Effects of social and psychosocial factors on risk of preterm birth in black women

Summary

Our objective was to examine how social and psychosocial factors may influence the risk of preterm birth. The design of the study was a hybrid retrospective and prospective cohort. African-American women residing in Baltimore, Maryland, were enrolled prenatally if they received care at one of three Johns Hopkins Medical Institution prenatal clinics (n = 384) or enrolled post-partum if they delivered at Johns Hopkins Medical Institution with late, none or intermittent prenatal care (N = 459). Preterm birth was defined as less than 37 weeks completed gestation. Interview data were collected on 832 enrolled women delivering singletons between March 2001 and July 2004. The preterm birth rate was 16.4%. In both unadjusted and adjusted models, exposure to racism over a woman’s lifetime had no effect on risk of preterm birth in our sample. However, we found evidence of a three-way interaction between reported lifetime experiences of racism, depressive symptoms during pregnancy and stress during pregnancy on preterm birth risk. Racism scores above the median (more racism) were associated with an increased risk of preterm birth in three subgroups with the effect moderated by depressive symptoms and stress. Social and psychosocial factors may operate in a complex manner related to risk of preterm birth.

Introduction

Within the US, substantial racial disparities in preterm birth (PTB) have persisted for as long as we have maintained records.¹ While a portion of this gap appears to be attributable to racial differences in known risk factors, there is growing research to suggest that exposure to racism may explain these disparities.^1,2 As succinctly described by Camara Jones and others, racism may be institutionalised, personally mediated, or internalised.³ There are a range of mechanisms by which racism across any of these domains may influence preterm delivery. The stress resulting from the experience of racism might potentially trigger preterm delivery. Psychosocial stress has received considerable attention with regard both to the aetiology of PTB as well as an explanation for black–white disparities.^4,5 and racism could be conceptualised as a stressor. Stress could increase the risk of infections,⁶ a well-established trigger of preterm delivery.⁷

The response to the experiences of racism could also include engaging in negative health behaviours and thereby could indirectly increase the risk of preterm delivery. Chronic disease is another risk factor for PTB that has also been linked to racism.^8–11 Racism may also influence where women reside and there is growing evidence of a link between residential neighbourhood environment and birth outcomes.^12–17

While exposure to racism has been shown to be associated with a wide range of adverse health outcomes, racism has rarely been examined with regard to birth outcomes in large epidemiological studies. Two studies have investigated this relationship using data in cohorts focused on other health outcomes in which retrospective data were obtained on women’s pregnancies. Rosenberg et al.¹⁸ used the Black Women’s Health study cohort to examine effects of racism and found no significant association between perceived lifetime racism and risk of PTB. However, unlike our most recent work, items were not constructed as a scale and effects of individual items were the basis of the analysis.¹⁸ Mustillo and colleagues studied PTB and effects of lifetime racial discrimination in the CARDIA cohort study.¹⁹ Women reporting one to two experiences of lifetime racial discrimination were at twofold higher PTB risk while those reporting three or more experiences were at threefold higher risk. Further-more, adding racial discrimination to other social factors in the models led to the elimination of the PTB racial disparity in this sample.¹⁹ While these results are provocative, the long period of recall (from months to several years) and the inability to include data beyond that self reported by women allowed only rather sparse valid covariable data to be collected on the index pregnancy (e.g. smoking).

Four other publications with designs focusing on pregnancy have examined acute and/or lifetime experiences of racism. In the first of their two published reports, Collins et al. reported a statistically non-significant increase in risk of very low birthweight (VLBW) associated with racism experienced during pregnancy.²⁰ The second study was much larger and found a statistically significant tripling of risk of VLBW associated with experiencing lifetime racism in three or more of the five domains.²¹ VLBW is comprised largely of preterm infants and so provides some indication of effects on preterm risk. In a large cohort study of over 1000 pregnant women from six ethnic groups (n = 1150), Shiono and colleagues²² reported no significant adverse effect of racial discrimination during pregnancy on mean birthweight. Finally, in a more recent cohort study of nearly 2000 pregnant women based in North Carolina, Dole et al. reported a small but statistically significant increase in risk of preterm birth associated with higher scores on a scale measuring lifetime racial discrimination.²³ In analyses stratified on race, this effect was demonstrated to be restricted to the approximately 600 Black women in the sample (OR = 1.8 in African-American women only vs. 1.6 overall) with too few white women reporting racial discrimination to assess its impact in the stratum.²⁴

Our study of preterm birth focused on how social and psychosocial factors, including racism, relate to risk of preterm birth among low-income African-American women. We focused on lifetime exposure to racism but also considered acute experiences (6 months prior to and during pregnancy.) Our study collected detailed data on a wide range of risk factors, thus enabling us to consider and adjust for important confounders and effect modifiers.

