Continuous Versus Grouping Approaches to Resilience: Predicting Mental Health Following Adversity in Two Longitudinal Datasets.

Continuous Versus Grouping Approaches to Resilience: Predicting Mental Health Following Adversity in Two Longitudinal Datasets - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice JAACAP Open . 2026 Jan 19;4(3):442–453. doi: 10.1016/j.jaacop.2026.01.002 Continuous Versus Grouping Approaches to Resilience: Predicting Mental Health Following Adversity in Two Longitudinal Datasets Lauren R Borchers a, #, ∗ , Chase Antonacci a, # , William E Copeland b , Michael J Meaney c, d , Ian H Gotlib a Author information Article notes Copyright and License information a Stanford University, Stanford, California b University of Vermont, Burlington, Vermont c McGill University, Montreal, Quebec, Canada d Agency for Science, Technology & Research, Singapore, Republic of Singapore ∗ Correspondence to Lauren R. Borchers, PhD, Department of Psychology, Stanford University, 450 Jane Stanford Way, Building 420, Stanford, CA 94305 lrborchers@stanford.edu # Lauren R. Borchers and Chase Antonacci shared equal contribution of this work. Accepted 2026 Jan 12; Collection date 2026 Jun. © 2026 The Authors This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13221814  PMID: 42220638 Abstract Objective Resilience in psychiatry is often defined as the absence of psychopathology following early adversity. Resilience may be better operationalized as a continuous construct reflecting the degree to which individuals are functioning better than would be predicted given their adversity exposure. We used a continuous metric of resilience to predict mental health problems in 2 richly phenotyped samples studied longitudinally, and compared findings to those of conventional grouping approaches. Method Exposure to early adversity was assessed at baseline (Study I: N = 1,265; Study II: N = 191), and mental health was assessed several times over a decade (9-21 years of age). Resilience was defined as the residual of participants’ mental health symptoms accounting for early adversity. We then used generalized additive models to examine mental health symptoms across adolescence as a function of resilience to adversity. Finally, we compared continuous approaches with grouping approaches to assess resilience. Results In both studies, greater resilience to adversity in childhood predicted fewer mental health problems in adolescence. Although both approaches to modeling resilience predicted unique variance in subsequent mental health symptoms when modeled simultaneously, the grouping approaches identified fewer individuals as resilient than did the continuous approach. Conclusion A continuous approach to operationalizing resilience significantly predicted mental health symptoms across adolescence. Furthermore, both continuous and categorical operationalizations predicted unique variance in mental health outcomes. Importantly, however, grouping approaches may overlook a significant number of individuals who are, by a continuous assessment, resilient. Determining the optimal approach for identifying resilient individuals should be a priority for mental health research. Key words: resilience, adolescence, mental health, early adversity, longitudinal study Plain language summary This study examines how children respond to adversity over time. After assessing early adversity at baseline, the authors assessed resilience in terms of mental health over a decade in two studies of youth (n=1,265 and n=191). In both studies, greater resilience was related to fewer mental health problems over time. The authors also investigated two approaches to assess resilience over time and found that a grouping approach identified fewer individuals as resilient than a continuous approach, suggesting a continuous approach may better identify resilient individuals in mental health research. Stressful experiences during childhood are common; indeed, more than half of youth report experiencing at least one form of adversity. 1 Youth exposed to early adversity are at heightened risk for developing mental health difficulties during adolescence and early adulthood 2 , 3 ; furthermore, exposure to multiple stressors in childhood has been related to even poorer outcomes. 4 It is important to recognize, however, that not all youth who experience adversity develop maladaptive outcomes. In fact, a significant proportion appear to show resilience in the face of such experiences. 5 Indeed, although prevalence estimates vary across studies and methods, developmental research often describes resilience as common (“ordinary magic” 6 ), whereas other studies have concluded that resilience is relatively rare. 4 A key challenge in adversity science is parsing heterogeneity in developmental trajectories. Understanding why some children develop psychopathology whereas others do not requires examining how best to define and operationalize the construct of resilience. Although researchers generally agree that a range of factors can influence resilience, 7 it is not clear how best to operationalize and measure this construct. 8 , 9 Some researchers have operationalized resilience through individuals’ self-reports of their ability to handle challenges, 10 whereas others have defined it as a measurable state and have examined individuals’ functioning immediately following exposure to a significant stressor. 11 In addition, some investigators have conceptualized resilience as a trait 12 and have assessed functioning decades after the stressor occurred. 13 In fact, Galatzer-Levy et al. 14 suggested that most individuals exhibit trajectories of resilience following traumatic events, although they also observed patterns of recovery, chronicity, and delayed onset of symptoms. Recent research has suggested that trajectory-based approaches are confounded by background stressors, potentially leading to inaccurate identification of factors related to resilience. To address this issue, Ahrens et al. 15 proposed a residual-based method that models stressor reactivity, accounting for individual differences in stress exposure. Finally, in their review of the literature, Kalisch et al. 16 argued that resilience is better conceptualized as a dynamic process of adaptation to stress exposure than as a trait-like characteristic, emphasizing the importance of assessing the construct within individuals’ specific contexts. Consistent with Ioannidis et al. , 17 we consider individuals to be resilient if they are functioning better than would be predicted given the severity of adversity to which they have been exposed. By this definition, even individuals with a psychiatric diagnosis can be resilient if their symptoms are less severe than would be predicted by the magnitude of their exposure. Currently, in classifying individuals as resilient, most researchers do not consider the severity of adversity experienced; indeed, in most categorical models of resilience, the presence of a psychiatric diagnosis precludes an individual from being identified as resilient despite exposure to even severe trauma. For example, using a categorical approach, an individual with 4 symptoms of depression (ie, one fewer than needed for a diagnosis of major depressive disorder) would be placed in the same “resilient” category as an individual with no symptoms, regardless of their exposure context. Conversely, an individual with 5 symptoms who just meets diagnostic criteria for depression but who had experienced severe trauma in childhood would be categorized as not resilient. Essentially, current grouping approaches to resilience classify individuals who score above a certain threshold in one group and below that threshold in another group, ignoring both their developmental context and the likelihood