The influence of maternal life course exposure to neighborhood affluence and disadvantage on Black/White disparities in adverse birth outcomes in South Carolina

Abstract

Exposure of mothers to neighborhood affluence or disadvantage during their childhood or adulthood may influence infant birth outcomes. Concurrently, these neighborhood exposures may be differently prevalent and impactful on health by race. Using a multigenerational dataset of maternally linked birth certificates from South Carolina (1989-2020), we investigated associations between maternal life course neighborhood exposure to affluence (neighborhood college completion) and disadvantage (neighborhood poverty) and infant low birth weight (LBW) and preterm birth (PTB), and differences by maternal Black/White race. Black women had a higher prevalence of LBW and PTB than White women, overall and within every exposure category. Black women were more often exposed to neighborhood disadvantage (high poverty) and less often exposed to neighborhood affluence (high college completion), in childhood and in adulthood than White women. Life course high (vs low) affluence and low (vs high) disadvantage neighborhood exposures were protectively associated with LBW, though only the latter was protective for PTB. Though we did not find evidence of differential vulnerability by maternal race to life course neighborhood exposures, the greater life course exposure of Black mothers to less affluent and more disadvantaged neighborhoods explained up to 9% of racial disparities even without effect modification present.

Introduction

Neighborhoods impact health through multiple pathways, as physical and social conditions such as poverty, community-level education, unemployment, and racial residential segregation impact the psychosocial and physiological stress responses of their residents.¹ Particularly, neighborhood social disadvantage (eg, high neighborhood poverty) may dysregulate the stress pathways of mothers, inf luencing adverse birth outcomes like low birth weight (LBW) and preterm birth (PTB).^2-6 Meanwhile, neighborhood social affluence (eg, high neighborhood college completion) may be protective against adverse birth outcomes through mechanisms of resource concentration and greater collective efficacy.⁷ Though prior work has established the influence of neighborhood social context during pregnancy on birth outcomes,^4,8 early programming and cumulative stress hypotheses argue that neighborhood exposures from outside of the pregnancy period, and from across a mother’s life course, may additionally influence the health of her and her infant.⁹

Maternal life course neighborhood exposures may also provide insight into persistent Black/White disparities in birth outcomes. As racial differences in exposure to neighborhood social affluence and disadvantage during pregnancy explain some proportion of racial disparities in adverse birth outcomes,^10,11 a greater portion of this disparity may be attributable to life course differential exposure to neighborhood social conditions by race.¹² Structural racism systematically and disproportionately excludes Black individuals from accessing capital through housing, education, occupation, and wealth,^13-16 which may result in the exposure (and confinement) of Black mothers to more disadvantaged and less affluent neighborhoods across the life course compared to White mothers.¹⁷ This racial difference in life course neighborhood exposures may then contribute to racial disparities in adverse birth outcomes.

Black women may also be differentially vulnerable to neighborhood social conditions across the life course than White women. Prior literature suggests diminished returns to individual-level socioeconomic factors, such as education¹⁸ and family income,¹² for the birth outcomes of Black vs White women. Expansion of the theory of marginalization-related diminished returns to the neighborhood context argues that for Black women exposed to more affluent and less disadvantaged neighborhoods in adulthood than in early life, the positive impact of upward mobility on health may be limited due to concurrent exposure of Black women to racial discrimination, hypervisibility, and a loss of cultural security in more affluent (and often, more White) neighborhoods.^19,20 Further, similar neighborhood affluence for Black and White women may not translate to similar individual-level affluence.²¹ However, empirical evidence to support the theory of diminished returns at the neighborhood level is limited, as prior studies of life course trajectories have investigated Black^22,23 and White women separately,²⁴ or simply controlled for race.²⁵ One study in California conducted a race-stratified analysis of maternal lifetime exposure to neighborhood poverty and found a persistently elevated risk of PTB for Black, but not White, upwardly mobile women.⁸ However, this result may not generalize to areas with distinct histories of structural racism.²⁶ Among downwardly mobile women, Black women may fare worse than White women, particularly if ties to protective childhood resources are weaker for Black women than White women.¹⁸ And for women who remain always disadvantaged or always affluent, differential vulnerability to neighborhood social conditions by race may be exponentiated over time.

In the present study, we adopted the approach offered by Ward and coauthors²⁷ to study racial disparities in health outcomes by investigating Black/White differences in exposure to neighborhood affluence and disadvantage over the life course and Black/White differences in the association between these exposures and adverse birth outcomes. To do so, we utilized a multigenerationally linked dataset of birth certificates from South Carolina from 1989-2020 to measure, for each mother, her exposures in early life and in adulthood to neighborhood affluence and disadvantage, from which we then constructed life course trajectories. We first hypothesized that Black women, compared to White women, would be exposed to less affluent and more disadvantaged neighborhoods across both time points and would experience less life course upward neighborhood mobility (improvement in neighborhood conditions). Further, we hypothesized that while upward neighborhood mobility would be generally protective for birth outcomes compared to remaining in never affluent/always disadvantaged neighborhoods, the association between upward mobility and adverse birth outcomes would be less protective for Black women than White women.

