After decades of intense basic science, epidemiologic, and clinical research, racial disparities in preterm birth continue to be marked with the rate in non-Hispanic black women (13.4%) being much higher compared with non-Hispanic white (8.8%) women.1 The trend toward decreasing preterm birth in the past decade has made only little effect in narrowing this disparity.1 Although socioeconomic inequalities are often cited, the biological underpinnings of the higher risk of preterm birth of non-Hispanic black women remains elusive because it continues to persist after adjusting for access to health care, medical and behavioral risks, and for educational, social, and economic status.2
It is well recognized by numerous measures that immigrants and refugees to the United States are healthier than native-born individuals. This is likely because the immigration process requires good health, and immigrants tend to have healthier lifestyles.3,4 This observation was first noted in women from Latin American countries and was termed the Latina epidemiologic paradox because the overall protective effect was found to diminish over time and across generations with acculturation.5 Studies in the Hispanic population have shown that recent immigrants also have lower rates of preterm birth.6,7 This observation has now extended to sub-Saharan Africa and Ethiopian immigrants.8,9
Somalia has a large diaspora as a result of the civil war that broke out in the 1990s with large numbers of ethnic Somalis now living in the United States. Analysis of deliveries in Washington state showed that the preterm birth risk in Somali women was comparable with that in native-born white women, yet Somalis had higher rates of obstetric complications and postterm delivery.10 Although seemingly at odds with the Latina paradox, the authors suggested Somali women might be less likely to engage with the health system or to comply with obstetric care providers' evidence-based recommendations.10
Ohio has one of the largest Somali populations in the United States, second after Minnesota. In addition, Ohio's overall higher rate of prematurity and black-to-white disparity in preterm birth is more representative for the entire United States compared with Washington state, which has significantly better birth outcomes.11 We took advantage of Ohio's racial diversity and presence of a Somali population to study gestational age at delivery and the patterns in spontaneous compared with health care provider-induced prematurity among different demographic groups in a single geographic region.
MATERIALS AND METHODS
This is a retrospective cohort study of birth certificates of neonates born in Ohio between January 1, 2000, and December 31, 2015. The study was approved by the institutional review boards of the Ohio Department of Health, The Research Institute at Nationwide Children's Hospital, The Ohio State University, and Mount Carmel Health System.
We classified births into four mutually exclusive groups based on maternal race, ethnicity, and birthplace (nativity): 1) U.S.-born, non-Hispanic white (USBW); 2) U.S.-born, non-Hispanic black (USBB); 3) Somalia-born (SB); and 4) African-born black (ABB) women of African nativity, excluding Somalia and countries with populations of predominantly Arab descent, as described in Appendix 1, available online at http://links.lww.com/AOG/B51. This last group is culturally and genetically diverse; however, we expected its members to be mostly of West African birthplace and to acquire similar immigration-related experiences. Exclusion criteria included multiple gestation, Hispanic ethnicity, multiple reported races, and unknown or ambiguous maternal birthplace. Preterm birth was defined as delivery before 37 completed weeks of gestation. Length of pregnancy was based on the best clinical–obstetric estimate12,13 and analyzed either as an ordinal variable based on completed weeks of gestation (integer-valued) or aggregated in clinically relevant intervals: early preterm birth (20–33 weeks of gestation), late preterm birth (34–36 weeks of gestation), term (37–41 weeks of gestation), and postterm (42 weeks of gestation or greater) birth.1 Term births were further analyzed as early term (37–38 weeks of gestation), full term (39–40 weeks of gestation), and late term (41 weeks of gestation).
Information on maternal birthplace was derived from the maternal country of birth field coded on the 2003 U.S. Standard Certificate of Live Birth, which Ohio adopted only in 2006. Information on maternal birthplace in 2000–2005 Ohio birth certificates exists but only as free text. This required one of our team members (E.A.O.) to manually code this data field for each birth with a non-U.S. maternal birthplace.
Accuracy of the gestational age in the birth certificate was evaluated and verified by manual medical chart abstraction in a sample of 683 births at four central Ohio hospitals selected as described in Appendix 2, available online at http://links.lww.com/AOG/B51. The average difference between the two gestational ages was 0.1 week with less than 3% of cases outside the limits of Bland-Altman agreement interval of −2.0 to +2.2 weeks. Lin's concordance correlation coefficient (ρc)14 of 0.979 (95% CI 0.976–0.982) between gestational age entered in the birth certificate and medical record qualified as “substantial.” Additional data obtained from the birth certificate were maternal age, marital status, parity, education level, smoking status during the index pregnancy, birth interval (period from prior birth), gestational and pregestational (chronic) hypertension, diabetes, and meconium-stained amniotic fluid.