that people who are close to the threshold in both groups are more similar than they are different. Although this approach may overlook important variability in experiences of adversity and the presentation of symptoms, it might also be useful for capturing clinically salient thresholds or interactions, such as the presence or absence of a diagnosis. Similar to transdiagnostic models of psychopathology, 14 , 15 a continuous operationalization of resilience avoids these definitional difficulties. However, few studies have examined the viability of a residual approach to operationalizing resilience, especially in richly phenotyped, longitudinal samples. Cahill et al. 18 used a continuous operationalization of resilience to study the functioning of 14,893 adolescents. These researchers measured binary exposure to different types of adversity and summed these scores to derive a cumulative measure of adversity (see Houtepen et al. 19 ). They then computed a resilience score by regressing scores obtained with self- and parent-reports of functioning on these cumulative adversity scores, using the residuals as indicators of how well individuals were functioning above or below what would be expected given their cumulative adversity exposure (eg, “resilience”). Combining this approach with growth mixture modeling, Cahill et al. found 7 distinct trajectories of resilience, 5 of which were defined by stable high or improving resilience. Moreover, they reported that most participants would be considered resilient from the prenatal period through adolescence. 4 Furthermore, Sheerin et al. 20 measured psychological resilience as the residual of symptoms after consideration of recent stressful life events, and found that individuals who were functioning better than projected had fewer alcohol use dependence problems when assessed 5 years later. Similarly, Kalisch et al. 21 emphasized the need for longitudinal, repeated-measures designs to evaluate individual changes in resilience over time, whereas van Harmelen et al. 22 explored social support as a factor that promotes long-term resilient functioning. Finally, Cahill et al. 23 found that trajectories of resilience, assessed during childhood using a residualized approach, predicted fewer depressive symptoms in adolescence and a lower likelihood of being placed in a “not in employment, education or training” category. Building on this perspective, alongside prior studies that have assessed categorical and continuous operationalizations of resilience—either cross-sectionally in relation to psychiatric diagnoses 24 or sociodemographic factors, 25 or longitudinally during the COVID-19 pandemic 26 —our study incorporates repeated mental health assessments and measures of lifetime history of stress exposure in youth. Importantly, no previous studies have explicitly compared continuous and grouping approaches to operationalizing resilience in large, longitudinal, well-characterized samples of youth. This paper builds on findings by Cahill et al. 18 by leveraging data from 2 large, diverse, longitudinal datasets in which participants were recruited as children and were followed through early adulthood. Our main objectives were to test a continuous metric of resilience and to examine whether this measure, assessed in childhood, predicts mental health symptoms longitudinally through adolescence and into young adulthood, and to compare this metric to existing categorical models of resilience. In doing so, we also examined changes in mental health symptoms with age given that emotional disorders typically peak in adolescence, 27 and differences as a function of biological sex given the well-documented sex differences both in mental health symptoms 28 and in the relation between early adversity and risk for psychopathology. 29 Furthermore, we compared the continuous operationalization of resilience in childhood to group-based categorizations of this construct in predicting mental health symptoms in late adolescence/early adulthood. Specifically, in Study I, we compared the continuous measure of resilience to a previously published grouping of resilience 4 using this dataset as “low risk/no disorder,” “resilient,” and “childhood disorder” (ie, not resilient). In Study II, we compared the continuous measure to a dichotomous 2 by 2 (+/− early adversity by +/− psychiatric diagnosis) approach to assessing resilience. Finally, we binarized the continuous metric to examine the number of individuals who were faring better or worse than projected, and compared that to results obtained using the grouping approach. Method Participants and Study Design Study I Study I was designed to examine how a continuous measure of resilience would map onto the groupings reported previously by Copeland et al. 4 In addition, we examined how mental health symptoms in this sample changed over time and whether the continuous measure of childhood resilience predicted mental health symptoms across adolescence. Finally, we compared the original groupings of resilience to the continuous measure of this construct in predicting mental health symptoms in late adolescence/early adulthood to examine whether grouping approaches identify a different number of individuals as resilient. Data in Study I came from the Great Smoky Mountains Study (GSMS) began in 1992, which was designed to examine the development of psychiatric disorders in early life. 30 , 31 The GSMS originally sampled children from 11 counties in the southeastern United States in 1993. Sampling weights were applied to adjust for differential probability of selection and to allow results to generalize to the broader population of youth in this area. The study oversampled children at high risk for developing mental health symptoms and individuals from an American Indian background (more information regarding the sample can be found in Supplement 1a ; Figure S1 , available online). Participants were included in the study if they (and their parents) reported on their early adversity, completed measures of mental health during childhood, and reported on their mental health at least once during adolescence/adulthood (N = 1,265). The GSMS was approved by the Duke University Medical Center Institutional Review Board. Participants’ parents provided written informed consent, and, when appropriate, participants provided assent. At baseline, participants (3 cohorts recruited at ages 9, 11, and 13 years) reported the number of adverse events that they experienced and were assessed for demographic and psychosocial risk factors for psychopathology. At baseline and follow-up assessments, participants completed a structured interview assessing symptoms associated with 5 psychiatric diagnoses (conduct disorder, oppositional defiant disorder, depression, anxiety, and substance abuse/dependence; participants did not self-report ADHD symptoms, given research indicating that self-reported symptoms of ADHD are not reliable 32 ). Participants were invited to complete follow-up assessments every year until age 16 years (with the exception of the youngest cohort, which was not interviewed at age 13), and then again at ages 19 and 21 years; participants completed an average of 8 additional assessments (mean = 7.91; range = 1-12). Supplement 1b , available online, provides details about missing data. Study II Study II was designed to examine how a continuous metric of resilience maps onto the conventional 2 by 2 framework (ie, an individual did or did not experience adversity and did or did not develop a mental health diagnosis). As in Study I, we assessed both the continuous and categorical measures of resilience in relation to longitudinal mental health problems in adolescence. Data in Study II were sourced from the Stanford Early Life Stress Study (ELS), which began in 2013 and was designed to examine the effects of early life stress on psychobiological development through adolescence. 