Challenges and Opportunities of Using a National Database to Evaluate Racial/Ethnic Disparities and Breastfeeding Effects on Sudden Unexpected Infant Death

Abstract

Background:

Sudden unexpected infant death (SUID) rates remain higher in American Indian/Alaska Native (AI/AN) and non-Hispanic Black (NHB) infants than other demographic groups. Racial disparities are also evident in breastfeeding, which is associated with reduced risk of SUID.

Objective:

To assess the relationship between racial/ethnic disparities in SUID and breastfeeding beyond the newborn period using U.S. nationally reported public databases.

Methods:

Data were extracted from Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research (WONDER) and the National Immunization Surveys (NISs) 2009–2017. WONDER data were restricted to full-term infants and sorted by death year, race/ethnicity, and other characteristics. NIS breastfeeding data included ever breastfed, breastfed at 6 months, and exclusive breastfeeding at 3 and 6 months. Breastfeeding rates and mortality data were aggregated based on race/ethnicity, and mortality rates were analyzed by weighted (number of births) multivariable linear regression.

Results:

SUID rates were highest among NHB and AI/AN infants who also had the lowest breastfeeding rates. When breastfeeding and race/ethnicity were included in the analyses, race/ethnicity confounded the relationship between breastfeeding and SUID. When race was excluded, ever breastfeeding and any breastfeeding at 6 months were associated with significantly decreased SUID rates.

Conclusion:

Race/ethnicity confounded the relationship between breastfeeding and SUID. Analysis was limited because individual SUID rates were available for maternal/birth characteristics but not for breastfeeding. Our study showed a need for adding additional data points to other national databases to better understand the role that breastfeeding plays in the racial/ethnic disparities in SUID.

Introduction

Racial and ethnic disparities in sudden unexpected infant deaths (SUIDs) have been persistent and well documented in the United States. SUID rates, which include Sudden Infant Death Syndrome (SIDS) (ICD-19 R95), other ill-defined and unspecific causes of mortality (R99), and accidental suffocation or strangulation in bed (W75), have decreased significantly overall since the 1990s.¹ Despite this decline, rates have remained significantly higher in American Indians/Alaska Natives (AI/AN) and non-Hispanic Blacks (NHBs) than in other ethnic groups, and overall higher than most other high-income countries.^2,3

Influencing these trends are racial and ethnic disparities that exist in breastfeeding.⁴ Breastfeeding has been associated with a decreased rate of SIDS, and exclusive breastfeeding through 6 months is part of the American Academy of Pediatrics recommendations for SIDS prevention.^5,6 Given that approximately one-half of SUID cases are SIDS, breastfeeding is also associated with lower SUID rates.⁷ The rates of exclusive breastfeeding are significantly lower in NHB, AI/AN, and Hispanic groups, and it is possible that these lower rates may contribute to the racial and ethnic disparities seen in SUID.^8,9 Previous literature examining breastfeeding initiation using U.S. birth certificates has shown that breastfeeding initiation is associated with a significant reduction in infant mortality⁶ and SUID in particular.⁹

However, breastfeeding initiation does not play a substantial role in mediating disparities,⁹ and breastfeeding duration, beyond initiation, has not been examined. Therefore, further analysis of the extent to which any and exclusive breastfeeding beyond initiation impacts SUID rates will contribute to an understanding of existing racial and ethnic disparities and further help guide future interventions. Unfortunately, the potential for such analysis is limited, because breastfeeding data beyond initiation are not included in current linked birth/infant death records, but are collected separately through the National Immunization Survey (NIS).¹⁰ Despite this limitation, given the ample evidence of the protective effects of breastfeeding on infant mortality, it merits further analysis within the context of SUID.

With these factors in mind, the goal of this ecological study is to examine if there was an association between racial/ethnic disparities in breastfeeding with racial/ethnic disparities seen in SUID, adjusting for confounders among maternal and infant characteristics using publicly available databases. Our hypothesis was that there was an association between breastfeeding continuation and the racial/ethnic disparities seen in SUID.

