Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jun 14.
Published in final edited form as: Res Hum Dev. 2021 Jun 14;18(3):181–196. doi: 10.1080/15427609.2021.1935607

Identifying Patterns of Youth Resilience to Neighborhood Disadvantage

S Alexandra Burt 1, Kelly L Klump 1, Alexandra Y Vazquez 1, Elizabeth A Shewark 1, Luke W Hyde 2
PMCID: PMC8562464  NIHMSID: NIHMS1710268  PMID: 34737681

Abstract

The present report describes the motivation for the Michigan Twin Neurogenetic Study (MTwiNS), which seeks to illuminate underlying biological mechanisms through which familial and community factors support resilience (i.e., adaptive competence in the face of adversity) in youth exposed to neighborhood disadvantage. To accomplish these goals, we must first understand how resilience manifests in this cohort. The current study uncovers evidence of three domains of youth resilience: psychiatric health, social engagement, and scholastic success. Although all three domains were relatively stable across a one-to-two year period, variability in this stability was observed. Additionally, although resilience in one domain was quite common, resilience across all 3 domains was less common. Finally, we show substantial variability in resilience within and across families, with substantial co-twin discordances that can be leveraged in future analyses that examine promotive contexts that are environmental in origin.

Keywords: Resilience, Neighborhood disadvantage, Discordant Twin Design

The rapidly growing resilience literature has emphasized that studying “health” need not mean only studying illness, but also how individuals and systems adapt and can thrive, even in the face of adversity. Though the term resilience is used across an array of fields and with different connotations across contexts, a major uniting focus of the study of resilience is to understand processes and mechanisms that maintain function, support recovery, or enhance function in the face of challenges. Understanding these processes and mechanisms can help identify potential targets for behavioral or biomedical interventions to promote lifelong health (NIH PAR-16–326). Given the potential of resilience research, the National Institutes of Health (NIH) Basic Behavioral and Social Science Opportunity Network (OppNet), in collaboration with multiple institutes and offices, created a group of studies (via the UG3/UH3 mechanism) to elucidate processes and mechanisms of resilience across multiple different types of adversities, contexts, stressors, and populations. As part of this Special Issue describing these studies, the goal of the present report was to briefly describe the background motivating the Michigan Twin Neurogenetics Study (MTwiNS) as well as the foundational empirical findings in our data that lay the groundwork for future studies of resilience in our sample and others. In doing so, we explicitly defined resilience as adaptive competence in the face of adversity, with a particular focus on resilience in the context of neighborhood adversity. Using this definition, we sought to establish the prevalence and pattern of interrelations of various domains of resilience. We then sought to illuminate the degree of co-twin discordance for these outcomes, as a preliminary way to indicate the extent to which there are environmental influences on resilient outcomes.

MTwiNS

Neighborhood disadvantage can be deadly, increasing the incidence of cardiovascular events, stress-related cancers, and pre-term birth (Cubbin, Hadden, & Winkleby, 2000; Diez Roux & Mair, 2010; Gomez, D O’Malley, Stroup, Shema, & Satariano, 2007; O’Campo et al., 2008). However, even non-lethal experiences of disadvantage can derail positive development, at least in part, by undermining the acquisition of core self-regulatory skills such as impulse control, emotion regulation, and decision making (Leventhal & Brooks-Gunn, 2000; Sampson, Raudenbush, & Earls, 1997). One key pathway through which disadvantaged neighborhoods are hypothesized to undermine the development of self-regulation is by disrupting neural development (Hackman, Farah, & Meaney, 2010; Hyde et al., 2020; Tomlinson et al., 2020).

Critically, however, as many as 60–70% of youth reared in disadvantaged neighborhood contexts do not evidence self-regulatory deficits (Masten, 2001; Masten & Coatsworth, 1998). What explains these unexpectedly adaptive outcomes? The goal of the MTwiNS is to begin to answer this question by illuminating the neural and social processes underlying resilience to neighborhood disadvantage in youth. We specifically postulate that neighborhood disadvantage undermines youth outcomes by compromising development of the neural circuits subserving regulatory control (emotion regulation, impulse control, decision making), and that it does so by attenuating normative genetic influences on neural network development (see Figure 1). We further hypothesize that, by buffering youth from the stressful effects of disadvantage, protective family- and community-level processes facilitate behavioral and neural self-regulatory development by allowing normative genetic influences to more fully manifest. Such findings would reflect a bioecological model of genotype-environment interaction (GxE) (Bronfenbrenner & Ceci, 1994; Lewontin, 1995), in which adverse environments restrict optimum development by suppressing normative genetic influences on that development. The logic of the bioecological model was best illustrated by Lewontin (1995) through the analogy of genetically variable seeds that are planted in either a nutrient-rich or a nutrient-deprived field (Lewontin, 1995). The environmental adversity conferred by the deprived soil should eventuate in a field populated largely by short plants, regardless of their genetic predisposition for height. By contrast, because all plants received adequate nutrition in the nutrient-rich soil, the plants would be able to fully express their genetic endowment for height, making height more heritable in this environment. In short, by protecting youth from the stressors present in disadvantaged contexts, positive parenting and communities may buffer the impact of stressors in the neighborhood and facilitate normative genetic influences on youth neural and behavioral development.