We previously validated an algorithm to distinguish between spontaneous and health care provider–initiated delivery using 11 fields in the 2003 U.S. Standard Certificate of Live Birth.15 We used this algorithm to categorize births into these two major phenotypes for a 10-year period starting in 2006 (when Ohio adopted the 2003 revision of the birth certificate).
An algorithm based on a Naïve-Bayes classifier using maternal names recorded on the birth certificate (first and last name fields) was used to capture possible differences in gestational ages between births to women of Somali descent born in Somalia and those born elsewhere including the United States. The classifier was trained in the 2000–2011 data set and tested in the 2012–2014 data set yielding greater than 99% accuracy.
The analysis was performed with STATA 13.0, SigmaPlot 12.5, GraphPad Prism 6.0a, and MedCalc. P<.05 was considered statistically significant unless specified. Differences in proportions were compared using χ2 tests and P values corrected for multiple comparisons using the Marascuilo method.16 The distributions of gestational length were evaluated for departure from unimodality using Hartigan's Dip Test.17 Because a health care provider–initiated delivery is a competing outcome for spontaneous preterm birth, we calculated phenotype-specific subhazard ratios using a competing regression analysis model in the manner of Fine and Gray,18 using the USBW group as the referent.
There were 2,233,589 singleton live births recorded in Ohio during the 16-year study period. Births to women in the four designated race–nativity groups accounted for 1,960,693 births (87.8%; Table 1). As a group, Somalia-born mothers were more likely to be multiparous, married, and to lack formal high school education. They were less likely to smoke, less likely to start prenatal care in the first trimester, and had shorter interpregnancy intervals if multiparous. Their pregnancies were less often complicated by hypertension yet more often complicated by diabetes.
Singleton preterm birth rates were significantly lower in the SB group (5.9%) compared with the USBB (13.0%), ABB (8.4%), and USBW (7.9%) groups (Fig. 1A; Appendix 3, available online at http://links.lww.com/AOG/B51). Rates of preterm birth for women in the ABB group were not significantly different from women in the USBW group. Women in the SB group maintained the lowest preterm birth rate for most years captured in our study (Fig. 1B). Similar levels of significance were maintained if analyses were limited to either primiparous or multiparous women (Appendix 3, http://links.lww.com/AOG/B51). Because less than 1% of women in the SB or ABB groups smoked, we further restricted the analysis to nonsmoking women. This resulted in women in the ABB group having higher preterm birth rates compared with women in the USBW group (Appendix 3, http://links.lww.com/AOG/B51). Because the SB group retained the lowest risk for preterm birth, this suggests protective factors other than being nonsmokers.
By analyzing preterm birth by clinically relevant periods, we noted that the early preterm birth rate of women in the SB group (2.1%) was lower than that of women in the USBB and ABB groups but higher than that of women in the USBW group (Fig. 2A). However, rates of late preterm birth were lower in the SB group (3.7%) compared with all other groups (Fig. 2B). A high postterm birth rate (5.8%) was unique among women in the SB group, 7- to 10-fold higher than in the other groups (Fig. 2C). The adjusted odds ratios for preterm birth and postterm birth for each of the race–nativity groups are displayed in Appendix 4, available online at http://links.lww.com/AOG/B51. The lower risk of preterm birth and higher risk of postterm pregnancies for women in the SB group remained significant after adjustment for the demographic and clinical characteristics in Table 1 and for birth year.
Grouping term births by clinically relevant periods (Figs. 2D–F) revealed that both the SB and USBB groups had a lower proportion of full-term births compared with the USBW and ABB groups (Fig. 2E). However, if most term births in USBB group were early term (Fig. 2D), the majority of term births to women in the SB group were late term (Fig. 2F) suggesting that the disparity among the four race–nativity groups in preterm and postterm births extends into the term period in a similar pattern.
The higher rate of postterm birth in the SB group was associated with a higher incidence of meconium-stained amniotic fluid at the time of birth but not of meconium aspiration syndrome (Appendix 5, available online at http://links.lww.com/AOG/B51). Newborns of women in the SB group had more frequent 5-minute Apgar scores less than 7 when compared with newborns of women in the USBW group, but not when compared with newborns of women in the USBB or ABB groups. We saw no differences in birth injuries or in the need for assisted ventilation greater than 30 minutes for women in the SB group compared with those in the other groups.