33 Participants (9-12 years of age at study entry) were a community sample of children from the San Francisco Bay Area. Additional information regarding the eligibility of the sample can be found in Supplement 2a , available online. At baseline, participants completed a comprehensive stress interview. At baseline and at follow-up assessments conducted at approximately 2-year intervals, participants completed measures of their mental health. Participants completed an average of 2 additional assessments (mean = 2.32; range = 1-3). Study II included those individuals with baseline data and at least one additional assessment (n = 191). Information about missing data is presented in Supplement 2b , available online. Measures Study I Trained interviewers administered the Child and Adolescent Psychiatric Assessment (CAPA) to participants to index both adversity and mental health. Early adversity was operationalized as the sum of binarized scores of exposure to stressors on the Life Events module (eg, threat, deprivation, loss, unpredictability) and psychosocial risk factors (eg, low socioeconomic status, mental health services for a parent). Notably, the adversity metric captures exposure without the potential confound of participants’ subjective appraisals of event severity. 34 Cumulative adversity scores were winsorized at 96% or greater and used in all analyses for Study I. More information regarding the administration of the CAPA is presented in Supplement 1c , available online. Study II Trained interviewers administered the Traumatic Events Screening Inventory for Children (TESI-C) 35 at baseline to assess early adversity. A panel of 3 or 4 coders blinded to the interviews used a modified version of the UCLA Life Stress Interview coding system 28 to calculate a severity score for each stressor endorsed. There was high agreement among coders (intracorrelation coefficient [ICC] = 0.99). 36 The TESI-C indexes 30 different stressors in addition to one open-ended question in the event that a particular stressor was not captured in the interview. Early adversity was defined as the sum of the maximum severity scores for each type of event reported (in order not to give disproportionate weight to the scores of individuals who reported many events). 37 Mental health was measured using the Youth Self-Report (YSR) 38 Total Problems scale. The YSR indexes the severity of mental health symptoms over the last 6 months and has strong psychometric properties. The Total Problems subscale of the YSR, comprising 102 items, was analyzed using raw scores to model symptom changes across age and biological sex. In addition, trained interviewers administered the Kiddie Schedule for Affective Disorders and Schizophrenia—Present and Lifetime (K-SADS-PL) 39 to examine psychiatric diagnoses. Consistent with the 2 by 2 approach to operationalizing resilience, if a participant had probable, partially remitted, or definite criteria for a diagnosis at any assessment, they received a “1,” hereafter referred to as “diagnosis”; all other participants received a “0.” We used a median split to categorize the severity of early adversity: participants below the median were coded as “1” and participants above the median were coded as “2,” hereafter referred to as “binary early adversity.” Statistical Analysis All analyses were conducted in R version 4.0.2. 40 In computing the continuous measure of resilience across both studies, we first calculated a residual score for each participant by regressing scores on a measure of early adversity on scores indexing mental health symptoms assessed at baseline (additional details in Supplement 3 , Figure S2 A and B, available online). Participants with no early adversity exposure received a score of 0, consistent with our formulation that resilience cannot be reliably measured or identified in the absence of adversity (ie, we think that is not conceptually valid to classify individuals as “resilient” if they have not been tested by adversity). Nevertheless, the pattern of findings held when we excluded participants with a resilience score of 0 (see Results). Finally, values were multiplied by −1 so that individuals with a positive residual were assigned negative values, reflecting lower resilience to adversity, and those with a negative residual were assigned positive values, reflecting greater resilience to adversity. By accounting for early adversity, we assessed the continuous nature of risk exposure in childhood. Study I In Study I, we conducted a one-way analysis of variance (ANOVA) to compare the continuous resilience measure across categorical groupings of resilience 4 : “low risk/no disorder” (1 or fewer early adversities and no mental health symptoms); “resilient” (2 or more early adversities but no symptoms); and “childhood disorder” (ie, not resilient; clinically significant symptoms prior to age 18 years regardless of adversity exposure). Next, we conducted generalized additive models (GAMs) using the mcgv package 41 to test whether age, biological sex, and the continuous measure of resilience predicted mental health symptoms over a decade (see Supplement 4 , available online, for code resources). Importantly, GAMs offer flexibility in capturing potentially complex non-linear relations between resilience and mental health symptoms; although we did not have strong a priori hypotheses about the specific shape of these associations, we used penalized smooth splines within the GAM framework to empirically estimate the functional form directly from the data without imposing potentially restrictive assumptions inherent in linear models. All continuous variables included in the model were standardized, and smoothing parameters were estimated using restricted maximum likelihood (REML) to reduce degrees of freedom and to minimize overfitting. Importantly, GAM models leverage all available data and do not drop cases listwise. Next, we compared categorical groupings to the continuous operationalization of resilience in predicting symptoms assessed in late adolescence. We first conducted an ANOVA to examine differences in mental health symptoms in late adolescence across groups. Then we used linear regression to model both the continuous and categorical metrics of adversity in relation to adolescent mental health problems and conducted formal model comparisons, assessing the combined model relative to models with each predictor entered individually. Finally, the continuous measure of resilience was binarized (ie, those individuals faring better and those faring more poorly than projected), and we conducted a χ 2 analysis to compare the overlap in identifying individuals as “resilient” across continuous and grouping approaches. Study II As in Study I, we first conducted GAMs to examine the effects of age, biological sex, and the continuous measure of resilience on mental health symptoms across adolescents. Next, we conducted a 2-way ANOVA to examine whether a 2 by 2 approach to classifying resilience in childhood predicted mental health symptoms in adolescence. Then, we used linear regression to model both the continuous metric of resilience and 2 by 2 classification simultaneously and conducted formal model comparisons. Next, we conducted a χ 2 analysis to compare the classification of individuals using the 2 different operationalizations of resilience. Finally, because residual-based indices of resilience can be strongly correlated with baseline symptoms when adversity explains a modest share of variance, as is typical in population cohorts, 17 , 42 , 43 we conducted an independent out-of-sample comparison to examine the robustness of our findings and to harmonize results across