Life Stressors in Young Americans are Linked via Asymmetric U-Shapes to Mental Health Severity

Life Stressors in Young Americans are Linked via Asymmetric U-Shapes to Mental Health Severity | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Short Report Life Stressors in Young Americans are Linked via Asymmetric U-Shapes to Mental Health Severity Dmitry Scherbakov, Nina de Lacy, Olga Barg, Jihad S. Obeid, Alexander V. Alekseyenko This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9464871/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective To assess whether adolescent and young adult life stressors predict midlife depression/anxiety symptoms. Methods Using The National Longitudinal Survey of Youth 1997 (n = 8,984; ages 35 outcomes), we modeled maximum scores on CES‑D/GAD‑7 scales with stratified random forests and visualized via accumulated local effects. The U-shaped association was confirmed with quadratic linear regression. Results Lower income, more cohabitations, earlier marriage, and ≥ 6 children predicted ≥ 1‑point higher symptoms; many predictors (income, education, children, marriages) demonstrated variations of U‑shaped association with the outcome. Conclusion Early stressors were associated with midlife mental health symptoms, demonstrating non-linear patterns, which are often not identified by traditional statistical analysis. Psychology NLSY97 depression anxiety life-course machine learning accumulated local effects Figures Figure 1 INTRODUCTION Depression and anxiety remain prevalent mental health concerns [ 1 , 2 ], with well‑documented disparities across socioeconomic, racial, ethnic, and gender groups [ 3 , 4 ]. Stressful life events are recognized as significant mental health factors [ 5 ]. Among those, adverse childhood experiences (ACEs) have long-lasting impact and affect outcomes later in life [ 6 ]. Less is known about gender and race-based differences in these experiences. A potential limitation of ACE studies is that they often focus on the childhood period (< 18 years), and do not cover early adulthood [ 7 ]. This study leverages longitudinal data from the National Longitudinal Survey of Youth 1997 (NLSY97) [ 8 ] to investigate how various life experiences pertaining to income, family structure, and relationship dynamics affect depression and anxiety symptom severity. The CES-D (Center for Epidemiologic Studies Depression) score [ 9 ] and GAD-7 (Generalized Anxiety Disorder 7-item) [ 10 ] serve as the primary measures of mental health symptoms in this analysis. We used random forest, a machine learning method capable of capturing non-linear relationships in data without specifying them beforehand [ 11 ]. By identifying key predictors and how they interact with the outcome across sex (female/male) and two income groups (high/low income), this associational study aims to inform strategies for prevention and intervention. The supplemental analysis presents results from analogue models stratified by race and prior mental health status instead of sex. METHODS Data Source The NLSY97 survey study provides comprehensive multi-wave longitudinal data on a representative sample of U.S. individuals born between 1980 and 1984. Variable selection Life stressors were selected by mapping NLSY97 variables to domains from the Stressful Life Events Schedule (SLES) [ 12 ]. The following domains were included: education, money (income), reproduction (number of children), romantic relationships, and miscellaneous events (deaths). Supplemental Table 1 Lists the variables selected for analysis. Variable Adjustments, Imputation, and Missing Values Handling Variables from NLSY97 were processed as follows, median imputation was used for missing values in the predictors [ 13 ]: Outcome: CES-D/GAD7 Score (0–21). Maximum of CV_CESD_SCORE, GAD7_SCORE variables, with individuals with missing scores in all three variables excluded from analysis. Personal income (continuous) : Mean across selected survey years [ 13 ]. The mean value reflects the average dollar value across 2007–2022, which is lower than the 2025-dollar value due to inflation. Personal income (dichotomized) : Mean of income across selected survey years, dichotomized as Low (< $ 30,000) and High (≥ $ 30,000). Education Level : Replaced with an ordinal scale ranging from 0 (none) to 7 ( “Professional degree (DDS, JD, MD)"). Number of Children : Sum of the two variables (CV_BIO_CHILD_HH_2015, CV_BIO_CHILD_NR_2015), reflecting all biological children respondent had up until 2015. Age at Dating Debut : Calculated from first reported sexual partner, cohabitation, or marriage year minus birth year. Age at First Marriage : Year of first marriage minus birth year. Number of Relationship Breakdowns : Sum across years. Number of Cohabitations : The last of the sequence of cohabitation partner numbers as reported in this categorical variable at any survey year/month selected for analysis. Number of Marriages : The last of the sequence of marital partner numbers as reported in this categorical variable at any survey year/month selected for analysis. Number of Deceased Close Relatives : Sum across selected years. Sex at Birth : No processing applied, but only used two sexes – female and male. Machine Learning Analysis Two random forest models (for low and high income), implemented via the caret package in R with 5-fold cross-validation to select the mtry parameter, predicted CES-D scores with all other variables as predictors in the model. As the goal was associative rather than predictive, no hold‑out test set was used; performance was assessed via cross‑validation. Accumulated local effects grouped by sex were computed using the DALEX package and a bootstrap of 100 iterations to estimate the spread of the predicted values and confidence intervals. Results were visualized using ggplot2. A similar separate analyses looked at 1) racial disparities using three random forest models (for