Figure 1.

Figure 1.

Open in a new tab

Guiding theoretical model

To test our hypothesized model, we are re-recruiting 500 adolescent twin pairs residing in modestly-to-severely disadvantaged neighborhoods across Michigan (the only at-risk twin study in the United States) to engage in neuroimaging using MRI. This unique sample will allow us to identify the neural regulatory control architecture (RCA) that predicts resilience in the face of adversity (i.e., “neuroresilience”). We will then leverage the genetically-informed nature of our twin sample to illuminate the etiologic origins of these neural RCA, evaluating whether the protective familial- (parenting) and community-level (informal social control and social cohesion) processes known to predict behavioral resilience do so by facilitating normative genetic influences on neural RCA via biometric GxE interaction analyses. Finally, in an exploratory set of analyses, we will attempt to identify a specific mechanism through which protective processes facilitate the expression of normative genetic influences. One possible mechanism relates to the biological embedding of stress, via the methylation (i.e., silencing) of genes contributing to neural development (McGowan et al., 2009; Meaney, 2010; Roth & Sweatt, 2011). By buffering the child from stress, we hypothesize that protective factors reduce the extent of methylation, increasing expression of genes that promote development of self-regulatory neural RCA.

Necessary baseline analyses

Conducting this study of neuroresilience in a meaningful way hinges on two key aspects of our data collection. The first centers on our ability to identify multiple domains of resilience in our participants, a requirement made somewhat more challenging by the many complexities inherent to measuring resilience as an outcome. Although initially conceptualized as a trait of the child, resilience has since been reconceptualized as a multidimensional dynamic outcome (Curtis & Cicchetti, 2003; Luthar, Cicchetti, & Becker, 2000; Rutter, 2006) shaped both by characteristics of the child and by familial- and community-level functioning (Kaplan, 2013; Masten & Curtis, 2000). Specific domains of resilience include both the presence of specific developmental competencies (e.g., success in school relative to grade level, effective interpersonal functioning) and the presence of overall psychological health (i.e., the absence of a clinically-significant mental disorder despite the presence of adversity; the presence of well-being and life satisfaction). Recent theoretical work on resilience outcomes has cogently argued that these domains are separable (Miller-Graff, 2020), although empirical interrelationships within and among them have yet to be empirically clarified. The first goal of the current study is thus to clarify the prevalence of resilience across multiple domains in these data, while also illuminating the interrelations among them.

The second key aspect of our data collection relates to our focus on a genetically-informed sample. A number of protective factors have already been shown to promote adaptive self-regulatory skills in the face of adversity, including protective parenting (Bernier, Carlson, & Whipple, 2010) and the neighborhood social processes of informal social control and social cohesion (Sampson et al., 1997). However, to understand resilience, we would argue that it is not enough to simply uncover correlations between protective factors and resilient outcomes. We need to understand how protective factors promote resilience (Luthar et al., 2000; Masten, 2001), a far more complicated issue given the possibility of genetic confounds, or genotype-environment correlations (rGE). For example, the ability to successfully support one’s child is likely related to the parent’s own self-regulation and underlying neural RCA. Should these pathways be inherited from parent to child, they could also contribute to a given child’s resilience to adversity. There is thus a clear need for studies able to examine environmental influences while explicitly controlling for genetic influences. The discordant monozygotic (MZ) twin design is ideal for such work, as MZ twins are genetically identical, and yet can and do have different outcomes as a result of their unique environmental experiences. For this design to offer meaningful answers, however, a portion of MZ pairs must actually demonstrate discordance in the outcome (i.e., differ in their level of resilience). The second goal of the current study is thus to clarify the extent to which co-twins demonstrate discordance within and across each of the various domains of resilience. We know of no other published study that has done this, augmenting the importance of these early analyses of our data.

METHODS

Participants

Families in the ongoing Michigan Twin Neurogenetics Study (MTwiNS) live in south-central Michigan and are part of the large-scale Michigan State University Twin Registry (see Burt & Klump, 2019). MTwiNS participants were initially recruited into the Twin Study of Behavioral and Emotional Development in Children (TBED-C) when they were 6 – 10 years old (Wave 1). A subset were then re-recruited into MTwiNS during adolescence (Wave 2; current N=708 twins from 354 families), with follow-up occurring 1 – 2 years later (Wave 3; current N = 360 twins in 180 families). For TBED-C (Wave 1), twins were recruited into one of two cohorts. The first, a population-based cohort, was sampled from birth records to represent families with twins living within 120 miles of Michigan State University, an area that includes much of Southern Michigan with families living in urban (e.g., Detroit, Flint), suburban, and rural areas. The second, at-risk cohort, was recruited from the same area, but only included families living in U.S. Census tracts where at least 10.5% of families lived below the poverty line (i.e., the mean for the state of Michigan at the onset of recruitment) (see Burt & Klump, 2019). As the goal of MTwiNS is to examine the effects of neighborhood disadvantage on brain and behavior, MTwiNS participants (i.e., TBED-C families that also participated at Wave 2) were drawn from TBED-C families recruited from neighborhoods with above average levels of poverty. We are thus recruiting all families who participated in the “at-risk” cohort, as well as those in the population sample that would have qualified for the second sample (i.e., they lived in neighborhood with above mean levels of poverty).

