Maternal Age Patterns of Preterm Birth: Exploring the Moderating Roles of Chronic Stress and Race/Ethnicity

Abstract

Background

Despite the suggested contribution of cumulative chronic stress to the racial/ethnic disparities in preterm birth (PTB), it is unclear how chronic stress, maternal age, and race/ethnicity are linked underlying PTB.

Purpose

We investigated the moderating effect of chronic stress on the maternal age–PTB association among non-Hispanic (N-H) White, N-H Black, Hispanic, and Asian women.

Methods

We analyzed the Washington State’s Pregnancy Risk Assessment Monitoring System data linked with birth certificates. The sample included women aged 18 years or older who birthed the first, singleton baby without birth defects. Chronic stress was measured by race/ethnicity-specific chronic stress indices. A maternal age–chronic stress interaction was modeled to predict PTB by logistic regression stratified by race/ethnicity. In subanalysis, the moderating role of racism was investigated in the maternal age–chronic stress interaction among three minority groups combined.

Results

Women’s maternal age trajectory of PTB varied by their race/ethnicity and chronic stress level. N-H White and N-H Black women showed a steeper maternal age-related increase in PTB (weathering) under higher chronic stress, indicating a chronic stress’ cumulative effect with maternal age. Besides, the extent of weathering was amplified by racism on top of chronic stress, particularly among N-H Black women.

Conclusions

These results show that both chronic stress and racism may develop accelerated PTB risk among minority women. Future research should use more objective and accurate chronic stress measures to ascertain the complex relationships among chronic stress, racial discrimination, and maternal age underlying the racial/ethnic differentials in PTB.

With chronic stress exposures and perceived racism, racial/ethnic minority pregnant women experience ‘weathering’ manifested as increasing preterm birth risk at advancing maternal age.

Introduction

Despite numerous efforts for decades, the racial/ethnic gaps in preterm birth (PTB; <37 weeks’ gestation) have not been greatly reduced in the USA. In 2016 non-Hispanic (N-H) Black and Hispanic women, respectively, were 50.5% and 4.3% more likely than N-H White women to experience PTB; nevertheless, the PTB rate has gradually increased from 2014 to 2016 among N-H White, N-H Black, and Hispanic women [1].

Reportedly, however, racial/ethnic groups of women show differences in not only the rates of adverse birth outcomes in aggregate but also maternal age trajectories of such outcomes. For example, Black women particularly in disadvantaged life conditions tend to show the lowest birth risk in their teens (around 18–20 years old) and then a steep increase in their 20s, presenting a linear upward age pattern of the risk. In contrast, White women experience higher birth risk in their teens than in their 20s, manifesting as a well-known U-shaped age pattern of the risk. Geronimus [2] first observed and named Black womens’ unique maternal age pattern of adverse birth risk “weathering,” which is defined as deterioration of health potential throughout the reproductive period as a result of exposure to chronic stress over the life course and the great effort to cope with stress [3]. From the weathering perspective, a maternal age for underprivileged women could be redefined as the duration of their exposure to lifelong stressful conditions [4]. By regarding maternal age as not only biological but also a psychosocial characteristic, the weathering framework could provide a useful conceptual map to explore the life course progression in women’s health culminating into their birth outcomes and the underlying chronic stress mechanisms [5].

Over the past two decades, researchers have empirically tested the weathering hypothesis among different populations in varying geographic locations in the USA and with different adverse birth outcomes. The study findings, however, are inconsistent on if weathering exists and which populations are subject to weathering [3, 4, 6–12].

The past research is limited to demonstrate weathering and unveil its underlying causes (suggestively, cumulative chronic stress over the life course) mainly in two ways: First, the weathering framework has been mostly applied to Black women, although other racial/ethnic groups of women (e.g., Hispanic and Asian women) could also experience cumulative stress and thereby weathering of their health potential over time. Despite the disproportionally concentrated socioeconomic disadvantages in Black communities, the studies’ focus on Black women may circumscribe the scope or variations in the relationships among contributing factors to weathering.

Second, few studies empirically demonstrated the role of chronic stress in weathering although theoretically chronic stress has been suggested to explain the Black women’s accelerated inclination in adverse birth outcomes with maternal age. With repeated and cumulative exposures, chronic stress could take a toll on women’s health, manifesting the dysfunctional bodily systems that could trigger early parturition, known as allostatic load, such as blunting of the normal corticotropin-releasing hormone (the cortisol feedback loop that gives rise to hypothalamus-pituitary-adrenal dysfunction in pregnancy); altering cellular immunity and immune control of inflammatory response; and cardiovascular/metabolic conditions [13]. Reportedly, Black women show higher total, inflammatory-marker, and cardiovascular-marker allostatic load scores than do White women, particularly at older ages (e.g., 35–64 years) [14].