studies, allowing for more direct comparison between Study I and Study II. In both studies, we fixed the adversity–symptom association to that obtained in a large, representative cohort 44 (N = 9,498, 8-21 years of age) by pooling standardized mean differences for more than 2 traumatic events across sex and symptom domains (pooled d ≈ 1.04), converted this value to r = 0.46 using the d_to_r() function in the effectsize package, 45 and recomputed residuals in Studies I and II relative to this external reference. We then re-ran all models using these independent out-of-sample residuals. Full details of this calibration and analyses are provided in Supplement 5a-c , available online. Results Study I Participant Characteristics The sample was 11.13 (SD = 1.63) years of age at baseline, 48% female, 74% white, and endorsed mean = 2.91 (SD = 2.50, range = 0-14) adversities and/or risk factors and mean = 1.86 (SD = 2.57) symptoms on the CAPA at baseline ( Table 1 ). Adversity scores were significantly associated with baseline mental health problems ( r = 0.19, p < .001, R 2 = 3.67%) ( Table 2 ). In all, 94.5% of participants were retained; participants included here did not differ from those with missing data ( Supplement 1b , available online, provides details). Study I oversampled for American Indian youth, and we observed a main effect of race on resilience scores ( F 2,1320 = 6.54, p = .001). Tukey Honest Significance Difference (HSD) tests indicated that American Indian youth had higher resilience scores than did White youth ( p = .001) but not Black youth ( p = .066), whereas Black and White youth did not differ ( p = .896). Therefore, we applied sampling weights in computing resilience scores. Furthermore, participants from an American Indian background reported fewer mental health problems ( t = 3.87, p < .001) but comparable early adversity ( t = 1.79, p = .074) compared to participants from a non–American Indian background. Table 1. Sample Characteristics Sample Characteristics Study I: GSMS Study II: Stanford ELS N = 1,265 N = 191 Biological sex 45% Female 58% Female Mean (SD) or % Baseline age, y 11.13 (1.63) 11.36 (1.05) Race Asian — 11% Biracial — 21% Black/African American 6% 7% Hispanic/Latino — 9% Native American 25% — White 69% 45% Other — 6% No. of follow-up assessments 7.91 (2.49) 2.05 (1.07) Early adversity 2.91 (2.50) 6.73 (5.28) Baseline mental health 1.86 (2.59) 39.76 (23.79) Outcome mental health 2.02 (3.06) 44.56 (27.57) Open in a new tab Note: The 2 studies measured early adversity and mental health differently, so scores should not be compared between samples. Early adversity scores computed by summing the count of endorsed stressors and psychosocial risk factors on the Child and Adolescent Psychiatric Assessment (CAPA) interview in Study 1, and summing the maximum panel-rated event severity scores across endorsed events in the Traumatic Events Screening Inventory for Children (TESI-C) in Study II (see Methods). Mental health symptoms assessed using the CAPA in Study I and the Youth Self-Report (YSR) Total Problems scale in Study II (see Methods). Outcome mental health represents symptoms at the fourth timepoint in Study II (mean age = 17.5 years) and from the participants’ assessments in Study I when they were closest in age to 17.5 years. GSMS = Great Smoky Mountains Study; Stanford ELS = Stanford Early Life Stress Study. Table 2. Summary of Results Study I: GSMS Study II: Stanford ELS Association between adversity scores and baseline mental health r = 0.19∗∗∗ r = 0.25∗∗ Difference in continuous resilience score across categorical groupings Group effect: F = 13.02∗∗∗ — Childhood disorder: mean = −0.27 — Resilient: mean = 0.59 — Low risk/no disorder: mean = 0.58 — GAM: Predicting mental health across adolescence from continuous resilience metric R 2 = 44.3% R 2 = 70.3% Continuous resilience: B = −0.11∗∗∗ Resilience: B = −0.35∗∗∗ Age: B = −0.18∗∗∗ Age: B = 0.20∗∗∗ Sex: B = −0.07 (NS) Sex: B = 0.22 (NS) ANOVA: Predicting mental health in late adolescence from categorical resilience metric Group effect: F = 21.24∗∗∗ Main effect of adversity: F = 4.42∗ Low risk/no disorder: mean = 0.98 Main effect of diagnosis: F = 15.81∗∗∗ Resilient: mean = 1.34 Interaction: F = 0.83 (NS) Childhood disorder: mean = 2.41 — Linear regression: Combined model with continuous and categorical resilience metrics Continuous: B = −0.14∗∗∗ Continuous: B = −8.54∗∗∗ Categorical (ref. group: childhood disorder) Categorical Low risk/no disorder: B = −1.30∗∗∗ Adversity: B = 3.80 (NS) Resilient: B = −0.94∗ Lifetime diagnosis: B = 9.03 (NS) — Interaction: B = 4.93 (NS) Model comparison: Adding continuous or categorical resilience metrics to improve fit Add continuous: F = 15.44∗∗∗ Add continuous: F = 14.52∗∗∗ Add categorical: F = 17.39∗∗∗ Adding categorical: F = 5.47∗∗ McNemar χ 2 : Difference in number of participants classified as resilient Continuous > Categorical: χ 2 = 641.37∗∗∗ Continuous > Categorical: χ 2 = 31.24∗∗∗ Open in a new tab Note: Note that the 2 studies measured early adversity and mental health differently, so results should not be compared directly between samples. ANOVA = analysis of variance: GAM = generalized additive model; GSMS = Great Smoky Mountains Study; NS = not significant; ref. = reference; Stanford ELS = Stanford Early Life Stress Study. ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001. Validation of the Continuous Measure of Resilience to Adversity We examined the relations between the continuous measure of resilience and the 3 groups defined by Copeland et al. 4 An ANOVA indicated that the 3 groups differed significantly on the continuous measure of resilience ( F 2,1180 = 13.02, p < .001). Post hoc Tukey HSD tests revealed that participants in the childhood disorder group (mean = −0.27, SD = 2.80) had significantly lower scores on the continuous measure of resilience than did participants in the resilient group (mean = 0.59, SD = 1.56; p = .034) and the low risk/no disorder group (mean = 0.58, SD = 1.36; p < .001); participants in the low risk/no disorder group did not differ significantly from those in the resilient group ( p = 1.000). Resilience to Adversity and Mental Health Study I included 6,057 observations for 1,265 unique participants. GAMs analyses showed that, on average, mental health worsened with age (B = 0.18, p < .001) ( Figure 1 A), but there was no main effect of biological sex (B = −0.07, p = .063). Greater baseline resilience, measured continuously, predicted fewer mental health symptoms across adolescence (B = −0.18, p < .001). The model had an adjusted R 2 value of 44.3%, indicating a good fit ( Figure 2 A). The continuous measure of resilience also predicted adolescent mental health functioning after excluding participants with a resilience score of 0 (B resilience to adversity = −0.17, p < .001; adjusted R 2 = 43.9%). The pattern of findings were unchanged when using out-of-sample residuals based on a large, representative cohort 44 : mental health symptoms significantly increased with age (B = 0.18, p < .001), and the continuous measure of resilience predicted fewer symptoms across adolescence (B = −0.13, p < .001). Full results for this out-of-sample analysis are provided in Supplement 5a , available online. Figure 1. Open in a new tab Mental Health Symptoms Increase Across Adolescence Note: Trajectories of mental health symptoms across adolescence. Study I measured mental health symptoms using the Child and Adolescent Psychiatric Assessment (CAPA) interview. Study II measured mental health symptoms using the Youth Self-Report (YSR) (see Methods). Some cases were dropped for ease of visualization. Figure 2. Open in a new tab Resilience to Adversity Predicts Mental Health Across Adolescence in Study I Note: Results from generalized additive