This recruitment strategy resulted in a sample containing a wide range of family incomes with substantial enrichment for families exposed to poverty. The average reported household income within MTwiNS at Wave 2 was between $60,000 and $69,999, ranging from less than $4,999 to greater than $90,000. A third (33%) of MTwiNS families recruited thus far reported an annual income below the living wage for a family of 3 in Michigan (http://livingwage.mit.edu/states/26). Additionally, because of structural inequalities in our society, enrichment for neighborhood disadvantage also increased representation of those from marginalized racial and ethnic identities. Parent-reported race of the twins included at Wave 2 was as follows: 78.5% White, 12.7% Black/African American, 5.3% Biracial, 0.8% Native American, 1.1% Hispanic, 0.8% Pacific Islander, and 0.6% Asian-American. Participants at Wave 2 were primarily in adolescence though the sample ranged in age from 7 to 19 years (mean age = 14.58 years; SD 2.23; 54.7% male; only 3.4% of the sample was 10 or younger). Follow-up at Wave 3 occurred between 6 months and 4 years after Wave 2 (average Wave 3 assessment was 1.36 years after the Wave 2 assessment), when twins were between 10 and 20 years old (mean age = 15.64; SD 2.29). The study protocol was approved by the Institutional Review Board at the University of Michigan. Children provided informed assent, while parents provided informed consent for themselves and their children.

Measures

Child and Youth Resilience.

Youth reported on their own resilience outcomes via the 17-item Child and Youth Resilience Measure-Revised (CYRM-R) (Jefferies, McGarrigle, & Ungar, 2019). The CYRM was developed using mixed-methods in youth age 13–23 across 11 countries (Ungar & Liebenberg, 2011). The measure contains 2 highly correlated factors – personal (e.g., I know how to behave in different situations, I cooperate with people around me) and relational (e.g., I think my family cares about me when times are hard, I like the way my family celebrates things) – which have shown test-retest reliability over time (Liebenberg, Ungar, & Van de Vijver, 2012). In the current sample, we examined all 17 items as a single factor, which showed strong internal consistency (α = .90 at Wave 2 and .93 at Wave 3). To determine the frequency of resilience as reported using this measure, we made use of the recommended cut-point of 63 (i.e., 63+ indicated at least modest levels of resilience where 62 or less did not).

Satisfaction with Life Survey.

Youth reported on their satisfaction with life via the Satisfaction with Life Scale. This 5 item scale (with a 7-point response option) is designed to measure global life satisfaction (e.g., I am satisfied with my life; So far I have gotten the important things I want in life) as a component of subjective wellbeing, and has a one factor structure (Diener, Emmons, Larsen, & Griffin, 1985). In the current sample, the items showed high levels of internal consistency (α = .89 at Wave 2 and .88 at Wave 3). To determine the frequency of resilience as reported using this measure, we made use of the recommended cut-point of 20 (i.e., 21+ indicated at least modest levels of resilience where 20 or less did not).

Achenbach System of Empirically Based Assessment (ASEBA).

Youth and parents reported on psychopathology, engagement in activities, and social competence via the Youth Self-Report (YSR) and the Child Behavior Checklist (CBCL) (Achenbach & Rescorla, 2000). Parents and teachers reported on the child’s academic competence via the CBCL and the Teacher Report Form (TRF), respectively (Achenbach & Rescorla, 2000). These measures are one of the most commonly used set of instruments for assessing academic and social competence, as well as internalizing and externalizing problems prior to adulthood (Nakamura, Ebesutani, Bernstein, & Chorpita, 2009). Informants rated the extent to which a series of statements described each child’s behavior during the past 6 months; responses were made on a 3-point scale ranging from 0 (never) to 2 (often/mostly true).

To measure psychiatric resilience, we examined the eight psychopathology subscales from the CBCL and YSR: Anxious/Depressed (e.g. Fears certain animals, situations, or places, other than school), Withdrawn/Depressed (e.g. There is very little he/she enjoys), Somatic Complaints (e.g. Constipated, doesn’t move bowels), Social Problems (e.g. Complains of loneliness), Thought Problems (e.g. Hears sounds or voices that aren’t there), Attention Problems (e.g. Can’t concentrate, can’t pay attention for long), Rule-Breaking (e.g. Breaks rules at home, school, or elsewhere), and Aggressive Behavior (e.g. Destroys things belonging to his/her family or others). We first recoded each scale as a dichotomous variable that indicated whether the child was at or above (0) or below (1) the “borderline” cut-point for that subscale (Achenbach & Rescorla, 2001). The eight dichotomous variables were then summed to serve as our psychiatric resilience indicator, ranging from 0 to 8 (where 8 indicates a lack of psychopathology or resilience).