Materials and Methods

Infant mortality data

Data were extracted from the Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research (CDC WONDER) database of Linked Birth/Infant Death Records.¹¹ Linked Birth/Infant Death Records data were compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative; the data were produced by the CDC, U.S. Department of Health and Human Services, and National Center for Health Statistics. SUID cases and clinical characteristics between 2009 and 2017 were included in the data from CDC WONDER.

Data were restricted to full-term infants aged 29–364 days to avoid confounding with preterm and early neonatal causes of death, as infants ≤28 days may have died from other causes.¹² The postneonatal SUID rates were limited to SIDS (ICD-10 R95), other ill-defined and unspecified causes of mortality (R99), and accidental suffocation or strangulation in bed (W75). Rates were then calculated by year of death and maternal race/ethnicity, using CDC WONDER assigned categories, as well as selected maternal and birth characteristics. Maternal characteristics selected for inclusion in analysis were maternal age (<20 versus >20), marital status (unmarried versus married), education level (less than or equal to high school degree or equivalent versus greater than high school degree), prenatal care (no prenatal care versus any prenatal care), birth weight (low birth weight versus normal birth weight or greater). Low birth weight was classified by weight <2,500 g, and normal birth weight or greater was ≥2,500 g.

Breastfeeding data

NIS breastfeeding data by race/ethnicity and time period were extracted from the 2009–2017 data set. NIS data are collected through random-digit dialing to survey households with children (ages 5–13) and teens (ages 13–17).¹⁰ These specifications were selected based on available data from the NIS sorted by race/ethnicity and breastfeeding time frames: ever breastfed (initiation), any breastfeeding at 6 months (exclusive or in combination with infant formula), and exclusive breastfeeding at 3 and 6 months.

Breastfeeding rates were converted into rates per 1,000 births to match the CDC WONDER reporting format by direct standardization using the number of births in the data included in each NIS race/ethnicity category. Rates for non-Hispanic Asian and non-Hispanic Hawaiian/Pacific Islander were combined to match the Asian Pacific Islander (API) designation in the CDC WONDER data set. The exception was AI/AN and Hawaiian/Pacific Islander designations in 2016 because they were excluded in that year's NIS data. AI/AN was excluded in NIS data post-2017, so 2018 and onward were excluded in this study's data aggregation.

Compilation and data analysis

SUID death and total birth data were extracted from CDC WONDER by year, race/ethnicity, and the maternal and birth characteristics listed earlier. Individual maternal and infant characteristics from WONDER, which uses birth certificate data, could not be matched with the NIS data set, which uses survey data. Therefore, data from NIS and CDC WONDER were combined by aligning breastfeeding rates based on year and race/ethnicity. Data sets were analyzed using weighted (by number of births) multivariable linear regression with calculated regression coefficients and 95% confidence limits for SUID rates. Analysis was conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA). Comparative analysis of SUID rates were presented with non-Hispanic White (NHW), and additional maternal/birth characteristics with lowest estimate of SUID rate, set as the referent groups. Afterward, separate multivariable linear regression analyses of maternal/birth characteristics and SUID were conducted for each breastfeeding time frame.

This study was reviewed by the Institutional Review Board of Cooper Medical School Rowan University and determined to be exempt.

Results

In total, 28,500 infant deaths were assembled from the CDC WONDER death certificates from 2009 to 2017, ∼3,167 deaths per year (Table 1). Aggregate SUID mortality rates per 1,000 births between 2009 and 2017 were lowest in infants born to mothers of API race and Hispanic ethnicity. Rates were highest in those of NHB or AI/AN at 1.62 and 1.89 per 1,000 deaths, respectively. Breastfeeding information was then gathered using NIS sample size from the same years—ever breastfed percentages stratified by race/ethnicity ranged from 67.0% to 85.3%. Any breastfeeding and exclusive breastfeeding at 3 and 6 months showed lower percentages, with exclusive breastfeeding at 6 months ranging from 13.3% to 23.1%. Percentages of ever breastfeeding were high for NHW and API, and comparably low for AI/AN and lowest for NHB. Percentages for any and exclusive breastfeeding among different races/ethnicities showed similar trends.