Methods

Study design and sample

The study was reviewed and approved by institutional review boards at the study institutions. The sample eligibility was restricted to African-American women residing in Baltimore City (Maryland). We approached eligible women for enrolment if they were receiving prenatal care at one of three selected clinics or had delivered at the affiliated hospital with late, sporadic or no prenatal care. Women were interviewed in person to collect information regarding sociodemographic factors, psychosocial factors and health behaviours. Women enrolled prenatally were interviewed twice, the first interview at 22–28 weeks gestation and the second, a short post-partum interview. Women enrolled post-partum with late or no prenatal care were interviewed only once, during the post-partum hospitalisation, to measure factors from all three trimesters of pregnancy. We collected interview data on a total of 872 women (enrolling 485 prenatally and 387 post-partum) over a 41-month period (March 2001 through July 2004) with a 68% response rate. Excluding multiple births (n = 24) and losses to follow up (n = 16), 832 participants were included in the analyses.

Measurement of key variables

The primary measure of racism experiences was a lifetime exposure measure, the Racism and Lifetime Experiences Scale (RALES).^25–29 Secondarily, we also included three items from the RALES Daily Life Experiences Scale.²⁵ We had used the full Daily Life Experiences Scale in our pilot study, but its poor psychometrics in our sample led us to include only a few of the most variable items with two separate recall periods queried: 6 months prior to learning of the pregnancy and during pregnancy. We also used the Racism-Related Experiences (RRE) Scale to capture women’s responses to racism.^25,30,31

In addition to these measures of racism, we collected data on several psychosocial constructs that we hypothesised may mediate or moderate the effects of racism. These included scales assessing stress (the 12-item hassles scale, Cronbach’s α = 0.81),⁴ depression symptoms (CES-D, Cronbach’s α = 0.89),^32,33 pregnancy locus of control (Cronbach’s α = 0.70),⁴ mastery (Cronbach’s α = 0.77),⁴ anxiety,³⁴ and social support (Cronbach’s α = 0.94).⁴ Sociodemographic data included maternal age, education, income, and the Family Resources Scale⁴ (a measure of extent to which resources meet needs) and socioeconomic status. Health behaviours queried included cigarette smoking, alcohol and illicit drug use, and vaginal douching.

The medical record was abstracted to collect data on component measures of gestational age (e.g. last menstrual period, ultrasound estimated date of confinement), parity, multiple gestation, initiation of prenatal care, number of prenatal visits, chronic diseases and complications of pregnancy.³⁵ Preterm was defined as birth prior to 37 completed weeks of gestation. Gestational age based on last menstrual period as reported on the medical record was systematically compared with other estimates of gestational age in a hierarchical fashion.³⁵

Analysis

There is no previous research to determine definitive cut points to use for the racism scale (RALES). We explored the effect of the racism in a number of ways, with a focus on treating the score as a continuous variable, comparing effects of quartiles, and with categorisation at the median. Covariables were explored as both continuous and categorical variables and in most cases were treated as dichotomous factors in the final models. Cut points for all covariables were created based on the term deliveries only. Covariables were considered as potential confounders/mediating factors if they were associated with either racism or preterm birth in our sample, or had been identified as confounders in the literature on racism and/or preterm birth.³⁶ They included: psychosocial factors [pregnancy locus of control (individuals’ belief of the degree to which their pregnancy outcome is under their internal control); stress during pregnancy; social support, depressive symptoms, anxiety]; health behaviours (alcohol use in pregnancy; drug use in pregnancy); health (prepregnancy weight; weight gain; infections, chronic disease); additional measures of socioeconomic status (rent/mortgage, income, housing quality). We retained covariables in models if the coefficient for the exposure (RALES) changed more than 10% when the covariable was included alone or in combination with other covariables. We hypothesised a priori that the social and psychosocial factors were likely to interact with one another and that racism’s effects might be moderated by other factors. To determine the significance of the interaction terms, we relied on the likelihood ratio chi-square test, rather than z-test statistics for the coefficients, as the standard errors of the latter are subject to substantial collinearity when all possible interactions involving some variables are added to the model. Where alpha was <0.10, we judged the model including the interactions to better explain the data.