The Academic Performance subscale on the TRF (α = .94 at Wave 2 and .93 at Wave 3) and the School Competency subscale of the CBCL (α = .55 at Wave 2 and .60 at Wave 3) served as our measures of academic resilience. These measures assess school performance across subject domains, special education services received, repeated classes, and academic or other school related problems (e.g., Does your child receive special education or remedial services?). Note that because these scales are designed to measure minimum competencies in clinical settings, only minimal levels of competence are required (e.g., the child can be failing a class and still be considered competent overall). To determine the frequency of academic resilience, we made use of the recommended CBCL t-score cut-point of 35 (i.e., 36+ indicated at least modest levels of resilience where 35 or less did not). To determine the frequency of academic resilience, we required a TRF raw score of at least 3 (i.e., 3–5 indicated at least modest levels of resilience where less than 3 did not).

The Social Competency subscale of the CBCL (α = .64 at Wave 2 and .68 at Wave 3) and the YSR (α = .56 at Wave 2 and .57 at Wave 3) served as our measures of social resilience. This scale assesses the child’s involvement in organizations, number of friends, contact with friends, behavior with others, and behavior alone (e.g. about how many times a week does your child do things with any friends outside of regular school hours?). To determine the frequency of social resilience, we made use of the recommended t-score cut-point of 35 (i.e., 36+ indicated at least modest levels of resilience where 35 or less did not).

Analyses

All analyses focus on the Wave 2 assessment, except where explicitly noted. We first present descriptive statistics for each measure, both continuously and after dichotomization (resilient vs. non-resilient). We then evaluated correlations across the continuous measures, with a focus on associations across informants and over time. To formally evaluate whether and how the various measures clustered into separable domains of resilience, we conducted a series of exploratory factor analyses (EFA) with promax rotation of all ten measures using ML estimation in Mplus 8.4. We evaluated the fit and loadings of one-, two-, three-, and four-factor models, respectively.

For our final step, we evaluated the proportion of co-twins discordant for a given domain of resilience, separately by zygosity. Unlike traditional twin models, which focus on the origins of twin similarity, discordant twin models focus on twin differences. The logic is as follows: MZ twins result from a single human conception, and so are effectively genetically-identical. As such, should twin differences in resilience have environmental contributions, we would expect to observe at least some differences between MZ twins. Dizygotic or DZ twin discordance is also a function of twin-specific exposures, as well as the ~50% of segregating genes not shared by DZ twins. Should associations between twin differences be larger in DZ pairs than in MZ pairs, we would (also) infer genetic contributions to that association. As we do not yet have the requisite statistical power to test differences in the magnitude of MZ and DZ twin differences, we simply present them here separately by zygosity.

RESULTS

Descriptive statistics.

The proportions of youth demonstrating resilient outcomes (as defined in the Methods section) are presented in Table 1, along with means and variance for each measure when operationalized continuously. As seen there, resilience in the presence of disadvantage was quite normative. Most of the sample (68–96%) were demonstrating at least some modest level of adaptive competence in the various measures of resilience. This pattern of results held even when we restricted the sample to those experiencing more extreme levels of neighborhood disadvantage (e.g., their Area Deprivation Index (Kind & Buckingham, 2018), a composite of 17 Census measures of disadvantage, was 40+%) or familial poverty (indexed by an annual family income at or below the living wage in the state of Michigan). We then calculated the proportion of twin pairs discordant for modest levels of resilience on these measures. As shown in Table 2, roughly 10–30% of pairs were discordant for specific measures of resilience. The only exception to this pattern of results was seen for reports of academic resilience, which were rarely discordant in MZ pairs.

Table 1.

Descriptive statistics.

Overall sample (N=708) Twins residing in severely disadvantaged neighborhoods (N=366) Twins living at or below the living wage (N=204)
Measure Mean (SD) Min Max % with at least modest levels of resilience Mean (SD) % with at least modest levels of resilience Mean (SD) % with at least modest levels of resilience
Child and youth resilience, twin report 71.18 (9.76) 34 85 82.4 70.36 (10.32) 79.0 69.43 (10.32) 77.8
Satisfaction with Life Survey, twin report 26.55 (6.01) 5 35 83.9 26.37 (6.12) 82.3 25.29 (6.42) 74.0
Absence of psychopathology, twin report 7.43 (1.26) 0 8 73.8 7.37 (1.35) 72.0 7.41 (1.25) 71.1
Absence of psychopathology, mom report 7.63 (0.97) 0 8 80.3 7.57 (1.11) 79.7 7.51 (1.07) 75.9
Social competence, twin report 8.59 (2.55) 2 13.5 82.3 8.17 (2.50) 77.9 7.47 (2.54) 66.7
Social competence, mom report 8.58 (2.63) 0 14 87.0 8.26 (2.62) 83.7 7.75 (2.75) 78.9
Engagement in activities, twin report 8.61 (2.46) 0 13 68.3 8.45 (2.53) 65.1 8.08 (2.49) 55.6
Engagement in activities, mom report 8.63 (2.46) 0 13 72.4 8.49 (2.56) 70.5 8.11 (2.82) 65.1
Academic competence, mom report 5.05 (1.05) 1 6 95.8 4.94 (1.07) 96.2 4.76 (1.15) 92.7
Academic competence, teacher report 3.76 (0.99) 1 5 82.0 3.58 (1.03) 75.9 3.45 (1.02) 71.3
Open in a new tab

Note. Resilience was defined here using cut-points for modest levels of competence, as defined using manual recommendations or based on prior literature. Specifically, resilience was defined as a score of 63+ on the Child and Youth Resilience Measure, 21+ on the Satisfaction with Life Survey, an absence of clinically-significant psychopathology by either mother or twin report, a raw score of 3 or higher on teacher-report of academic competence, or a t-score of 35 or higher on mother or child-reports of social competence, mother or child-reports of engagement in activities, or mother-report of academic competence.