Although a few studies analyzed if proxies for chronic stress altered the maternal age patterns of adverse birth outcomes, such proxy variables were usually limited to women’s neighborhood poverty and smoking [3, 8, 10, 11]. Neighborhood poverty and smoking were found to increase the magnitude of weathering within Black women such that only smokers or those with lifelong residence in impoverished neighborhoods experienced a maternal age-related increase in adverse birth outcomes or exhibited more rapid increase while those with lifelong residence in affluent neighborhoods did not show a sign of weathering; in contrast, White women did not experience weathering, regardless of their neighborhood poverty and smoking status [3, 8, 10, 11]. Some argue that racial discrimination is also a chronic stressor that predisposes Black women to adverse birth outcomes, contributing to the Black–White disparities in adverse birth outcomes [15, 16]. Evidence, albeit few, also exists that the effect of racial discrimination on adverse birth outcomes differs across maternal age. For example, Collins et al. [15] observed that the positive association between maternal lifetime exposure to interpersonal racial discrimination and very low birth weight (birth weight < 1,500 g) was stronger among Black women aged 20–29 than those aged <20 and 30 or older. Although these results imply that women’s race/ethnicity and chronic stress could moderate the effect of maternal age on adverse birth outcomes, more empirical studies should be followed to demonstrate the conceptual relations among race/ethnicity, chronic stress, maternal age, and birth outcomes by using other chronic stress measures with higher predictive power beyond a single proxy measure of either smoking or neighborhood poverty among various racial/ethnic groups.

Admittedly, however, measuring chronic stress is challenging, evidenced by its various operational definitions in the existing literature [17]. Accuracy of those chronic stress measurement is often compromised since chronic stressors of interest are conceptualized as independent or isolated, rather than multifaceted and cumulative entities although different components of stress are not randomly distributed but tend to cluster or even generate synergies with one another [17, 18]. More importantly, chronic stress measurement is rarely culturally relevant; hence, it may not appropriately reflect women’s stressful life conditions that could be shaped differently by their racial/ethnic background [19]. Some argued that different racial/ethnic groups could experience and report various types of stress or the same stress to varying degrees [20, 21]. In the same vein, Giscombé and Lobel [22] argued that prenatal stress might have interactive effects with race/ethnicity.

Therefore, the current study attempts to address the identified limitations in ascertaining the chronic stress mechanisms of weathering by investigating the moderating effect of chronic stress on the maternal age–PTB association among N-H White, N-H Black, Hispanic, and Asian women with a cumulative measure of chronic stress before and during pregnancy specific to each racial/ethnic group of women. We hypothesize that the effect of maternal age on PTB will differ significantly by both the level of chronic stress and race/ethnicity. Such a cumulative chronic stress measure developed for each race/ethnicity could capture each subgroup’s unique chronic stress exposures, enhance the accuracy and predictive power of their experienced chronic stress, and accordingly contribute to explaining the role of chronic stress in weathering.

Materials and Methods

Pregnancy Risk Assessment Monitoring System (PRAMS) data linked with birth certificates for Washington State (WA) in 2004–2007 were used for analysis. PRAMS is an ongoing collaborative surveillance system designed to monitor maternal experiences and behaviors before, during, and shortly after pregnancy. Every month each participating state selects a sample of newly delivered mothers from live birth certificates by stratified random sampling without replacement (1,300–3,400 women each year for an individual state) to send a questionnaire [23]. In addition to the core questions adopted across the participating states, WA PRAMS questionnaire includes pretested standard questions developed by the Centers for Disease Control and Prevention or the state, such as racism and social support [23]. Despite the significance of these two factors to understand racial/ethnic minority women’s chronic stress experience, racism and social support were originally collected only by WA and New York City PRAMS during 2004–2007 survey years. New York City PRAMS, however, had to be excluded from analysis because birth order information was not available in the data while the current study was limited to the firstborn. WA mails questionnaires 2–6 months after delivery and follows up with a telephone interview for nonrespondents. The final PRAMS data are weighted for sample design, nonresponse, and noncoverage [23].

This study used multiple imputation (MI) to handle missing data on dependent and independent variables. The regression method was selected for MI assuming that the data had a monotone missing data pattern [24]. The proportion of missing data on the dependent and independent variables ranged from 0% to 17.9%. Although racial/ethnic minority groups tended to have more missing information than N-H White women, this might not have generated biased study findings considering analyses were stratified by race/ethnicity. Among 3,626 women without missing information after MI, 59 subjects were excluded because they gave birth to infant(s) that were second or third birth, twins+, or with birth defects. Additional 112 subjects were excluded because they were younger than 18 years (n = 78) or had an abnormal record of gestational age (n = 34); gestational age whose value was greater than 47.3208 (99th percentile) was deleted. As a result, a total of 3,455 women were in analysis, who were aged 18 years or older and gave birth to their first singleton infant without birth defects in WA between 2004 and 2007 (1,342 N-H White, 549 N-H Black, 892 Hispanic, and 672 Asian women).