Methods

Data sources and construction of study sample

This study utilized the South Carolina Multigenerational Birth Dataset, a collaboration between this research team and the South Carolina Department of Health. The dataset included maternally linked singleton birth certificates of mothers and infants from 1989 to 2020 to capture data on three generations: G1 grandmothers, G2 mothers, and G3 infants.²⁸ To capture the births of both G2 mothers and their offspring during our study period, G2 mothers in the multigenerational dataset could not be older than 31 years of age. As the average age of Black and White mothers at first birth in SC has historically been under this limit,^29-31 we included in our sample only the firstborn G3 births of first G2 daughters to avoid further issues of selection on fertility (Appendix S1, Figure S1). We also excluded G3 births with missing explanatory covariates (<1% of race-restricted sample). Our analytic sample included 82 332 total births, 39 741 to Black G2 mothers (48.3%) and 42 591 to White G2 mothers (51.7%) (Table 1).

Table 1.

Descriptive characteristics of the sample of G3 infants born to Black and White G2 mothers in South Carolina, 2004-2020.

			Mother’s race
Characteristic	Overall N = 82 332		Non-Hispanic White N = 42 591		Non-Hispanic Black N = 39 741		P-valuea
	n	%	N	%	n	%
G3 infant characteristics
LBW, n (%)	8358	10.2	3034	7.1	5324	13.4	<.001*
PTB, n (%) b	7960	9.7	3616	8.5	4344	10.9	<.001*
Female infant sex, n (%)	40 306	49.0	20 673	48.5	19 633	49.4	.01*
Year of birth	Range: 2004-2020		Range: 2004-2020		Range: 2004-2020		NA
G2 early life (G1 adulthood) neighborhood exposures
Year of birth	Range: 1989-2009		Range: 1989-2007		Range: 1989-2009		NA
G2 early life neighborhood college completionc, n (%)							All <.001*
Median (IQR) value	10.51 (7.45-16.40)		10.97 (7.55-17.58)		10.01 (7.29-15.37)
<median (less affluent)	41 174	50.0	19 921	46.8	21 253	53.5
≥median (more affluent)	41 158	50.0	22 670	53.2	18 488	46.5
G2 early life neighborhood povertyd, n (%)						All <.001*
Median (IQR) value	15.82 (9.86-23.38)		12.27 (8.26-17.92)		21.02 (13.99-28.60)
≥median (more disadvantaged)	41 154	50.0	13 938	32.7	27 216	68.5
<median (less disadvantaged)	41 178	50.0	28 653	67.3	12 525	31.5
G2 age, adulthood neighborhood exposures, and life course mobility trajectories
G2 maternal age (years)	Mean (SD): 20.9 (3.4)		Mean (SD): 21.3 (3.6)		Mean (SD): 20.4 (3.2)		<.001*
G2 adulthood neighborhood college completionc, n (%)							All <.001*
Median (IQR) value	16.11 (10.93-25.46)		16.84 (11.38-26.26)		15.32 (10.51-24.45)
<median (less affluent)	41 167	50.0	19 909	46.7	21 258	53.5
≥median (more affluent)	41 165	50.0	22 682	53.3	18 483	46.5
G2 adulthood neighborhood povertyd, n (%)						All <.001*
Median (IQR) value	18.28 (11.89-25.85)		15.82 (10.25-22.12)		21.47 (14.44-30.16)
≥median (more disadvantaged)	41 156	50.0	16 810	39.5	24 346	61.3
<median (less disadvantaged)	41 176	50.0	25 781	60.5	15 395	38.7
Neighborhood college completion trajectory, n (%)							<.001*
Never affluent (low-low)	24 934	30.4	11 694	27.6	13 240	33.4
Downward mobility (high-low)	16 233	19.7	8215	19.3	8018	20.2
Upward mobility (low-high)	16 240	19.8	8227	19.3	8013	20.2
Always affluent (high-high)	24 925	30.1	14 455	33.8	10 470	26.2
Neighborhood poverty level trajectory, n(%)						<.001*
Always disadvantaged (high-high)	25 573	31.1	7309	17.2	18 264	46.0
Downward mobility (low-high)	15 583	18.9	9501	22.3	6082	15.3
Upward mobility (high-low)	15 581	19.0	6629	15.6	8952	22.5
Never disadvantaged (low-low)	25 595	31.0	19 152	44.9	6443	16.2
Neighborhood college completion mobility, n (%)
Proportion upwardly mobilee	16 240	39.4	8227	41.3	8013	37.7	<.001*
Proportion downwardly mobilef	16 233	39.4	8215	36.2	8018	43.4	<.001*
Neighborhood poverty level mobility, n (%)
Proportion upwardly mobile	15 581	37.9	6629	47.6	8952	32.9	<.001*
Proportion downwardly mobile	15 581	37.9	6629	33.2	8952	48.6	<.001*

Abbreviations: IQR, interquartile range; LBW, low birth weight; PTB, preterm birth.

^aComparing characteristics of Black and White G2 mothers, where * indicates P < .05 (two-sided).