Table 1.

Postnatal Sudden Unexpected Infant Death (Per 1,000 Live Births) and Average Breastfeeding Rates by Race/Ethnicity, 2009–2017

Maternal race/ethnicity	Births (n)	Deaths (n)	SUID rate (per 1,000)	Ever breastfed (%)	Any breastfeeding at 6 months (%)	Exclusive breastfeeding at 3 months (%)	Exclusive breastfeeding at 6 months (%)
Asian/Pacific Islander	2,342,516	669	0.29	85.3	66.0	43.4	23.1
Hispanic	8,326,891	3,901	0.47	82.7	50.6	39	18.9
Non-Hispanic White	19,375,288	14,535	0.75	83.3	56.6	45.6	21.9
Non-Hispanic Black	5,344,292	8,676	1.62	67.0	40.2	28.3	13.3
American Indian/Alaska Native^a	347,624	657	1.89	73.8	42.7	36.9	16.2

SUID rates from CDC WONDER data 2009–2017, and breastfeeding rates from CDC National Immunization Study data for births in 2009–2017.⁹

American Indian/Alaska Native breastfeeding data are unavailable for 2016.

SUID, sudden unexpected infant death.

Regression analyses were conducted to determine if breastfeeding, independent of race/ethnicity, affected SUID rates (Table 2). The multivariable model including breastfeeding and race/ethnicity together demonstrated confounding and resulted in estimates of mortality for all types of breastfeeding duration/exclusivity that were not statistically significant. When excluding race/ethnicity from the analysis, all breastfeeding categories were significantly associated with decreased mortality from SUID (Table 2). Ever breastfeeding was associated with the greatest reduction of mortality, with 3.89 decreased deaths per 1,000 live births. Any breastfeeding at 6 months and exclusive breastfeeding at 6 months were associated with similar reduction rates of 2.20 and 2.03 decreased deaths per 1,000 live births, respectively.

Table 2.

Sudden Unexpected Infant Death Rates and Breastfeeding Duration/Exclusivity

	Univariable analysis ^a : mortality with breastfeeding without race			Multivariable analysis ^b : mortality with breastfeeding and race
Breastfeeding duration/exclusivity	Estimate of mortality	95% CI	p	Estimate of mortality	95% CI	p
Never breastfed	0	0	N/A	0	0	N/A
Ever breastfed	−3.895	−4.435 to −3.356	<0.0001	0.322	−0.716 to 1.360	0.543
Not any breastfed, 6 months	0	0	N/A	0	0	N/A
Any breastfed, 6 months	−2.199	−2.643 to −1.755	<0.0001	0.224	−0.477 to 0.925	0.531
Not exclusively breastfed, 3 months	0	0	N/A	0	0	N/A
Exclusively breastfed, 3 months	−1.688	−2.178 to −1.198	<0.0001	0.110	−0.632 to 0.851	0.772
Not exclusively breastfed, 6 months	0	0	N/A	0	0	N/A
Exclusively breastfed, 6 months	−2.030	−2.714 to −1.346	<0.0001	0.088	−0.735 to 0.910	0.835

Units are in numbers of excess or decreased deaths per 1,000 live births.

Univariable analysis of SUID estimates of mortality including only breastfeeding.

Multivariable analysis of SUID estimates of mortality including race and breastfeeding.

CI, confidence interval; N/A, not applicable; SUID, sudden unexpected infant death.