Dependent and Independent Variables

The primary outcome variable was PTB, defined as gestational age less than 37 weeks at birth. Predictor variables were race/ethnicity, maternal age, chronic stress, and racism (the reason for including racism will be explained shortly). Race/ethnicity was determined based on women’s self-report of their race and Hispanic ethnicity. Four racial/ethnic groups (N-H White, N-H Black, Hispanic, and Asian) were included in this study. Maternal age was available only as a categorical variable with seven groups (<18, 18–19, 20–24, 25–29, 30–34, 35–39, and 40+). However, those younger than 18 were excluded from the sample, and the 35–39 and 40+ age groups were combined due to small cell size in the latter, yielding five age groups (18–19, 20–24, 25–29, 30–34, and 35+). Racism was answered yes or no to the question “During the 12 months before your new baby was born, did you feel emotionally upset (for example angry, sad, or frustrated) as a result of how you were treated based on your race?” This question measured not the exposure, but distress from experienced racial discrimination. Chronic stress before and during pregnancy was measured based on the race/ethnicity-specific factor structures of chronic stress where the identified factors within each racial/ethnic group were summarized into one continuous mean score for every individual. Each step was briefly described below. Detailed information can be found elsewhere [17, 25].

Step 1: selection of chronic stress proxy variables

The authors conducted a systematic literature review to identify proxy measures of women’s chronic stress before and during pregnancy across multiple stress domains (e.g., external stressors, buffers of stress, enhancers of stress, and perceived stress) used to predict adverse birth outcomes among N-H White, N-H Black, Hispanic, and Asian women in the USA. Based on the study findings, 11 proxy variables for chronic stress were derived from the WA PRAMS data: racism, total household income, maternal educational attainment, health insurance before pregnancy, Medicaid before pregnancy, unaffordable prenatal care, special supplemental nutrition program for Women, Infants, and Children (WIC) during pregnancy, physical abuse before pregnancy, physical abuse during pregnancy, 13 items of stressful life events, and four items of social support (e.g., money, care, ride, and talk). As stated earlier, the WA PRAMS collected the information of each chronic stressor retrospectively after delivery, reflecting women’s past exposure to stress before (just before or during the past 12 months before pregnancy) or during pregnancy.

Step 2: identification of underlying constructs of chronic stress

With a total of 26 chronic stress items from the 11 variables mentioned above, exploratory factor analysis using a maximum likelihood extraction method identified three common latent factors of chronic stress among four racial/ethnic groups—financial hardship, perceived isolation, and physical violence—with notable racial/ethnic variations [25]. Besides, one thing to note is that racism did not correlate with other chronic stress items to form a common latent factor; hence, racism was analyzed in the model as an effect moderator, independent of other chronic stressors.

Step 3: development of race/ethnicity-specific chronic stress index

Due to the differential factor structures of chronic stress between the racial/ethnic groups, it was logical to invent a scoring system for chronic stress specific to each race/ethnicity. By different factor structures, we mean the difference in the order of each factor by importance (or the extent of model variance explained by a respective factor) and the chronic stress items under each of the three factors (Appendix). Within the group, the chronic stress items loaded to each factor were averaged for the subscale score (e.g., financial hardship, perceived isolation, and physical violence scores), and the three subscale scores were averaged for the full-scale score: that is, the chronic stress index. Here is an example of the scoring scheme for N-H Black women:

• Financial hardship score (1) = 1/4 × (insurance before pregnancy + WIC during pregnancy + maternal education + household income)

• Physical violence score (2) = 1/3 × (being in a physical flight + physical abuse before pregnancy + physical abuse during pregnancy)

• Perceived isolation score (3) = 1/3 × (help for money + help for ride + help for talk + help for care when sick)

• Chronic stress score = 1/3 × {(1) + (2) + (3)}

All variables were coded in a manner that a higher value reflected a greater exposure to the individual chronic stressor, which entailed reverse coding for some variables (e.g., maternal education, household income, and social support). Next, the chronic stress index was standardized, stratified by race/ethnicity with a mean of zero and SD of 1. Chronic stress was treated as a continuous variable in the model. The higher the index value was, the higher the level of chronic stress was before and during pregnancy.

Statistical Analysis

Descriptive analysis was conducted for the predetermined variables by race/ethnicity, using frequencies and percentages for categorical variables. Chi-square test (Mantel–Haenszel chi-square test for maternal age with multiple categories) determined the statistical significance of the differences in participants’ characteristics within and between racial/ethnic groups. Multivariable logistic regression was implemented to examine the hypothesized moderating effects of chronic stress, race/ethnicity, and later racism on the maternal age–PTB association. First, a model tested a three-way interaction among maternal age, chronic stress, and race/ethnicity in predicting PTB, which was statistically significant (p < .0001). Then, to make the model interpretation simpler, a separate model was built for each racial/ethnic group, which included maternal age, chronic stress, and their interaction term. Next, the moderating role of racism in the maternal age–chronic stress interaction was investigated among three racial/ethnic minority groups combined without N-H White women. All three racial/ethnic minority groups had to be combined in order to yield sufficient statistical power to obtain valid interaction estimates when stratified by racism status. N-H White women were omitted from the analysis because their racial discrimination experience would fundamentally differ in its nature from the experience of the racial/ethnic minority women in the USA. Again, a model tested a three-way interaction first among maternal age, chronic stress, and racism in predicting PTB of racial/ethnic minority women, which was statistically significant (p < .0001). Then, a separate model was built according to the women’s racism status, which included maternal age, chronic stress, and their interaction term.

