Measures of Racism and Discrimination in Preterm Birth Studies

Abstract

Objective:

Preterm birth, any birth < 37 weeks gestation, disproportionally affects Black birthing people and is associated with adverse perinatal and fetal health outcomes. Racism increases the risk of preterm birth, but standardized measurement metrics are elusive. This narrative synthesis examines literature on measures of racial discrimination used in preterm birth research.

Data Sources:

Six databases (CINAHL, Cochrane, Embase, PubMed (MEDLINE), Scopus, Web of Science) and ClinicalTrials.gov were searched. Search terms were categorized into three groups (racism terms, measurement terms, preterm birth terms) to identify original research articles that explored associations between racism and preterm birth. English-language, original research articles with US populations were included.

Methods of Study Selection:

Studies were excluded if conducted in only White populations, if only paternal factors were included, or if only racial differences in preterm birth were described. Articles were independently reviewed by two blinded researchers for inclusion at every stage of screening and data extraction; a third reviewer resolved discrepancies.

Tabulation, Integration, and Results:

Sixty studies were included in the final analysis. Articles primarily included measures examining interpersonal forms of racism (n=17) through Experiences of Discrimination and Everyday Discrimination scales; neighborhood composition (n=22) with Neighborhood Deprivation Index (NDI) and Index of Concentration and the Extremes (ICE); policy level racism (n = 12) through institutions like residential racial segregation, or policy inequities; or multiple forms (n=9).

Conclusions:

Among studies, assessment methods and application of constructs varied. This heterogeneity poses significant challenges to understanding associations between racial discrimination and preterm birth, and describing potential etiologic pathways of preterm birth, which ultimately hinders development of effective intervention. Strategies to capture multilevel exposures to racism require the development and expansion of metrics that are culturally-inclusive, empirically valid and reliable among Black pregnant populations.

Systematic Review Registration:

PROSPERO, CRD42022327484.

Precis:

Studies of racism and preterm birth use inconsistent, unstandardized metrics of racial discrimination; standard measures should be developed for application in perinatal research.

INTRODUCTION

Black women in the United States experience higher rates of preterm birth compared to White women (14% vs 9%)(1-3), likely influenced by multiple social and structural factors (4-6). During pregnancy, experiences of gendered racism may cause trauma, stress, and disruptions to the physiology of pregnancy (7, 8). Racism has been linked to oxidative stress and antioxidant imbalance that can reduce telomere length in gestational tissues resulting in cellular senescence (9). And, while Kramer and Hogue (10) caution that current literature cannot support direct casual links between racism and preterm birth, they suggest several plausible proximal (maternal inflammation, vascular, and neuroendocrine dysfunction) and distal (health care, genetics, and epigenetics) pathways to this adverse birth outcome due to stress. Thus, there is urgency to both clarify the physiological implications of racism on pregnancy, and to quantify and address racism at multiple levels of perinatal research (11). Without contextualizing one’s adverse birth outcomes in environmental exposures to discrimination, there will be little progress in reducing racial disparities in maternal and fetal morbidity and mortality (12).

A broad call for standardized methods, measures, and approaches to measuring racism and racial discrimination in epidemiologic studies serves as the impetus for the purpose of this work (13). Explorations of racism in preterm birth studies are not new. However, measurement is inconsistent, and there is variability in the forms of racism examined. Racism has been identified in three forms: (1) internalized, (2) interpersonal, and (3) structural (14). To date, preterm birth literature has largely captured individual-level exposure to racism. For instance, perceived discrimination was independently associated with low birthweight, preterm birth, and small for gestational age in an integrated review (15). Among 1,410 Black women with mild to moderate depression, daily experiences of racist microaggressions were significantly associated with preterm birth (6). Consistently worrying about treatment due to race was found to increase risk of preterm birth among another cohort of 2,201 Black women (16). Of the existing reviews that have examined racism in preterm birth (15, 17-19), most have focused on individual forms with the main objective to determine if associations with preterm birth exist. This emphasis on individual-level racism in preterm birth literature, without considering structural discrimination, which is much more pervasive and impactful on access to resources and opportunities, is a disservice to the experiences of Black birthing people. To build towards standardized measures of structural racism and racial discrimination in the field, it is imperative to benchmark how racism is currently assessed in preterm birth research. Therefore, our objective was to conduct a narrative synthesis examining measures of racism in preterm birth studies.

