeptical that the very small black population in Mexico could give rise to the 1317 observations that reported Mexican ancestry and black race, and because of these validity concerns, we excluded them from this analysis.
To estimate risk ratios (RRs) for LBW (weight at birth <2500 g) and preterm birth (gestational age at delivery <37 weeks, based on clinical estimate), we ran models using negative binomial regression and Poisson regression with robust SE estimation. Negative binomial regression and Poisson regression with robust SE estimation are methods that allow for the direct estimation of relative risks15 and were, thus, preferred to the logistic regression procedure for this study. Blacks reporting American ancestry were chosen as the referent ethnic subgroup in all of our multivariate models.
The first model included indicator variables for ancestry as predictors of birth outcome in addition to covariates, the second model added nativity (foreign- or US-born), and the third model added interaction terms between ancestry and nativity. The covariates in the multivariate models included the following maternal risk factors: maternal age (represented as a linear spline transformed continuous variable with a single knot at 20 years), parity (represented by single dummy variable indicating nulliparity), prepregnancy weight (represented as a linear spline transformed variable with a single knot at 150 pounds), presence (yes/no) of medical risk factors (uterine bleeding, previous preterm, or small-for-gestational-age birth), education (completion of <12 years, 12–15 years, or >16 years of school), insurance status (Medicaid, private insurance, or no health insurance), tobacco use (yes/no) during pregnancy, and employment status (employed/unemployed) during pregnancy. Paternal education was also included as a covariate as were variables indicating the parent race combination: mother and father of same race (because our sample was >80% black, this group was predominantly black couples), infant's mother black and infant's father white, infant's mother black and infant's father of other race, infant's mother white and infant's father black, and infant's mother white and infant's father of other race. We also created a separate indicator variable indicating whether or not the father's race was reported by the mother. Finally, we included terms representing the following interaction effects between covariates, which we thought could be supported based on the existing literature: prepregnancy weight and tobacco consumption, employment during pregnancy and parity, and employment during pregnancy and marital status. Because we included maternal education as a covariate, we ran the models twice: once including all of the women in the sample and the second time excluding women aged ≤20 years. This exclusion did not change our findings; the results presented here include all of the women in the sample.
Our sample consisted of 168039 births to women who reported their race as “black,” plus 45233 births to US-born American white women. The ancestry groups and distributions of selected maternal risk factors are listed in Table 1.
Table 2 shows crude RRs of preterm birth and LBW for each ancestry group, using American blacks as the referent, as well as adjusted RRs estimated in models 1 and 2. Results of the models presented in Table 2 and Figs 1 and 2 suggested substantial heterogeneity between black subgroups, and, with the exception of Cubans, all of the ancestry subgroups had lower risks of PTD and LBW than the American black reference group. Maternal ancestry and nativity seemed to be more predictive of LBW than preterm birth, with crude risk RRs ranging from 0.55 (95% confidence interval [CI]: 0.52–0.59) among South/Central Americans blacks to 0.91 (95% CI: 0.59–1.39) among Cubans. The range of RRs was narrower for preterm birth but showed a similar pattern, with South/Central Americans and Africans having the lowest LBW and PTD risks of the black subgroups and Cubans and US-born Americans having the highest.
Table 2 also presents RRs associated with selected covariates. The associations between covariates and birth outcomes were consistent with expectations based on previous research. Uterine bleeding and having had a previous preterm or small-for-gestational age-infant were both strongly related to preterm birth and LBW. Higher levels of education for both mothers and fathers were associated with lower risks of PTD and LBW; however, ∼25% of observations had missing data on educational attainment of the father. Being on Medicaid or having no health insurance was associated with higher risks of preterm birth and LBW than was having private health insurance. Paternal race also seemed to have an effect on birth outcomes. For example, relative to women partnered with men of the same race, births to black women whose infant's father was white (n = 7784) had a lower risk of LBW delivery (RR = 0.84; 95% CI: 0.77–0.92), although the same effect was not seen for preterm birth. Conversely, relative to women partnered with men of the same race, white women whose infant's father was black (n = 661) had an elevated risk of LBW delivery (RR = 1.53; 95% CI: 1.12–2.10) and PTD (RR = 1.97; 95% CI: 1.52–2.55). As with paternal education, almost a quarter of the observations were missing data on paternal race. In contrast, the data on maternal risk factors were fairly complete; prepregnancy weight was missing on ∼6% of records, but this was the only maternal covariate for which 5% or more of the observations were missing data.