^bMissing PTB outcome, n = 30.

^cNeighborhood college completion is measured as the percent of adults aged 25 and older with a college education within a given census tract.³² The metric is measured in the year of birth of the G2 (G2 early life) and the year of birth of the G3 (G2 adulthood), based on the respective residence of the G1 or G2 at the time.

^dNeighborhood poverty level is measured as the percent of residents with family income below the poverty line within a given census tract.³³ The metric is measured in the year of birth of the G2 (G2 early life) and the year of birth of the G3 (G2 adulthood), based on the respective residence of the G1 or G2 at the time.

^eThe proportion of upwardly mobile individuals is calculated as the number of individuals with a trajectory of upward mobility, out of the number of individuals with less affluent/more disadvantaged early life exposure.

^fThe proportion of downwardly mobile individuals is calculated as the number of individuals with a trajectory of downward mobility, out of the number of individuals with more affluent/less disadvantaged early life exposure.

From G2 birth certificates we drew information on G1 grandmothers’ geocoded census tract of residence in the G2 year of birth, ie, G2 “early life” (1989-2009). From G3 birth certificates we drew information on G3 birth outcomes, G2 adulthood characteristics (eg, maternal age, race, educational attainment), and the G2′s census tract of residence in the G3 year of birth, ie, G2 adulthood (2004-2020). We then integrated data from the U.S. Census (1990 and 2000) and American Community Survey (ACS) 5-Year Estimates (2005-2009 through 2020) to measure neighborhood exposures at the census tract level for G2 mothers in early life and in adulthood. Intercensal years were interpolated.³⁴

Exposure

We utilized ACS-derived data to measure neighborhood exposures of each G2 mother in early life and in adulthood. We conceptualized the census tract as a sociocultural proxy for the neighborhood.⁶ To represent affluence we used the percent completion of college among people aged 25 years and older in the census tract.³² To represent disadvantage we used the percent of residents with family income below the poverty threshold in the census tract.³³ We chose these measures as they each represent neighborhood social conditions that are mechanistically linked to birth outcomes through causal pathways including resource concentration and psychosocial stress.^{1,3,4,6,7,35-37}

We dichotomized both neighborhood measures around generation-specific median levels of each exposure^22,38 to represent neighborhood high/low affluence (ie, median level of college education or higher, or less than median college education) and high/low disadvantage (ie, median level of poverty or higher, or less than median poverty level) in G2 early life and G2 adulthood. A median split of neighborhood exposures emphasized a relative approach to change (ie, movement between the top and bottom halves of the distribution) over a mother’s life course. Dichotomization of each exposure allowed construction, for each measure, of four trajectories representing the life course exposures of the G2 mothers: never affluent/always disadvantaged (ie, <median college completion or ≥median poverty at both time points); upwardly mobile (improvement in position in the distribution of either measure over time); downwardly mobile (worsening of position in the distribution of either measure over time); and always affluent/never disadvantaged (ie, ≥median college completion or <median poverty at both time points).

Outcome

We studied G3 infant LBW, defined as live birth at a weight <2500 grams, and PTB, defined as live birth at a gestational age <37 weeks. Both LBW and PTB were categorized using the Alexander method and the birth weight and gestational age listed on G3 birth certificates.³⁹ Both LBW and PTB were evaluated in separate models.

Covariates

We used a basic adjustment set of variables from G3 birth certificates: G2 maternal age (in years; mean-centered linear and quadratic forms); G3 infant sex (male, female); and G3 infant year of birth. Our choice of covariates was based on our expectation that other characteristics of G2 mothers in adulthood (eg, educational attainment) would mediate the associations between early life and adulthood neighborhood social conditions, and thus would be inappropriate to include as controls in our life course model (Appendix S1; Figure S2).

Effect modifier

We defined race as a social, not a biologic, construct⁴⁰ that determines differential exposure to interpersonal and structural racism over time and subsequently impacts access to resources and health.⁴¹ On SC birth certificates, race and ethnicity (Hispanic origin) are defined separately for the mother, father, and infant by the informant, usually the mother.^42,43 We utilized G2 maternal race, as listed on G3 birth certificates, as the effect modifier. Only non-Hispanic Black and non-Hispanic White G2 mothers were included to both focus on the Black/White disparity in adverse birth outcomes and because of the small number of births to G2 women of other racial or ethnic groups in our multigenerational sample (n = 3018; Appendix S1, Figure S1).

Statistical analysis

We first performed Chi square tests and Student’s t-tests to describe the characteristics (including outcome and exposure prevalence) of Black and White G2 women in our sample.

We next utilized logistic regression models to estimate associations between G2 life course trajectories of neighborhood social conditions and G3 adverse birth outcomes. We evaluated neighborhood affluence and neighborhood disadvantage separately for LBW and PTB. Models were first unadjusted, then adjusted for the covariates listed above (Model 1). We first hypothesized that, relative to the never affluent/always disadvantaged reference trajectories, the downward neighborhood mobility, upward neighborhood mobility, and always affluent/never disadvantaged trajectories would be protective for birth outcomes, with the always affluent/never disadvantaged trajectory demonstrating the strongest protective association.