SOURCES

This study was a narrative synthesis of literature examining racism as an exposure and preterm birth (defined as a live birth at <37 weeks completed gestation) as an outcome. This systematic review was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO ID# CRD42022327484).

Six databases (CINAHL, Cochrane, Embase, PubMed (MEDLINE), Scopus, Web of Science) and ClinicalTrials.gov were used to identify original research articles, published in English, that measured associations between racism and preterm birth. Databases were searched from March 10, 2022, to March 18, 2022, for articles published from inception to the date of search. Search terms were categorized into three groups to capture articles that used metrics of racism in preterm birth research: racism terms, measurement terms, preterm birth terms (Appendix 1, available online at http://links.lww.com/xxx).

STUDY SELECTION

Articles were included if they: (1) summarized an original research article (2) included preterm birth or gestational age at delivery as an outcome of interest; and (3) measured aggregate or individual-level exposure to racism or racial discrimination. Articles were excluded if they: (1) were restricted to a non-US population (2) included only White race participants; (3) evaluated only paternal (non-birthing people) factors as exposures; or (4) examined racial or ethnic differences in preterm birth without measurement of racism.

Upon completion of searches, all titles and abstracts were uploaded to Covidence, an online systematic review tool (20). At each stage, review was completed by two independent, blinded reviewers. Discrepancies were resolved by a third reviewer. Reasons for exclusion were noted in a full-text review. Data extraction was performed by two independent, blinded study team members; individuals were trained in a protocol to extract data similarly. In cases of conflicts, final data were chosen by a third team member.

Study characteristics extracted included: article title, first author, publication year, DOI, objective, hypothesis, study design and timing, sample size overall and by race or ethnicity, study region and state, and years of data collection. Studies examining preterm birth, defined as all births < 37 weeks gestation, as a primary outcome were included. Secondary birth outcomes included: gestational age (weeks of pregnancy), low birth weight (<2500 grams), birthweight (first weight of infant after birth), small for gestational age (birthweight < 10^th percentile for gestational age), and infant mortality (infant death occurring <1 year).

For each metric of racial discrimination, we extracted: name of the metric, original publication of metric, investigator-written description, dataset used, level of racism (perceived, neighborhood, structural), dimension(s) of racism (housing, disorder, education, employment, socioeconomic status, criminal legal, psychological, other), and previous assessments of validity or reliability. The study team prepared a narrative summary of the 60 included articles by level and dimension of the metric of racial discrimination.

RESULTS

The literature search resulted in 13,176 articles (Figure 1); 3,846 were deemed duplicates and removed. Of the remaining 9,330 titles and abstracts that were screened, 9,203 were excluded for not including measures of racism or for not reporting preterm birth as an outcome. The remaining 127 articles were included for full-text review. Sixty-six articles were excluded during full-text review: 19 were not original research articles, 13 did not include preterm birth as an outcome, 24 did not discuss racial discrimination, 10 only considered racial disparities and did not explicitly measure racism (Figure 1). Data were extracted from the remaining 60 articles; 85 metrics of racial discrimination were counted.

Figure 1.

Articles included are displayed by primary exposure measures examining neighborhood composition (n=22), interpersonal (perceived or self-reported) (n=17), policy level inequities (n=12), or multiple forms (n=9) (Figure 2). Of the nine that examined multiple levels of racism, one considered perceived racism and structural forms through neighborhood composition, seven measured neighborhood composition only as a form of structural racism, and one included metrics of all three levels.

Figure 2.

Twenty-two articles (36.7%) (21-42) considered exclusively neighborhood factors, and six of these articles used more than one metric to assess neighborhood measures. In total, 40 metrics of neighborhood level racial discrimination were used (Table 1). Studies of neighborhood composition included dimensions of housing, socioeconomic status, social disorder, built environment, and general environment.