Finally, multivariable analyses exploring mortality and breastfeeding including covariates of maternal and birth characteristics, without race/ethnicity, presented in Table 3, were associated with significant decreases in SUID rates for ever breastfed and any breastfeeding at 6 months. Ever breastfed was associated with a 2.31 decrease in mortality per 1,000 deaths, and any breastfeeding at 6 months was associated with a 0.57 decrease (p-value <0.0001 and 0.0011, respectively). Mortality was not statistically associated with exclusive breastfeeding in this multivariable model.

Table 3.

Multivariable Analysis of Sudden Unexpected Infant Death Rates with Breastfeeding and Other Covariates Without Race/Ethnicity

SUID rates with breastfeeding and covariates without race/ethnicity
Parameter	Estimate of SUID (per 1,000 live births)	95% CI	p
Never breastfed	0	0	N/A
Ever breastfed	−2.309	−2.713 to −1.905	<0.0001
Unmarried	0	0	N/A
Married	−0.557	−0.613 to −0.500	<0.0001
Age <20	0	0	N/A
Age ≥20	−0.141	−0.243 to −0.039	0.007
High school education or less	0	0	N/A
Greater than high school education	−0.354	−0.408 to −0.299	<0.0001
No prenatal care	0	0	N/A
Any prenatal care	−0.874	−1.091 to −0.657	<0.0001
Low birth weight (<2,500 g)	0	0	N/A
Birth weight ≥2,500 g	−0.937	−1.078 to −0.796	<0.0001
Not any breastfed, 6 months	0	0	N/A
Any breastfed, 6 months	−0.567	−0.906 to −0.227	0.0011
Unmarried	0	0	N/A
Married	−0.603	−0.662 to −0.544	<0.0001
Age <20	0	0	N/A
Age ≥20	−0.145	−0.252 to −0.038	0.008
High school education or less	0	0	N/A
Greater than high school education	−0.348	−0.405 to −0.290	<0.0001
No prenatal care	0	0	N/A
Any prenatal care	−0.897	−1.125 to −0.669	<0.0001
Low birth weight (<2,500 g)	0	0	N/A
Birth weight ≥2,500 g	−0.962	−1.111 to −0.814	<0.0001
Not exclusively breastfed, 3 months	0	0	N/A
Exclusively breastfed, 3 months	0.108	−0.260 to 0.477	0.564
Unmarried	0	0	N/A
Married	−0.628	−0.687 to −0.569	<0.0001
Age <20	0	0	N/A
Age ≥20	−0.150	−0.258 to −0.042	0.006
High school education or less	0	0	N/A
Greater than high school education	−0.360	−0.418 to −0.302	<0.0001
No prenatal care	0	0	N/A
Prenatal care	−0.906	−1.135 to −0.678	<0.0001
Low birth weight	0	0	N/A
Birth weight ≥2,500 g	−0.967	−1.116 to −0.818	<0.0001
Not exclusively breastfed, 6 months	0	0	N/A
Exclusively breastfed, 6 months	0.215	−0.292 to 0.723	0.406
Unmarried	0	0	N/A
Married	−0.628	−0.687 to −0.570	<0.0001
Age <20	0	0	N/A
Age ≥20	−0.151	−0.258 to −0.043	0.0061
High school education or less	0	0	N/A
Greater than high school education	−0.361	−0.419 to −0.303	<0.0001
No prenatal care	0	0	N/A
Any prenatal care	−0.907	−1.135 to −0.678	<0.0001
Low birth weight (<2,500 g)	0	0	N/A
Birth weight ≥2,500 g	−0.967	−1.116 to −0.818	<0.0001

The values in bold are the SUID rate.

CI, confidence interval; N/A, not applicable; SUID, sudden unexpected infant death.