Black, Hispanic, and Non-Black/Non-Hispanic cohorts, provided as Supplemental Fig. 2); 2) by prior mental health risk (High vs Low risk) defined as early depression (CES-D score in years 2000 and 2002) score > = 4.5 and neuroticism (2-items from Ten-Item Personality Inventory Scale, TIPI: calm/emotionally stable and anxious/easily upset) score > = 2 as “High risk”, and early depression < = 3 and neuroticism < = 2 defined as “Low risk” (Supplemental Fig. 3). U-Shape statistical test To confirm that the relationship between predictors and outcome forms a U-shape, we fitted separate quadratic linear regression for each predictor and checked whether the quadratic term is statistically significant after Holm adjustment for multiple comparison. RESULTS 7,314 out of 8,984 individuals had a non-missing CES-D/GAD score. The accumulated local effects profiles of the random forest models for females and males across 100 bootstrap iterations are presented in Fig. 1 . At the optimal mtry , cross‑validated performance (RMSE/MAE/R²) was 4.28/3.21/0.05 for the high‑income model and 5.22/4.13/0.05 for the low‑income model. We tested personal income, education level, number of marriages, and number of children for a U-shaped association with the outcome, and in all respective models found quadratic terms (responsible for curvature) statistically significant. When comparing the results of the two models based personal income (dashed vs solid lines across all facets) persistent difference of approximately 2 points in CES-D/GAD score is seen: higher personal income was associated with reduced CES-D scores across all predictors. Furthermore, income as a predictor in both models showed a clear gradient: low earners had substantially higher symptom severity (over 3 points vs. the $ 60– $ 80k range), while very high earners (> $ 150k) had slightly elevated CES-D scores than the middle-income group. Education displayed another shallow U-shape—both very low and very high levels were linked to small increases (< 0.5 points), though among higher-earning women, more education reduced scores by about 1 point. Family size also followed a U-pattern: participants without children had modestly higher CES-D/GAD scores, scores were lowest with 1–3 children in high income group or with 1–5 children in low income group, while families with ≥ 6 children experienced higher symptoms by almost2 points in low-income households. Dating before age 20 predicted roughly a 1-point increase in symptoms among high-income participants, whereas later first marriage was protective (≈ 2 points lower in low-income; ≈1 point in high-income). More relationship breakdowns added ≈ 1 point. Cohabitation count was a strong adverse marker, particularly for women (≈ 2.5-point increase). Marriages showed a U-shape: never-married individuals had ≈ 0.5-point higher scores; multiple marriages similarly resulted in elevated symptoms, with low-income participants showing ≈2-point increases at high counts. Parental divorce had minimal impact, and bereavement modestly raised scores, with a steeper ≈1-point rise for high-income men. Supplemental Fig. 2 shows that the above patterns are consistent when prediction is stratified by race/ethnicity. Supplemental Fig. 3 shows similar patterns in analysis grouped by prior mental health risk, confirming that results hold even when accounting for earlier depression and neuroticism. DISCUSSION Our study identified differences of 1–2 points in CES-D/GAD score based on analyzed life variable While this might appear minor at a population scale, it can nonetheless represent a clinically important difference [ 14 ]. We observed a shallow U‑shaped income–mental health curve: symptoms were highest near poverty and rose slightly at very high incomes (< 0.5 points), patterns that linear models often miss [ 15 ]. Low income appeared more detrimental for men, with some elevation at very high incomes, consistent with social comparison effects [ 16 ]. Education was generally protective [ 17 ], but among low‑income individuals, very high education predicted higher symptoms, plausibly reflecting effects of under‑/unemployment [ 18 ]. Family size also resembled a U-shape, consistent with resource-strain mechanisms [ 19 ] [ 20 ]. Later dating debut and later marriage suggests a protective effect (without implying causality), aligning with evidence that early marriage can be an adversity [ 21 , 22 ]. Yet very late marriage may increase divorce risk, and marital dissolution is linked to higher likelihood of depression [ 23 , 24 ]. Relationship breakdowns predicted symptoms more strongly than bereavement counts; this pattern persisted after adjusting for early symptoms and neuroticism, though NLSY97 lacks reliable adolescent depression measures. Cohabitation and marriage counts showed similar associations with later symptoms; marriages followed a U‑shape, with never‑married status functioning as adversity. Prior work underscores the importance of early relationship histories [ 25 ], links higher partner counts and relationship transitions to elevated depression [ 21 ] [ 26 ], and identifies bereavement as a major risk—especially for females—while also refining these links (e.g., breakups predict depression chiefly when ties lack emotional commitment, are early versus peers, or are socially embedded; stable cohabitation may not increase depression) [ 27 ] [ 28 ] [ 29 ]. Parental divorce was not a strong predictor of adversity here despite meta‑analytic findings about resource disruption [ 30 ]. Racial/ethnic stratification broadly mirrored these patterns across White, Black, and Hispanic groups, with nuances: early marriage (< 25) related more strongly to elevated midlife symptoms among White participants, and increases in cohabitations had stronger effects for White and Black participants [ 31 ]. Limitations We did not include all SLES domains (e.g., crime/legal, housing, employment, broader social ties, comorbid