Table 1:

Neighborhood Composition as a Form of Structural Racism in Preterm Birth Studies

Dimension(s)	Metric of Racism	Metric Reference	First Author, Year	Study design	Region	State	Racial composition	Dataset to generate metric	Study team subjective rank
Housing; Disorder; Built environment	4-factor neighborhood characteristics model*	Evenson 2009; Laraia 2006	Vinikoor-Imler 2011 † ‡	Retrospective cohort study	South	NC	Black and white	2000 Census	Medium
Criminal-legal	Annual county-level jail incarceration rate		Jahn 2020 †	Retrospective case-control study	National	National	Black and white	Bureau of Justice Statistics Census of Jails and Annual Survey of Jails.	Medium
Built environment	Built environment index	Miranda 2012a; Miranda 2012b; Kroeger 2012	Anthopolos 2014 †	Retrospective cohort study	South	NC	Black and white	2008 Duke Community Assessment Project, 2006-2007 Durham PD Crime Analysis Lab; 2008 Durham county tax assessor data	High
Environment	California Communities Environmental Health Screening Tool		Padula 2018	Retrospective cohort study	West	CA	Black, white, Latine, and Asian	CalEnviroScreen 2.0	High
Criminal-legal	County-level homicide		Mark 2022	Retrospective case-control study	National	National	Black, white, and Latine	1991-2002 FBI UCR	High
Housing	Gi* statistic	Getis & Ord 1992	Salow 2018	Retrospective case-control study	Midwest	Multiple	Black only	2010 Census	Medium
Housing; SES	Index of Concentration at the Extremes	Massey 2001; Krieger 2015	Chambers 2019 †	Retrospective analysis of vital statistics	West	CA	Black only	2011 ACS	High
			Huynh 2018 †	Cross sectional case-control study	Northeast	NY	Black, white, Latine, and Asian	2010 ACS, 5 year	High
			Jahn 2020 †	Retrospective case-control study	National	National	Black and white	2000 Census; ACS (no year)	Medium
			Shrimali 2020 †	Retrospective cohort study	West	CA	Black, white, and Latine	1980, 1990, 2000 Census; 2007 ACS, 5 year	High
			Thoma 2019 †	Retrospective case-control study	National	National	Black and white	2016 ACS, 5 year	Medium
Housing; Education; Employment; SES	Neighborhood deprivation index	Messer 2006	Blebu 2021	Cross sectional survey	West	CA	Black only	2010 ACS	Low
			Janevic 2010	Retrospective analysis of vital statistics	Northeast	NY	None of the above	2000 Census	High
			Ma 2015	Retrospective case-control study	South	SC	Black and white	2000 Census	High
			Messer 2010	Retrospective case-control study	South	NC	Black and white	2000 Census	Medium
			O'Campo 2008	Retrospective cohort study	Multi-Site	Multiple	Black and white	2000 Census	Medium
			Vinikoor-Imler 2011 † ‡	Retrospective cohort study	South	NC	Black and white	2000 Census	Medium
Housing; Education; Employment; SES	Neighborhood deprivation index§	Messer 2006	Kramer 2014	Retrospective cohort study	South	GA	Black, white, and Latine	1990, 2000 Census; 2005 ACS, 5 year	High
Housing; Education; Employment; SES	Neighborhood disadvantage index∥*	Sealy-Jefferson 2017	Sealy-Jefferson 2019 †	Retrospective cohort study	Midwest	MI	Black only	ACS (no year)	Medium
SES	Neighborhood economic trajectories (GINI index)		Cubbin 2020 †	Retrospective cohort study	South	TX	Black and white	1990, 2000 Census; 2006 ACS, 5 year	Medium
Housing; Disorder; Built environment	Neighborhood environment scales∥*	Mujahid 2007; Echeverria 2004	Sealy-Jefferson 2016	Retrospective cohort study	Midwest	MI	Black only	LIFE	High
SES	Neighborhood poverty trajectories	Margerison-Zilko 2015	Cubbin 2020 †	Retrospective cohort study	South	TX	Black and white	1990, 2000 Census; 2006 ACS, 5 year	Medkum
Disorder	Neighborhood quality	Coulton 1996; Elo 2009	Mendez 2014 †	Retrospective cohort study	Northeast	PA	Black, white, and Latine	SPEAC	High
Housing	Neighborhood racial isolation & dissimilarity spatial index	Massey & Denton 1988; Reardon & O'Sullivan 2004	Kramer 2010	Retrospective cohort study	National	National	Black and white	2000 Census	Medium
Housing	Neighborhood racial isolation spatial index	Reardon & O’Sullivan 2004	Anthopolos 2011	Retrospective cohort study	South	NC	Black amd white	2000 Census	Medium
Housing	Neighborhood racial isolation spatial index	Anthopolos 2011	Anthopolos 2014 †	Retrospective cohort study	South	NC	Black and white	2000 Census	High
Disorder; Built environment	Perceived physical and social residential environment ∥*	Sealy-Jefferson 2015	Sealy-Jefferson 2019 †	Retrospective cohort study	Midwest	MI	Black only	LIFE	Medium
SES	Poverty Exposure	Massey & Fischer 2000	Britton 2013 † ¶	Retrospective analysis of vital statistics	National	National	Black and Latine	2000 Census	Medium
Housing	Residential segregation	Massey & Denton 1988	Bell 2006	Cross sectional survey	National	National	Black only	2000 Census	Medium
			Britton 2013 † ¶	Retrospective analysis of vital statistics	National	National	Black and Latine	2000 Census	Medium
			Chambers 2018†#	Cross sectional case-control study	West	CA	Black and white	2009 ACS, 5 year	High
			Grady 2010	Retrospective cohort study	South	MI	Black and other	2000 Census	Medium
			Margerison-Zilko 2017 †	Retrospective analysis of vital statistics	National	National	Black only	2010 Census	Medium
			Osypuk 2008	Retrospective case-control study	Multi-Site	Multiple	Black and white	2000 Census	High
			Mendez 2014 †	Retrospective cohort study	Northeast	PA	Black, white, and Latine	2000 Census	High
Disorder	Ross Neighborhood Disorder Scale∥*	Ross & Mirowsky 1999	Dove-Meadows 2020 †	Mixed methods	Midwest	Multiple	Black only	Unnamed study	Medium
Housing; Built environment	Neighborhood physical characteristics		Zuberi 2016 †	Retrospective ecological study	Northeast	PA	None of the above	PNCIS, USPS, HUD	Low
Housing; Disorder	Neighborhood social environment		Zuberi 2016 †	Retrospective ecological study	Northeast	PA	None of the above	ACS (no year); PNCIS	Low
Environment	Urban green space		Tvina 2021	Retrospective case-control study	Midwest	WI	Black only	National Land Cover Database	High
Housing; Built environment	Tax foreclosure		Sealy-Jefferson 2019 †	Retrospective cohort study	Midwest	MI	Black only	Wayne County Treasurer; Data Driven Detroit	Medium