Statistical significance was determined as p < 0.05. The PRAMS weight statement was included throughout the modeling process to account for sample selection and responses and to reflect the population of mothers giving birth to live infants in WA during the 2004–2007 survey periods. All analyses were conducted using SAS Version 9.4 statistical software (SAS Institute, Inc., Cary, NC).

Results

Table 1 shows the maternal characteristics by race/ethnicity. Relative to N-H White women, N-H Black, Hispanic, and Asian women were approximately 48%, 2%, and 13% more likely to experience PTB, respectively. The timing of the first childbirth significantly differed by race/ethnicity. Asian women delayed their childbearing to the extent that more than half of them gave birth in their 30s or older, followed by N-H White women. On the other hand, N-H Black and Hispanic women showed higher teen birth rates than the others, and about 60% of these groups of women gave birth in their 20s. Also, chronic stress and racism experiences were not equally shared by all racial/ethnic groups. Specifically, women in N-H Black and Hispanic communities were more likely to experience higher chronic stress than the average (median > 0) while women in N-H White and Asian communities were more likely to experience lower chronic stress than the average (median < 0). Additionally, except for N-H White women, N-H Black women were more likely than Hispanic and Asian women to perceive racism during the 12 months before the recent delivery.

Table 1.

Maternal Characteristics by Race/Ethnicity

Characteristic	N-H White	N-H Black	Hispanic	Asian	Total
	(n = 1,342)	(n = 549)	(n = 892)	(n = 672)	(n = 3,455)
Preterm birth, n (%)
No	1,202 (89.58)	465 (84.60)	796 (89.37)	594 (88.23)	3,057 (89.27)
Yes	140 (10.42)	84 (15.40)	96 (10.63)	78 (11.77)	398 (10.73)
Maternal age, n (%)
18–19	73 (6.21)	42 (8.14)	73 (8.46)	17 (3.21)	205 (6.39)
20–24	258 (20.54)	147 (27.24)	273 (32.15)	82 (14.06)	760 (22.15)
25–29	379 (28.58)	174 (31.62)	257 (28.91)	169 (25.59)	979 (28.48)
30–34	406 (28.81)	100 (17.96)	176 (18.41)	256 (36.07)	938 (27.33)
35+	226 (15.87)	86 (15.03)	113 (12.07)	148 (21.07)	573 (15.65)
Chronic stressa, median (range)	−0.04 (−0.08, 0.34)	0.07 (−0.45, 0.60)	0.04 (−0.35, 0.38)	−0.05 (−0.19, 0.63)	−0.03 (−0.45, 0.63)
Perceived racism, n (%)
No	1,310 (97.40)	471 (85.80)	784 (88.15)	605 (89.88)	3,170 (94.83)
Yes	32 (2.60)	78 (14.20)	108 (11.85)	67 (10.12)	285(5.17)

Raw frequencies and weighted percentages are reported for each characteristic by race/ethnicity.

^aChronic stress scores were standardized stratified by race/ethnicity (mean = 0 and SD = 1); as such, each racial/ethnic group’s median and range are reported, instead of mean. A range is shown in parentheses.

Table 2 shows a maternal age pattern of PTB by race/ethnicity. Interestingly, not only N-H Black women but also N-H White women presented a maternal age-related increase in PTB (p < .0013 and p < .0001, respectively, for a linear trend) with greater Black–White gaps in PTB among older women (risk difference of 3.54% for women aged 18–19 years vs. 7.69% for those aged 35+ years). Hispanic women, however, presented a W-shaped pattern with dips in their 20–24 and 30–34 years. Contrary to N-H White and N-H Black women, Asian women presented a downward linear trend with advancing maternal age (p < .0001 for a linear trend).

Table 2.

Prevalence^a of Preterm Birth by Maternal Characteristics Among Racial/Ethnic Groups

Characteristic	N-H White (n = 1,342)	N-H Black (n = 549)	Hispanic (n = 892)	Asian (n = 672)	Total (n = 3,455)
Maternal age
18–19	4 (5.73)	4 (9.27)	11 (17.17)	3 (17.45)	22 (8.91)
20–24	29 (11.46)	23 (15.74)	27 (9.01)	9 (11.31)	88 (11.03)
25–29	40 (10.26)	27 (15.98)	28 (10.41)	19 (11.70)	114 (10.61)
30–34	44 (10.86)	15 (14.39)	11 (6.34)	34 (13.21)	104 (10.69)
35+	23 (10.41)	15 (18.10)	19 (17.41)	13 (8.83)	70 (11.37)
p-valueb	<.0001	0.0013	0.1380	<.0001	<.0001

^aWeighted percentage.

^bEach p-value comes from Mantel–Haenszel chi-square test.