For women in the USBW, USBB, and ABB groups, the majority of pregnancies ended at 39 completed weeks of gestation as marked by overlapping unimodal peaks (Fig. 3). The rightward shift in the distribution of gestational age at delivery in women in the SB group was consistent with their lower preterm birth rate and higher frequency of postterm pregnancies. The departure from the unimodal pattern of distribution was largest for the SB group.
Overall, 99.5% of singleton births (2006–2015) with a gestational age 20 weeks or greater (n=1,195,917) could be categorized into either spontaneous (n=598,521) or health care provider–initiated (n=591,676) deliveries (Appendix 6, available online at http://links.lww.com/AOG/B51). Women in the SB group had the lowest preterm birth rates for both categories (spontaneous 3.8% and health care provider–initiated 2.1%) followed by women in the USBW (4.7% [3.1%]), ABB (4.5% [3.9%]), and USBB (8.0% [4.8%]) groups, respectively (Appendix 3, http://links.lww.com/AOG/B51).
The frequency distribution of gestational length for pregnancies with spontaneous delivery onset showed a pattern similar to unsegregated births (Fig. 4A). Overall, 33.8% of spontaneous term deliveries to women in the SB group occurred at 40 weeks of gestation as compared with 26.8% for the USBW and ABB groups and 24.2% for the USBB group. Concurrently, the rate of spontaneous term deliveries that occurred late term was 18.5% in the SB group as compared with less than half of this number in all other groups (USBB: 4.8%, USBW: 5.5%, ABB: 6.9%).
The frequency distribution of gestational length for women in the SB group who had a health care provider–initiated term delivery (Fig. 4B) had an accentuated bimodal pattern resulting from more deliveries at either 39 weeks of gestation (30.5%) or 41 weeks of gestation (27.2%) as opposed to 40 weeks of gestation (25%). This pattern agrees with obstetric practice regarding indications for full term (likely representing the 39 weeks of gestation peak) or late term (likely representing the 41 weeks of gestation peak) deliveries. These data were supported by the higher incidence of repeat cesarean deliveries among health care provider–initiated deliveries at 39 weeks of gestation (39.8% [215/540]) compared with 41 weeks of gestation (12.3% [59/481]). By comparison, all the other groups had a unimodal distribution with most health care provider–initiated deliveries occurring at 39 weeks of gestation.
Singleton preterm births were segregated by delivery phenotype into having spontaneous (n=62,106) or health care provider–initiated (n=39,790) (Appendix 7, available online at http://links.lww.com/AOG/B51) onset and analyzed using a competing risk model. Relative to the USBW group, women in the SB group had reduced risks for both spontaneous (subhazard ratio [sub-HR] 0.799, 95% CI 0.703–0.908, P=.001; Fig. 4C) and health care provider–initiated preterm birth (sub-HR 0.673, 95% CI 0.566–0.799, P<.001; Fig. 4D). In contrast, women in the USBB group had higher risk for both preterm birth phenotypes (spontaneous: sub-HR 1.734, 95% CI 1.703–1.766, P<.001; health care provider–initiated: sub-HR 1.590, 95% CI 1.554–1.628, P<.001). Women in the ABB group had a mixed profile with moderate risk of health care provider–initiated preterm birth (sub-HR 1.284, 95% CI 1.150–1.433, P<.001), but no elevated risk for spontaneous preterm birth (sub-HR 0.969, 95% CI 0.875–1.074, P=.552) compared with women in the USBW group.
To strengthen the validity of our observations, we developed a classifier that predicted Somali ethnicity from maternal names. Among all Ohio singleton births during the 2000–2014 period (n=2,109,732), 7,821 births were to women predicted by the classifier as ethnic Somali. These further segregated by maternal birthplace into: Somali women born in Somalia (n=6,801) and Somali women born outside Somalia (n=826 of whom 265 were U.S.-born). Of women predicted by the algorithm as ethnic Somali, 97.3% were recorded on their newborn's birth certificate as black non-Hispanic.
This analytic approach allowed us to confirm the lower rate of late preterm birth (Appendices 8 and 9, available online at http://links.lww.com/AOG/B51) and the significantly higher frequency of postterm pregnancies among ethnic Somali women. Similarly, among term births, we confirmed the lower proportion of early- and full-term births with a higher proportion of late-term births in women of Somali ethnicity. Interestingly, these features appeared attenuated in Somali women born outside Somalia compared with Somali women born in Somalia. The peak of the empiric distribution function of gestational length (Fig. 5) was maintained at 40 weeks of gestation only for Somali women born in Somalia; the peak of the group of Somali women born outside of Somalia overlapped with the USBW group peak at 39 weeks of gestation.