models (GAMs). Positive values on the y-axis indicate more severe mental health problems; positive values on the x-axis indicate greater resilience to adversity. Mental health symptoms and resilience to adversity values are z scored. Results suggest overall reduction in mental health symptoms with increasing resilience to adversity. The slight upturn in the fit line at the far right of the distribution likely reflects inherent variability at the extremes of the data distribution that is driven by a small number of individuals. Some cases were dropped for ease of visualization. Next, we compared the grouping approach of Copeland et al. 4 to the continuous measure of resilience in predicting mental health symptoms in late adolescence. For the categorical approach, a one-way ANOVA yielded a main effect of group ( F 2,1115 = 21.24, p < .001) in predicting mental health symptoms. The Tukey HSD test revealed that participants in the resilient group (mean = 1.34, SD = 2.27) had fewer symptoms in late adolescence than participants in the childhood disorders group (mean = 2.41, SD = 3.37; p = .031). The low risk/no disorder group (mean = 0.98, SD = 1.72) had fewer symptoms than participants in the childhood disorders group ( p < .001). The resilient and the low risk/no disorder groups did not differ significantly from each other ( p = .713). Next, we used linear regression to examine both the continuous and categorical resilience variables simultaneously. The contribution of group (low risk/no disorder B = −1.30, p < .001, resilient B = −0.94, p = .025; reference group = childhood disorder) and the continuous resilience score ( B = −0.14, p < .001) indicated that both operationalizations of resilience contributed unique variance to the prediction of mental health symptoms in late adolescence. Next, we conducted formal model comparisons examining whether the addition of the continuous or categorical resilience metrics significantly improved model fit over either predictor on its own. We found that adding the continuous measure significantly improved model fit ( F 1,1114 = 15.44, p < .001) relative to only including the categorical variable, and likewise, adding the categorical variable to the model including only the continuous metric improved the model fit ( F 2,1114 = 17.39, p < .001). To assess multicollinearity between the continuous and categorical metrics, we computed variance inflation factors, which indicated acceptable collinearity (ie, <5 46 ; VIF continuous resilience = 1.01; VIF categorical resilience = 1.01). Finally, we conducted a McNemar χ 2 analysis examining the proportion of participants considered resilient across the 2 approaches. For this analysis, the continuous resilience score was binarized to classify individuals as faring better or faring more poorly than projected based on the line of best fit. Fewer participants were considered resilient in the grouping framework than in the continuous approach (χ 2 [1] = 641.37, p < .001). Specifically, 61% of participants who were classified as having a “childhood disorder” (ie, not resilient) were functioning better than would have been projected given the magnitude of their exposure to early adversity ( Figure 3 A). Figure 3. Open in a new tab Comparison of Continuous and Categorical Operationalizations of Resilience Note: Comparison of continuous, residual-based resilience metric to categorical groupings in Study I and Study II. Categorical groupings are derived from Copeland et al. 4 in Study I and a standard 2 by 2 (adversity by lifetime diagnosis) approach in Study II. Blue datapoints indicate lower resilience; red datapoints indicate greater resilience. Participants whose datapoints are in blue-shaded areas would not be considered resilient using a categorical approach but are considered to be resilient using the continuous approach. Some cases were dropped for ease of visualization. Study II Participant Characteristics The participants were 11.36 (SD = 1.05) years of age at baseline, 58% female, 45% white, amd had a mean adversity score of 6.73 (SD = 5.28, range = 0-31.5) and a mean YSR Total Problems raw score of 39.76 (SD = 23.79) at baseline ( Table 1 ). Adversity scores were significantly associated with baseline mental health problems ( r = 0.25, p = .004, R 2 = 6.35%) ( Table 2 ). We retained 84.9% of participants; those individuals included here did not differ from those with missing data ( Supplement 2b , available online). Resilience to Adversity and Mental Health Similar to findings from Study I, on average, participants reported greater mental health symptoms with increasing age (B = 0.20, p < .001) ( Figure 1 B); furthermore, there was not a main effect of biological sex (B = 0.22, p = .065). The GAM approach showed that greater resilience predicted lower mental health problems over time (B = −0.35, p < .001). The GAM had an adjusted R 2 value of 70.3%, indicating a good fit to the data ( Figure 2 B). The continuous measure of resilience also predicted mental health across adolescence after excluding 0s (B resilience to adversity = −0.34, p < .001; adjusted R 2 = 69.9%). The pattern of findings were unchanged when using out-of-sample residuals based on a large, representative cohort 44 : mental health symptoms significantly increased with age (B = 0.20, p < .001), and the continuous measure of resilience predicted fewer symptoms across adolescence (B = −0.32, p < .001). Full results for this out-of-sample analysis are provided in Supplement 5b , available online. A 2-way (adversity by lifetime diagnosis) ANOVA predicting mental health symptoms at the final assessment when participants were mean = 17.5 years of age yielded main effects of early adversity ( F 1,117 = 4.42, p = .038) and diagnosis ( F 1,117 = 15.81, p < .001); the interaction was not significant ( F 1,117 = 0.83, p = .365). Participants with more severe early adversity (mean = 50.64, SD = 25.35) and participants with a psychiatric diagnosis (mean = 54.31, SD = 28.17) had higher symptoms in adolescence than did participants with less severe early adversity (mean = 40.69, SD = 28.84) and no diagnosis (mean = 34.30, SD = 23.07), respectively. In a combined model including both the continuous and categorical metrics as predictors, the continuous resilience score significantly predicted symptoms in late adolescence (B = −8.54, p < .001), but the main effects and interaction of adversity and lifetime diagnosis (ie, the 2 by 2 categorical features) did not (interaction B = 4.93, p = .593; adversity B = 3.80, p = .579; lifetime diagnosis B = 9.03, p = .513). Model comparisons examining whether the continuous measure of resilience improved the prediction of mental health symptoms at the final assessment beyond that obtained using a 2 by 2 approach indicated that adding the continuous measure significantly improved model fit ( F 1,116 = 14.52, p < .001). However, adding the 2 by 2 categorical predictor to the continuous metric of childhood resilience also improved the model fit ( F 1,116 = 5.47, p = .001), indicating that both assessments of resilience yield unique variance in mental health symptoms in adolescence. At the same time, this model indicated evidence of multicollinearity (continuous resilience VIF = 1.04; adversity binary VIF = 2.33; lifetime diagnosis = 9.54; adversity binary∗lifetime diagnosis interaction = 11.94) and should be interpreted with caution. Finally, we conducted a McNemar χ 2 analysis to compare the 2 by 2 approach to the continuous operationalization of resilience. Fewer individuals were considered resilient in the 2 by 2 approach than in the binarized continuous approach (χ 2 [1] = 31.24, p < .001). Specifically, among participants with a psychiatric diagnosis (n = 87), 46% were identified as resilient in the continuous framework, none of