Discussion

We found that breastfeeding was associated with decreased SUID rates when excluding other variables in the analysis. However, combining this finding with race/ethnicity led to confounding of the relationship, likely due to the method of combing the two data sets based on race/ethnicity. Using these methods, we were unable to show that continued breastfeeding, and exclusive breastfeeding at 3 and 6 months were associated with lower rates of SUID once race/ethnicity and other variables were added to the equation. Individual patient data using linked data sets for continued breastfeeding would provide a more robust analysis of the degree to which breastfeeding beyond initiation may mediate the racial/ethnic disparities seen in SUID. Our findings support those of Li et al that improving breastfeeding can assist in infant mortality reduction, as her article found that breastfeeding initiation is associated with significant adjusted reduced odds ratio of overall infant deaths in race/ethnicity-stratified analysis.⁶

Many hypotheses have been put forward regarding the source of racial/ethnic differences in SUID, including cultural differences with regard to implementation of safe sleep practices.¹³ Examining causes of such disparities requires acknowledgement that many factors, such as poverty and access to care, are also associated with race/ethnicity. Although we explored major risk factors, the variables in our study do not capture the full extent of the socioeconomic as well as social vulnerability factors that influence health outcomes, health behaviors such as breastfeeding, and infant mortality. Acknowledging race/ethnicity as social constructs enables a deeper understanding of factors that may relate to health outcomes.^14,15

Limitations

There are several limitations to this study. The primary limitation is the use of aggregate data from the NIS applied to CDC WONDER data. Owing to the separation of the CDC WONDER data set and the methods used in the NIS data set, we cannot directly assess the relationship between breastfeeding and race/ethnicity at the aggregate level. Our study was ecological in nature, yet this approach is the only approach that is possible with the current publicly available national data. A national data set of SUID with individual patient data on breastfeeding continuation and exclusivity would aid in a large-scale more accurate analysis of the effects of breastfeeding on infant mortality, which would permit a multivariable analysis and further exploration of mediator and moderator effects.

The racial/ethnic groups used in these databases may mask immense diversity within any population. Hispanic ethnicity alone can be divided into numerous subgroups, each of which has its own cultural practices and norms that may contribute to differing health outcomes such as SUID. Studies of Hispanic ethnicity indicate that there is a great deal of variability in SUID rates; for example, SUID rates in Hispanics varied from 0.48 in Mexican infants to 0.18 in Central and South American infants.¹⁶ It would be misguided to make specific conclusions about what cultural differences may be contributing to differences in mortality rates through our study methodology, as we can only comment on rates and directionality of relations within each large subgroup. Furthermore, for the years selected in this data set we could only sort variables by maternal race/ethnicity, and could not account for paternal race. This points to the complexity of assigning such designations and using race and ethnicity as categories of study.

Conclusions

When combining large publicly available national databases to explore the association between breastfeeding and SUID, we found that race/ethnicity confounded the relationship because race/ethnicity was used as the linking factor of the aggregate data. When excluding race/ethnicity from the analysis, all breastfeeding categories were significantly associated with decreased mortality. Our study emphasizes the need to add breastfeeding duration and exclusivity (and other factors) as additional time data points in other large-scale national multistate birth and infant death databases, to study them in the context of race/ethnicity as well as other social determinants of health that play a role in racial disparities in SUID. Collecting such data could aid in other studying other disparities in infant mortality.

Footnotes

Authors' Contributions

Writing—review and editing (equal) by S.Y. Conceptualization (lead), writing—original draft (lead), data curation (lead), and writing—review and editing (equal) by M.D. Software (lead) and writing—review and editing (equal) by J.G. Data curation (supporting) and writing—review and editing (equal) by S.D. and N.B.B. Conceptualization (supporting), writing—original draft (supporting), and writing—review and editing (supporting) by M.C.B. and L.F.-W. All authors approved the final article as submitted and agree to be accountable for all aspects of the study.

Data Sharing Statement

Deidentified individual participant data are not publicly available due to institutional restrictions.

Author Disclosure Statement

No authors have financial conflicts of interest to disclose. L.F.-W. works as a consultant and physician lead for Communities and Hospitals Advancing Maternity Practices (CHAMPS) at Boston Medical Center in Boston, MA, USA and consultant to National Institute for Children's Health Quality. M.C.B. and L.F.-W. serve as unpaid positions on the Board of the Academy of Breastfeeding Medicine.