Table 3 presents logistic regression coefficients of the maternal age–chronic stress interaction by race/ethnicity to determine if the maternal age effect on PTB changes with increasing chronic stress. As an essential premise of the weathering hypothesis, we hypothesized that the higher the level of chronic stress, the higher the effect of maternal age on PTB, expecting a linearly or curvilinearly increasing impact of maternal age when chronic stress increases by one unit, which may be indicative of stress accumulation over time. Despite some fluctuations across maternal age categories, this study generally supported the premise by showing a curvilinear increase in the maternal age effect on PTB when chronic stress heightens by one unit only in N-H White, N-H Black, and Hispanic communities. Under higher chronic stress, the magnitude of increase in the maternal age effect was the highest among N-H White women, approximately four and seven times the magnitude of N-H Black and Hispanic women, respectively. However, such stark gaps between the groups are an incomplete picture until racism experienced by racial/ethnic minority women is taken into account, which will be discussed below. Interestingly, Asian women exhibited the opposite pattern where the magnitude of maternal age effect linearly declined with one-unit increase in chronic stress. According to the sensitivity test, the study findings did not show major differences between the data with and without MI.

Table 3.

Logistic Regression Models Examining the Interaction Between Maternal Age and Chronic Stress in Predicting PTB by Race/Ethnicity, WA PRAMS, 2004–2007

	N-H White (n = 1,342)			N-H Black (n = 549)			Hispanic (n = 892)			Asian (n = 672)
	β	95% CI		β	95% CI		β	95% CI		β	95% CI
Maternal age
18–19	–	–	–	–	–	–	–	–	–	–	–	–
20–24	−1.05***	−1.20	−0.90	−0.03	−0.35	0.30	−0.89***	−1.01	−0.77	4.63***	3.32	5.94
25–29	−1.12***	−1.26	−0.97	0.07	−0.24	0.39	−0.64***	−0.76	−0.52	4.94***	3.64	6.24
30–34	−0.89***	−1.03	−0.75	0.10	−0.23	0.44	−1.33***	−1.48	−1.18	5.12***	3.82	6.42
35+	−1.11***	−1.26	−0.97	0.31	−0.02	0.64	0.00	−0.12	0.13	4.58***	3.27	5.88
Chronic stress	−20.76***	−22.63	−18.88	−4.54***	−6.07	−3.00	−0.96*	−1.88	−0.04	18.30***	14.19	22.41
Maternal age × chronic stress
18–19	–	–	–	–	–	–	–	–	–	–	–	–
20–24	22.20***	20.29	24.12	5.51***	3.88	7.14	3.10***	2.02	4.17	−16.42***	−20.58	−12.25
25–29	18.60***	16.67	20.52	5.41***	3.81	7.00	3.10***	2.09	4.12	−18.56***	−22.70	−14.42
30–34	24.60***	22.69	26.52	5.97***	4.33	7.61	−1.04	−2.11	0.03	−17.27***	−21.39	−13.15
35+	19.61***	17.64	21.58	5.17***	3.55	6.79	3.20***	2.18	4.23	−19.28***	−23.45	−15.11

CI confidence interval; N-H non-Hispanic; PTB preterm birth; WA PRAMS Washington State Pregnancy Risk Assessment Monitoring System.

*p < .05, **p < .01, ***p < .0001.

Racism as an Effect Moderator

Table 4 presents logistic regression coefficients of the maternal age–chronic stress interaction by racism in predicting PTB among three minority groups combined. An increasing rate of the maternal age effect attenuated among older women without perceived racism while such a rate escalated among older women with perceived racism, demonstrating racism as a moderating factor of weathering. Among those with perceived racism, a sudden decrease in the increasing rate in 30–34 years might result from the small cell size; Hispanic women with perceived racism, in particular, did not report having any PTB at all in their 30–34 years (data not shown).

Table 4.

Logistic Regression Models Examining the Interaction Between Maternal Age and Chronic Stress in Predicting PTB by Perceived Racism, WA PRAMS, 2004–2007

		No perceived racism (n = 1,860)			Perceived racism (n = 253)
		β	95% CI		β	95% CI
All minorities (n = 2,113)	Maternal age
	18–19	–	–	–	–	–	–
	20–24	−0.84***	−0.95	−0.72	−0.41	−0.98	0.17
	25–29	−0.60***	−0.71	−0.50	0.01	−0.54	0.55
	30–34	−0.62***	−0.73	−0.51	−0.83**	−1.39	−0.26
	35+	−0.45***	−0.57	−0.34	0.57*	0.03	1.12
	Chronic stress	−0.48	−1.06	0.11	−12.68***	−18.40	−6.96
	Maternal age × chronic stress
	18–19	–	–	–	–	–	–
	20–24	2.22***	1.53	2.91	15.68***	9.89	21.48
	25–29	1.32***	0.67	1.96	15.85***	10.07	21.62
	30–34	0.94**	0.28	1.60	13.47***	7.66	19.28
	35+	0.48	−0.20	1.15	16.92***	11.16	22.69
		No perceived racism (n = 471)			Perceived racism (n = 78)
		β	95% CI		β	95% CI
N-H Black (n = 549)	Maternal age
	18–19	–	–	–	–	–	–
	20–29	0.50**	0.12	0.88	−2.87***	−4.22	−1.52
	30+	0.68**	0.29	1.06	−3.24***	−4.63	−1.85
	Chronic stress	−4.14***	−6.05	−2.23	−14.46***	−21.37	−7.56
	Maternal age × chronic stress
	18–19	–	–	–	–	–	–
	20–29	4.77***	2.82	6.72	16.43***	9.47	23.38
	30+	4.47***	2.51	6.43	21.48***	14.35	28.62

CI confidence interval; N-H non-Hispanic; PTB preterm birth; WA PRAMS Washington State Pregnancy Risk Assessment Monitoring System.