^*indicates metric has been assessed for reliability

^†indicates study listed more than once

^‡indicates PTB as secondary outcome

^§indicates metric referenced was modified by investigator

^∥indicates metric has been assessed for validity

^¶indicates very preterm birth (VPTB) as primary outcome

^#indicates gestational age as primary outcome

Seven articles (11.7%)(23, 25, 27, 35, 43-45) included measures of neighborhood racial segregation through Massey & Denton’s 1988 conceptualization of five distinct measures: clustering, centralization, evenness, exposure, and concentration (46). Segregation was also approximated through the Neighborhood Racial Isolation Spatial Index, derived from Massey & Denton and expanded by Reardon and O’Sullivan (47). For example, Anthopolos and colleagues (21, 22) approximated a measure of racial segregation based on the regional-scale spatial isolation index (47) and included modifications to account for population interaction within neighborhoods and application to Census block groups (21). These measures were applied at the county or metropolitan area level and included studies that examined Black-White differences using Census data as the primary residential data source.

The Neighborhood Deprivation Index (NDI) proposed by Messer and colleagues (48) developed and standardized an index of contextual variables to examine impacts of neighborhood factors and associated perinatal health outcomes. The NDI incorporates measures of poverty, education, employment, housing, and occupation from Census data through a principal component analysis (PCA). Six studies (10%) deployed the NDI as described in the original proposal and validation (24, 28, 31, 33, 34, 41). For instance, Blebu et al. (24) conducted a PCA to determine how each measure of deprivation contributed to variance in the sample and created a z-score transformed index score, with higher scores representing higher neighborhood deprivation. Modified versions of the NDI were also included. Kramer, Dunlap & Hogue (30) performed the suggested analysis of eight measures that deviated from the recommended and included prevalence of “female-headed households” and “adults with management or professional positions.”

The Index of Concentration at the Extremes (ICE) introduced by Massey in 2001 (49) quantifies measures of concentrated privilege and deprivation by approximating geographic distributions of income inequality by race. Krieger and colleagues (50) expanded Massey’s original concept to three scores: ICE race, ICE income, and ICE race + income. ICE has been applied to perinatal outcomes. For example, Chambers et al. (43) included ICE as an approximation of structural racism in a California birth cohort, creating scores that ranged from −1 to 1 and grouping participants into quintiles ranging from least to most privileged; women living in least privileged zip codes were more likely to experience preterm birth compared to women living in the most privileged zip codes, and effects remained after adjusting for maternal factors (43). Five studies (8.3%) used the ICE in their study as a primary exposure and approximation for neighborhood and structural racism (43, 51-54).