Funding Information

No funding was received for this article.

References

CDC. Data and Statistics for SIDS and SUID. Available from: https://www.cdc.gov/sids/data.htm [Last accessed: April 28, 2021].

Bartick

, Tomori

. Sudden infant death and social justice: A syndemics approach. Matern Child Nutr, 2019; 15(1):e12652; doi: 10.1111/mcn.12652

Parks

, Erek-Lambert

, Shapiro-Mendoza

. Racial and ethnic trends in sudden unexpected infant deaths: United States, 1995–2013. Pediatrics, 2017; 139(6):e20163844; doi: 10.1542/peds.2016-3844

Chiang

, Li

, Anstey

, et al. Racial and ethnic disparities in breastfeeding initiation horizontal line United States, 2019. MMWR Morb Mortal Wkly Rep, 2021; 70(21):769–774; doi: 10.15585/mmwr.mm7021a1

Eidelman

, Schanler

, Johnston

, et al. Breastfeeding and the use of human milk. Pediatrics, 2012; 129(3):e827–e841; doi: 10.1542/peds.2011-3552

, Ware

, Chen

, et al. Breastfeeding and post-perinatal infant deaths in the United States, A national prospective cohort analysis. Lancet Reg Health Am, 2022; 5:100094; doi: 10.1016/j.lana.2021.100094

NIH. SIDS Is Not...Safe Infant Sleep Basics; 2022. Available from: https://safetosleep.nichd.nih.gov/safesleepbasics/SIDS/SIDSisNot [Last accessed: March 15, 2022].

Jones

, Power

, Queenan

, et al. Racial and ethnic disparities in breastfeeding. Breastfeed Med, 2015; 10(4):186–196; doi: 10.1089/bfm.2014.0152

Bartick

, Barr

, Feldman-Winter

, et al. The role of breastfeeding in racial and ethnic disparities in sudden unexpected infant death: A population-based study of 13 million infants the United States. Am J Epidemiol, 2022; 191(7):1190–1201; doi: 10.1093/aje/kwac050

10.

National Immunization Surveys (NIS). CDC: Centers for Disease Control and Prevention; 2018. Available from: https://www.cdc.gov/vaccines/imz-managers/nis/index.html [Last accessed: March 15, 2022].

11.

CDC

WONDER.

CDC: Centers for Disease Control and Prevention; 2022. Available from: https://wonder.cdc.gov [Last accessed: March 15, 2022].

12.

Lavista-Ferres

, Anderson

, Johnston

, et al. Distinct populations of sudden unexpected infant death based on age. Pediatrics, 2020; 145(1):e20191637; doi: 10.1542/peds.2019-1637

13.

MacDorman

, Hoyert

, Mathews

. Recent declines in infant mortality in the United States, 2005–2011. NCHS Data Brief 2013(120):1–8.

14.

American Academy of Pediatrics Board of Directors and Executive Committee. AAP Perspective: Race-based medicine. Pediatrics, 2021; 148(4):e2021053829; doi: 10.1542/peds.2021-053829

15.

Mpody

, Ghimire

, Nafiu

. Looking through race-conscious or race-neutral lenses in pediatric research. Pediatrics, 2022; 149(3):408–423; doi: 10.1086/233900

16.

Shapiro-Mendoza

, Parks

, Lambert

, et al. The Epidemiology of Sudden Infant Death Syndrome and Sudden Unexpected Infant Deaths: Diagnostic Shift and other Temporal Changes. In: SIDS Sudden Infant and Early Childhood Death: The Past, the Present and the Future. (Duncan JR, Byard RW. eds.) University of Adelaide Press ©: Adelaide, Australia; 2018 The Contributors, with the exception of which is by Federal United States employees and is therefore in the public domain.; 2018.