health issues) in our analysis. A limitation of the random forest approach is that it adapts closely to the specifics of the observed data, which may limit the generalizability of the results beyond the study sample. Other imputation methods besides the median could have been used [ 13 ]. Conclusion Socioeconomic conditions, family size, relationship timing and instability, and bereavement shape midlife depression/anxiety, often nonlinearly. U‑shaped associations were common, underscoring the value of machine learning models, such as Random Forest, that capture curvature in the data. Declarations Funding This publication was supported by grant T15 LM013977, Biomedical Informatics and Data Science for Health Equity Research (SC BIDS4Health). Competing interests. Authors have no competing interests to declare Ethics approval This study analyzed publicly available, de‑identified data from the National Longitudinal Survey of Youth 1997 (NLSY97). In accordance with institutional and national guidelines, the study was exempt from ethics committee review; all procedures adhered to the principles of the Declaration of Helsinki. Consent to participate Informed consent was obtained by the original data collectors from all participants. Consent to publish Not applicable. No individual, identifiable data or images are included. Data availability NLSY97 data are available via the U.S. Bureau of Labor Statistics National Longitudinal Surveys (https://www.nlsinfo.org). Analysis code and derived materials will be shared upon reasonable request. Author contributions Conceptualization: D.A.S., J.S.O., A.V.A.; Methodology: D.A.S., N.dL., A.V.A.; Formal analysis: D.A.S.; Visualization: D.A.S.; Writing—original draft: D.A.S.; Writing—review & editing: all authors; Supervision: A.V.A.; Correspondence: A.V.A. References Baxter AJ, Scott KM, Vos T, Whiteford HA (2013) Global prevalence of anxiety disorders: a systematic review and meta-regression. Psychol Med 43(5):897–910 Greenberg PE, Fournier A-A, Sisitsky T, Pike CT, Kessler RC (2015) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):5356 Williams DR, Mohammed SA (2009) Discrimination and racial disparities in health: evidence and needed research. J Behav Med 32:20–47 Zimmerman F, Katon W (2005) Socioeconomic status, depression disparities, and financial strain: what lies behind the income-depression relationship? Health Econ 14 12:1197–1215 Kendler KS, Karkowski LM, Prescott CA (1999) Causal relationship between stressful life events and the onset of major depression. Am J Psychiatry 156(6):837–841 Felitti VJ, Anda RF, Nordenberg D, Williamson DF, Spitz AM, Edwards V, Marks JS (1998) Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. Am J Prev Med 14(4):245–258 Evans GW, Li D, Whipple SS (2013) Cumulative risk and child development. Psychol Bull 139(6):1342 Cooksey EC Using the National Longitudinal Surveys of Youth (NLSY) to conduct life course analyses. Handb life course health Dev 2018:561–577 Eaton WW, Muntaner C, Smith C, Tien A, Ybarra M Center for epidemiologic studies depression scale: Review and revision. The use of psychological testing for treatment planning and outcomes assessment 2004 Mossman SA, Luft MJ, Schroeder HK, Varney ST, Fleck DE, Barzman DH, Gilman R, DelBello MP, Strawn JR (2017) The Generalized Anxiety Disorder 7-item (GAD-7) scale in adolescents with generalized anxiety disorder: signal detection and validation. Annals Clin psychiatry: official J Am Acad Clin Psychiatrists 29(4):227 Breiman L (2001) Random forests. Mach Learn 45:5–32 Williamson DE, Birmaher B, Ryan ND, Shiffrin TP, Lusky JA, Protopapa J, Dahl RE, Brent DA (2003) The stressful life events schedule for children and adolescents: development and validation. Psychiatry Res 119(3):225–241 Jadhav A, Pramod D, Ramanathan K (2019) Comparison of performance of data imputation methods for numeric dataset. Appl Artif Intell 33(10):913–933 Kounali D, Button KS, Lewis G, Gilbody S, Kessler D, Araya R, Duffy L, Lanham P, Peters TJ, Wiles N (2022) How much change is enough? Evidence from a longitudinal study on depression in UK primary care. Psychol Med 52(10):1875–1882 Li C, Guangjie N, Lin W, Feier C (2022) More income, less depression? Revisiting the nonlinear and heterogeneous relationship between income and mental health. Front Psychol 13 Courtenay WH (2000) Constructions of masculinity and their influence on men's well-being: a theory of gender and health. Soc Sci Med 50(10):1385–1401 Bjelland I, Krokstad S, Mykletun A, Dahl AA, Tell GS, Tambs K (2008) Does a higher educational level protect against anxiety and depression? The HUNT study. Soc Sci Med 66(6):1334–1345 Allan BA, Kim T, Shein B (2022) Underemployment and mental health: A longitudinal study. J Couns Psychol 69(5):578 Nomaguchi KM, Milkie MA (2003) Costs and rewards of children: The effects of becoming a parent on adults' lives. J marriage family 65(2):356–374 Hodgkinson S, Godoy L, Beers LS, Lewin A (2017) Improving mental health access for low-income children and families in the primary care setting. Pediatrics 139(1) Uecker JE (2012) Marriage and mental health among young adults. J Health Soc Behav 53(1):67–83 Uecker JE, Stokes CE (2008) Early marriage in the United States. J Marriage Family 70(4):835–846 Wolfinger NH (2015) Want to avoid divorce? Wait to get married, but not too long. Inst Family Stud Amato PR (2000) The consequences of divorce for adults and children. J marriage family 62(4):1269–1287 Teachman J (2003) Premarital sex, premarital cohabitation, and the risk of subsequent marital dissolution among women. J Marriage Family 65(2):444–455 Amato PR (2010) Research on divorce: Continuing trends and new developments. J marriage family 72(3):650–666 Kessler RC The effects of stressful life events on depression. Depression 2013:67–90 Meier AM (2007) Adolescent first sex and subsequent mental health. Am J Sociol 112(6):1811–1847 Brown SL, Bulanda JR, Lee GR (2005) The significance of nonmarital cohabitation: Marital status and mental health benefits among middle-aged and older adults. Journals Gerontol Ser B: Psychol Sci Social Sci 60(1):S21–S29 Amato PR (2001) Children of divorce in the 1990s: an update of the Amato and Keith (1991) meta-analysis. J Fam Psychol 15(3):355 Jackson JS, Knight KM, Rafferty JA (2010) Race and unhealthy behaviors: chronic stress, the HPA axis, and physical and mental health disparities over the life course. Am J Public Health 100(5):933–939 Additional Declarations The authors declare no competing interests. Supplementary Files APPENDIX.