The most frequent dimension of structural racism assessed was housing. The most common metric was the Index of Concentration at the Extremes (ICE), used by five investigators. Other studies operationalized redlining and residential segregation policies by combining Home Owners’ Loan Corporation (HOLC) historical maps (55-57) or Home Mortgage Disclosure Act (HMDA) records (45, 58) with contemporary birth data. Three of these studies were conducted with state-level data from the Northeast.

Neither socioeconomic status nor poverty alone is a metric of racism. Still, three studies (5.0%) exclusively evaluated neighborhood poverty as a proxy for racial discrimination via area and neighborhood deprivation. Cubbin et al. (26) examined trajectories of neighborhood wealth and poverty in relationship to Black-White disparities in preterm birth over time. They calculated a Gini index, a measure of neighborhood income inequality, to categorize census tracts into four poverty trajectories (59). Salow and colleagues (37) calculated a Getis-Ord Gi* statistic (60), a spatial z-score that estimates how racial composition of a neighborhood differs from a regional mean, to examine associations of segregation among non-Hispanic Black women with diagnosed preterm birth from medical records review. .(60)Britton et al. (25) examined the influence of racial segregation on preterm birth and determine differences across geographic exposures to poverty by calculating the natural log of population size and median household income (in $1000s) and examining the Black or Hispanic poverty rate.

Six studies (10.0%) described social disorder as an indicator of neighborhood-level racial discrimination (38, 39, 41, 42, 45, 61). Mendez and colleagues (45) examined residential redlining, crime and safety, physical disorder, and social disorder in association with preterm birth. Of the six studies describing social disorder, four of the scales also included the built environment. Vinikoor-Imler et al. (41) described social disorder and built environment through Census block level indices of several domains: Physical Incivilities, Walkability, Decoration, Arterial or Thoroughfare, and Social Spaces. Other studies used objective neighborhood physical characteristics like property maintenance, housing vacancy, social spaces, and tax foreclosure to describe the built environment (22, 41, 42). Sealy-Jefferson et al. (39) examined the influence of neighborhood tax foreclosures - considered a measure of social disorder - on longitudinal risk of preterm birth among African American women; these researchers also implemented a metric of participant-reported perceptions of neighborhood characteristics: walkability, access to healthy foods, safety, and social cohesion (38, 39). Two studies described environmental racism in association with preterm birth. Padula et al. (36) used the California Communities Environmental Health Screening Tool to regress individual pollution indicators to birth outcomes in Fresno County. Tvina et al. (40) quantified proximity of urban green space to maternal residence in Milwaukee.

Seventeen articles (28.3%) (4, 6, 16, 45, 62-74) examined forms of individual or interpersonal discrimination as a risk factor for preterm birth (Table 2). Three articles used two metrics to assess interpersonal racism. Most commonly used were the validated Experiences of Discrimination Scale (75) and the Everyday Discrimination Scale (76). Additional scales used were the Daily Life Experiences of Racism and Bother (77), the John Henryism Active Coping Scale (78), the Perceived Racism Scale (79), Racism and Life Experiences Scale and the Racism-Related Experiences Scale (80). Short items from the Pregnancy Risk Assessment Monitoring Survey (PRAMS) inquired whether the respondent had been upset by experiences of racism; three studies used these indicators (26, 62, 63). One study cited use of nine questions measuring racial discrimination included in the 1997 Black Women’s Health Study (81). Chronic worry about race related treatment were included in two studies (16, 26). Nine studies included Black participants only. Studies were often cross-sectional, and about half contained data from prospective cohort studies. Studies were most frequently conducted at the state-level.

Table 2:

Self-reported Racial Discrimination in Preterm Birth Studies

Dimension	Metric of Racism	Metric Reference	First Author, Year	Study design	Region	State	Racial composition	Dataset to generate metric	Study team subjective rank
Psychosocial	Daily Life Experiences of Racism and Bother∥*	Bond 2008	Slaughter-Acey 2016	Retrospective cohort study	Midwest	MI	Black only	LIFE	High
Psychosocial	Everyday Discrimination Scale∥*	Williams 1997	Fryer 2020	Secondary analysis of prospective cohort	Multi-Site	Multiple	Black and Latine	CCHN	Medium
			Mendez 2014 †	Retrospective cohort study	Northeast	PA	Black, white, and Latine	SPEAC	High
Psychosocial	Everyday Discrimination Scale§	Dominguez 2008	Daniels 2020	Cross sectional survey	West	CA	Black only	AAWHHS	Low
Psychosocial	Everyday Discrimination Scale§	Williams 1997; Krieger 2005	Dominguez 2008 ‡	Prospective cohort study	West	CA	Black and white	Unnamed study	Low
Psychosocial	Experiences of Discrimination∥*	Krieger 1990; Krieger 2005	Dove-Meadows 2020 †	Mixed methods	Midwest	Multiple	Black only	Unnamed study	Medium
			Giurgescu 2012	Cross sectional case-control study	Midwest	IL	Black only	Unnamed study	Low
			Grobman 2018	Prospective cohort study	Multi-Site	Multiple	Black, white, Latine, and Asian	nuMoM2b	Medium
Psychosocial	Experiences of Discrimination§	Krieger 1990; Krieger 2005	Dole 2004	Prospective cohort study	South	NC	Black and white	PINS	Low
Public sentiment	Implicit Association Test	Greenwald 1998	Orchand 2017	Retrospective analysis of vital statistics	National	National	Black and white	Project Implicit	Low
Psychosocial	John Henryism Active Coping Scale	James 1996	Wheeler 2018 †	Secondary analysis of prospective cohort	South	NC	Black and white	HPHB	Low
Psychosocial	Life Experiences Survey§	Sarason 1978	Dole 2003	Prospective cohort study	South	NC	Black, white, and none of the above	PINS	Low
Psychosocial	Perceived Racism Scale	Dole 2004	Wheeler 2018 †	Secondary analysis of prospective cohort	South	NC	Black and white	HPHB	Low
Psychosocial	Perceived Racism Scale§	McNeilly 1996	Rankin 2011	Cross sectional case-control study	Midwest	IL	Black only	Unnamed study	Low
Psychosocial	PRAMS: upset by experiences of racism		Barber 2021	Cross sectional survey	South	VA	Black, white, and none of the above	PRAMS	Low
Psychosocial			Bower 2018	Cross sectional survey	Multi-Site	Multiple	Black only	PRAMS	Low
Psychosocial			Kim 2020†	Cross sectional survey	West	WA	Black, white, Latine, and Asian	PRAMS	Medium
Psychosocial	Racism and Lifetime Experiences Scale (RALES)∥*	Harrell 1997	Misra 2010 †	Qualitative	South	MD	Black only	Unnamed study	Medium
Psychosocial	Racism-Related Experiences (RRE) Scale	Harrell 1997	Misra 2010 †	Qualitative	South	MD	Black only	Unnamed study	Medium
Psychosocial	1997 BWHS questionnaire		Rosenberg 2002	Cross sectional survey	National	National	Black only	BWHS	Low
Psychosocial	Chronic worry about race-based unfair treatment		Braveman 2017	Cross sectional survey	West	CA	Black and white	CA MIHA	Low
Psychosocial	Race-specific chronic stress index		Kim 2020†	Cross sectional survey	West	WA	Black, white, Latine, and Asian	PRAMS	Medium

^*indicates metric has been assessed for reliability

^†indicates study listed more than once

^‡indicates PTB as secondary outcome

^§indicates metric referenced was modified by investigator

^∥indicates metric has been assessed for validity

^¶indicates very preterm birth (VPTB) as primary outcome

^#indicates gestational age as primary outcome

Structural policies were categorized into three themes: criminal-legal system encounters, public sentiment, and policy and political representation (Table 3). “Criminal-legal system encounters” considered racial inequities in policing, jails, prisons, and courts as predictors of preterm birth. “Public sentiment” describes novel, internet-based measures of racial zeitgeist that were associated with preterm birth. “Policy and political representation” features studies that considered analyses of legislation or racial composition of political power as correlates of preterm birth.

