Introduction

A large body of evidence has recorded striking racial disparities in physical and mental health [1,2,3,4,5,6]. Therefore, examining how different manifestations of racism, including effects of institutionalized systems and interpersonal social dynamics in which individuals are “racialized”, affects health across the lifespan is a priority to improving population health [7,8,9]. For instance, heightened daily life stress and vigilance stemming from ongoing racialization may amplify the risk of poorer mental health and contribute to higher levels of chronic inflammation and accelerated multi-system biological aging [10,11,12,13,14,15]. Biological aging can be defined as the progressive loss of system integrity that occurs with advancing chronological age, including changes in DNA-methylation (DNAm; [16, 17]). DNAm measures of biological aging can be applied early in the life course to study the etiology of socially organized health disparities, decades before differences in disease and mortality are measurable [18].

Racial marginalization and low socioeconomic status has been associated with more advanced and faster biological aging as measured in DNAm in both adults and children, and, in adults, these differences in biological aging partially account for health disparities between and within racial groups [19,20,21,22,23]. A few studies have found DNAm measures of biological aging to be associated with mental health [21, 24, 25]. Yet, previous research has largely been cross-sectional in design and, thus, does not address the dynamic interplay between racialization, mental health, and biological aging.

Here we leverage a prospective birth cohort study to examine whether the emergence of racial disparities in mental health is linked to racial disparities in DNA-methylation measures of biological aging across childhood and adolescence. While previous studies have shown that individual- and family-level factors, such as family instability, are associated with child developmental and health outcomes [26,27,28,29], our analysis addresses an important gap in the literature by underscoring that macro-level racism reinforces health inequities at an early age. We conceptualized racial and ethnic disparities as outcomes of structural racism, more specifically, as “racialization”, which emphasizes the social processes and institutionalized systems in which individuals are positioned. First, we used self-identified race and ethnicity as an indicator of individuals’ racial social position. Importantly, race is a social construct, not a biological category, and the ways in which race is socially defined and experienced shape individuals’ access to resources and opportunities [30,31,32,33]. Second, we utilized the Thiel Index as a statistical measure that captures the extent of neighborhood racial segregation [34]. Racial segregation in US neighborhoods is largely a result of historical policies and practices such as redlining, discriminatory lending, and exclusionary zoning, which systematically marginalized communities [35]. Third, for racially marginalized youth who did not identify as solely White, we included skin tone as a proxy for colorism, recognizing that variations in skin color can influence social experiences and opportunities [34, 36,37,38]. Colorism perpetuates the racial hierarchy within marginalized communities as a consequence of historical conditions including colonialism, slavery, and the preferential treatment of lighter skin tones by dominant groups [36]. Moreover, we considered family-level and neighborhood-level socioeconomic conditions as well as perinatal (e.g., birthweight) factors, police interactions and parenting stress, as potential mediators of associations between structural racism and child development.

We probe these biosocial dynamics in preregistered analyses of N = 4898 participants from the Future of Families & Child Wellbeing Study, which intentionally oversampled financially under resourced families. Among the N = 2039 participants who had repeated saliva DNAm at ages 9 and 15 years, most participants racially positioned themselves as African-American/Black (n = 901, 47%), followed by Hispanic/Latinx (n = 511, 26%), White (n = 366, 19%), Multiracial (n = 99, 5%), and “Other” (n = 52, 3%).

Methods

Participants

The Future of Families and Child Wellbeing Study (FFCWS) follows a sample of 4898 children born in large US cities during 1998–2000. FFCWS oversampled children born to unmarried parents and interviewed parents at birth and ages 1, 3, 5, 9 and 15. During home visits, saliva DNA was collected the Illumina 450 K and EPIC methylation arrays with ages 9 (n = 1971) and 15 years (n = 1974) assayed on the same plate (for further information on DNA preprocessing see Supplemental Material). DNAm study participants self-identified race/ethnicity defined by study protocol as African-American/Black only (n = 901, 47%), “Other” (n = 52, 3%), Hispanic/Latinx (n = 511, 26%), Multiracial (n = 99, 5%), White (n = 366, 19%). The University of Michigan and Princeton University Institutional Review board granted ethical approval. Informed written consent was obtained from all participants and study participants’ legal guardians. All methods were performed in accordance with the relevant guidelines and regulations.