We report a systematic comparative analysis of gestational length in four racially diverse populations in Ohio including Somali Americans, an immigrant group with considerable growth in the past two decades. We found that Somali women have overall lower preterm birth rates compared with U.S.-born, non-Hispanic black women. Rates of late preterm birth in Somali women were lower compared with U.S.-born, non-Hispanic white and African-born black women, an immigrant group from primarily West African birth countries. Interestingly, compared with all other groups, women of Somali ethnicity tended to have prolonged pregnancies. Clinical observation has suggested that Somalia-born women may be more reluctant to follow recommendations for health care provider–initiated delivery.10 However, based on our previously validated algorithm that is able to differentiate spontaneous from health care provider–initiated births,15 the lower rate of preterm birth and higher rate of longer gestation in women in the SB group were also noted for births with spontaneous onset. This unbiased observation excludes as the only explanation the refusal of Somalia-born women as a group to follow medical advice regarding health care provider–initiated delivery. The lower preterm birth rates of women in the SB group occurred in the context of both negative (lower rates of maternal age younger than 18 years, smoking, and hypertension) and positive (less education, higher diabetes rates, shorter interpregnancy intervals) risk factors, which concurred with previous studies on Somalia-born women.9,10,19
The lack of a significant difference in early preterm birth rates between ethnic Somali women segregated by nativity supports a genetic resilience to the underlying etiologies responsible for early preterm birth. In contrast, the attenuation in positive disparity with respect to late preterm birth and in the tendency for prolonged pregnancies favors an environmental determinism for these traits. From a genetic perspective, Somali women are Afro-Asiatic with distinct clustering of microsatellite markers compared with African American women who are Niger-Kordofanian.20 Of particular relevance to this study is that Somalis originate in East Africa, an area distinguished by high genetic diversity.20 Extreme environments, infectious disease pressure, new diets driven by transition to pastoralism, and cultural and socioeconomic contexts contributed to well-documented gene–environment interactions that favored selection of mutations that conferred survival and reproduction advantages.21 This may explain the differences in birth phenotypes among East African populations. A recent study comparing perinatal outcomes of East African immigrants in Australia found that only Somali women had lower odds of preterm birth compared with Australian women, whereas Eritrean or Sudanese women did not. Paradoxically, Ethiopians had higher odds.22
Assessing preterm birth and gestational length as variable traits in population studies requires uniform obstetric practice guidelines across various ethnic groups including the ability to accurately date pregnancies by ultrasonography, which is impractical in Somalia, at least at this time. Based on these practical observations, performance of this analysis in the United States minimizes potential biases based on inaccurate dating. Our study is unique because of the extent of additional analyses (segregation of health care provider–initiated births from those with spontaneous onset, verification of Somali ethnicity based on names, and of gestational age accuracy based on chart reviews) that all add weight to our observations.
The high rate of preterm birth in the United States compared with other developed countries constitutes a public health concern given the implication for neonatal outcomes and risk of quality-of-life impairment for the affected children and families. A series of measures implemented after 2006 have halted the rising trends and have resulted in some decrease in preterm birth rates.1,23 Of note, the preterm birth rate for Somali women remained low throughout our 16-year period of study with only 1 year exceeding 7%. This is in contrast to a study by Flynn et al19 in which preterm birth rates for Somali women delivering at a hospital in Minnesota doubled in the latter half of the 14-year study period. Unlike the Minnesota study in which gestational age was based on late menstrual period, which overestimates preterm birth,1,24 our analysis was based on obstetric estimates and was statewide, thus minimizing selection bias related to changes in referral patterns for anticipated preterm birth.
Based on the magnitude of the difference, the observed “positive disparity” of the immigrant groups in terms of preterm birth and especially of Somali women is likely multifactorial. Nevertheless, the temporal stability of a low preterm birth phenotype makes the Somali population in Ohio an invaluable resource to untangle the underlying etiology of their particular pregnancy phenotype and to detect changing trends relative to quantitative measures of acculturation and in second-generation Somali Americans.25 If dietary patterns preferred by less acculturated Somali women decrease susceptibility to preterm birth, there is benefit to understanding the underlying mechanism using current genetic and molecular tools. Conversely, a more direct effect on the growing Somali community will be the determination of whether current U.S. guidelines of pregnancy and labor induction are generalizable to Somali American women, thus highlighting the need for obstetric studies in this subpopulation.