To test our second hypothesis regarding differences in the associations between life course neighborhood trajectories and adverse birth outcomes by G2 maternal race, we introduced interaction terms between trajectories and race (Model 2). We first tested for effect modification on the log-odds scale (multiplicative interaction) using F-tests; if interactions between trajectory and race were statistically significant, models were subsequently stratified by race. We next tested for effect modification on the probability scale (additive interaction). To do so, we estimated the race-specific predicted prevalences of adverse birth outcomes overall and within each trajectory using marginal effects from race-stratified models, from which we estimated Black/White contrasts of trajectory differences.

Finally, to quantify how much of the Black/White disparity in adverse birth outcomes was attributable to maternal life course trajectories of neighborhood exposure, we calculated the percent change in the race odds ratio (ie, Black vs White G2 maternal race) from the unadjusted to the adjusted model for each of our four outcome-exposure combinations.

$$\begin{gathered} \text{Percent}(\%)\text{disparity explained}=\frac{(\text{Race OR}-1)-(\text{Race aOR}-1)}{\text{Race OR}-1} \\ \times 100 \end{gathered}$$

We tested both single-level and multilevel versions of our model to represent the nesting of G3 births within neighborhoods. The ICC value of the multilevel model was low (0.004), and AIC values of the two models were comparable, suggesting that use of the single-level model was more parsimonious. In our final models, standard errors were clustered at the census tract of G2 maternal adulthood residence. For all results, we reported odds ratios (OR) and 95% CIs as well as two-tailed P-values with significance at α < .05.

Robustness checks

Our first robustness check (Check 1) assessed whether geographic mobility over the life course might introduce selection bias. We stratified the sample by geographic mobility: “stayers” (G2 mothers that remained in the same county over time) and “movers” (mothers whose county of residence changed from birth to adulthood). We repeated all analyses within each subgroup. Our second robustness check (Check 2) followed Liu and coauthors²⁵ and utilized an absolute specification of mobility (vs the relative specification in our main analyses), defined as whether either absolute poverty level decreased (decrease in disadvantage) or absolute college completion rate increased (increase in affluence) between G2 early life and adulthood.²⁵ We evaluated upward mobility vs any other neighborhood change. In the third robustness check (Check 3) we repeated all analyses on a non-firstborn-restricted sample (Appendix S1, Figure S1). All G3 infants of the same mother were assigned the same exposure based on her adulthood neighborhood at the firstborn’s birth. We repeated all analyses using a hierarchical logistic regression that clustered G3 infants within G2 mothers. Our fourth robustness check (Check 4) repeated all analyses using a full adjustment set, including covariates that potentially mediated the associations between maternal life course neighborhood trajectories and birth outcomes (Appendix S1, Figure S2). Our fifth robustness check (Check 5) used quartiles of early life and adulthood neighborhood exposure to define 16 life course trajectories, with the purpose of considering additional sensitivities along the distribution of neighborhood conditions.

This study was approved by the Institutional Review Board at the University of Michigan under IRB HUM00207601 and the South Carolina Department of Health and Human Services (IRB.22-002).

Results

Table 1 describes the sample, including prevalences of exposure to early life, adulthood, and life course trajectory neighborhood social conditions. Table S1 (Appendix S1) provides data on the prevalence of adverse birth outcomes overall and by maternal race across early life, adulthood, and life course trajectory exposures. Figure 1 follows the framework of Ward et al.²⁷ to provide a visual representation of Black/White disparities in the unadjusted prevalence of LBW and PTB across maternal trajectories of life course neighborhood affluence and disadvantage.

Figure 1.

Evaluation of Black/White disparity in the unadjusted prevalence (%) of adverse birth outcomes (LBW, low birth weight; PTB, preterm birth) across trajectories of G2 maternal neighborhood affluence (high/low college completion) and neighborhood disadvantage (high/low poverty level), by trajectory prevalence and maternal Black/White race, among G3 infants born in South Carolina (2004-2020). Panel A (top) depicts LBW and life course neighborhood trajectories. Panel B (bottom) depicts PTB and life course neighborhood trajectories. Life course neighborhood trajectories include never affluent/always disadvantaged (low college completion or high poverty in both childhood and adulthood), upward mobility, downward mobility, and always affluent/never disadvantaged. The racial disparity in a birth outcome can be measured as the difference along the y-axis between Black and White outcomes. The racial disparity in an exposure can be measured as the difference in outer circle size in the same x-axis category between Black and White women. The lack of effect modification by Black/White race is depicted as a similar trend in birth outcome prevalence across levels of exposure for Black and White women.

Black/White differences in the prevalence of adverse birth outcomes

Both LBW and PTB were both more prevalent overall among G3 infants born to Black G2 women than among those born to White G2 women (LBW: Black, 13.4% vs White, 7.1%; PTB: Black, 10.9% vs White, 8.5%; both P < .001; Table 1). This disparity persisted in every level and type of neighborhood exposure studied (Appendix S1, Table S1), including across trajectories (Figure 1; Panel A, LBW and Panel B, PTB).