Measures

Mental health

Parent-reported internalizing and externalizing behaviors were assessed at ages 3, 5, 9, and 15 years using the Child Behavior Checklist (CBCL) [39]. In addition, children self-reported symptoms of depression and anxiety at 15 years [40]. See Table 1 for a detailed description of these measures.

Table 1 Description of measures.
Full size table

DNA methylation measures of biological aging

The pace of biological aging was measured using DunedinPACE at 9, 15 years [41]. Accelerated biological aging was measured with GrimAge [42] and PhenoAge Acceleration [43]. See Table 1 for a detailed description.

Structural racism

We conducted a comprehensive analysis to quantify racism by employing three measures: self-identified race/ethnicity, the Thiel Index as a statistical measure that captures the extent of racial segregation within neighborhoods, and, for racially marginalized youth who did not identify as solely White, we included skin tone as a proxy for colorism, Moreover, we considered family-level and neighborhood-level socioeconomic factors, as well as police interactions and parenting stress, and perinatal factors as potential mediators of associations between racialization processes and child development. See Table 1 for a detailed description of these measures. See Supplemental Table 1 for descriptive statistics of measures of interest.

Statistical analyses

Our preregistered analyses (https://osf.io/xbgzu) and results are categorized into three objectives. Supplemental Material Table 1 lists preregistered analyses and analytic deviations.

Racial and ethnic disparities in mental health from early childhood through adolescence

We examined racial and ethnic disparities in parent-reported internalizing and externalizing behaviors longitudinally across early childhood through adolescence, and in cross-sectional analyses of self-reported anxiety and depressive symptoms at age 15. Internalizing and externalizing scores were log-transformed. We applied a two-slope latent growth curve model (LGCM) depicted in Fig. 1A that estimated initial levels (intercepts) and rates of change over time (slopes) [44]. Latent factors were added for ages 7, 11, and 13 years to capture the trajectory of change in two-year intervals, accounting for missing observations [45, 46]. A two-slope LGCM is preferred over a one-slope model when the data exhibits different growth patterns over distinct periods and was found to improve model fit estimates (see Supplemental Table 3 and 4). The intercept (I) represents the estimated starting point or initial level of the outcome variable, for instance internalizing scores at the baseline age (i.e., at 3 years). Slinear represents the average linear change in the internalizing scores over childhood, i.e., as children age from 3 to 15 years. A negative mean Slinear indicates decreasing internalizing symptoms as children get older, on average. Sadolescence represents additional change in internalizing scores from 9 to 15 years. A positive mean Sadolescence indicates increasing internalizing symptoms over adolescence, on average. We regressed latent intercepts and slopes on measures of structural racism to examine how racial disparities in internalizing and externalizing trajectories evolved over time. Missing data was accounted for using full information maximum likelihood, and no participant was excluded from our analysis. We used robust maximum likelihood estimation to account for non-normality.

Fig. 1: Longitudinal structural equation models.
figure 1

A depicts a latent growth curve model of mental health and race. Squares represent observed variables. Circles represent latent factors, including missing waves, intercepts, and slopes. Triangles denote constants, i.e. mean intercept and slopes (S1 = Slinear; S2 = Sadolescence). Single headed arrows denote regressions, and double headed arrows denote covariances. B depicts a latent change score model of mental health and DNA methylation measures of biological aging across adolescence. Squares represent observed variables. Circles represent latent factors. Single headed arrows denote regressions, and double headed arrows denote correlations. MH = Mental health measures of parent-reported externalizing and internalizing behaviors. BioAge = DNA-methylation measures of biological aging. I = Intercept. LC = Longitudinal change from 9-to-15-years. The estimated means of intercepts, longitudinal change, and covariates (e.g., prenatal factors) are not illustrated here, but were included in the model.