As our preterm birth rate (~16%) does not meet the rare disease assumption of logistic regression, we used Cox proportional hazards analysis in which preterm birth was the event and gestational age in days was treated as the time variable.³⁷ None of the variables were treated as time-varying as we did not have information on the timing and could only identify the prenatal period generally. We note, however, that the hazard ratios and P-values were virtually identical to the odds ratios and P-values produced by logistic regression models. We evaluated the residuals and looked at an interaction with log(time); we found that none of the main variables of interest deviated from the assumption of proportional hazards. Missing values for nearly all variables in these analyses were <3% individually and in total so we excluded those study subjects from the relevant analyses. The exception was the adequacy of prenatal care variable for which approximately 25% of the subjects could not be categorised due to missing data. Based on findings of previous studies about missing data on this variable,³⁸ missing comprised its own category.

Results

Table 1 describes the characteristics of the study cohort. Consistent with our eligibility criteria and the population served by the clinical sites, our study sample reflects a population of disadvantaged women at high risk of preterm birth.

Table 1.

Descriptive characteristics of the study sample, Baltimore, Maryland, 2001–04

Descriptive characteristics	Total sample (N = 832) n (%)
Race-self report: Black or African American	832 (100%)
Ethnicity – self report: Hispanic	10 (1.2%)
Rent/mortgage: mean (std dev) range	$372.77 (219.99) 0–950
Age: mean (std dev) range	23.1 (5.6) 12–43
Public housing
Yes	317 (38.1%)
No	491 (59.0%)
Education:
<12 years of school, <21 years old	309 (37.1%)
<12 years of school, 21+ years old	440 (52.9%)
≥12 years of school	83 (10.0%)
Recruitment:
Prenatal	456 (54.8%)
Post-partum	376 (45.2%)
Previous preterm delivery
≥1 prior preterm delivery	103 (12.4%)
No prior preterm deliveries	532 (63.9%)
Primiparae	197 (23.7%)
Smoked cigarettes in second trimester of pregnancy
Yes	157 (18.9%)
No	669 (80.4%)
Family Resource Scale
mean (std dev) range	46.9 (14.9) 25–107
top quartile 57+	188 (23.8%)
lower three quartiles <57	601 (76.2%)
Racism and Lifetime Experiences Scale (RALES)
mean (std dev) range	10.3 (5.5) 0–30
above median 10+	425 (51.1%)
below median <10	407 (48.9%)
Response to Racism Experiences (RRE) Scale
mean (std dev) range	41.8 (11.3) 0–60
above median 44+	405 (48.7%)
below median <44	427 (51.3%)
Stress (hassles) scale
mean (std dev) range	18.1 (5.7) 8–41
top quartile (≥21)	220 (26.4%)
lower three quartiles (<21)	610 (73.3%)
Depressive symptoms (CES-D)
mean (std dev) range	16.1 (10.5) 0–53
≥l6	351 (42.2%)
<16	481 (57.8%)
Locus of control (pregnancy-specific)
mean (std dev) range	22.3 (3.3) 13–28
top quartile (≥26)	169 (20.3%)
lower 3 quartiles (<26)	647 (77.8%)
Mastery
mean (std dev) range	21.6 (3.6) 8–28
above median (≥22)	394 (47.5%)
below median (<22)	436 (52.5%)
Social support
mean (std dev) range	45 (9.8) 11–55
above median (≥48)	412 (49.5%)
below median (<48)	420 (50.5%)
Adequacy of prenatal care
Missing	209 (25.1%)
Inadequate (little or no prenatal care)	302 (36.3%)
Adequate	178 (21.4%)
Adequate plus	143 (17.2%)
Pregnancy outcome
Term birth	693 (83.3%)
Preterm birth	139 (16.7%)
Gestational age (days): mean (std dev) range	270.2 (20.3) 148–316
Birthweight (grams): mean (std dev) range	3127 (618) 614–5077

We found no association between individual items representing acute experiences of racism, either in the 6 months prior to learning of the pregnancy or during the pregnancy. Lifetime experiences of racism, analysed as RALES scores in the top quartile or above the median, also had no effect overall on the risk of preterm birth (Table 2). There was no significant main effect for stress but there was a significantly increased risk associated with a score of 16 or higher on the CES-D (Table 2). Further analyses examining hypothesised interactions revealed a more complex relationship for lifetime racism (Figure 1, Table 3). For RALES scores above or at the median, there was evidence of interaction with the stress scale ‘hassles in the week prior to birth’ such that there was a stronger adverse effect of RALES for women with higher stress scores (top quartile of stress scale). In this model, adjustment for smoking decreased the interaction with stress. In a second model, an interaction was also seen for the RALES variable with scores above 16 on the CES-D (depressive symptoms) so that the adverse effect of RALES was stronger for women with a CES-D >16. (The cut point of 16 is frequently used in a number of CES-D studies.³³) In this model, the inclusion of locus of control magnified the interaction of racism with the CES-D. No interactions were observed for models with the following hypothesised effect modifiers: response to racism experiences (RRE) and its subscales, social support, the family resource scale (FRS) and its subscales, locus of control and anxiety during pregnancy.