There was considerable variation in the distribution of covariates between ancestry subgroups, as shown in Table 1. However, there seems to be little correlation between rates of PTD and LBW and the frequency with which risk factors were reported. For example, American ancestry blacks, who consistently had the worst birth outcomes, were most likely to report tobacco use during pregnancy but were less likely than Puerto Rican, South/Central American, and African blacks to have low education or be unemployed and were less likely than South/Central Americans, Africans, Puerto Ricans, and West Indians/Brazilians to lack private health insurance. Conversely, South/Central American blacks fared best in terms of birth outcomes, but this group also had a high prevalence of low education and unemployment, and ∼83% lacked private health insurance.
Model 2 considered the effect of nativity on birth outcomes, suggesting that foreign-born women have considerably lower risk of LBW (RR = 0.81; 95% CI: 0.77–0.86) and PTD (RR = 0.86; 95% CI: 0.81–0.91) than do US-born women. Model 3 investigates whether the effects of nativity vary by ancestry group. Table 3 presents the LBW and PTD RRs comparing foreign-born to US-born women in each ancestry category. Among blacks reporting South/Central American ancestry, there seemed to be a substantial nativity effect, with foreign-born women faring better than US-born women of the same ancestry (RR for LBW = 0.73; 95% CI: 0.61–0.88; RR for PTD = 0.77; 95% CI: 0.65–0.90). In contrast, among blacks reporting West Indian or Brazilian ancestry, this effect was not seen, with US-born women having rates of LBW and PTD similar to foreign-born women (RR for LBW = 0.96; 95% CI: 0.83–1.11; RR for PTD = 0.99; 95% CI: 0.85–1.13). Results in most of the other ancestry categories were too imprecise to allow for interpretation.
We found considerable heterogeneity in the black population of New York City with respect to risk of LBW and preterm birth, with American blacks generally having the highest risks of poor birth outcomes even after adjusting for maternal risk factors and other covariates. We also found that being US born was associated with higher risks of preterm birth and LBW than was being foreign born. These findings are consistent with previous research, which has demonstrated inverse associations between poor birth outcomes and non-American ancestry and foreign-born status.
Previous studies have found that American blacks in Massachusetts deliver infants with lower mean birth weights than Haitian, West Indian, Cape Verdan, Hispanic, and “other” blacks7 and that African-born and Caribbean-born black women in Illinois have lower rates of LBW than their US-born counterparts.5,6 However, when births to black women in California were divided into US-born and foreign-born groups, regardless of country of origin, there was no detectable nativity effect on birth outcomes.14 The inconsistency in these results may be explained in part by the findings of this study, which suggest that ancestry or country of origin is more predictive of birth outcomes than foreign-born status and that being foreign born is associated with lower rates of LBW and preterm birth only within certain ancestry groups, such as South/Central American blacks. This hypothesis is supported by previous examinations of national linked birth and infant death records13 and Detailed Natality Data12 that found that the magnitude of maternal nativity effects on LBW varied significantly across racial/ethnic groups.
Our study explored the role of several maternal risk factors in the relationship between ancestry/nativity and birth outcomes, but the results of our multivariate models indicate that these risk factors do not explain the heterogeneity in birth outcomes; in particular, South/Central Americans have among the lowest risks of poor birth outcomes despite a high prevalence of several maternal risk factors. That low education and high levels of unemployment among foreign-born South/Central Americans do not result in high rates of PTD and LBW is consistent with phenomena such as the “Hispanic paradox,”11 and suggests that culture and social context mediate the impact of such exposures.
An additional potential source of variation in the distribution of birth outcomes among black women is the race of the father, which has been only rarely explored.16 Our study indicates that there may be some interaction between paternal and maternal race in predicting birth outcomes. Because our reference group consisted primarily of single-race black couples, our results likely resemble a comparison of biracial infants to infants of 2 black parents. Our finding that infants born to black mothers and white fathers fared better than the referent may not be surprising, because interracial marriage is most common among the most highly educated,17 and thus the lowest risk, couples. What is, perhaps, less expected is the finding that infants of white mothers and black fathers are more likely to be preterm or LBW than the infants of the predominantly black reference group. This suggests that these infants would do far worse than infants of white couples and that paternal race is potentially an important factor in predicting the LBW/PTD risk of these infants.
