Not-breastfeeding is a risk factor for SUID, most notably among Hispanic infants, NHW, and NHB infants. Our results also confirm the known disparities in SUID. The mediation analysis revealed the complex but not very substantial role that breastfeeding plays in the racial disparities of SUID, suggesting that breastfeeding plays a part in the higher SUID risk in NHB infants and the lower SUID risk in Hispanic infants. To our knowledge, this is the first study to formally examine the extent to which breastfeeding mediates the disparities in sudden infant death, although there have been several descriptive studies of the disparities in breastfeeding and SIDS/SUID rates,[7,25] and there is literature on disparities in either breastfeeding[42–45] or SIDS/SUID.[18,46]
The overall absolute risk difference, 0.12 per 1,000 live births, demonstrates that not-breastfeeding has at least a moderately strong association with SUID. This finding suggests that breastfeeding interventions may have an impact on SUID, assuming that the association between not-breastfeeding and SUID is causal and that interventions to increase breastfeeding are successful. Absolute risk differences in not-breastfeeding on SUID were only found for NHB, NHW, and Hispanic infants. Although the adjRD was also very high in AI/AN infants, it did not reach significance, because the number of SUID cases was relatively small.
Despite the higher risk of SUID relative to NHW infants, the mediation analysis showed that 2.3% of this racial disparity of SUID risk in NHB infants was mediated by breastfeeding, which is lower than we may have expected from our hypothesis. The aOR of not-breastfeeding on the risk of SUID was also lower than that of most other racial/ethnic groups. The low aOR (and low proportion mediated) is likely due to competing risks for which we did not completely adjust. Such competing risks would affect both breastfeeding and not-breastfeeding NHB populations and thereby dilute the effect of breastfeeding. These risks likely include social and structural determinants of health such as high US child poverty rates; transportation issues; lack of generational wealth; insecurity in housing, food, and income;[49,50] and other legacies of historical trauma, including ongoing structural racism. Additionally, risk factors that become evident beyond the immediate postpartum period may contribute to disparities and are unaccounted for in this analysis, particularly the effects of early cessation of breastfeeding. Higher breastfeeding rates might reduce only a small amount of the racial disparity seen in SUID relative to NHW infants.
AI/AN infants have a higher adjusted risk of SUID than NHW. The disparity in SUID relative to NHW infants is likely related to unmeasured confounders, including high levels of social adversity, which are almost certainly related to their legacy of historical trauma. Bedsharing in hazardous circumstances, such as with tobacco exposure or with adult alcohol use, particularly without breastfeeding, may be part of this legacy; these were associations with risk factors for which we could not completely adjust. For example, we could not measure parental alcohol use in this study, although it is known that there are high rates of alcohol use disorder in AI/AN communities. However, we were able to adjust for high rates of AI/AN maternal antenatal smoking, a potent SIDS risk factor.[8,52] Compared with other racial/ethnic groups, AI/AN infants had a very high absolute risk difference for not-breastfeeding on SUID, in addition to a relatively high OR of not-breastfeeding on SUID, although both effect estimates had very broad CIs that included unity. The negative mediation effect is explained by the fact that AI/AN mothers were likely to breastfeed at a higher rate than NHW mothers, once adjusted for background risks.
Thus, breastfeeding does not mediate the disparities in SUID risk relative to NHW infants, because AI/AN mothers are more likely to breastfeed than similarly disadvantaged NHW mothers. This does not mean breastfeeding is not important to reducing AI/AN SUID risk; indeed, SUID risk in AI/AN infants would be even higher if it were not for their higher adjusted likelihood of breastfeeding relative to NHW infants, and the disparities would be even greater.
Asian infants have a markedly lower SUID risk than NHW infants. Our data do not show that not-breastfeeding was associated with SUID risk in Asian infants. This finding must be interpreted very cautiously, given the broad CI around the effect estimate. However, the rarity of not-breastfeeding (9.9%) means that there may be too few exposed SUID cases to obtain precise effect estimates. We can only speculate that a less strong or absent association also reflects less residual confounding in the respective racial/ethnic group. Indeed, Asian mothers and infants had much lower rates of background risk factors, such as low rates of antenatal smoking, young maternal age, low education, and unmarried status. Such a pattern may possibly reflect a low prevalence of unmeasured confounders such as alcohol use or mental health problems. The small negative mediation effect by Asian race occurred due to lower likelihood of mothers to breastfeed relative to NHW mothers, once background risk was considered.
Breastfeeding mediated 2.1% of the difference in SUID between Hispanic and NHW infants. This finding partially explains the lower rate of SUID in Hispanic infants, because Hispanic mothers are more likely to breastfeed than NHW mothers and their infants have lower SUID risk. Our findings are consistent with the "Hispanic paradox" a phenomenon in which Hispanic Americans have some better health outcomes than do NHW Americans despite lower socioeconomic status, and better health outcomes than NHB Americans despite similar socioeconomic status. The mediation analysis suggests that breastfeeding is a small part of the reason why SUID rates are low relative to NHW and NHB infants. The high aOR may reflect that not-breastfeeding is a particularly rare behavior (highest breastfeeding rates) in Hispanic mothers and thus likely is both a risk factor and an indicator of certain background risks (technically unmeasured confounders) that contributes substantially to SUID occurrence.
Limitations and Strengths
One of the biggest limitations of our study is that the data set contains breastfeeding data only from hospital discharge, whereas SUID risk is affected by breastfeeding duration in a dose–response fashion, even beyond its peak incidence of 2–4 months of age. Thus, we likely overestimated breastfeeding rates and possibly biased our results toward the null. Next, missing data for California, with its large Hispanic population, may have resulted in an overestimation of overall SUID rates. Thus, our sample is not representative of the US population. We also lacked data on sleep position. However, sleep position should not be adjusted for, because it is 1 of the many components on the causal pathway between not-breastfeeding and sudden infant death. Our study lacked data on sleep environment, but presumed unsafe sleep environments (particularly bedsharing) often do not correlate with high sudden infant death rates, especially in populations with high breastfeeding rates.[7,18,55] Lack of data on alcohol consumption is also a limitation.
The strengths of our study are its large sample size, presence of several years of data, the ability to analyze breastfeeding as a mediator, and that we were able to adjust for a large number of covariates. We included covariates that have not received much attention in the SIDS/SUID literature, including maternal obesity,[56–60] parity, and maternal nativity. Maternal nativity appears to be an important covariate to include in future analyses and may have clinical and policy implications.
The role of social and structural determinants of health is particularly important for NHB and likely for AI/AN infants. These social and structural determinants of SUID risk, such as legacies of historical trauma, are likely present in both the breastfeeding and non-breastfeeding NHB and AI/AN populations and hence dilute the effect of breastfeeding. This is consistent with findings from Ware et al., who suggested that lack of effect of breastfeeding initiation on Black post-neonatal deaths could be related to multiple other social disadvantages.
Am J Epidemiol. 2022;191(7):1190-1201. © 2022 Oxford University Press