Table 2.

Racism Lifetime Experiences (RALES), stress and depressive symptoms (CES-D): crude risk of preterm birth, Baltimore, Maryland 2001–04

	n	% Preterm	Unadjusted hazard ratio [95% CI]
Racism Lifetime Experiences
Top quartile	214	15.4%	0.88 [0.58, 1.35]
Bottom three quartiles	618	17.2%	1.00 Reference
Racism Lifetime Experiences
Above median	425	16.7%	1.00 [0.70, 1.44]
Below median	407	16.7%	1.00 Reference
Stress
Top quartile (≥21)	220	19.6%	1.28 [0.89, 1.84]
Lower three quartiles (<21)	610	15.7%	1.00 Reference
Depressive symptoms
CES-D ≥ 16	351	19.9%	1.46 [1.05, 2.04]
CES-D < 16	481	14.4%	1.00 Reference

Figure 1.

Risk of preterm birth: hazard ratios (HR) for racism stratified by stress and CES-D on log scale (depressive symptoms) with 95% confidence intervals.

Table 3.

Racism and Lifetime Experiences and risk of preterm birth, Baltimore, Maryland 2001–04: interaction models

Model terms	Stratum	Total births	Per cent preterm	Racism HR unadjusted	Interaction P-value	Racism HR adjusteda	Interaction P-value
Two way interactions: individual models
Racism, Stress,	Lower stress	610	15.7%	0.88 [0.59, 1.32]	Racism*Stress	0.92 [0.61,1.38]	Racism*Stress
Racism*Stressb	High stress	220	19.5%	1.29 [0.83, 2.01]	P = 0.41	1.30 [0.83, 2.04]	P = 0.30
Racism, CES-D,	High CES-D	351	19.9%	1.41 [0.93, 2.14]	Racism*CES-D	1.40 [0.92, 2.14]	Racism*CES-D
Racism*CES-Dc	Lower CES-D	481	14.3%	0.71 [0.43, 1.15]	P = 0.07	0.70 [0.42, 1.14]	P = 0.04
Two way interactions: joint model
Racism, Stress, CES-D	High stress, high CES-D	182	18.7%	1.43 [0.90, 2.30]	LR interaction test	1.47 [0.91, 2.39]	LR interaction test
Racism*Stress					P = 0.19		P = 0.10
	High stress, lower CES-D	42	21.4%	0.74 [0.38, 1.45]		0.79 [0.40, 1.57]
Racism*CES-Da					Racism*Stress		Racism*Stress
	Lower stress, high CES-D	177	20.3%	1.35 [0.80, 2.29]	P = 0.96	1.34 [0.78, 2.29]	P = 0.54
					Racism*CES-D		Racism*CES-D
	Lower stress, lower CES-D	445	13.5%	0.70 [0.43, 1.14]		0.72 [0.43, 1.20]
					P = 0.12		P = 0.12
Two- and three-way interactions: joint model
Racism, Stress, CES-D	High stress, high CES-D	182	18.7%	1.29 [0.78, 2.12]	LR interaction test	1.32 [0.79, 2.20]	LR interaction test
Racism*Stress					P = 0.10		P = 0.05
Racism*CES-D	High stress, lower CES-D	42	21.4%	1.37 [0.58, 3.23]	Racism*Stress	1.51 [0.64, 3.57]	Racism*Stress
Racism*Stress*CES-Da					P = 0.18		P = 0.08
	Lower stress, high CES-D	177	20.3%	1.56 [0.92, 2.63]	Racism*CES-D	1.55 [0.90, 2.64]	Racism*CES-D
					P = 0.03		P = 0.04
	Lower stress, lower CES-D	445	13.5%	0.61 [0.36, 1.04]	Racism*Stress*CES-D	0.63 [0.36, 1.08]	Racism*Stress*CES-D
					P = 0.07		P = 0.06

^aAdjusted for cigarette smoking second trimester and high locus of control.

^bAdjusted for second trimester cigarette smoking.

^cAdjusted for locus of control.

Racism: Racism Life Experiences scale, above median racism experiences is exposure.

CES-D: Center for Epidemiologic Studies Depressive symptoms scale, score ≥ 16 is exposure.

Stress: Hassles scale, top quartile of stress is exposure.

LR interaction test P-value: compares model with interaction(s) to model with main effects only.

HR: hazard ratio; LR: likelihood ratio.