Black/White differences in the prevalence of neighborhood exposures

Consistent with our first hypothesis, Black and White women were differentially exposed to neighborhood conditions across the life course. Black G2 mothers were more likely than White G2 mothers to have been born into neighborhoods with low affluence and high disadvantage (early life low college completion: Black 53.5%, White 46.8%; early life high poverty level: Black 68.5%, White 32.7%; both P < .001; Table 1) and were less likely to be upwardly mobile out of these neighborhoods (college completion proportion upwardly mobile, of those with less affluent early life exposure: Black 37.7%, White 41.3%; poverty level proportion upwardly mobile: Black 32.9%, White 47.6%; both P < .001; Table 1). For women born into neighborhoods with high affluence/low disadvantage, Black G2 women were more likely to be downwardly mobile than White women (college completion proportion downwardly mobile, of those with more affluent early life exposure: Black 43.4%, White 36.2%; poverty level proportion downwardly mobile: Black 48.6%, White 33.2%; both P < .001; Table 1). In total, Black G2 mothers were more likely to experience life course neighborhood never affluent/always disadvantaged trajectories compared to White G2 mothers (neighborhood college completion, never affluent trajectory prevalence: Black 33.4%, White 27.6%; neighborhood poverty, always disadvantaged trajectory prevalence: Black 46.0%, White 17.2%; both P < .001).

Life course neighborhood exposures and adverse birth outcomes

Using G2 maternal life course neighborhood trajectories as the exposure, only the always affluent/never disadvantaged trajectories of neighborhood college completion and neighborhood poverty (compared to the never affluent and always disadvantaged trajectories, respectively) were statistically associated with lower adjusted odds of LBW (college always (vs never) affluent, aOR 0.86 [95% CI, 0.80-0.91]; poverty never (vs always) disadvantaged, aOR 0.89 [95% CI, 0.83-0.94]; Table 2, Model 1). Full model output appears in Tables S2 and S3 (Appendix S1). Null associations were found between upward mobility and LBW (college upward mobility vs never affluent, aOR 0.97 [95% CI, 0.90-1.04]; poverty upward mobility vs always disadvantaged, aOR 0.98 [95% CI, 0.92-1.04]) as well as between downward mobility and LBW (college downward mobility vs never affluent, aOR 1.00 [95% CI, 0.94-1.06]; poverty downward mobility vs always disadvantaged, aOR 0.97 [95% CI, 0.91-1.04]).

Table 2.

Regression results from multivariable logistic regressions of low birth weight (LBW) among G3 infants (n = 82 332) born in South Carolina, 2004-2020, and G2 maternal (A) neighborhood college completion and (B) neighborhood poverty level across trajectories.

Risk factors	Unadjusted models Unadj. OR [95% CI]	Model 1: Life course trajectorya Adj. OR [95% CI]	Model 2: race and trajectory interacteda Adj. OR [95% CI]
(A) Neighborhood college completion (affluence)
Black (vs White) G2 maternal race	2.02 [1.92, 2.12]b	1.98 [1.88, 2.08]b	1.91 [1.74, 2.09]b
Life course neighborhood trajectory
Never affluent (Ref)	1.00	1.00	1.00
Downward mobility	0.97 [0.91, 1.04]	1.00 [0.94, 1.06]	1.00 [0.90, 1.11]
Upward mobility	0.94 [0.87, 1.01]	0.97 [0.90, 1.04]	0.96 [0.86, 1.07]
Always affluent	0.79 [0.74, 0.84]b	0.86 [0.80, 0.91]b	0.80 [0.72, 0.88]b
Black race × life course neighborhood trajectory
Black × never affluent (Ref)	NA	NA	1.00
Black × downward mobility	NA	NA	0.99 [0.86, 1.14]
Black × upward mobility	NA	NA	1.01 [0.88, 1.16]
Black × always affluent	NA	NA	1.13 [0.99, 1.29]
(B) Neighborhood poverty level (disadvantage)
Black (vs White) G2 maternal race	2.02 [1.92, 2.12]b	1.93 [1.82, 2.04]b	2.02 [1.82, 2.24]b
Life course neighborhood trajectory
Always disadvantaged (Ref)	1.00	1.00	1.00
Downward mobility	0.79 [0.74, 0.84]b	0.97 [0.91, 1.04]	1.02 [0.91, 1.14]
Upward mobility	0.90 [0.84, 0.96]b	0.98 [0.92, 1.04]	1.03 [0.90, 1.17]
Never disadvantaged	0.65 [0.61, 0.69]b	0.89 [0.83, 0.94]b	0.93 [0.83, 1.03]
Black race × life course neighborhood trajectory
Black × always disadvantaged (Ref)	NA	NA	1.00
Black × downward mobility	NA	NA	0.94 [0.82, 1.09]
Black × upward mobility	NA	NA	0.94 [0.81, 1.09]
Black × never disadvantaged	NA	NA	0.93 [0.82, 1.07]
F-test results: College, P = .30; Poverty, P = .82

Abbreviations: NA, not applicable; OR, odds ratio.