Table 2.

Twin pairs discordant for modest levels of resilience

Measure % of MZ pairs % of DZ pairs
Child and youth resilience, twin report 16.8 20.2
Satisfaction with Life Survey, twin report 11.9 18.4
Absence of psychopathology, twin report 22.0 34.4
Absence of psychopathology, mom report 15.4 26.7
Social competence, twin report 14.0 23.6
Social competence, mom report 12.5 17.4
Engagement in activities, twin report 23.3 32.9
Engagement in activities, mom report 20.5 25.9
Academic competence, mom report 3.4 8.6
Academic competence, teacher report 3.7 30.7
Open in a new tab

Note. Co-twins were identified as discordant if one twin met criteria for modest levels of resilience, as defined in Table 1, and their co-twin did not. MZ and DZ indicated monozygotic and dizygotic, respectively.

Correlations.

Correlations across measures are presented in Table 3. As seen there, informants demonstrated relatively high levels of agreement on the various aspects of adaptive competence, with a correlation of .67 between mothers and teachers in their reports of youth academic success, and .51 and .55 between mothers and their children in their reports of engagement in activities and social competence, respectively. Reporters differed significantly more (p<.001) in their assessment of the absence of psychopathology, with mothers and their children correlating only r =.30 on that measure.

Table 3.

Correlations among measures and over time

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Stability over time
1. Child and youth resilience measure, twin report -- .64**
2. Satisfaction with Life Survey, twin report .60** -- .59**
3. Absence of psychopathology, twin report .42** .41** -- .61**
4. Absence of psychopathology, mom report .18** .34** .30** -- .62**
5. Engagement in activities, twin report .22** .23** .07 .05 -- .50**
6. Social competence, twin report .36** .24** .16** .08 .44** -- .64**
7. Engagement in activities, mom report .19** .23** .15** .13** .51** .29** -- .38**
8. Social competence, mom report .28** .22** .13** .16** .34** .55** .44** -- .61**
9. Academic competence, mom report .17** .12** .20** .29** .20** .28** .18** .34** -- .77**
10. Academic competence, teacher report .13* .13 .15** .18** .24** .34** .23** .35** .67** -- .81**
Open in a new tab

Note. Stability over time represent within-measure correlations across Waves 2 and 3, which were conducted 1.36 years apart, on average.

** and * indicated that correlation was significantly larger than zero at p <.01 and p <.05, respectively.

Longitudinal correlations within measure across Waves 2 and 3 are also presented in Table 3. These associations indicate that most forms of adaptive competence and the absence of psychopathology are quite stable over time, with most correlations hovering around .60. There were two exceptions, however. Regardless of informant, academic competence was significantly more stable (p<.001) than were the other forms of adaptive competence. By contrast, engagement in activities was less stable over time than the other forms of adaptive competence, although these differences were only consistently significant when examining maternal report of engagement in activities.

Exploratory factor analysis.

We evaluated the fit and loadings of one-, two-, three-, and four-factor models, respectively. The RMSEAs for these models were .13, .12, .09, and .06, respectively. Although the four-factor model thus provided the best fit to the data, this model was likely over-fitted, in that two factor loadings were estimated to be greater than 1.0 and one estimated residual variance (for mother-reported Activities) was negatively-signed. When this occurs, it is recommended that investigators either select a solution with one less factor or drop items (Brown, 2015). We chose the former, given that an RMSEA of .09 is still within an acceptable range. Given this ambiguity, we examined other aspects of the model as well. The scree plot, for instance, yielded evidence of a reasonably clean break between the three- and four-factor solutions, and three eigenvalues were above 1.0 (3.301, 1.538, 1.385; the fourth was 0.884). We thus proceeded with the three-factor solution.

Factor loadings for the three-factor solution are presented in Table 4. As seen there, the three factors appeared to nicely capture all three commonly discussed domains of resilience, including psychiatric health and life satisfaction, social engagement, and scholastic success. Also as expected, these factors were correlated with one another, with cross-factor correlations ranging from .23 to .45 across all pairs of factors. In general, measures loaded quite cleanly onto their respective factor, with primary loadings of .60 through .90 and cross-loadings of .16 to .19. The primary exception to this general pattern was seen for maternal report of psychiatric resilience, which loaded only moderately on the psychiatric health and life satisfaction factor (.42) and demonstrated a somewhat larger cross-loading on the academic success factor (.25).

Table 4.