After identification of these two (separate) two-way interactions, we fitted a single regression model that included both of the two-way interactions (RALES-stress, RALES-CES-D). Based on a priori hypotheses, we then fitted a model with a three-way interaction and two two-way interactions (Table 3, Figure 1). Based on the improvement in the likelihood ratio test (P = 0.10) and the P-values for all three interactions, this model was markedly better than the prior model without the three-way interaction. Upon further adjustment for cigarette smoking in the second trimester and high locus of control, the P-values for all three interactions in the fully adjusted model were <0.10 and the P-value was <0.05 for the LR test (Table 3). A score above the median on the RALES was associated with an increased risk of preterm birth in three of the four strata; reported perceived racism had a slightly protective effect in the stratum of women with low scores on both the stress and depressive symptoms scales.

Discussion

Despite many exhortations on the potential health impact of racism, the literature on racism and birth outcomes has remained sparse. In our study of racism and preterm birth, we sought to add to this literature by addressing limitations of past work and expanding the conceptual framework. This included going beyond the prenatal period to consider exposures across a woman’s life course. The reports in the literature discussed in the introduction suggested that racism may influence preterm birth through exposures and experiences that occur across a woman’s lifetime and before conception and not just through incidents occurring during her pregnancy. This is also consistent with the ‘weathering hypothesis’ put forth by Geronimus.³⁹ Accordingly, we measured lifetime exposure to perceived racism. Also, in contrast to some studies of racism and birth outcomes, we assessed lifetime exposure to perceived racism using a scale demonstrating excellent psychometric properties for our sample cohort rather than a single item.

While exposure to perceived racism may be a risk factor for preterm birth, a woman’s response to racism could explain differential impacts of similar experiences.³¹ In the smaller Collins study²⁰ as well as in the Mustillo study of the CARDIA cohort,¹⁹ women who reported racism were also asked two follow-up questions about their response to the experiences. No direct effect or role as a mediator or moderator of the racism effect was reported in either paper. In our analyses, response to the racism experiences (RRE scale) also failed to play a measurable role as either a mediator or moderator of the racism experiences.

Other social and psychosocial factors, including stress and depressive symptoms, may also be an important part of the context in which racism is experienced and influences health. Numerous authors have theorised that understanding the effect of racism on health is complex and requires consideration of a broad range of factors.^40,41 Yet few of the prior published studies of racism and birth outcomes collected information on other social or psychosocial factors. Even when data were available, no positive or negative findings with regard to mediation or moderation of the racism effects were reported.^19,20,23 A strength of our work is the inclusion of a wide range of factors that may modify the effects of racism experiences, including a woman’s response to racism experiences. We found that higher levels of lifetime racism influenced preterm birth risk, but in a complex manner with significant interactions with stress and prenatal depressive symptoms. It may be that these lifetime experiences of racism make a woman more vulnerable to other triggers of preterm birth (stress, depressive symptoms). The younger age distribution of our sample might also limit detection of effects of racism as such women may not have a sufficiently long period of ‘lifetime’ experiences of racism to influence birth outcomes and exposure may be more likely to exert its greatest toll on older women. As might be expected from such a nascent area of research, many questions remain unanswered. Race is not the only factor that may lead to prejudice and discrimination. Ultimately, the intersectionality of the oppression stemming from race as well as other factors such as class and gender must be appreciated to understand the experiences of black women and why their health and that of their children is jeopardised. The Ontario Human Rights Commission defined the phenomenon of intersectionality as ‘… multiple forms of discrimination occurring simultaneously’.⁴² The concept of intersectionality has been advanced by a number of leading scholars in recent years.⁴³ However, we still lack valid and reliable measures to assess exposures and experiences in this manner. Studies on health outcomes have generally focused on each form of oppression separately (each of the ‘isms’, racism and sexism). Future work should endeavour to find ways to measure and study the complexity of intersectionality as it relates to health outcomes.

Our work and that of others in this area provide support for continued investigation of the links between racism and preterm birth. While understanding how racism may relate to the risk of preterm birth is an important area for future study, our results also suggest that other factors may moderate the effects of racism and may point to possible areas in which to develop interventions. Although we identified two important psychosocial factors as effect-modifying (stress and depressive symptoms), the scale we used to assess a woman’s response to racism did not appear to play a role in the link between racism and preterm birth. It may be that this scale did not capture important coping responses. Future studies need to examine racism in the context of the social, psychosocial and biological environment.

In our sample of low-income African American women living in Baltimore City, lifetime experiences of racism appear to have an impact on risk of preterm birth after considering other social and psychosocial moderating factors. The adverse effects of racism may be missed if vulnerable subgroups are not identified and the context is not considered.