Full size image

Racial and ethnic disparities in biological aging across adolescence

Next, we tested for racial and ethnic disparities in repeated measures of saliva DNAm quantifications of biological aging [25, 47, 48] in Latent Change Score (LCS) models depicted in Fig. 1B [49, 50]. Biological aging measures were standardized to baseline age levels at age 9. The model estimated initial levels, for instance racial disparities in mean biological age scores at 9 years, and latent changes, for instance racial disparities in mean longitudinal changes in biological aging from 9-to-15 years. We regressed initial levels and latent change on measures of structural racism to examine racial disparities in biological aging across adolescence.

Associations of mental health and biological aging

Lastly, we assessed if changes in biological aging from 9-to-15-years were correlated with changes in internalizing and externalizing behaviors from 9-to-15-years in LCS models (see Fig. 1B). We correlated initial levels in biological aging and mental health, which, for instance, may indicate that higher biological aging at age 9 years is associated with higher externalizing symptoms at age 9 years. We also correlated initial levels in biological aging and mental health with latent changes, which, for instance, may indicate that higher externalizing symptoms at age 9 years is associated with higher changes in biological aging from age 9-to-15 years. We also correlated latent changes which, for instance, may indicate that higher changes in biological aging from age 9-to-15-years is associated with higher externalizing symptoms from age 9-to-15-years.

We report nominal p-values with an alpha < 0.05 threshold and note if results remain significant after Benjamini-Hochberg False-Discovery-Rate correction (FDR, [51]. An overview of all our FDR corrections can be found in Supplemental Table 16 and 17. All data used in the present study is available to eligible researchers via FFCWS management (https://ffcws.princeton.edu/documentation). Examples of Mplus scripts for the main models can be found in the supplemental material, with pre-processing scripts in R available through our GitLab repository upon request.

Results

Racial and ethnic disparities in mental health from early childhood through adolescence

Externalizing behaviors

First, we examined associations of self-identified race/ethnicity, neighborhood segregation, and skin tone with externalizing behaviors in the full sample of N = 4898. As depicted in Fig. 2A, we found that Black and Multiracial children compared to White identifying children had higher parent-reported externalizing levels across early childhood through adolescence (Black: b = 0.09, 95%CI [0.03, 0.16], p < 0.01, Multiracial: b = 0.06, 95%CI [0.01, 0.11], p < 0.05, significant after FDR correction). Racial disparities in slopes and interactions of race with gender did not remain significant after FDR correction (Supplemental Table 3; Supplemental Material Fig. 1).

Fig. 2: Racial and ethnic disparities in externalizing and internalizing behaviors from early childhood through adolescence.
figure 2

A and B depict model-predicted trajectories in log-transformed parent-reported externalizing and internalizing behaviors, respectively, for Black, Latinx and White children. Data for the smaller subsamples of Multiracial and Other are not shown for visualization purposes but were included in the analyses. Error bars depict the sum of the standard errors of the intercept and slope estimates. See Supplemental Material Fig. 3 for similar plots of observed mean scores.

Full size image

Further, children living in more racially segregated neighborhoods, (who were more likely to identify as Black and Multiracial, Supplemental Material Figure 4), had a higher externalizing intercept (b = 0.09, 95%CI [0.04, 0.13], p < 0.05, significant after FDR correction) and a steeper decline over early childhood (b = –0.08, 95%CI [–0.15, –0.01], p < 0.05, significant after FDR correction, Supplemental Table 5). Amongst Black, Latinx and Multiracial children, darker skin tone was associated with a higher externalizing intercept (b = 0.09, 95%CI [0.01, 0.18], p < 0.05, significant after FDR correction, Supplemental Table 6).