F-tests report two-sided P-values (α < .05).

^aModels 1 and 2 adjust for G2 maternal age (in years, mean-centered and quadratic), infant sex (female/male), infant year of birth, and clustering of births at the G2 adulthood census tract of residence. All models were conducted separately for (a) neighborhood college completion and (b) neighborhood poverty level. The reference group for Black maternal race is White maternal race. The reference group for life course trajectory is never affluent/always disadvantaged (ie, worse than median neighborhood conditions in G2 childhood and adulthood).

^bStatistical significance of the OR at P < .05 (two-sided).

For PTB, only the always affluent trajectory of neighborhood college completion was associated with lower adjusted odds of PTB (college always vs never affluent, aOR 0.90 [95% CI, 0.85-0.96]; Table 3, Model 1); null associations were observed for all trajectories of neighborhood poverty (eg, poverty never (vs always) disadvantaged, aOR 0.98 [0.92-1.05]; Table 3, Model 1). Full model output appears in Tables S4 and S5 (Appendix S1). Again, we observed no statistically significant associations between the upward mobility trajectories and PTB (college upward mobility vs never affluent, aOR 0.97 [95% CI, 0.91-1.04]; poverty upward mobility vs always disadvantaged, aOR 0.99 [95% CI, 0.93-1.06]). This finding for both LBW and PTB was counter to our second hypothesis that the upward mobility trajectory, compared to never affluent/always disadvantaged trajectories, would have a protective association with adverse birth outcomes.

Table 3.

Regression results from multivariable logistic regressions of PTB among G3 infants (n = 82 302) born in South Carolina, 2004-2020, and G2 maternal (A) neighborhood college completion and (B) neighborhood poverty level across trajectories.

Risk factors	Unadjusted models Unadj. OR [95% CI]	Model 1: life course trajectorya Adj. OR [95% CI]	Model 2: race and trajectory interacteda Adj. OR [95% CI]
(A) Neighborhood college completion (affluence)
Black (vs White) G2 maternal race	1.32 [1.26, 1.39]b	1.32 [1.25, 1.38]b	1.32 [1.21, 1.44]b
Life course neighborhood trajectory
Never affluent (Ref)	1.00	1.00	1.00
Downward mobility	0.94 [0.88, 1.00]	0.95 [0.89, 1.01]	0.97 [0.87, 1.07]
Upward mobility	0.96 [0.90, 1.03]	0.97 [0.91, 1.04]	0.97 [0.88, 1.07]
Always affluent	0.88 [0.83, 0.93]b	0.90 [0.85, 0.96]b	0.90 [0.82, 0.99]b
Black race × Life course neighborhood trajectory
Black × never affluent (Ref)	NA	NA	1.00
Black × downward mobility	NA	NA	0.97 [0.84, 1.11]
Black × upward mobility	NA	NA	1.01 [0.88, 1.15]
Black × always affluent	NA	NA	1.00 [0.89, 1.14]
(B) Neighborhood poverty level (disadvantage)
Black (vs White) G2 maternal race	1.32 [1.26, 1.39]b	1.31 [1.25, 1.38]b	1.36 [1.24, 1.50]b
Life course neighborhood trajectory
Always disadvantaged (Ref)	1.00	1.00	1.00
Downward mobility	0.86 [0.81, 0.93]b	0.95 [0.88, 1.02]	0.99 [0.88, 1.10]
Upward mobility	0.96 [0.90, 1.03]	0.99 [0.93, 1.06]	0.97 [0.86, 1.09]
Never disadvantaged	0.87 [0.82, 0.93]b	0.98 [0.92, 1.05]	1.03 [0.94, 1.13]
Black race × Life course neighborhood trajectory
Black × always disadvantaged (Ref)	NA	NA	1.00
Black × downward mobility	NA	NA	0.94 [0.81, 1.09]
Black × upward mobility	NA	NA	1.04 [0.90, 1.21]
Black × never disadvantaged	NA	NA	0.90 [0.79, 1.03]
F-test results: College, P = .91; Poverty, P = .25

Abbreviations: NA, not applicable; OR, odds ratio; PTB, preterm birth.

F-tests report two-sided P-values (α < .05).

^aModels 1 and 2 adjust for G2 maternal age (in years, mean-centered and quadratic), infant sex (female/male), infant year of birth, and clustering of births at the G2 adulthood census tract of residence. All models were conducted separately for (a) neighborhood college completion and (b) neighborhood poverty level. The reference group for Black maternal race is White maternal race. The reference group for life course trajectory is never affluent/always disadvantaged (ie, worse than median neighborhood conditions in G2 childhood and adulthood).

^bStatistical significance of the odds ratio at P < .05 (two-sided).