Results of the Exploratory Factor Analyses

Factor 1:
Psychiatric health		 Factor 2:
Social engagement		 Factor 3:
Scholastic success
Child and youth resilience measure, twin report .70 .19 --
Satisfaction with life survey, twin report .80 -- --
Absence of psychopathology, twin report .60 .15 --
Absence of psychopathology, mom report .42 -- .25
Social competence,
twin report		 -- .63 --
Social competence, mom report -- .61 .16
Engagement in activities,
twin report		 -- .71 --
Engagement in activities,
mom report		 -- .62 --
Academic competence, mom report -- -- .90
Academic competence, teacher report -- -- .69
Open in a new tab

Note. Only factor loadings of 0.15 or higher are reported. Those lower than .15 are indicated with a double dash. Note that, because this was an exploratory factor analysis, we were not able to adjust for nesting within families.

Resilience domains.

As our final step, we evaluated the proportion of twins who met criteria for at least modest levels of resilience across all measures within a given domain (see Table 5). In the full sample, 55.3% of twins were categorized as at least modestly resilient on all four indices of psychiatric health and 52.0% were categorized as resilient across all four measures of social engagement. In short, a notably smaller proportion of twins were considered “fully resilient” when we examined a more complete, extensive, and nuanced set of indicators of a given domain than when we examined only a single measure. This finding did not persist to academic success, however, for which 87.4% of twins continued to be identified as at least modestly resilient. This overall pattern of results persisted when we restricted the sample to those experiencing more extreme levels of neighborhood disadvantage or familial poverty. When resilience was defined multidimensionally (i.e., across all measures in all three domains), the proportion of twins who could be categorized as resilient contracted to less than 40%. In short, although most participants demonstrated modest resilience according to any one measure, far fewer (though still a substantial number) were considered resilient when resilience was defined collectively via the presence of social engagement and academic success, and the absence of psychopathology.

Table 5.

Prevalence of resilience at the broader domain level

% resilient across the domain % discordant MZ pairs % discordant DZ pairs
Full sample Most disadvantaged neighborhoods At or below living wage
Psychiatric health 55.3 51.6 47.8 42.2 54.4
Social engagement 52.0 47.5 35.8 46.6 60.3
Academic success 87.4 84.6 78.0 4.6 16.8
Cross-domain resilience 39.2 34.9 30.5 60.1 71.8
Open in a new tab

Note. Twins were identified as resilient if they met or exceeded the criteria for modest resilience across all measures in a given domain.

We also observed considerably higher levels of discordance across co-twins at the level of the broader domains, with the exception of academic success (for which twins remained quite similar to one another). Nearly half of MZ pairs (42.2% and 46.6%, respectively) and more than half of DZ pairs (54.4% and 60.3%, respectively) were discordant for resilience across the overall domains of psychiatric health and social engagement. As above, when resilience was defined multidimensionally (i.e., across all three domains), the majority of twins (60.1% MZ pairs and 71.8% of DZ pairs) were discordant for resilience.

DISCUSSION

The empirical goals of the present study were to establish the prevalence and pattern of interrelations of various forms of resilience in youth exposed to neighborhood adversity, and illuminate the degree of co-twin discordance for these outcomes. Analyses revealed high levels of informant-agreement across specific measures, particularly for adaptive competencies, as well as high levels of stability over time. We also found that, at the measure-specific level, modest levels of resilience were the most likely outcome. EFA suggested that there were three multi-informant domains of resilience in these data: psychological health, social engagement, and academic success. When examined at the level of the broader domain, resilience in the domains of psychological health and social engagement, as defined by endorsements of resilience across measures and reporters, were observed in roughly half of the sample, whereas modest levels of academic success remained very high. When examined across all three domains, however, the prevalence of resilience was less than 40%. We similarly observed that 40–60% of twin pairs were discordant for psychological health or social engagement at the broader domain level, while 60–70% were discordant across all three domains. In sum, these analyses suggest that resilience in the MTwiNS can be separated into 3 domains, that resilience is relatively common, and that co-twins do show substantial differences in resilience when measured at the broader domain level. Such findings suggest that subsequent discordant twin analyses using these data should make use of the broader domains rather than individual measures within a given domain. Moreover, the substantial variation in resilience domains within and across families suggests that there is substantial individual variation to map on to neural RCA and to be predicted by social factors.

Our findings regarding the prevalence of resilience are consistent with those from prior studies. Infurna and Luthar (2017) noted that the high prevalence of resilience typically observed is, in part, a function of the many studies of resilience focusing only on a few indicators within one domain of resilience. This is problematic for establishing the “prevalence” of resilience since successful adaption to adversity may be present in some domains (or by a single reporter) even as significant difficulties occur in others. For example, Luthar and colleagues (1993) found that ~74% of adolescents in their urban sample were categorized as resilient on specific measures of academic performance or peer ratings. However, when examining broad resilience across all domains, the prevalence of resilience decreased to 29%. Our results add to this conclusion, clearly demonstrating that the operationalization of resilience dramatically affects its prevalence and, that while resilience in any one domain (and by a single reporter) is relatively common, resilience across multiple domains is somewhat less common (though still relatively frequent).