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9464871","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Short Report","associatedPublications":[],"authors":[{"id":626397503,"identity":"8dd0f608-0f5c-48c6-90b1-445dbb832635","order_by":0,"name":"Dmitry Scherbakov","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxUlEQVRIiWNgGAWjYFACHgYGxgYGA34oF8gmVotkA8laDA4Qq0W+/ezBx5U77IyNr50x/vCDwUZ2wwECWgzO5CUbnj2TbGZ2O8dMsochzZiwFgkeM8nGNmYbkBZmBobDiQS1yM/gMf/Z2FZvYzw7x/gzA8N/wloYbvCYMTa2HTYzkM4xkGZgOEBYi8GZHGPJxjPHjSVup5VJ9hgkG88k6LD2M4YfG3dUG/bPTt784UeFnWwfQYehWUqa8lEwCkbBKBgFOAAA6ONBnBwhq1oAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0005-7274-0934","institution":"Medical University of South Carolina","correspondingAuthor":true,"prefix":"","firstName":"Dmitry","middleName":"","lastName":"Scherbakov","suffix":""},{"id":626398887,"identity":"80dbc334-b96b-44bc-b856-3b362a3d53ac","order_by":1,"name":"Nina de Lacy","email":"","orcid":"","institution":"University of Utah","correspondingAuthor":false,"prefix":"","firstName":"Nina","middleName":"","lastName":"de Lacy","suffix":""},{"id":626398889,"identity":"4be050df-9bd2-4c34-bf3f-d086e0def981","order_by":2,"name":"Olga Barg","email":"","orcid":"","institution":"University of Pennsylvania School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Olga","middleName":"","lastName":"Barg","suffix":""},{"id":626398891,"identity":"a9f2617d-5b50-4925-97b2-aa58c9b2ba96","order_by":3,"name":"Jihad S. Obeid","email":"","orcid":"","institution":"Medical University of South Carolina","correspondingAuthor":false,"prefix":"","firstName":"Jihad","middleName":"S.","lastName":"Obeid","suffix":""},{"id":626398894,"identity":"5403612d-f9e1-4fb5-97e2-66956caef6cf","order_by":4,"name":"Alexander V. Alekseyenko","email":"","orcid":"","institution":"Medical University of South Carolina","correspondingAuthor":false,"prefix":"","firstName":"Alexander","middleName":"V.","lastName":"Alekseyenko","suffix":""}],"badges":[],"createdAt":"2026-04-19 21:36:16","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9464871/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9464871/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109118170,"identity":"66991f4b-f87b-4263-b38c-5b54331c1760","added_by":"auto","created_at":"2026-05-12 16:51:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":250506,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 1. Accumulated local effect (ALE) profiles from the two random forest models for low income population (dashed line) and high income (solid line), grouped by sex at birth (red for female), with life stressors as predictors and CES-D/GAD maximum score in later life as outcome (displayed as predicted score points on y scale, with possible values from 0 to 21, where 21 is the highest symptom severity – only subset of values is shown to improve figure readability). Spline smoothing applied. Shaded areas represent bootstrap‑based uncertainty in the accumulated local effect estimates across 100 resampled datasets.\u003c/p\u003e","description":"","filename":"Screenshot20260420at10.28.40AM.png","url":"https://assets-eu.researchsquare.com/files/rs-9464871/v1/9aa66b924f77607fd2b7fef8.png"},{"id":109204871,"identity":"eabd909e-c03f-4a76-bcf3-3a0050665d71","added_by":"auto","created_at":"2026-05-13 15:02:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":393448,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9464871/v1/3113342f-a583-4f35-805e-7edb254a52cf.pdf"},{"id":109118168,"identity":"e19785d2-4879-4a8c-af96-0c0598b1c970","added_by":"auto","created_at":"2026-05-12 16:51:14","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":829944,"visible":true,"origin":"","legend":"","description":"","filename":"APPENDIX.docx","url":"https://assets-eu.researchsquare.com/files/rs-9464871/v1/30099d69fdc77c5a748f7df4.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eLife Stressors in Young Americans are Linked via Asymmetric U-Shapes to Mental Health Severity\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eDepression and anxiety remain prevalent mental health concerns [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], with well‑documented disparities across socioeconomic, racial, ethnic, and gender groups [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Stressful life events are recognized as significant mental health factors [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Among those, adverse childhood experiences (ACEs) have long-lasting impact and affect outcomes later in life [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Less is known about gender and race-based differences in these experiences. A potential limitation of ACE studies is that they often focus on the childhood period (\u0026lt;\u0026thinsp;18 years), and do not cover early adulthood [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study leverages longitudinal data from the National Longitudinal Survey of Youth 1997 (NLSY97) [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] to investigate how various life experiences pertaining to income, family structure, and relationship dynamics affect depression and anxiety symptom severity. The CES-D (Center for Epidemiologic Studies Depression) score [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] and GAD-7 (Generalized Anxiety Disorder 7-item) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] serve as the primary measures of mental health symptoms in this analysis. We used random forest, a machine learning method capable of capturing non-linear relationships in data without specifying them beforehand [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. By identifying key predictors and how they interact with the outcome across sex (female/male) and two income groups (high/low income), this associational study aims to inform strategies for prevention and intervention. The supplemental analysis presents results from analogue models stratified by race and prior mental health status instead of sex.