Black/White differences in the association between life course neighborhood exposures and adverse birth outcomes

We found no statistical evidence of effect modification by G2 maternal race of the association between life course neighborhood trajectories of affluence or disadvantage and adverse birth outcomes on either the multiplicative scale (eg, no statistical significance of the interaction term between Black vs White maternal race and upward mobility vs never affluent trajectories; LBW, Table 2, Panel A, Model 2, aOR 1.01 [95% CI, 0.88-1.16]; PTB, Table 3, Panel A, Model 2, aOR 1.01 [0.88, 1.15]; all F-test P > .05) or the additive scale (eg, difference in the predicted prevalence of LBW, always affluent vs upward mobility trajectories, White: −1.2 [95% CI, −2.2 to −0.2], Black: −0.8 [−2.3, 0.16]; P > .05; Appendix S1, Table S6). A lack of evidence for differential vulnerability to life course neighborhood trajectories for adverse birth outcomes by G2 maternal race ran counter to our hypothesis.

Proportion of Black/White disparities explained by life course neighborhood exposures

Adjustment for maternal life course neighborhood trajectory of affluence in models of LBW lowered the odds ratio comparing Black vs White mothers from 2.02 (unadjusted OR, 95% CI [1.92-2.12]) to 1.98 (adjusted OR, 95% CI [1.88-2.08]), representing a 3.9% reduction in the Black-White disparity in LBW (Table 2). However, adjustment for neighborhood affluence did not meaningfully change the association between maternal race and PTB (unadjusted OR, 1.32 [95% CI, 1.26-1.39]; aOR, 1.32 [1.25-1.38]; 0% of disparity explained; Table 3). Adjustment for maternal life course neighborhood trajectory of disadvantage explained 8.8% of the Black-White disparity in LBW (aOR, 1.93 [95% CI, 1.82-2.04]; Table 2) and 3.1% of the disparity in PTB (aOR, 1.31 [95% CI, 1.25-1.38]; Table 3).

Robustness checks

Results from robustness checks appear in Appendices S2-S6. For Check 1 (Appendix S2, Tables S7-S14), results were generally robust among stayers (n = 57 151; ie, stayers exhibited a protective association for the always affluent/never disadvantaged trajectories for LBW, and the always affluent trajectory for PTB) but were not consistently robust among movers (n = 25 181; associations were weaker and no longer statistically significant for the never disadvantaged trajectory for LBW or for any trajectories for PTB). For Check 2 (Appendix S3, Tables S15-S18), we observed no statistically significant associations between absolute upward mobility in neighborhood conditions and LBW or PTB, or evidence of effect modification by maternal race. For Check 3 (Appendix S4, Tables S19-S22), all results were qualitatively consistent with our main analyses and differences in precision are likely attributable to the increased sample size (N = 148 217). For Check 4 (Appendix S5, Table S23), results moderately attenuated with control for the full set of covariates but remained consistent with main results. For Check 5 (Appendix S6, Tables S24 and S25), results supported our median split approach as most differences appeared across the 50th percentile and we found no evidence of effect modification by maternal race.

Discussion

We utilized a multigenerational dataset of maternally linked birth certificates (1989-2020) from South Carolina to evaluate racial disparities in associations between maternal life course exposure to neighborhood affluence (defined by college completion) and neighborhood disadvantage (defined by poverty) and infant LBW and PTB. Following the approach of Ward and coauthors²⁷, we first found that Black women in our sample had a higher prevalence of both LBW and PTB than White women overall and within every level of exposure, including a substantial racial disparity in adverse birth outcomes even among women with always affluent and never disadvantaged life course neighborhood exposures. Second, in support of our first hypothesis we found that Black women were exposed to less affluent neighborhoods (those with lower rates of college completion) and more disadvantaged neighborhoods (those with higher poverty levels) in early life, in adulthood, and across the life course compared to White women. Third, we found no evidence of effect modification by maternal race of the association between life course neighborhood exposures and birth outcomes. Nonetheless, our findings suggest moderate contributions of mothers’ life course neighborhood trajectories to racial disparities in birth outcomes. For LBW, both neighborhood college completion and neighborhood poverty contributed to Black-White disparities, with poverty explaining a larger portion (8.8%) than education (3.9%). In contrast, for PTB, neither neighborhood affluence nor disadvantage across the life course explained more than 3% of the Black-White disparity, suggesting that other, unmeasured factors may be more inf luential.

We additionally investigated how maternal life course neighborhood exposures were associated with adverse birth outcomes. Though prior studies have linked both early-life and adulthood maternal exposures to birth outcomes using individual socioeconomic indicators,^18,44 few have analyzed women’s neighborhood mobility across the life course. Existing research has focused on neighborhood’s poverty histories,⁴⁵ interpregnancy exposure changes,^46,47 race-specific analyses,^22-24 or settings not generalizable to the US South.⁸ In our work, only the always affluent/never disadvantaged trajectories of life course neighborhood exposure were protective compared to the never affluent/always disadvantaged trajectories. In their longitudinal study of California births (1982-2011), Pearl and coauthors⁸ similarly found that women with consistently low (<20%) neighborhood poverty (analogous to our “never disadvantaged” trajectory) had the lowest prevalence of PTB.⁸ However, while their results showed a gradient in outcomes across trajectories, we found similar outcomes in the always disadvantaged, upwardly mobile, and downwardly mobile trajectories. One interpretation of our findings of lower prevalences of LBW and PTB in the never disadvantaged trajectories vs all others is that any life course exposure to neighborhood disadvantage in SC—in early life or in adulthood, and regardless of duration—is detrimental to birth outcomes. In this case, upward mobility may not be protective if lower neighborhood disadvantage in adulthood cannot compensate for early life high disadvantage. Another interpretation is of differences in approach: Pearl and coauthors focused on movement across the 20% neighborhood poverty level cutoff, while we focused on movement across the 50th percentile. As a result, their threshold may capture different transitions in neighborhood context than ours does. Nonetheless, lower-than-median neighborhood disadvantage in South Carolina may also be qualitatively different from low neighborhood disadvantage in CA, as SC counties rank more highly overall on the Index of Deep Disadvantage⁴⁸—this may also account for much smaller (or negligible) health benefits from upward mobility in SC.