There are a few limitations worth noting. First, though we used a strong sampling frame (identifying families via birth records), the sample is specific to the State of Michigan and therefore findings may not generalize to those in different regions or countries where the context of neighborhood poverty may differ. Second, though this is one of the only “at-risk” twin samples in the world, it is still a community sample, albeit one over-sampled for disadvantage. Thus, the majority of families are not facing extreme disadvantage, nor are many facing more extreme forms of adversity (e.g., exposure to war) that have been the focus of other studies of resilience (Miller-Graff, 2020). Third, though we aimed to identify multiple aspects of resilience, these measures are not exhaustive. For example, other adaptive outcomes could include prosocial behavior, engagement in civic activities, and volunteering. Finally, as noted in theoretical models of resilience, we defined resilience here as an outcome rather than a process. Our future work will also aim to understand the mechanisms through which these outcomes occur and can thus touch on resilience as a process.

Despite these limitations, the current study has a number of key implications, both for the likely success of the project, but also for the field as a whole. First, the current results meaningfully inform recent theory regarding the factor structure of resilience outcomes, the Multidimensional Taxonomy of Resilience (MTIR; Miller-Graff, 2020). In this taxonomy, Miller-Graff organizes resilience into two branches: manifested resilience, which captures observed resilient outcomes, and generative resilience, or the process of becoming resilient. Manifested resilience is further divided into specific domains of resilience (developmental competence, psychological health, and character) based largely on prior theory, but also on some empirical work. Our current study did not include assessments of ‘character’ per se. However, we did have several measures of psychological health, academic success, social competence, and activities. Factor analyses of these data generally supported the MTIR, albeit imperfectly. The MTIR conceptualizes psychological health via both the absence of psychopathology and the presence of well-being and life satisfaction, a conceptualization that was fully supported herein (i.e., Factor 1 in the current study included measures of the absence of psychopathology and the presence of life satisfaction). By contrast, our factor analysis suggested that social engagement and academic competence (two elements of Developmental Competence in the MTIR) do not form a single factor, but rather two separate correlated factors. Of course, the number of indicators and reporters that we had for each measure of these domains can influence the resulting factor structure. However, the relatively “clean” factor structure (with few cross-loadings) suggests that there was a relatively clear separation of these two factors. Future studies should seek to replicate this factor structure with the goal of refining the MTIR.

Second, our finding of robust twin discordance for domains of resilience is quite novel, as we know of no published study of twin discordance for resilience in the presence of a common adversity. The relative absence of twin differences for academic success point to the likelihood that genetic or shared familial effects are especially important for those outcomes, an interpretation consistent with results from prior studies (Knopik, Neiderhiser, DeFries, & Plomin, 2016). By contrast, the moderate-to-high levels of discordance observed for social engagement, psychological health, and resilience overall point to the likely presence of environmental influences on those outcomes. Such findings are important, both because they provide a window into the etiologic origins of resilience, but also because they strongly suggest that our future plans to leverage the discordant twin design to clarify the role of methylation in behavioral resilience and its neural RCA should be quite informative. That is, these data provide initial evidence that there are sufficient differences between co-twins that can be leveraged to identify environmental influences on resilience. By eliminating genetic confounds that otherwise undermine the translational pipeline from basic studies to intervention, these data from twin differences should better inform resilience-supporting interventions.

Acknowledgments:

We are grateful to the staff of the TBED-C and MTwiNS studies for their hard work. This work would not be possible without the contributions of the families who participated in TBED-C and MTwiNS. We thank them for sharing their lives with us.

Funding:

Research reported in this publication was supported by the National Institute of Mental Health of the National Institutes of Health (NIMH) and the Office of the Director National Institute of Health (OD), under Award Number UG3MH114249 and R01MH081813, as well as the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R01HD093334 (awarded to SAB and LWH), R01HD066040 (awarded to SAB), and F32HD098780 (awarded to EAS). The content is solely the responsibility of the authors and does not necessarily represent the views of the National Institutes of Health. Funding was also provided by a NARSAD young Investigator Grant from the Brain and Behavior Foundation (to LWH) and by a Graduate Research Fellowship from the National Science Foundation (to AYV).

Footnotes

Conflict of interest statement: No conflicts declared.