Internalizing behaviors

Second, we examined associations of race/ethnicity, neighborhood segregation, and skin tone with internalizing behaviors. As shown in Fig. 2B we found that all groups of racially marginalized children showed higher internalizing levels across childhood compared to White children (Black b = 0.27, 95%CI [0.20, 0.34], p < 0.001; Latinx b = 0.28, 95%CI [0.21, 0.35], p < 0.001; Other b = 0.06, 95%CI [0.01, 0.12], p < 0.05; Multiracial b = 0.12, 95%CI [0.06, 0.18], p < 0.001, significant after FDR correction, Supplemental Table 4). Moreover, racially marginalized children had a steeper decrease in internalizing behavior compared to White children (Slinear in Supplemental Table 4; significant after FDR correction). The subsequent increase in internalizing behaviors from age 9-to-15 years did not differ significantly across groups (all p > 0.05).

There was some evidence that racial disparities in internalizing symptoms differed by gender (interaction race and gender Sadolescence b = –0.44, 95% CI [–0.77, –0.12], p < 0.01, significant after FDR correction). Accordingly, Black compared to White boys had a steeper increase in internalizing symptoms across adolescence. Conversely, Black and Hispanic girls displayed an attenuated increase in internalizing symptoms compared to White girls across adolescence (Supplemental Table 4; Supplemental Material Fig. 2B).

Further, children living in more racially segregated neighborhoods had a higher internalizing intercept (b = 0.07, 95%CI [0.02,0.12], p < 0.001, significant after FDR correction), and a steeper decline in internalizing behavior across childhood (Slinear b = –0.12, 95%CI [–0.20, –0.04], p < 0.01, significant after FDR correction, Supplemental Table 5). Amongst Black, Latinx and Multiracial children, darker skin tone was not significantly associated with internalizing behaviors (Supplemental Table 6).

Covariate analyses

Next, we assessed to what extent these racial/ethnic disparities in child mental health were statistically accounted for by covariate adjustment for proximal contextual factors related to structural racism (e.g., family socioeconomic status, neighborhood socioeconomic disadvantage, police interactions, parenting stress and closeness).

Racial and ethnic disparities in externalizing and internalizing behaviors were largely statistically accounted for by covariate control for family socioeconomic status and neighborhood socioeconomic disadvantage, whereas covariate control for police interactions and parenting had little effect (Supplemental Tables 3–6). Importantly, all groups of racially marginalized children were far more likely to live in socioeconomically under resourced families and neighborhoods, whereas age-15 adolescent reports of police interactions and parenting stress showed race differences for Black compared to White children only, with Black children reporting higher levels of direct police interaction and their parents reporting higher stress (Supplemental Table 2 and Supplemental Material Figure 4).

Anxiety and depressive symptoms

We did not find evidence of racial and ethnic disparities in self-reported anxiety or depressive symptoms at age 15 (Supplemental table 7). Amongst marginalized youth only, darker skin tone was associated with more anxiety symptoms (b = –0.07, 95%CI [–0.14 to –0.01], p < 0.05), but this result did not survive FDR correction (Supplemental table 7).

Racial and ethnic disparities in biological aging across adolescence

We tested for associations of race/ethnicity, neighborhood segregation, and skin tone with biological aging measured at age 9 and 15 years in N = 2039 children with DNAm (see Fig. 1B for a graphical model illustration). We found that Black and Latinx compared to White youth tended to have a higher intercept and higher longitudinal increase in the pace of aging (DunedinPACE) across adolescence (Fig. 3A & B, Supplemental Table 8, significant after FDR correction). Black compared to White children also had a more advanced biological age intercept and higher longitudinal increase in biological age (GrimAge and PhenoAge Acceleration; Supplemental Table 8, Fig. 3, significant after FDR correction). Further, children living in more racially segregated neighborhoods had a higher intercept in all measures of biological aging, and a higher longitudinal increase as indicated by DunedinPACE and PhenoAge Acceleration (Fig. 3C, D; Supplemental Table 9, significant after FDR correction).