Interpretation of the results of this study should take into account some important limitations. First, as with much self-reported data, the validity of the data on ancestry is uncertain. For example, a portion of the African ancestry group was composed of women reporting that they were born in the United States, and these were assumed to be women of recent African ancestry. Because all American blacks can claim African ancestry in a fundamental sense, there may have been women included in the African group who, for the purposes of our analysis, would have been more suitably grouped with those of American ancestry. However, the number of women who reported African ancestry and being born in the United States was very small (n = 82) in comparison to both the US-born African group (n = 10793) and blacks reporting American ancestry (n = 88966), and any misclassification would not have affected the results. Likewise, there were >9000 women in the foreign-born American black group, making it unlikely that this group truly and entirely consists of women born overseas to black parents with American ancestry. It is more likely, in our estimation, that a nontrivial proportion of the women in this group are actually women born overseas to non-American parents but who have chosen to identify themselves as American blacks, perhaps indicating longer residence time in the United States and a higher degree of cultural assimilation.
Second, the NYC DOH coding scheme grouped women reporting Brazilian ancestry with those reporting West Indian/Caribbean ancestry instead of with those with South/Central American ancestry. It is not clear how the birth outcome patterns observed for the West Indian ancestry group would differ if those with Brazilian ancestry were removed. Third, although missing data were minimal for maternal covariates, nearly a quarter of observations were missing data on paternal race and education; the association found between parent race combination and birth outcomes should be interpreted with recognition of this limitation.
Fourth, this study treats ancestry and nativity as proxies for social and environmental contexts; thus, the associations found in these analyses do not suggest an etiologic role for ancestry or nativity but rather indicate that causal factors are unevenly distributed between black women of different ancestries and nativities. However, with the exception of a few measures, such as maternal smoking, the details of the social and environmental exposures that may have a causal impact on birth outcomes are not available. Finally, the results presented here are specific to New York City from 1998 to 2002. It is not clear to what extent these findings can be extrapolated to other contexts and time periods.
There seems to be significant variation in birth outcomes, by both ethnicity and nativity, among black women. Future studies that attempt to look at race in relation to birth outcomes should not only avoid, if possible, grouping US-born and foreign-born women together, but should also be careful not ignore ethnic heterogeneity among black women, because important health differences may be masked in the process. Furthermore, future research should attempt to capture those aspects of the social/environmental context associated with a given ancestry or nativity that may be causally related to preterm birth and LBW.
This research was funded in part by National Institute of Child Health and Human Development grant R21 HD050739 and by National Institute of Child Health and Human Development training grant T32 HD07168.
- Accepted May 24, 2006.
- Address correspondence to David L. Howard, PhD, Carolina Population Center, 211 W Cameron Ave, Chapel Hill, NC 27516. E-mail:
The funding agency was not involved in the design or conduct of the study; collection, management or analysis or interpretation of the data; or preparation, review, or approval of the manuscript.
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Martin JA, Kochanek KD Strobino DM, Guyer B, MacDorman MF. Annual summary of vital statistics, 2003. Pediatrics.2005;115 :619– 634
- ↵Fang J, Madhavan S, Alderman MH. Low birth weight: race and maternal nativity–impact of community income. Pediatrics.1999;103(1) . Available at: www.pediatrics.org/cgi/content/full/103/1/e5
- ↵Palotto EK, Collins JW Jr, David RJ. Enigma of maternal race and infant birth weight: a population-based study of US-born Black and Caribbean-born Black women. Am J Epidemiol.2000;151 :1080– 1085
- ↵Acevedo-Garcia D, Soobader M, Berkman LF. The differential effect of foreign-born status on low birth weight by race/ethnicity and education. Pediatrics.2005;115(1) . Available at: www.pediatrics.org/cgi/content/full/115/1/e20
- ↵Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol.2004;159 :702– 706
- Copyright © 2006 by the American Academy of Pediatrics