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eData Source\u003c/h2\u003e \u003cp\u003eThe NLSY97 survey study provides comprehensive multi-wave longitudinal data on a representative sample of U.S. individuals born between 1980 and 1984.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eVariable selection\u003c/h3\u003e\n\u003cp\u003eLife stressors were selected by mapping NLSY97 variables to domains from the Stressful Life Events Schedule (SLES) [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. The following domains were included: education, money (income), reproduction (number of children), romantic relationships, and miscellaneous events (deaths). Supplemental Table\u0026nbsp;1 Lists the variables selected for analysis.\u003c/p\u003e\n\u003ch3\u003eVariable Adjustments, Imputation, and Missing Values Handling\u003c/h3\u003e\n\u003cp\u003eVariables from NLSY97 were processed as follows, median imputation was used for missing values in the predictors [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eOutcome: CES-D/GAD7 Score (0\u0026ndash;21).\u003c/b\u003e Maximum of CV_CESD_SCORE, GAD7_SCORE variables, with individuals with missing scores in all three variables excluded from analysis.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003ePersonal income (continuous)\u003c/b\u003e: Mean across selected survey years [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. The mean value reflects the average dollar value across 2007\u0026ndash;2022, which is lower than the 2025-dollar value due to inflation.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003ePersonal income (dichotomized)\u003c/b\u003e: Mean of income across selected survey years, dichotomized as Low (\u0026lt; \u003cspan\u003e$\u003c/span\u003e30,000) and High (\u0026ge; \u003cspan\u003e$\u003c/span\u003e30,000).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eEducation Level\u003c/b\u003e: Replaced with an ordinal scale ranging from 0 (none) to 7 (\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e\u0026ldquo;Professional degree (DDS, JD, MD)\").\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eNumber of Children\u003c/b\u003e: Sum of the two variables (CV_BIO_CHILD_HH_2015, CV_BIO_CHILD_NR_2015), reflecting all biological children respondent had up until 2015.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eAge at Dating Debut\u003c/b\u003e: Calculated from first reported sexual partner, cohabitation, or marriage year minus birth year.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eAge at First Marriage\u003c/b\u003e: Year of first marriage minus birth year.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eNumber of Relationship Breakdowns\u003c/b\u003e: Sum across years.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eNumber of Cohabitations\u003c/b\u003e: The last of the sequence of cohabitation partner numbers as reported in this categorical variable at any survey year/month selected for analysis.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eNumber of Marriages\u003c/b\u003e: The last of the sequence of marital partner numbers as reported in this categorical variable at any survey year/month selected for analysis.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eNumber of Deceased Close Relatives\u003c/b\u003e: Sum across selected years.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eSex at Birth\u003c/b\u003e: No processing applied, but only used two sexes \u0026ndash; female and male.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e\n\u003ch3\u003eMachine Learning Analysis\u003c/h3\u003e\n\u003cp\u003eTwo random forest models (for low and high income), implemented via the \u003cem\u003ecaret\u003c/em\u003e package in \u003cem\u003eR\u003c/em\u003e with 5-fold cross-validation to select the \u003cem\u003emtry\u003c/em\u003e parameter, predicted CES-D scores with all other variables as predictors in the model. As the goal was associative rather than predictive, no hold‑out test set was used; performance was assessed via cross‑validation. Accumulated local effects grouped by sex were computed using the DALEX package and a bootstrap of 100 iterations to estimate the spread of the predicted values and confidence intervals. Results were visualized using ggplot2.\u003c/p\u003e \u003cp\u003eA similar separate analyses looked at 1) racial disparities using three random forest models (for Black, Hispanic, and Non-Black/Non-Hispanic cohorts, provided as Supplemental Fig.\u0026nbsp;2); 2) by prior mental health risk (High vs Low risk) defined as early depression (CES-D score in years 2000 and 2002) score\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;4.5 and neuroticism (2-items from Ten-Item Personality Inventory Scale, TIPI: calm/emotionally stable and anxious/easily upset) score\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;2 as \u0026ldquo;High risk\u0026rdquo;, and early depression\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;3 and neuroticism\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;2 defined as \u0026ldquo;Low risk\u0026rdquo; (Supplemental Fig.\u0026nbsp;3).