Fig. 3: Racial and ethnic disparities in DNA-methylation measures of biological aging across adolescence.
figure 3

A depicts the difference in biological aging age-9 intercepts (DunedinPACE, GrimAge Acceleration, PhenoAge Acceleration) between Black compared to White identifying children without (i.e., Main Effect) and with covariate adjustment (covariates: postnatal factors, perinatal factors, family socioeconomic status [SES], and neighborhood disadvantage [Neighb.]). B depicts the difference in longitudinal change from age-9-to-15 in biological aging between Black compared to White children without and with covariate adjustment. C and (D) depict associations of neighborhood racial segregation with biological aging intercepts and longitudinal change, respectively. E and (F) depict associations of skin tone with biological aging intercepts and longitudinal change, respectively. Plotting the main effect and the covariates shows that the associations between racialization (self-identified race/ethnicity, racial segregation and skin tone, respectively) and biological aging were only slightly reduced (change in standardized effect sizes +/− 0.01–0.05) and remained statistically significant after accounting for perinatal (gestational age, birthweight, substance use during pregnancy) and postnatal (BMI, smoking, puberty status) covariates. Associations between racialization and biological aging were more substantially reduced after accounting for SES and neighborhood disadvantage.

Full size image

Amongst Black, Latinx and Multiracial children, darker skin tone was associated with a faster pace of aging (DunedinPACE) intercept and more advanced biological age intercept and higher longitudinal increase in biological age as indicated by PhenoAge Acceleration (Supplemental Table 10, Fig. 3E, F, significant after FDR correction). No significant associations were found for skin tone and GrimAge Acceleration (Supplemental Table 10).

Subsequently, we tested to what extent these associations were statistically accounted for by covariate adjustment for factors previously associated with DNAm measures of biological aging and/or structural racism, such as postnatal covariates (BMI, smoking, puberty status), perinatal birth factors (gestational age, birthweight, substance use during pregnancy), as well as proximal contextual factors (family socioeconomic status, neighborhood disadvantage, parental stress, parental closeness, and police interactions). The difference in biological aging between Black and White youth in biological aging largely persisted after accounting for postnatal covariates, perinatal birth factors, as well as proximal contextual factors (Fig. 3A, B, Supplemental Table 8). Associations of neighborhood racial segregation with biological aging largely persisted after accounting for postnatal, perinatal, police and parenting factors, but associations with DunedinPACE and GrimAge Acceleration were fully accounted for by socioeconomic status and neighborhood disadvantage (Fig. 3C, D, Supplemental Table 9). The association between skin tone and PhenoAge Acceleration remained significant after accounting for all covariates including socioeconomic status and neighborhood disadvantage, while its association with DunedinPACE was accounted for by postnatal, SES and neighborhood effects (Supplemental Table 10, Fig. 3E, F).

Associations of mental health and biological aging

First, we examined whether longitudinal change from age-9-to-15-years in externalizing behaviors was associated with longitudinal changes from age-9-to-15-years in biological aging (see Fig. 1B for a graphical model and Table 2 for parameter estimates). We found that higher longitudinal increases in externalizing behavior were positively correlated with in pace of aging and biological age acceleration (DunedinPACE: r = 0.06, 95%CI [0.03, 0.13], p < 0.01; GrimAge Acceleration: r = 0.06, 95%CI [0.02, 0.11], p < 0.01; PhenoAge Acceleration: r = 0.05, 95%CI [0.01, 0.10], p < 0.05; significant after FDR correction). These associations largely persisted after accounting for perinatal, postnatal covariates, SES and neighborhood disadvantage (Supplemental Table 11; see Supplemental Table 12 for longitudinal correlations from subgroup analysis of White, Black and Latinx groups).

Table 2 Parameter Estimates.
Full size table

Second, we tested whether longitudinal changes in internalizing behaviors were associated with longitudinal changes in biological aging. Higher longitudinal increases in internalizing behavior were correlated with higher longitudinal increases in pace of aging and biological age acceleration (DunedinPACE: r = 0.06, 95%CI [0.01, 0.10], p < 0.05, PhenoAge Acceleration: r = 0.06, 95%CI [0.01, 0.10], p < 0.05, significant after FDR correction; GrimAge Acceleration: r = 0.04, 95%CI [0.01, 0.08], p = 0.08). These associations largely persisted after accounting for postnatal covariates, perinatal birth factors, SES and neighborhood disadvantage (Supplemental Table 11).