whom would be considered resilient based on the categorical framework ( Figure 3 B). Discussion To understand why some youth develop psychopathology following exposure to adversity whereas others do not, it is critical that we gain a more comprehensive understanding of resilience, which we posit involves moving beyond traditional categorical models. Here, we operationalized resilience as a construct that takes into account the continuous nature of risk exposure and developmental outcomes. Analyzing data from 2 distinct longitudinal studies in which mental health was assessed repeatedly over the course of a decade, we found that symptoms increased from childhood to early adulthood; however, in participants who were characterized by greater resilience to adversity in childhood, mental health symptoms decreased over time across adolescence. Thus, greater resilience to adversity was associated with a reduction in the age-related increase in psychopathology. In this context, in both studies we found that over 40% of individuals who were classified as resilient based on the continuous approach were not categorized as resilient using a grouping approach; this finding is consistent the assertion by Cahill et al. 18 that resilience is a common response to early adversity. Study I was designed to test the residualized operationalization of resilience in childhood and to compare this measure to a grouping approach used in a previous study 4 with this sample. We found that mental health symptoms increased with age, that girls had more severe mental health symptoms than boys, and that the continuous measure of resilience predicted mental health symptoms across adolescence. Furthermore, the previously defined groups differed on the continuous measure of resilience to adversity. Importantly, these 2 approaches yielded unique information relevant for the prediction of mental health problems; specifically, both the categorical and continuous measures of resilience improved the prediction of symptoms in late adolescence relative to either metric on its own. Finally, a significant proportion of individuals who had been characterized as having a “childhood disorder” (ie, not resilient) in the categorical framework were identified as resilient using the continuous framework. Thus, a continuous operationalization of resilience to adversity not only appears to be more inclusive than are traditional categorizations of resilience, but also raises the possibility that individuals who were not previously identified as resilient may, in fact, exhibit meaningful adaptation when they are assessed relative to others and within the context of their specific experiences. Although these individuals have elevated levels of symptoms, it is possible that they nevertheless exhibit resilient functioning that is not captured by grouping methods that have been used in other studies. Study II was designed to extend findings from Study I in an independent dataset and to compare the conventional 2 by 2 approach to operationalizing resilience with the continuous approach. Here, we replicated the findings from Study I that mental health symptoms increased from childhood to early adulthood, and that a continuous approach to modeling resilience predicted mental health symptoms longitudinally. Furthermore, model comparisons revealed that including the continuous measure of resilience improved the fit of the 2 by 2 model in predicting mental health symptoms in adolescence. Importantly, therefore, the continuous approach to modeling resilience captures variability that may be overlooked in categorical 2 by 2 designs, providing a more nuanced and inclusive characterization that identifies more individuals as resilient. Across the 2 datasets, we found that although both continuous and grouping operationalizations of resilience predict mental health outcomes, grouping approaches appear to overlook a significant proportion of individuals who are characterized as resilient when they are assessed using the continuous approach. Because grouping approaches do not consider important contextual information and also typically include a reference group of healthy control participants with limited or no exposure to early adversity, these approaches implicitly assume that, by definition, resilient individuals should have the same low levels of symptoms following adversity as their non-exposed counterparts We do not think that this is the optimal operationalization of resilience, as it overlooks important heterogeneity and requires individuals to exhibit absolute low levels of symptoms following even severe adversity exposure in order to be considered resilient. We believe that this approach is shortsighted and is too strict a criterion for defining resilience, given the high prevalence of early adversity in the population 1 along with relatively high rates of psychopathology in adolescence. 27 At the very least, incorporating both grouping and continuous metrics of resilience is important in assessing risk, given our findings that each contributed unique variance in the prediction of future psychopathology. Importantly, although the continuous approach offers a fine-grained, contextualized measure of resilient functioning by capturing deviations from expected outcomes, categorical models may still capture clinically salient thresholds or interactions (eg, presence vs absence of diagnosis) that continuous methods may smooth over. Given that each approach explains unique variance and improves the prediction of longitudinal mental health problems, the 2 approaches may index overlapping but distinct facets of adaptation, with the continuous model reflecting subtle variability in functioning and the categorical model anchoring key diagnostic or exposure-defined risk profiles. Therefore, future research should examine their combined predictive utility, which may reflect the fact that resilience manifests in both gradations of functioning and clinically defined categorical outcomes. We should note 4 limitations of this study. First, the residual score was highly correlated with baseline mental health symptoms. In conceptualizing resilience continuously, we found that many individuals functioned better than expected given their adversity; indeed, 61% and 42% of individuals in Study I and Study II, respectively, who would be considered “not resilient” in other frameworks, were in ours. We think that our approach adds value in understanding degrees of resilience rather than absolutes or thresholds, and we believe that these findings advance our understanding of how resilience can be measured in a way that is both comprehensive and clinically meaningful. Second, we should note that the absence of mental health symptoms does not necessarily indicate a healthy or flourishing life, nor is resilience simply the absence of vulnerability to psychopathology. Thus, it will be important in future research to examine both how resilience to adversity predicts functioning in other domains, such as cognition and physical health, and why some individuals might be more resilient to adversity than are others. As a related point, because we used the same measures of mental health to construct the residuals and to measure outcomes, our ability to test whether the residual-based metric improves prediction over the component measures used to derive it was limited. Nevertheless, it is important to emphasize that the residual-based approach provides valuable insights beyond prediction, illustrating resilience as a dynamic process that captures deviations from expected functional outcomes and enriching our conceptualization of this construct. Third, we did not examine the developmental timing of adversity exposure or the interval between exposure and baseline assessment. The age at which