Table 3:

Policy and Governmental Practices as Forms of Structural Racism in Preterm Birth Studies

Dimension(s)	Metric of Racism	Metric Reference	First Author, Year	Study design	Region	State	Racial composition	Dataset to generate metric	Study team subjective rank
Criminal-legal	Annual county-level jail incarceration rate		Jahn 2020 †	Retrospective case-control study	National	National	Black and white	Bureau of Justice Statistics Census of Jails and Annual Survey of Jails.	Medium
Policy	Earned income tax credit	Markowitz 1982	Komro 2019 #	Quasi-experimental difference-in differences	National	National	Black and white	EITC policies	Medium
Criminal-legal	Incarceration prevalence		Dyer 2019	Cross sectional case-control study	South	LA	Black only	Vera Institute of Justice	Low
Public sentiment	Internet query-based area racism	Stephens-Davidowitz 2014; Chae 2015	Chae 2018	Retrospective analysis of vital statistics	National	National	Black only	Google Trends at Designated Market Area	Low
Housing	HOLC grades		Hollenbach 2021	Retrospective Cohort study	Northeast	NY	None of the above	1940s HOLC	High
Housing	HOLC grades		Krieger 2020	Retrospective Cohort study	Northeast	NY	Black, white, Latine, amd Asian	1938 HOLC	High
Housing	HOLC grades	Aaronson 2019; Jacoby 2018	Nardone 2020	Retrospective Cohort study	West	CA	Black, white, Latine, and Asian	1940s HOLC	High
Housing; SES	Index of Concentration at the Extremes	Massey 2001; Krieger 2015	Chambers 2019 †	Retrospective analysis of vital statistics	West	CA	Black only	2011 ACS	High
			Huynh 2018 †	Cross sectional case-control study	Northeast	NY	Black, white, Latine, and Asian	2010 ACS, 5 year	High
			Jahn 2020 †	Retrospective case-control study	National	National	Black and white	2000 Census; ACS (no year)	Medium
			Shrimali 2020 †	Retrospective Cohort study	West	CA	Black, white, and Latine	1980, 1990, 2000 Census; 2007 ACS, 5 year	High
			Thoma 2019 †	Retrospective case-control study	National	National	Black and white	2016 ACS, 5 year	Medium
Criminal-legal	Judicial treatment	Lukachako 2014	Chambers 2018†#	Cross sectional case-control study	West	CA	Black and white	2012 Center on Juvenile and Criminal Justice	High
Criminal-legal	Police contact		Hardeman 2021	Cross sectional Cohort study	Midwest	MN	Black and white	City of Minneapolis Police Incident Report	High
Policy	Political participation: elected office	Lukachako 2014	Chambers 2018†#	Cross sectional case-control study	West	CA	Black and white	2016 county board of supervisor website	High
Policy	Political participation: representation	LaVeist 1993	Margerison-Zilko 2017 †	Retrospective analysis of vital statistics	National	National	Black only	City council sites	Medium
Criminal-legal	Violent crimes; Neighborhood deprivation index	Messer 2006	Messina 2013	Retrospective analysis of vital statistics; crime data	South	GA	Black, white, and none of the above	Atlanta area crime data; GA vital statistics	High
Criminal-legal	PRAMS: Criminal justice involvement		Dallaire 2018	case-control study	South	VA	Black and none of the above	PRAMS	Medium
Housing	Redlining index	Mendez 2013	Matoba 2019	cross-sectional, retrospective population-based study	Midwest	IL	Black only	1990-1995 HMDA	High
Housing	Residential redlining	Mendez 2011	Mendez 2014 †	Retrospective Cohort study	Northeast	PA	Black and Latine	HMDA	High
Policy; Other: immigration	State immigrant policies (inclusive and criminalizing)		Sudhinaraset 2021	Retrospective analysis of vital statistics	National	National	Black, white, Latine, and Asian	Immigration policies	High
Public sentiment	State-level Twitter sentiments	Sanders 2011	Nguyen 2018	Cross-sectional analysis of vital statistics	National	National	None of the above	Twitter API	Low

^*indicates metric has been assessed for reliability

^†indicates study listed more than once

^‡indicates PTB as secondary outcome

^§indicates metric referenced was modified by investigator

^∥indicates metric has been assessed for validity

^¶indicates very preterm birth (VPTB) as primary outcome

^#indicates gestational age as primary outcome

Six studies considered encounters with the criminal-legal system as a form of structural racism. Messina and Kramer (82) examined area violent crimes in a city by Census block and vital records of preterm birth from 1989-2006. Two studies considered associations between jail incarceration and preterm birth, one nationwide and one in Louisiana (52, 83). Chambers et al. (43) defined judicial treatment as county-level ratio of Black to white people incarcerated for a felony. Dallaire et al. (84) investigated how recent incarceration of a birthing parent or partner was associated with preterm birth. Hardeman et al. (85) considered community-level police contact in Minneapolis as an exposure to preterm birth. Two studies used nationwide internet-based data sources to approximate area-level public sentiments of racial discrimination and associations with preterm birth (86, 87).