Lastly, we tested if age-15 biological aging was associated with mental health from early childhood through adolescence. We found more advanced biological age at age 15 years, as indicated by GrimAge and PhenoAge Acceleration, was associated with a higher externalizing intercept (GrimAge: b = 0.12, 95%CI [0.07, 0.18], p < 0.01; PhenoAge: b = 0.07, CI [0.13, 0.12], p < 0.05), stronger decrease in childhood externalizing (GrimAge b = –0.14, 95%CI [–0.23, –0.04], p < 0.01), and a subsequently stronger increase over adolescence (GrimAge b = –0.20, 95%CI [–0.06, –0.03], p < 0.01), which remained significant after FDR correction. While the association with PhenoAge Acceleration was fully accounted for by socioeconomic variables, the association with GrimAge Acceleration largely remained significant after accounting for covariates (Supplemental Table 13). More advanced biological age, as indicated by GrimAge Acceleration, was also correlated with a higher internalizing intercept (b = 0.08, 95%CI [0.02, 0.15], p < 0.05; significant after FDR correction). This association was largely accounted for by postnatal covariates as well as family and neighborhood socioeconomic factors (Supplemental Table 14). A faster DunedinPACE-pace of aging at age-15-years was associated with higher concurrent levels of anxiety and depressive symptoms, and the association with anxiety symptoms remained significant after FDR correction as well as after covariate controls (anxiety: b = 0.07, 95%CI [0.02, 0.11], p < 0.01; depression: b = 0.05, 95%CI [0.01, 0.09], p < 0.05, Supplemental Table 15). An overview of all our FDR corrections can be found in Supplemental Tables 16 and 17.

Discussion

We leveraged a prospective birth cohort study to examine whether the emergence of racial disparities in mental health is linked to the emergence of racial disparities in DNAm measures of biological aging across childhood and adolescence. We find that children who identify as part of racially marginalized groups and those residing in racially segregated neighborhoods exhibit higher levels of both externalizing and internalizing behaviors. Longitudinal trends in internalizing behaviors differed by race and gender. Moreover, Black children, compared to their White counterparts, as well as children from more racially segregated neighborhoods and those more strongly affected by colorism, tended to have higher levels of biological aging at age 9 and exhibit greater biological age acceleration during adolescence. Notably, increases in internalizing and externalizing behaviors over time were correlated with increases in biological aging. While socioeconomic factors largely statistically accounted for racial and ethnic disparities in mental health, racial differences in biological aging often persisted after controlling for these variables. This suggests that racial differences in mental health and biological aging manifest early in life and are interconnected.

Our findings are consistent with previous studies that have examined racial disparities in internalizing and externalizing behaviors among children and adolescents, as well as psychopathology in adults [26, 29]. While earlier research of this cohort has demonstrated that family-level factors, such as family instability and stress, are associated with child developmental and health outcomes [26,27,28,29], our analysis underscores the role of macro-level factors, particularly structural racism, in perpetuating health inequities from an early age. Additionally, our findings reveal that racial trajectories in mental health vary by gender, underscoring the need for future studies to consider the combined effects of race and gender on health [52].

Furthermore, our analysis of racial neighborhood segregation indicates that children living in more segregated neighborhoods, who were more likely to identify as Black and Multiracial, exhibit higher levels of externalizing and internalizing symptoms compared to their peers in more racially integrated neighborhoods. Our examination of skin tone amongst marginalized youth reveals that children facing greater social disadvantages related to skin tone (i.e., darker skin tone) tend to display higher externalizing behaviors. Individuals with darker skin tones often encounter a heightened risk of poor health outcomes compared to those with lighter skin tones [53, 54]. These racial health disparities arise from the racialization of darker-skinned phenotypes (colorism), a racial hierarchy that leads to marginalization and discrimination [36, 37]. This highlights the significant, yet understudied, role of skin tone as a social determinant of mental health.