stressors occur may influence resilient functioning by allowing more time for recovery or, conversely, by exerting stronger effects when exposures occur during particularly sensitive developmental periods. Future studies that document explicitly the ages at which stressors occur will be important for clarifying how developmental timing might shape resilient outcomes. Finally, although our findings were consistent across the 2 studies, they used different operationalizations of adversity (ie, event counts in Study 1 and summed panel-rated severity scores in Study 2). Although the consistency of results when using out-of-sample residuals strengthens confidence in the robustness of our approach, future studies should explore systematically how different approaches to measuring adversity affect the derived metrics of resilience. Despite these limitations, we obtained convergent findings across 2 independent datasets, each with repeated assessments of mental health symptoms over a decade, that highlight the robustness and predictive utility of the residualized approach to measuring resilience over and above conventional categorical approaches. Our modeling approach to studying resilience allowed us to view resilient functioning continuously and dynamically. In this context, we think that it is important that researchers consider the magnitude of adversity exposure, given that how we operationalize resilience shapes how we identify and promote factors involved in optimal functioning. Therefore, we urge investigators to adopt more nuanced, developmentally contextualized approaches that consider aspects of early adversity when they study resilience to mental health symptoms. CRediT authorship contribution statement Lauren R. Borchers: Writing – original draft, Visualization, Methodology, Formal analysis, Conceptualization. Chase Antonacci: Writing – review & editing, Writing – original draft. William E. Copeland: Writing – review & editing, Supervision, Methodology, Funding acquisition, Data curation, Conceptualization. Michael J. Meaney: Writing – review & editing, Supervision, Conceptualization. Ian H. Gotlib: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization. Acknowledgments The authors thank the children and families who participated in the study. Footnotes This article was reviewed under and accepted by Joel Stoddard, MD, MAS. This research was funded by the National Institute of Mental Health (grants R37MH101495 to IHG; R01MH117559 to WEC; and T32MH020016 to CA). Selected findings from the larger studies from which these data were drawn have been published. Data Sharing: Data collected for the study will not be made available to others. Disclosure: Lauren R. Borchers, Chase Antonacci, William E. Copeland, Michael J. Meaney, and Ian H. Gotlib have reported no biomedical financial interests or potential conflicts of interest. Supplemental Material Supplemental Material mmc1.docx (1.4MB, docx) References 1. McLaughlin K.A., Greif Green J., Gruber M.J., Sampson N.A., Zaslavsky A.M., Kessler R.C. Childhood adversities and first onset of psychiatric disorders in a national sample of US adolescents. Arch Gen Psychiatry. 2012;69(11):1151–1160. doi: 10.1001/archgenpsychiatry.2011.2277. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Healy C., Eaton A., Cotter I., Carter E., Dhondt N., Cannon M. Mediators of the longitudinal relationship between childhood adversity and late adolescent psychopathology. Psychol Med. 2021;52(15):1–9. doi: 10.1017/S0033291721000477. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. LeMoult J., Humphreys K.L., Tracy A., Hoffmeister J.A., Ip E., Gotlib I.H. Meta-analysis: exposure to early life stress and risk for depression in childhood and adolescence. J Am Acad Child Adolesc Psychiatry. 2020;59(7):842–855. doi: 10.1016/j.jaac.2019.10.011. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Copeland W.E., Halvorson-Phelan J., McGinnis E., Shanahan L. Adult mental health, substance use disorders, and functional outcomes of children resilient to early adversity. Am J Psychiatry. 2023;180(12):906–913. doi: 10.1176/appi.ajp.20230038. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Luthar S.S., Cicchetti D., Becker B. Research on resilience: response to commentaries. Child Dev. 2000;71(3):573–575. doi: 10.1111/1467-8624.00168. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Masten A.S. Ordinary magic: resilience processes in development. Am Psychol. 2001;56(3):227–238. doi: 10.1037/0003-066X.56.3.227. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Masten A.S., Lucke C.M., Nelson K.M., Stallworthy I.C. Resilience in development and psychopathology: multisystem perspectives. Annu Rev Clin Psychol. 2021;17:521–549. doi: 10.1146/annurev-clinpsy-081219-120307. [ DOI ] [ PubMed ] [ Google Scholar ] 8. Walker B.H. Resilience: what it is and is not. Ecol Soc. 2020;25(2):11. doi: 10.5751/ES-11647-250211. [ DOI ] [ Google Scholar ] 9. Kuranova A., Booij, Sanne H., et al. Reflections on psychological resilience: a comparison of three conceptually different operationalizations in predicting mental health. Eur J Psychotraumatol. 2021;12(1) doi: 10.1080/20008198.2021.1956802. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Connor K.M., Davidson J.R.T. Development of a new resilience scale: the Connor-Davidson Resilience Scale (CD-RISC) Depress Anxiety. 2003;18(2):76–82. doi: 10.1002/da.10113. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Child L.E., Medvedev O.N. Investigating state and trait aspects of resilience using generalizability theory. Curr Psychol. 2024;43(10):9469–9479. doi: 10.1007/s12144-023-05072-4. [ DOI ] [ Google Scholar ] 12. Hu T., Zhang D., Wang J. A meta-analysis of the trait resilience and mental health. Personal Individ Differ. 2015;76:18–27. doi: 10.1016/j.paid.2014.11.039. [ DOI ] [ Google Scholar ] 13. Pechtel P., Belleau E.L., Kaiser R.H., et al. Stress and reward: a multimodal assessment of childhood sexual abuse. Neurobiol Stress. 2022;21 doi: 10.1016/j.ynstr.2022.100498. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Galatzer-Levy I.R., Huang S.H., Bonanno G.A. Trajectories of resilience and dysfunction following potential trauma: a review and statistical evaluation. Clin Psychol Rev. 2018;63:41–55. doi: 10.1016/j.cpr.2018.05.008. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Ahrens K.F., Schenk C., Kollmann B., et al. Resilience to major life events: advancing trajectory modeling and resilience factor identification by controlling for background stressor exposure. Am Psychol. 2024;79(8):1076–1091. doi: 10.1037/amp0001315. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Kalisch R., Baker D.G., Basten U., et al. The resilience framework as a strategy to combat stress-related disorders. Nat Hum Behav. 2017;1(11):784–790. doi: 10.1038/s41562-017-0200-8. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Ioannidis K., Askelund A.D., Kievit R.A., van Harmelen A.L. The complex neurobiology of resilient functioning after childhood maltreatment. BMC Med. 2020;18(1):32. doi: 10.1186/s12916-020-1490-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Cahill S., Hager R., Shryane N. Patterns of resilient functioning in early life: identifying distinct groups and associated factors. Dev Psychopathol. 2024;36(4):1789–1809. doi: 10.1017/S0954579423001165. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Houtepen L.C., Heron J., Suderman M.J., Tilling K., Howe L.D. Adverse childhood experiences in the children of the Avon Longitudinal Study of Parents and Children (ALSPAC) Wellcome Open Res. 2018;3:106. doi: 10.12688/wellcomeopenres.14716.1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Sheerin C.M., Bountress K.E., Hicks T.A., et al. Longitudinal examination of the impact of resilience and stressful life events on alcohol use disorder outcomes. Subst Use Misuse. 