\u003c/p\u003e\n\u003ch3\u003eU-Shape statistical test\u003c/h3\u003e\n\u003cp\u003eTo confirm that the relationship between predictors and outcome forms a U-shape, we fitted separate quadratic linear regression for each predictor and checked whether the quadratic term is statistically significant after Holm adjustment for multiple comparison.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003e7,314 out of 8,984 individuals had a non-missing CES-D/GAD score. The accumulated local effects profiles of the random forest models for females and males across 100 bootstrap iterations are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. At the optimal \u003cem\u003emtry\u003c/em\u003e, cross‑validated performance (RMSE/MAE/R\u0026sup2;) was 4.28/3.21/0.05 for the high‑income model and 5.22/4.13/0.05 for the low‑income model. We tested personal income, education level, number of marriages, and number of children for a U-shaped association with the outcome, and in all respective models found quadratic terms (responsible for curvature) statistically significant.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen comparing the results of the two models based personal income (dashed vs solid lines across all facets) persistent difference of approximately 2 points in CES-D/GAD score is seen: higher personal income was associated with reduced CES-D scores across all predictors. Furthermore, income as a predictor in both models showed a clear gradient: low earners had substantially higher symptom severity (over 3 points vs. the \u003cspan\u003e$\u003c/span\u003e60\u0026ndash;\u003cspan\u003e$\u003c/span\u003e80k range), while very high earners (\u0026gt; \u003cspan\u003e$\u003c/span\u003e150k) had slightly elevated CES-D scores than the middle-income group. Education displayed another shallow U-shape\u0026mdash;both very low and very high levels were linked to small increases (\u0026lt;\u0026thinsp;0.5 points), though among higher-earning women, more education reduced scores by about 1 point. Family size also followed a U-pattern: participants without children had modestly higher CES-D/GAD scores, scores were lowest with 1\u0026ndash;3 children in high income group or with 1\u0026ndash;5 children in low income group, while families with \u0026ge;\u0026thinsp;6 children experienced higher symptoms by almost2 points in low-income households.\u003c/p\u003e \u003cp\u003eDating before age 20 predicted roughly a 1-point increase in symptoms among high-income participants, whereas later first marriage was protective (\u0026asymp;\u0026thinsp;2 points lower in low-income; \u0026asymp;1 point in high-income). More relationship breakdowns added\u0026thinsp;\u0026asymp;\u0026thinsp;1 point. Cohabitation count was a strong adverse marker, particularly for women (\u0026asymp;\u0026thinsp;2.5-point increase). Marriages showed a U-shape: never-married individuals had\u0026thinsp;\u0026asymp;\u0026thinsp;0.5-point higher scores; multiple marriages similarly resulted in elevated symptoms, with low-income participants showing \u0026asymp;2-point increases at high counts. Parental divorce had minimal impact, and bereavement modestly raised scores, with a steeper \u0026asymp;1-point rise for high-income men.\u003c/p\u003e \u003cp\u003eSupplemental Fig.\u0026nbsp;2 shows that the above patterns are consistent when prediction is stratified by race/ethnicity. Supplemental Fig.\u0026nbsp;3 shows similar patterns in analysis grouped by prior mental health risk, confirming that results hold even when accounting for earlier depression and neuroticism.\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eOur study identified differences of 1\u0026ndash;2 points in CES-D/GAD score based on analyzed life variable While this might appear minor at a population scale, it can nonetheless represent a clinically important difference [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe observed a shallow U‑shaped income\u0026ndash;mental health curve: symptoms were highest near poverty and rose slightly at very high incomes (\u0026lt;\u0026thinsp;0.5 points), patterns that linear models often miss [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Low income appeared more detrimental for men, with some elevation at very high incomes, consistent with social comparison effects [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Education was generally protective [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], but among low‑income individuals, very high education predicted higher symptoms, plausibly reflecting effects of under‑/unemployment [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Family size also resembled a U-shape, consistent with resource-strain mechanisms [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eLater dating debut and later marriage suggests a protective effect (without implying causality), aligning with evidence that early marriage can be an adversity [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Yet very late marriage may increase divorce risk, and marital dissolution is linked to higher likelihood of depression [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Relationship breakdowns predicted symptoms more strongly than bereavement counts; this pattern persisted after adjusting for early symptoms and neuroticism, though NLSY97 lacks reliable adolescent depression measures. Cohabitation and marriage counts showed similar associations with later symptoms; marriages followed a U‑shape, with never‑married status functioning as adversity. Prior work underscores the importance of early relationship histories [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], links higher partner counts and relationship transitions to elevated depression [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], and identifies bereavement as a major risk\u0026mdash;especially for females\u0026mdash;while also refining these links (e.g., breakups predict depression chiefly when ties lack emotional commitment, are early versus peers, or are socially embedded; stable cohabitation may