Moreover, our study provides the first comprehensive analysis of structural racism and DNAm measures of biological aging in children. We found that Black compared to White identifying children, children living in more racially segregated neighborhoods, and marginalized children with darker skin tones, tended to have higher age-9 levels of biological aging and, importantly, more biological age acceleration over adolescence. These findings extend previous cross-sectional studies that have examined disparities related racialization in both children and adults [48, 55,56,57]. For instance, Hicken and colleagues [58] find that Black compared to White identifying adults have higher blood-based GrimAge and PhenoAge Acceleration (GrimAge Acceleration: b = 0.42, 95%CI [0.20, 0.64], p < 0.001; PhenoAge Acceleration: b = 0.29, 95%CI [0.02, 0.57], p < 0.001). Our saliva-based biological aging findings for 9-year-old Black children compared to White children are partially of a similar magnitude to these reports in adults (GrimAge Acceleration: b = 0.13, 95%CI [0.07, 0.18], p < 0.001; PhenoAge Acceleration: b = 0.40, 95%CI [0.34, 0.45], p < 0.001).

Lastly, we found that longitudinal increases in biological aging across adolescence were correlated with increases in internalizing and externalizing behavior. This is consistent with the interpretation that poor well-being has negative physical health consequences and vice versa [59, 60]. Alternatively, other factors, such as heightened racialized daily life stress and vigilance could concurrently influence both within-person change in mental health and biological aging over adolescence [10, 13]. These findings support the notion that racial disparities in biological aging result from early-onset and accumulating racialized experiences linked to biological stressors, and potentially highlight adolescence as a sensitive developmental period for lifespan health trajectories [60,61,62,63,64]. Over time, an increased mental health burden could contribute to racial disparities in disease and mortality, alongside unequal access to healthcare and educational opportunities [58].

Associations between mental health and racialization as well as between biological aging and neighborhood racial segregation and skin tone were largely statistically accounted for by socioeconomic variables. In contrast, associations between racialization and both levels and change in biological aging were only slightly attenuated and largely remained statistically significant after accounting for perinatal and postnatal covariates (e.g. birthweight, BMI). Structural racism creates socioeconomic advantages for some racial groups and disadvantages for others [65]. Accordingly, marginalized children are more likely to live in under-resourced families and neighborhoods: Black and Latinx children were 82.8% and 43.2% more likely, respectively, to live in disadvantaged neighborhoods compared to White children. Hence, separating racialized and socioeconomic inequality in racially stratified populations is statistically challenging and perhaps theoretically futile. Progress in understanding structural racism and health will come from applying intersectional perspectives and collecting dynamic measurements on racialization, such as economic health benefits varying across racial groups and measures of experienced racial discrimination, which are currently lacking [66]. Regular exposure to discriminatory policies and actions, especially in low-income, racially segregated areas, contributes to the emergence of racial disparities in physiological and psychological burden [9, 67, 68]. Our results substantiate neighborhood racial segregation as a health-relevant measure in childhood that is correlated with race/ethnicity and socioeconomic factors.

We note three major limitations of this study. First, despite our efforts to provide comprehensive measures of structural racism, the available metrics lack the nuances needed to capture both macro-structural and individual-level processes of racialization. Our findings underscore the urgent need for differentiated and comprehensive measures of racism in child developmental cohorts. Second, the study of racial disparities in biological aging is challenged by a lack of diversity in the discovery studies used to develop biological aging algorithms [18]. These studies predominantly involve White adults from higher-income countries, potentially leading to biased results. However, previous research in adults suggests that DNAm measures of aging are broadly applicable across human populations. For example, they consistently correlate with health and mortality across various countries [69] and are not associated with the percentage of European genetic admixture in non-Hispanic Black veterans [70]. Despite these findings, it remains crucial to develop biological aging algorithms that are inclusive and representative of the aging processes in diverse populations, highlighting a significant area for improvement in this field.

By applying DNA methylation algorithms, originally developed for adult studies on multi-system health and mortality, to children, our study reveals that the connection between mental and physical health—both of which are influenced by racism—emerges within the first two decades of life. It is imperative that programs dedicated to advancing racial health equity confront the psychological and physical repercussions of structural racism on children and adolescents.