*p < .05, ***p < .0001.

Next, we conducted several initial analyses to confirm the moderating role of racism in weathering. First, race/ethnicity was controlled for in the model because of the different maternal age trajectories of PTB between three racial/ethnic minority groups of women. The results did not show significant differences from the initial findings (data not shown). Further, when stratified by race/ethnicity, N-H Black women with perceived racism showed even stronger weathering than three minority groups combined. Due to small cell sizes, we had to combine some of the age groups, which ended up with three categories: 18–19, 20–29, and 30 or older. Even with the fewer age categories, the maternal age–chronic stress interaction could not be modeled for each racism status among Hispanic and Asian women.

Discussion

This study empirically tested the conceptual pathways among maternal age, chronic stress, and racism underlying weathering about PTB among four major racial/ethnic groups of women in WA by analyzing the WA PRAMS data (2004–2007). Despite the data limitation, the study’s strength is to extend the weathering framework to Hispanic and Asian women and to use a cumulative chronic stress measure specific to each racial/ethnic group, which is a chronic stress measure deemed more accurate but never tested in previous studies.

As hypothesized, higher chronic stress tended to amplify a maternal age-related increase in PTB risk (weathering) among all racial/ethnic groups but Asian women; racism fortified weathering even further among racial/ethnic minority women (N-H Black women in particular) under higher chronic stress. The increasing maternal age effect among women in the face of chronic stress may indicate the cumulative impact of chronic stress on the body over time.

Among those under higher chronic stress, N-H White women showed a steeper maternal age-related increase in PTB risk than did N-H Black or Hispanic women although the prevalence of PTB was lower in the former than in the latter across maternal age. Existing studies observed weathering about various poor birth outcomes (e.g., low birth weight, PTB, or neonatal intensive care unit admission) among White women, but only those who were underprivileged, such as women who were unmarried, living in poor neighborhoods, smoking cigarettes, receiving inadequate prenatal care, or covered by public health insurance [4, 11, 26]. Generally, these unfavorable life conditions could be proxies for chronic stress among women in disadvantaged status [27, 28] or their unhealthy behavioral coping mechanisms [29]. Of note, many of these risks are endemic in impoverished neighborhoods as such smoking itself could be an indirect indicator of women’s disadvantaged condition [11]; indeed, Hibbs et al. [27] used women’s smoking status as a proxy for chronic stress. Moreover, the N-H White women’s greater reactivity to the given chronic stress is corroborated by the existing evidence that White women are prone to perceive stress to a greater extent than their Black counterparts particularly when stress stems from finance-, health-, and relationship-related problems [30, 31]. Even Geronimus et al. [32] reported a sign of biological aging in response to cumulative stress among poorer and less educated White but not Black and Mexican populations living under comparable conditions. The phenomenon could be explained by N-H White women’s low resilience, less habituation to socioeconomic disadvantage, or a sense of failure between the reality and expectation of White privilege [32].

Also, the three racial/ethnic minority groups showed potent weathering at advancing maternal age when they had experienced racism in the past 12 months before the delivery. Racial discrimination affects women’s health and well-being directly and indirectly through increasing their vulnerability to individual stressors [33]. A body of literature has documented a contribution of racism, on top of psychosocial risk factors, to poor birth outcomes and weathering about those outcomes [34, 35]. Indeed, Giscombé and Lobel [22] argued that racism could be conceptualized as a distinct form of stress, independent of other types of stress because racism involves stimuli (actions, events, or practices executed by individuals and organizations), appraisal of the stimuli as stressful, and negative emotional responses to the stimuli. In the 1900s, these three approaches to stress definition (i.e., environmental-, perceptual-, and response-based definitions) were adopted predominantly, but not jointly. These were suggested not as linear steps, rather as essential components of stress. In the early days, negative emotional states or responses (e.g., anxiety and depression) to stimuli were frequently equated with stress [36]. Later in the late 1990s, attention was shifted to stress as environmental stimuli, extricated from perceptions of events and emotional concomitants [37]. However, this idea was criticized by scholars including Lazarus emphasizing the centrality of individual perceptions or appraisals to conceptualizing stress since not all objective conditions or events cause distress for individuals experiencing them [38, 39]. Currently, the combined use of these three components is strongly encouraged to capture the synergistic effects among stimuli, appraisal, and emotion on health/disease [40]. Nevertheless, it is worth pointing out that the three components are not necessary conditions to be stress, particularly when considering racism as stress. More frequently occurring and covert forms of racism (e.g., microaggression) may not trigger significant emotional responses when experienced due to the ambiguous nature of and desensitization to microaggression [41, 42].