Policy-level associations with preterm birth were considered by four studies. State-level earned income tax credit (EITC) policies (88) and immigration policies (89) were operationalized as indicators of racism, and both articles used a national sample. Black political representation was assessed by two studies; Chambers et al. (43) created a ratio of Black to White individuals elected to a county’s board, while Margerison-Zilko et al. (44) created a ratio of percent Black city council members to percent Black voting-age residents in a city.

DISCUSSION

Measures varied widely in constructs used to define “racism.” Most commonly, literature utilizes metrics of neighborhood-level factors as a form of structural racism in association with preterm birth. Least commonly, literature considers policy measures and associations with preterm birth.

Methodological heterogeneity in exposure measurement, regardless of racial composition of a study, presents challenges to comprehensively understanding the impact of racism on preterm birth. Previous reviews have found associations between specific measures of discrimination and preterm birth. Crawford et al. (19) conducted a scoping review of 13 articles associating racial microaggressions and perinatal health outcomes; they found inconsistencies in definitions and measures of microaggressions. Kyung Lee et al. (90) examined 12 studies of redlined vs non-redlined neighborhoods and various adverse health outcomes and found significant associations with preterm birth. Mehra et al. (17) reviewed 42 studies of racial residential segregation and adverse birth outcomes where 18 measured preterm birth; authors found segregation measures varied widely. Ncube et al. (91) examined associations of neighborhood context with preterm birth and low birthweight, and determined that metrics of neighborhood disadvantage comprised of poverty, deprivation, racial residential segregation, racial composition, and crime. Of note, only seven studies identified were used in a meta-analysis because of variations in the racial composition of the included studies. Building on previous research, this current review uniquely focuses on the measures of racism across multiple domains.

No singular metric is likely to capture all sources or experiences with racism a birthing individual may face. Several investigators have suggested epidemiologic studies integrate multiple metrics across several domains to better measures of racism (13, 92, 93). In our subjective assessment, highly effective articles incorporated multiple measures across two or more domains to capture experiences of structural racism. Recently, ICE has emerged as the preferred measure to approximate structural racism in perinatal research as it incorporates multiple factors of race, income, and race + income. The use of ICE, in conjunction with other measures of structural racism (e.g., redlining) may strengthen a comprehensive assessment of structural racism in epidemiologic studies of preterm birth. A combination of metrics is likely necessary to thoroughly record experiences and environments of racism and its relation to preterm birth. Further, consensus on optimal metrics would improve comparability of research findings. For example, a standardized battery of validated and reliable self-report questionnaires to assess all three levels of racial discrimination – in conjunction with racial discrimination indices constructed from survey or Census data – may be ideal. Future efforts must identify valid and reproducible measures of racism, as standard metrics would benefit various studies of preterm birth and perinatal health.

Complexities of pregnancy and adverse birth outcomes must be acknowledged, as there is no single pathway. Kramer and Hogue (10) showed no consensus on how perceived stress from racism impacted mechanisms that in turn impacted preterm birth. Thus, research must advance methodological and analytic approaches to integrate multiple levels of social, environmental, and individual variables to describe impacts on pregnancy physiology and risk of adverse perinatal outcomes. Importantly, researchers must consider structural factors that impact perinatal health and move away from solely including race and ethnicity a covariate or effect modifier.

As with any review, bias may be introduced throughout article evaluation. To reduce bias, checkpoints were integrated to allow primary investigators with content expertise to review article inclusion.

Characteristics of studies (eg: geographic location, study design, racial composition) varied widely and were largely not generalizable or comparable. Many studies used self-reported survey measures as primary exposures; however, over half (n=34) assessed structural racism through neighborhood or environmental measures directly captured from publicly available sources. As notated as a limitation in another review (15), authors often used racism and discrimination interchangeably. It is understood that discrimination is a measurable conduit of racism, but does not broadly capture the multiple forms and levels within which racism permeates (14).

Racism is a known but inconsistently measured risk factor for preterm birth and assessment methods vary. This heterogeneity poses significant challenges to comprehensively describing the etiology of preterm birth and ultimately hinders effective intervention. Fully capturing impacts of racism may require multiple measures at multiple levels. Further, it is prudent to build consensus on optimal measures for consistency and comparability across perinatal research.