not increase depression) [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e] [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eParental divorce was not a strong predictor of adversity here despite meta‑analytic findings about resource disruption [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Racial/ethnic stratification broadly mirrored these patterns across White, Black, and Hispanic groups, with nuances: early marriage (\u0026lt;\u0026thinsp;25) related more strongly to elevated midlife symptoms among White participants, and increases in cohabitations had stronger effects for White and Black participants [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e"},{"header":"Limitations","content":"\u003cp\u003eWe did not include all SLES domains (e.g., crime/legal, housing, employment, broader social ties, comorbid health issues) in our analysis. A limitation of the random forest approach is that it adapts closely to the specifics of the observed data, which may limit the generalizability of the results beyond the study sample. Other imputation methods besides the median could have been used [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eSocioeconomic conditions, family size, relationship timing and instability, and bereavement shape midlife depression/anxiety, often nonlinearly. U‑shaped associations were common, underscoring the value of machine learning models, such as Random Forest, that capture curvature in the data.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis publication was supported by grant T15 LM013977, Biomedical Informatics and Data Science for Health Equity Research (SC BIDS4Health).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthors have no competing interests to declare\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study analyzed publicly available, de‑identified data from the National Longitudinal Survey of Youth 1997 (NLSY97). In accordance with institutional and national guidelines, the study was \u003cstrong\u003eexempt\u003c/strong\u003e from ethics committee review; all procedures adhered to the principles of the Declaration of Helsinki.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Informed consent was obtained by the original data collectors from all participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to publish\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Not applicable. No individual, identifiable data or images are included.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;NLSY97 data are available via the U.S. Bureau of Labor Statistics National Longitudinal Surveys (https://www.nlsinfo.org). Analysis code and derived materials will be shared upon reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Conceptualization: D.A.S., J.S.O., A.V.A.; Methodology: D.A.S., N.dL., A.V.A.; Formal analysis: D.A.S.; Visualization: D.A.S.; Writing—original draft: D.A.S.; Writing—review \u0026amp; editing: all authors; Supervision: A.V.A.; Correspondence: A.V.A.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBaxter AJ, Scott KM, Vos T, Whiteford HA (2013) Global prevalence of anxiety disorders: a systematic review and meta-regression. Psychol Med 43(5):897\u0026ndash;910\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreenberg PE, Fournier A-A, Sisitsky T, Pike CT, Kessler RC (2015) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):5356\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliams DR, Mohammed SA (2009) Discrimination and racial disparities in health: evidence and needed research. J Behav Med 32:20\u0026ndash;47\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZimmerman F, Katon W (2005) Socioeconomic status, depression disparities, and financial strain: what lies behind the income-depression relationship? Health Econ 14 12:1197\u0026ndash;1215\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKendler KS, Karkowski LM, Prescott CA (1999) Causal relationship between stressful life events and the onset of major depression. Am J Psychiatry 156(6):837\u0026ndash;841\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFelitti VJ, Anda RF, Nordenberg D, Williamson DF, Spitz AM, Edwards V, Marks JS (1998) Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. Am J Prev Med 14(4):245\u0026ndash;258\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEvans GW, Li D, Whipple SS (2013) Cumulative risk and child development. Psychol Bull 139(6):1342\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCooksey EC Using the National Longitudinal Surveys of Youth (NLSY) to conduct life course analyses. Handb life course health Dev 2018:561\u0026ndash;577\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEaton WW, Muntaner C, Smith C, Tien A, Ybarra M Center for epidemiologic studies depression scale: Review and revision. \u003cem\u003eThe use of psychological testing for treatment planning and outcomes assessment\u003c/em\u003e 2004\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMossman SA, Luft MJ, Schroeder HK, Varney ST, Fleck DE, Barzman DH, Gilman R, DelBello MP, Strawn JR (2017) The Generalized Anxiety Disorder 7-item (GAD-7) scale in adolescents with generalized anxiety disorder: signal detection and validation. Annals Clin psychiatry: official J Am Acad Clin Psychiatrists 29(4):227\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L (2001) Random forests. Mach Learn 45:5\u0026ndash;32\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliamson DE, Birmaher B, Ryan ND, Shiffrin TP, Lusky JA, Protopapa J, Dahl RE, Brent DA (2003) The stressful life events schedule for children and adolescents: development and validation. Psychiatry Res 119(3):225\u0026ndash;241\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJadhav A, Pramod D, Ramanathan K (2019) Comparison of performance of data imputation methods for numeric dataset. Appl Artif Intell 33(10):913\u0026ndash;933\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKounali D, Button KS, Lewis G, Gilbody S, Kessler D, Araya R, Duffy L, Lanham P, Peters TJ, Wiles N (2022) How much change is enough? Evidence from a longitudinal study on depression in UK primary care. Psychol Med 52(10):1875\u0026ndash;1882\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi C, Guangjie N, Lin W, Feier C (2022) More income, less depression? Revisiting the nonlinear and heterogeneous relationship between income and mental health. Front Psychol 13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCourtenay WH (2000) Constructions of masculinity and their influence on men's well-being: a theory of gender and health. Soc Sci Med 50(10):1385\u0026ndash;1401\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBjelland I, Krokstad S, Mykletun A, Dahl AA, Tell GS, Tambs K (2008) Does a higher educational level protect against anxiety and depression? The HUNT study. Soc Sci Med 66(6):1334\u0026ndash;1345\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllan BA, Kim T, Shein B (2022) Underemployment and mental health: A longitudinal study. J Couns Psychol 69(5):578\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNomaguchi KM, Milkie MA (2003) Costs and rewards of children: The effects of becoming a parent on adults' lives. J marriage family 65(2):356\u0026ndash;374\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHodgkinson S, Godoy L, Beers LS, Lewin A (2017) Improving mental health access for low-income children and families in the primary care setting. Pediatrics 139(1)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUecker JE (2012) Marriage and mental health among young adults. J Health Soc Behav 53(1):67\u0026ndash;83\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUecker JE, Stokes CE (2008) Early marriage in the United States. J Marriage Family 70(4):835\u0026ndash;846\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWolfinger NH (2015) Want to avoid divorce? Wait to get married, but not too long. Inst Family Stud\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmato PR (2000) The consequences of divorce for adults and children. J marriage family 62(4):1269\u0026ndash;1287\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeachman J (2003) Premarital sex, premarital cohabitation, and the risk of subsequent marital dissolution among women. J Marriage Family 65(2):444\u0026ndash;455\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmato PR (2010) Research on divorce: Continuing trends and new developments. J marriage family 72(3):650\u0026ndash;666\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKessler RC The effects of stressful life events on depression. Depression 2013:67\u0026ndash;90\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeier AM (2007) Adolescent first sex and subsequent mental health. Am J Sociol 112(6):1811\u0026ndash;1847\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrown SL, Bulanda JR, Lee GR (2005) The significance of nonmarital cohabitation: Marital status and mental health benefits among middle-aged and older adults. Journals Gerontol Ser B: Psychol Sci Social Sci 60(1):S21\u0026ndash;S29\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmato PR (2001) Children of divorce in the 1990s: an update of the Amato and Keith (1991) meta-analysis. J Fam Psychol 15(3):355\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJackson JS, Knight KM, Rafferty JA (2010) Race and unhealthy behaviors: chronic stress, the HPA axis, and physical and mental health disparities over the life course. Am J Public Health 100(5):933\u0026ndash;939\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[{"identity":"8acee987-8544-4e8c-acf5-a0f414a2db9b","identifier":"10.13039/100000002","name":"National Institutes of Health","awardNumber":"T15 LM013977","order_by":0}],"hasAdminPriorityOnWorkflow":true,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"NLSY97, depression, anxiety, life-course, machine learning, accumulated local effects","lastPublishedDoi":"10.21203/rs.3.rs-9464871/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9464871/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eTo assess whether adolescent and young adult life stressors predict midlife depression/anxiety symptoms.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eUsing The National Longitudinal Survey of Youth 1997 (n\u0026thinsp;=\u0026thinsp;8,984; ages\u0026thinsp;\u0026lt;\u0026thinsp;35 exposures; ages\u0026thinsp;\u0026gt;\u0026thinsp;35 outcomes), we modeled maximum scores on CES‑D/GAD‑7 scales with stratified random forests and visualized via accumulated local effects. The U-shaped association was confirmed with quadratic linear regression.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eLower income, more cohabitations, earlier marriage, and \u0026ge;\u0026thinsp;6 children predicted\u0026thinsp;\u0026ge;\u0026thinsp;1‑point higher symptoms; many predictors (income, education, children, marriages) demonstrated variations of U‑shaped association with the outcome.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eEarly stressors were associated with midlife mental health symptoms, demonstrating non-linear patterns, which are often not identified by traditional statistical analysis.\u003c/p\u003e","manuscriptTitle":"Life Stressors in Young Americans are Linked via Asymmetric U-Shapes to Mental Health Severity","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 16:51:05","doi":"10.21203/rs.3.rs-9464871/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3d1d9e0e-a22d-4cff-bfe5-475e8afb2f67","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":66607433,"name":"Psychology"}],"tags":[],"updatedAt":"2026-05-12T16:51:10+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 16:51:05","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9464871","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9464871","identity":"rs-9464871","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}