REFERENCES

  1. Achenbach TM, & Rescorla L (2000). ASEBA preschool forms & profiles: An integrated system of multi-informant assessment. University of Vermont, Department of Psychiatry. Burlington, VT. [Google Scholar]
  2. Bernier A, Carlson SM, & Whipple N (2010). From external regulation to self-regulation: Early parenting precursors of young children’s executive functioning. Child Development, 81(1), 326–339. [DOI] [PubMed] [Google Scholar]
  3. Bronfenbrenner U, & Ceci SJ (1994). Nature-nurture reconceptualized in developmental perspective: A bioecological model. Psychological Review, 101, 568–586. [DOI] [PubMed] [Google Scholar]
  4. Brown TA (2015). Confirmatory factor analysis for applied research: Guilford publications. [Google Scholar]
  5. Burt SA, & Klump KL (2019). The Michigan State University twin registry (MSUTR): 15 years of twin and family research. Twin Research and Human Genetics, 22(6), 741–745. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cubbin C, Hadden WC, & Winkleby MA (2000). Neighborhood context and cardiovascular disease risk factors: the contribution of material deprivation. Ethnicity & disease, 11(4), 687–700. [PubMed] [Google Scholar]
  7. Curtis W, & Cicchetti D (2003). Moving research on resilience into the 21st century: Theoretical and methodological considerations in examining the biological contributors to resilience. Development and Psychopathology, 15(03), 773–810. [DOI] [PubMed] [Google Scholar]
  8. Diener E, Emmons RA, Larsen RJ, & Griffin S (1985). The satisfaction with life scale. Journal of Personality Assessment, 49(1), 71–75. [DOI] [PubMed] [Google Scholar]
  9. Diez Roux AV, & Mair C (2010). Neighborhoods and health. Annals of the New York Academy of Sciences, 1186(1), 125–145. [DOI] [PubMed] [Google Scholar]
  10. Gomez SL, D O’Malley C, Stroup A, Shema SJ, & Satariano WA (2007). Longitudinal, population-based study of racial/ethnic differences in colorectal cancer survival: impact of neighborhood socioeconomic status, treatment and comorbidity. BMC cancer, 7(1), 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Hackman DA, Farah MJ, & Meaney MJ (2010). Socioeconomic status and the brain: mechanistic insights from human and animal research. Nature Reviews Neuroscience, 11(9), 651–659. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Hyde LW, Gard AM, Tomlinson RC, Burt SA, Mitchell C, & Monk CS (2020). An Ecological Approach to Understanding the Developing Brain: Examples Linking Poverty, Parenting, Neighborhoods, and the Brain. American Psychologist, 75(9), 1245–1259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Jefferies P, McGarrigle L, & Ungar M (2019). The CYRM-R: A Rasch-validated revision of the child and youth resilience measure. Journal of Evidence-Based Social Work, 16(1), 70–92. [DOI] [PubMed] [Google Scholar]
  14. Kind AJ, & Buckingham WR (2018). Making neighborhood-disadvantage metrics accessible—the neighborhood atlas. New England Journal of Medicine, 378(26), 2456–2458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Knopik VS, Neiderhiser JM, DeFries JC, & Plomin R (2016). Behavioral genetics: Macmillan Higher Education. [Google Scholar]
  16. Leventhal T, & Brooks-Gunn J (2000). The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126(2), 309–337. [DOI] [PubMed] [Google Scholar]
  17. Lewontin R (1995). Human Diversity. New York: Scientific American Library. [Google Scholar]
  18. Luthar SS, Cicchetti D, & Becker B (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71(3), 543–562. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Masten AS (2001). Ordinary magic. Resilience processes in development. American Psychologist, 56(3), 227–238. [DOI] [PubMed] [Google Scholar]
  20. Masten AS, & Coatsworth JD (1998). The development of competence in favorable and unfavorable environments. Lessons from research on successful children. American Psychologist, 53(2), 205–220. [DOI] [PubMed] [Google Scholar]
  21. McGowan PO, Sasaki A, D’Alessio AC, Dymov S, Labonté B, Szyf M, … Meaney MJ (2009). Epigenetic regulation of the glucocorticoid receptor in human brain associates with childhood abuse. Nature Neuroscience, 12(3), 342–348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Meaney MJ (2010). Epigenetics and the biological definition of gene× environment interactions. Child Development, 81(1), 41–79. [DOI] [PubMed] [Google Scholar]
  23. Miller-Graff LE (2020). The Multidimensional Taxonomy of Individual Resilience. Trauma, Violence, & Abuse, 1524838020967329. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. O’Campo P, Burke JG, Culhane J, Elo IT, Eyster J, Holzman C, … Laraia BA (2008). Neighborhood deprivation and preterm birth among non-Hispanic Black and White women in eight geographic areas in the United States. American journal of epidemiology, 167(2), 155–163. [DOI] [PubMed] [Google Scholar]
  25. Roth TL, & Sweatt JD (2011). Annual Research Review: Epigenetic mechanisms and environmental shaping of the brain during sensitive periods of development. Journal of Child Psychology and Psychiatry and Allied Disciplines, 52, 398–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Rutter M (2006). Implications of resilience concepts for scientific understanding. Annals of the New York Academy of Sciences, 1094, 1–12. doi:1094/1/1[pii]10.1196/annals.1376.002 [DOI] [PubMed] [Google Scholar]
  27. Sampson RJ, Raudenbush SW, & Earls FJ (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(August), 918–924. [DOI] [PubMed] [Google Scholar]
  28. Tomlinson RC, Burt SA, Waller R, Jonides J, Miller AL, Gearhardt AN, … Hyde LW (2020). Neighborhood poverty predicts altered neural and behavioral response inhibition. Neuroimage, 116536. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Ungar M, & Liebenberg L (2011). Assessing resilience across cultures using mixed methods: Construction of the child and youth resilience measure. Journal of Mixed Methods Research, 1558689811400607. [Google Scholar]

RESOURCES