2021;56(9):1346–1351. doi: 10.1080/10826084.2021.1922454. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Kalisch R., Köber G., Binder H., et al. The Frequent Stressor and Mental Health Monitoring-Paradigm: a proposal for the operationalization and measurement of resilience and the identification of resilience processes in longitudinal observational studies. Front Psychol. 2021;12 doi: 10.3389/fpsyg.2021.710493. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. van Harmelen A.L., Kievit R.A., Ioannidis K., et al. Adolescent friendships predict later resilient functioning across psychosocial domains in a healthy community cohort. Psychol Med. 2017;47(13):2312–2322. doi: 10.1017/S0033291717000836. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Cahill S., Hager R., Chandola T. The validity of the residuals approach to measuring resilience to adverse childhood experiences. Child Adolesc Psychiatry Ment Health. 2022;16(1):18. doi: 10.1186/s13034-022-00449-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Overstreet C., Stratton K.J., Berenz E., et al. Resilience to interpersonal trauma and decreased risk for psychopathology in an epidemiologic sample. J Psychopathol Behav Assess. 2017;39(3):506–513. doi: 10.1007/s10862-017-9601-2. [ DOI ] [ Google Scholar ] 25. Nishimi K., Bürgin D., O’Donovan A. Psychological resilience to lifetime trauma and risk for cardiometabolic disease and mortality in older adults: a longitudinal cohort study. J Psychosom Res. 2023;175 doi: 10.1016/j.jpsychores.2023.111539. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Choi K.W., Nishimi K., Jha S.C., et al. Pre-pandemic resilience to trauma and mental health outcomes during COVID-19. Soc Psychiatry Psychiatr Epidemiol. 2023;58(3):453–465. doi: 10.1007/s00127-022-02367-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Solmi M., Radua J., Olivola M., et al. Age at onset of mental disorders worldwide: large-scale meta-analysis of 192 epidemiological studies. Mol Psychiatry. 2022;27(1):281–295. doi: 10.1038/s41380-021-01161-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Riecher-Rössler A. Sex and gender differences in mental disorders. Lancet Psychiatry. 2017;4(1):8–9. doi: 10.1016/S2215-0366(16)30348-0. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Borchers L.R., Yuan J.P., Leong J.K., et al. Sex-specific vulnerability to externalizing problems: sensitivity to early stress and nucleus accumbens activation over adolescence. Biol Psychiatry. 2025;97(1):73–80. doi: 10.1016/j.biopsych.2024.01.011. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Copeland W.E., Angold A., Shanahan L., Costello E.J. Longitudinal patterns of anxiety from childhood to adulthood: the Great Smoky Mountains Study. J Am Acad Child Adolesc Psychiatry. 2014;53(1):21–33. doi: 10.1016/j.jaac.2013.09.017. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Costello E.J., Angold A., Burns B.J., et al. The Great Smoky Mountains Study of Youth. Goals, design, methods, and the prevalence of DSM-III-R disorders. Arch Gen Psychiatry. 1996;53(12):1129–1136. doi: 10.1001/archpsyc.1996.01830120067012. [ DOI ] [ PubMed ] [ Google Scholar ] 32. Smith B.H., Pelham W.E., Gnagy E., Molina B., Evans S. The reliability, validity, and unique contributions of self-report by adolescents receiving treatment for attention-deficit/hyperactivity disorder. J Consult Clin Psychol. 2000;68(3):489–499. doi: 10.1037/0022-006X.68.3.489. [ DOI ] [ PubMed ] [ Google Scholar ] 33. Humphreys K.L., Kircanski K., Colich N.L., Gotlib I.H. Attentional avoidance of fearful facial expressions following early life stress is associated with impaired social functioning. J Child Psychol Psychiatry. 2016;57(10):1174–1182. doi: 10.1111/jcpp.12607. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Christensen D.S., Dich N., Flensborg-Madsen T., Garde E., Hansen Å.M., Mortensen E.L. Objective and subjective stress, personality, and allostatic load. Brain Behav. 2019;9(9) doi: 10.1002/brb3.1386. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Ford J.D., Racusin R., Rogers K., et al. An interview for children: Traumatic Events Screening Inventory (TESI-C) Version 8.4. Natl Cent PTSD Dartm Child Psychiatry Res Group. Published online 2002. 2002 https://www.ptsd.va.gov/professional/assessment/documents/TESI-C.pdf Accessed February 4, 2026. [ Google Scholar ] 36. King L.S., Graber M.G., Colich N.L., Gotlib I.H. Associations of waking cortisol with DHEA and testosterone across the pubertal transition: effects of threat-related early life stress. Psychoneuroendocrinology. 2020;115 doi: 10.1016/j.psyneuen.2020.104651. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Buthmann J.L., Antonacci C., Uy J.P., Borchers L.R., Miller J.G., Gotlib I.H. Stress reactivity moderates the association between early experiences of unpredictability and emotional problems in adolescents. Neurobiol Stress. 2025;34 doi: 10.1016/j.ynstr.2024.100706. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Achenbach T.M., Achenbach T.M. University of Vermont; 1991. Manual for the Youth Self-Report and 1991 Profile. Department of Psychiatry. http://archive.org/details/manualforyouthse00ach_23j [ Google Scholar ] 39. Ambrosini P.J. Historical development and present status of the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS) J Am Acad Child Adolesc Psychiatry. 2000;39(1):49–58. doi: 10.1097/00004583-200001000-00016. [ DOI ] [ PubMed ] [ Google Scholar ] 40. R Core Team R: a language and environment for statistical computing. https://www.R-project.org/ 41. Wood S.N., Goude Y., Fasiolo M. In: Interpretability for Industry 4.0: Statistical and Machine Learning Approaches. Lepore A., Palumbo B., Poggi J.M., editors. Springer International Publishing; 2022. Interpretability in generalized additive models; pp. 85–123. [ DOI ] [ Google Scholar ] 42. Cohen J., Cohen P., West S.G., Aiken L.S. 3rd ed. Routledge; 2013. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. [ DOI ] [ Google Scholar ] 43. Elman J.A., Vogel J.W., Bocancea D.I., et al. Issues and recommendations for the residual approach to quantifying cognitive resilience and reserve. Alzheimers Res Ther. 2022;14:102. doi: 10.1186/s13195-022-01049-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Gur R.E., Moore T.M., Rosen A.F.G., et al. Burden of environmental adversity associated with psychopathology, maturation, and brain behavior parameters in youths. JAMA Psychiatry. 2019;76(9):966–975. doi: 10.1001/jamapsychiatry.2019.0943. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Ben-Shachar M.S., Lüdecke D., Makowski D. effectsize: Estimation of effect size indices and standardized parameters. J Open Source Softw. 2020;5(56):2815. doi: 10.21105/joss.02815. [ DOI ] [ Google Scholar ] 46. Daoud J.I. Multicollinearity and regression analysis. J Phys Conf Ser. 2017;949(1) doi: 10.1088/1742-6596/949/1/012009. [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplemental Material mmc1.docx (1.4MB, docx) Articles from JAACAP Open are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (3.2 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Resilience across time: Longitudinal measurement invariance of the Brief Resilience Scale in French university students.#833
• Resilience and social support as correlates of dementia risk: insights from a biopsychosocial analysis of the All of Us dataset.#678283
• Cohort profile: The Resilient Minds national study of mental health and cognitive resilience in community dwelling adults aged 18 to 93#197599
• Predictors of Academic Resilience Based on Demographic and Psychological Profiles among Undergraduate Nursing Students: A Multicentric Survey in India.#13413
• Allostatic load profiles, resilience, and psychosocial adaptation in young adults.#455136