1. Martin JA, Hamilton BE, Osterman MJ, Driscoll AK, Mathews TJ. Births: final data for 2015. Natl Vital Stat Rep 2017;66:1.
2. Institute of Medicine. Preterm birth: causes, consequences, and prevention. Washington, DC: National Academies Press; 2007.
3. Hill TD, Angel JL, Balistreri KS, Herrera AP. Immigrant status and cognitive functioning in late-life: an examination of gender variations in the healthy immigrant effect. Soc Sci Med 2012;75:2076–84.
4. Singh GK, Siahpush M. All-cause and cause-specific mortality of immigrants and native born in the United States. Am J Public Health 2001;91:392–9.
5. Flores ME, Simonsen SE, Manuck TA, Dyer JM, Turok DK. The “Latina epidemiologic paradox”: contrasting patterns of adverse birth outcomes in U.S.-born and foreign-born Latinas. Womens Health Issues 2012;22:e501–7.
6. Gould JB, Madan A, Qin C, Chavez G. Perinatal outcomes in two dissimilar immigrant populations in the United States: a dual epidemiologic paradox. Pediatrics 2003;111:e676–82.
7. Madan A, Palaniappan L, Urizar G, Wang Y, Fortmann SP, Gould JB. Sociocultural factors that affect pregnancy outcomes in two dissimilar immigrant groups in the United States. J Pediatr 2006;148:341–6.
8. Howard DL, Marshall SS, Kaufman JS, Savitz DA. Variations in low birth weight and preterm delivery among blacks in relation to ancestry and nativity: New York City, 1998-2002. Pediatrics 2006;118:e1399–405.
9. Elo IT, Vang Z, Culhane JF. Variation in birth outcomes by mother's country of birth among non-Hispanic black women in the United States. Matern Child Health J 2014;18:2371–81.
10. Johnson EB, Reed SD, Hitti J, Batra M. Increased risk of adverse pregnancy outcome among Somali immigrants in Washington state. Am J Obstet Gynecol 2005;193:475–82.
11. March of Dimes. 2016 premature birth report cards. Available at: http://www.marchofdimes.org/mission/prematurity-reportcard.aspx
. Retrieved February 18, 2017.
12. Martin JA, Osterman MJ, Kirmeyer SE, Gregory EC. Measuring gestational age in vital statistics data: transitioning to the obstetric estimate. Natl Vital Stat Rep 2015;64:1–20.
13. Wier ML, Pearl M, Kharrazi M. Gestational age estimation on United States livebirth certificates: a historical overview. Paediatr Perinat Epidemiol 2007;21(suppl 2):4–12.
14. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics 1989;45:255–68.
15. Klebanoff MA, Yossef-Salameh L, Latimer C, Oza-Frank R, Kachoria R, Reagan PB, et al. Development and validation of an algorithm to determine spontaneous versus provider-initiated preterm birth in US vital records. Paediatr Perinat Epidemiol 2016;30:134–40.
16. Marascuilo LA. Large-sample multiple comparisons. Psychol Bull 1966;65:280–90.
17. Hartigan JA, Hartigan PM. The dip test of unimodality. Ann Statist 1985;13:70–84.
18. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999;94:496–509.
19. Flynn PM, Foster EM, Brost BC. Indicators of acculturation related to Somali refugee women's birth outcomes in Minnesota. J Immigr Minor Health 2011;13:224–31.
20. Tishkoff SA, Reed FA, Friedlaender FR, Ehret C, Ranciaro A, Froment A, et al. The genetic structure and history of Africans and African Americans. Science 2009;324:1035–44.
21. Fan S, Hansen ME, Lo Y, Tishkoff SA. Going global by adapting local: a review of recent human adaptation. Science 2016;354:54–9.
22. Belihu FB, Davey MA, Small R. Perinatal health outcomes of East African immigrant populations in Victoria, Australia: a population based study. BMC Pregnancy Childbirth 2016;16:86.
23. Chang HH, Larson J, Blencowe H, Spong CY, Howson CP, Cairns-Smith S, et al. Preventing preterm births: analysis of trends and potential reductions with interventions in 39 countries with very high human development index. Lancet 2013;381:223–34.
24. Lynch CD, Zhang J. The research implications of the selection of a gestational age estimation method. Paediatr Perinat Epidemiol 2007;21(suppl 2):86–96.
25. Johnson-Agbakwu CE, Flynn P, Asiedu GB, Hedberg E, Breitkopf CR. Adaptation of an acculturation scale for African refugee women. J Immigr Minor Health 2016;18:252–62.