Simultaneously, perceived racism could induce “stress proliferation” where early and chronic stressors apt to spread outward into other domains giving rise to secondary stressors; through stress proliferation, racial/ethnic minority women with perceived racism could be at risk of cumulative stress over their life course [33]. In addition, racial discrimination could contribute to cumulative stress by intensifying the stress response to subsequent negative stimuli; persistent racial discrimination could also color racial/ethnic minority women’s interpretations of stressors, and thereby make them sensitive to perceiving even less overt race bias or discrimination [33]. Further, WA retains its unique context that could exacerbate the racial/ethnic minority residents’ exposure and negative appraisal of racial discrimination. Despite the continuous increase in the number of non-White populations over time, WA is still a White-dominant state in the USA. In 2016, N-H White population accounted for 69.5% of the State’s total population while Black, Hispanic, and Asian populations, respectively, accounted for 4.1%, 12.4%, and 8.6% [43]. Hence, a small proportion of racial/ethnic minorities in the region may have increased the chances for N-H Black, Hispanic, and Asian pregnant women to encounter frequent discrimination or perceive it as more stressful. Indeed, the more White individuals residing in a neighborhood, the higher the levels of experienced racial discrimination among Black neighborhood residents [44].

Moreover, this study adds to the limited literature on weathering among Hispanic women, which showed inconsistent or inconclusive findings [45–47]. Hispanic women in this study did not show as clear a maternal age gradient of PTB risk as observed in the racial/ethnic minorities combined with perceived racism. We suspect that weathering in this population may have been underestimated by not taking nativity, duration of U.S. residence, or acculturation into account in the predictive model due to data unavailability. Foreign-born residents tend to experience lower adverse birth risk than their U.S.-born counterparts upon arrival or in their early phase of immigration although such risk would eventually converge to the risk among their U.S.-born counterparts as the length of time in the USA increases [47]. Moreover, emerging evidence suggests not the monotonic and assimilative but a curvilinear pattern of acculturation where low birth weight rates among Hispanic immigrants declined over the first few years in the USA and increased thereafter [48]. Therefore, foreign-born Hispanic women’s health status, generally better off than their U.S.-born counterparts across maternal age, could dampen possibly stronger weathering among older, U.S.-born Hispanic women by mixing these two heterogeneous populations.

Finally, we observed an interesting reverse-weathering pattern among Asian women, which has not been reported in the literature by far. Such a finding, however, contrasts with several relevant studies directly or indirectly implicating weathering in this population [49, 50]. Three theories are possible to explain the reverse-weathering among Asian women: “salmon bias,” shift of the optimal maternal age to an older end, and paternal support as a resilience resource among older mothers. First, salmon bias refers to return migration among the foreign-born in the USA who are ill, unemployed, or otherwise unsuccessful, which produces better health outcomes among the foreign-born than among their U.S.-born or White counterparts. Existing literature, despite its focus predominantly on Hispanic populations, argued implausibility of returning migration among immigrant women of reproductive age [47, 48, 51, 52]. Thus, reverse weathering may not be an artifact of older Asian women with underlying conditions for PTB returning to their country of origin and leaving healthy Asian women in the USA. Alternatively, delayed childbearing as a social norm may marginalize younger pregnant women in Asian communities, putting them at higher PTB risk than their older counterparts. Maternal age (or desired age of childbearing) is in part socially determined based on the members’ agreement on when is normal and beneficial to have a baby, which is learned through the society’s culture or social norm [2, 53, 54]. When delayed childbearing is accepted as normal, the society and its members may be more likely to disapprove early pregnancy, isolate young pregnant women, or withhold necessary resources from them [55], which could eventually threaten their maternal and fetal health. Third, social support from a husband could help older women become more resilient to chronic stress in Asian communities. According to Ghosh et al. [56], women with moderate-to-high stress were at increased odds of delivering preterm (OR = 2.15 [95% CI 0.92, 5.03]), whereas women with greater paternal support during pregnancy had no increased risk even with moderate-to-high chronic stress (OR = 1.13 [95% CI 0.94, 1.35]). According to our subanalysis, Asian women under higher chronic stress, unlike the other racial/ethnic counterparts, showed a rapid decline in the unmarried rate at advancing maternal age (6.5% vs. 40.9%–44.5%, the proportion of those unmarried at age 35 years or older).

There are several limitations to consider. First, maternal age as a continuous variable was not available in the PRAMS data. Maternal age as categories could increase modeling errors because age intervals could not smoothly fit the curvilinear age pattern of PTB and reflect the gradual accumulation of chronic stress over time.

Second, the chronic stress measure in this study was based on women’s one-time responses to the survey in 2004–2007, which assessed chronic stressors experienced during only a limited period—12 months before pregnancy or during pregnancy—not over one’s life course. Although such a cumulative measure as the race/ethnicity-specific composite index of chronic stress could have more predictive power and better represent the nature of stressors interacting with one another [22], this measurement is still incomplete and distal to fully capture women’s chronic stress. Besides, the chronic stress measure in this study omitted key stressors, including acculturation stress, psychological distress, and neighborhood characteristics due to data unavailability, which may have led to an underestimation of the level of chronic stress, particularly among racial/ethnic minority women.