We did not find evidence to support our third hypothesis that maternal race modified the association between life course neighborhood exposure and birth outcomes, on either the multiplicative (Tables 2 and 3) or additive scale (Table S6). This contrasts with prior studies of maternal adulthood-only exposures that estimated weaker^4,49,50 (in line with the theory of marginalization-related diminished returns^19,20) or stronger⁶ neighborhood effects for Black women than White women. Pearl and coauthors⁸ found that upward mobility was protective for White mothers in CA but not for Black mothers,⁸ supporting the theory of diminished returns and suggesting that adulthood low disadvantage better compensated for early life high disadvantage for White women than Black women. Previously, the theory of “nonequivalence” of early life exposures between Black and White women has been used to explain racial disparities in the health impact of life course trajectories.²⁰ However, in our study, neighborhood exposures may actually have been more equivalent by race than in California: in South Carolina, 45% of White and 16% of Black mothers resided in low disadvantage (low poverty) neighborhoods (1989-2020) compared to 72% of White mothers and 24% of Black mothers in California (1982-2011).⁸ These narrower differences in exposure to neighborhood affluence and disadvantage by race in South Carolina compared to other settings may ref lect lower overall college attainment and median income in the Deep South^51,52 and may produce more similar returns across racial groups.¹²

Our results were robust to a number of checks. More consistently robust results for stayers than for movers (Check 1) suggested selection differences between the groups. No associations between absolute upward mobility and adverse birth outcomes (Check 2) may suggest that absolute neighborhood improvement is subject to a threshold effect. Results also remained robust with use of an alternate sample including siblings (Check 3), adjustment for potential mediators (Check 4), and an alternate exposure specification (Check 5).

Limitations

As a cohort study, our results should not be interpreted as causal. Construction of our multigenerational sample required linkage of G2 and G3 births in SC during our study period, rendering us unable to observe births to G2 mothers over age 31 or births to G2 women who engaged in interstate migration. Our results may only be generalizable to firstborn infants and mothers aged 31 and under in SC, though our results were robust to use of an alternate full sibling sample. If upwardly mobile mothers with a lower risk of adverse birth outcomes moved out-of-state or delayed pregnancy until later ages, our results may underestimate any protective association of the trajectory.

We do not measure duration of residence for G2 women, the timing of any geographic mobility, or any intervening residences.⁵³ We are unable to study the issue of selection into neighborhoods,⁴ though our first robustness check considers selection on geographic mobility. This all may result in underestimation of associations between neighborhood social affluence or disadvantage and adverse birth outcomes and in differential misclassification of the exposure, particularly if Black women were more likely to change residence throughout their life course than White women. That our data spanned only two time points also left us unable to adjust for potential time-varying confounders that may have occurred between these periods.

We limited our consideration of neighborhood social conditions to single measures of affluence and disadvantage, though other measures (eg, residential stability, racial residential segregation, neighborhood disorder) or metrics (eg, Index of Concentration at the Extremes) may be meaningful for birth outcomes across a mother’s life course.^1,7,54 Still, our quartile-based robustness check considered how alternate specifications of our exposures of interest could affect our results. Finally, SC, compared to other US states, has a notorious history of both de jure and de facto racial segregation, and associations between race and neighborhoods may be different in less racialized contexts.

Conclusion

We present evidence of a stark Black/White disparity in LBW and PTB in South Carolina, to which a glaring racial disparity in exposure to neighborhood affluence and disadvantage across the life course is an important contributor. The lack of evidence for modification by maternal race of the association between life course neighborhood exposures and adverse birth outcomes does not diminish from the reality that racial disparities in exposure prevalence accounted for up to 9% of identified racial disparities in outcome prevalence. We found protective associations between birth outcomes and the always affluent/never disadvantaged trajectories only, which may suggest that any exposure to neighborhood disadvantage across the life course is detrimental for infant health. Our work argues that cross-sectional approaches to neighborhood effects may overlook meaningful life course exposure to neighborhood social context, particularly as life course neighborhood exposure to affluence and disadvantage exhibits major differences by race—further encouraging the adoption of a life course perspective for the study of birth outcomes.

Supplementary Material

Supplementary material is available at the American Journal of Epidemiology online.

Data availability

The authors do not have permission to share data.