Third, it was not possible to generate separate models according to both nativity and racism status among Hispanic and Asian women due to data unavailability and small sample size, respectively. However, within the racial/ethnic minority groups with perceived racism, the confounding effect of foreign-born status on weathering may not be substantial since the foreign-born are more likely to perceive discrimination as their acculturation level (e.g., English usage) increases [57]. Generally, the foreign-born are more likely than the U.S.-born to be treated as substandard, mediocre, or inferior [58].

Fourth, this study analyzed only four racial/ethnic groups of women whose birth was registered in WA in 2004–2007, which may limit generalizability of the study findings to the racial/ethnic groups in other regions. Because the study’s primary focus was on four major racial/ethnic groups in the USA, other groups in the PRAMS data, such as Hawaiian, other non-Whites, American Indian, Alaska Native, and mixed race were excluded from the analysis. Also, women younger than 18 years were omitted from the analysis due to their small sample size and possible confounders (e.g., biological immaturity) that could obscure the chronic stress–maternal age–PTB relationships.

Lastly, this study could not address the heterogeneity of Black, Hispanic, and Asian women regarding subethnic origin due to data unavailability. Although PRAMS data categorized Asians into Chinese, Japanese, Filipino, and other Asian, it was not feasible to fit the data for each of the subethnic groups because of small sample size and resultant lack of statistical power.

Conclusion

In light of the limitations of the current study, we cannot conclude that cumulative chronic stress and racism are driving forces of weathering among racial/ethnic groups of women in the USA. However, our findings, at least among pregnant women in WA, are consistent with the hypothesis; as such, the effect of maternal age on PTB was augmented by chronic stress among N-H White, N-H Black, and Hispanic women, and the extent of maternal age effect also varied between the groups. Although not hypothesized in the first place, this study additionally observed that racism, on top of chronic stress, increased the positive association between maternal age and PTB among racial/ethnic minority women, particularly N-H Black women. A strength of this study is to tease out the entangled relationships among race/ethnicity, chronic stress, racism, maternal age, and PTB, which paved a way to test the pathways in future studies among other populations in different sociocultural and geographic contexts.

Future research should consider racial/ethnic differences in biological responses to chronic stress underlying the identified racial/ethnic variations in maternal age patterning of PTB under higher chronic stress. Race/ethnicity-specific profiles of the exposures and biological responses to chronic stress could help develop risk algorithms for PTB distinctive to racial/ethnic groups, and accordingly, guide a targeted approach to intervening with pregnant women with a high chronic stress burden to reduce their PTB risk. Further, it is necessary to investigate the resilience resources of Asian women that resulted in their maternal age-related decrease in PTB even under higher chronic stress. We suggest a possible protective effect of paternal support in marital ties among Asian women in the face of prolonged and cumulative stress, which warrants further investigation.

Appendix

Table A1.

Items Under Chronic Stress Latent Factors by Race/Ethnicity

	N-H White	N-H Black	Hispanic	N-H Asian
Factor 1	Can’t pay bills Medicaid before pregnancy Maternal education Insurance before pregnancy WIC during pregnancy Household income	Insurance before pregnancy WIC during pregnancy Maternal education Household income	Insurance before pregnancy Maternal education WIC during pregnancy Household income	Insurance before pregnancy Maternal education WIC during pregnancy Household income
Factor 2	Help for money ($50) Help for talk Help for ride Help for care when sick	Being in a physical fight Physical abuse before pregnancy Physical abuse during pregnancy	Help for money Help for talk Help for care when sick Help for ride	Help for money Help for care when sick Help for talk Help for ride
Factor 3	Being in a physical fight Physical abuse before pregnancy Physical abuse during pregnancy	Help for money Help for ride Help for talk Help for care when sick	Divorce/separation Physical abuse during pregnancy Physical abuse before pregnancy Increased argument with a partner	Physical abuse before pregnancy Unwanted pregnancy Being in a physical fight Imprisonment of self or husband

N-H non-Hispanic; WIC Women, Infants, and Children (the government-assisted special supplemental nutrition program). Factor 1: financial hardship. Factor 2: perceived isolation (N-H White, Hispanic, and Asian) or physical violence (N-H Black). Factor 3: physical violence (N-H White, Hispanic, and Asian) or perceived isolation (N-H Black).

Funding

This study was funded by the Sigma Theta Tau International: Xi Chapter (honor society) research grant.

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Sangmi Kim, Eun-Ok Im, Jianghong Liu, and Connie Ulrich declare that they have no conflict of interest. There was no informed consent because of the secondary data analysis. All procedures were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000.

Authors' Contributions: S.K. conceived of the presented idea, performed data analysis, and took the lead in writing the manuscript. All authors provided critical feedback and helped shape the research, analysis, and manuscript. Connie Ulrich supervised the study.

Ethical Approval: This study was exempt from full review by the Institutional Review Board where the authors are affiliated.