A systematic review of individual, social, and societal resilience factors in response to societal challenges and crises

The study addresses how individuals maintain or regain mental health in the face of large-scale societal stressors such as pandemics, economic crises, wars and conflicts, and natural disasters. Modern resilience research conceptualizes resilience as an outcome—maintenance or rapid recovery of mental health during/after stressor exposure—rather than a fixed personal trait. Within a multisystemic perspective, resilience factors span individual (e.g., optimism, self-efficacy), social (e.g., perceived social support), and societal levels (e.g., environmental resources). Prior work highlights potential higher-order resilience mechanisms (e.g., positive appraisal style, regulatory flexibility) that may mediate links between multilevel factors and resilient outcomes, though comprehensive tests are scarce. Trajectory modeling (e.g., GMM) commonly identifies heterogeneous mental health responses to stress (resilient, recovery, delayed, chronic). This review systematically synthesizes evidence on which multilevel resilience factors predict resilient trajectories to societal challenges in OECD countries, their incremental validity beyond sociodemographics and other factors, and how study, sample, and context features influence these associations.

Prior research often focused on individual resilience factors (e.g., optimism, hardiness, self-efficacy, control beliefs), but conceptual overlap and redundancy among factors are common, and incremental validity over other variables is rarely tested. Reviews using trajectory approaches suggest most stress-exposed individuals follow resilient or recovery trajectories, but methodological critiques of GMM remain. Social resilience research has emphasized perceived social support, which shows robust cross-sectional links to better mental health; however, longitudinal causal associations are debated and may suffer from modeling artifacts. Other social factors (family cohesion, connectedness, participation) have been less examined. Societal resilience factors are understudied and lack a consensus taxonomy; proposed distinctions include contextual (e.g., macroeconomic conditions, inequality, environmental access) versus target factors (e.g., perceived safety, infrastructure, job security). Conceptual frameworks posit mechanisms—such as positive appraisal and regulatory flexibility—may integrate effects of multiple lower-level factors, but empirical tests in societal stress contexts are rare.

Design and registration: Preregistered systematic review (OSF: 10.17605/OSF.IO/GWJVA) reported per PRISMA. Search strategy: Five databases (APA PsycNet [PsycInfo, PsycArticles, PsycExtra], Embase including PubMed and EmbaseCore, PTSDPubs, Scopus, Web of Science Core Collection) from 2004 to Aug 2, 2023. Reference lists of related reviews and included studies were also screened. Eligibility: Longitudinal observational studies of adults (≥18) from civil general populations (non-military, non-clinical) in OECD member states exposed to societal challenges (e.g., pandemics, wars/armed conflicts, climate crisis, natural disasters). Studies had to model mental health trajectories using growth mixture modeling (GMM) or comparable trajectory approaches and evaluate individual, social, or societal resilience factors as predictors in multinomial regression (including three-step approaches accounting for class assignment uncertainty). Minimum N ≥ 300; ≥3 assessment waves; first assessment within 4 years of stressor exposure; pre-stressor data not required. Resilience factor classification: Multilevel psychosocial resources were categorized as individual, social, or societal based on prior reviews; variables such as education, income, and family status were included as resilience factors if modifiable or proxies for established factors. Distinction between resilience factors and mechanisms followed author labeling and landmark reviews. Screening and selection: Duplicates removed via Zotero; two independent reviewers screened titles/abstracts and full texts in Rayyan (kappa=0.68 and 0.75, respectively), resolving disagreements by discussion/senior adjudication. Data extraction: Custom OSF extraction form captured sample characteristics, stressor type, trajectory solutions, predictors, and outcomes. Outcomes grouped into mental distress (general distress, depression, anxiety, PTSD, stress symptoms) and positive mental health (life satisfaction, personal growth, HRQoL, well-being). Factor-level assignments were double-checked by a second reviewer. Quality appraisal: Modified Newcastle-Ottawa Scale assessed selection, comparability, outcome assessment, reporting of methodological details, and trajectory model quality (constraints of variances/slopes). Overall study quality scored 0–100%. Synthesis and analysis: Used an effect size-informed vote-counting scheme rating evidence per factor and outcome: significant favorable associations coded as + (no controls), ++ (controlled for sociodemographics), +++ (controlled for sociodemographics and other resilience factors); non-significant as o/oo/ooo; significant unfavorable as -/--/---. Effect sizes categorized via OR thresholds (very small to large) per Cohen after appropriate transformations for dichotomized outcomes. Heterogeneity of covariate sets precluded standard meta-analysis; thus, non-parametric tests examined moderators: Fisher-Freeman-Halton, Kruskal–Wallis, Mann–Whitney U, and Spearman correlations (Monte Carlo 10,000 samples where applicable). Primary comparisons contrasted resilient trajectories vs less favorable (delayed, chronic, etc.); secondary compared recovery vs less favorable. Sensitivity analyses assessed links between evidence ratings and study quality, and timing intervals between stressor and assessments.

Search and study characteristics: 50 primary studies (55 reports) across 15 OECD countries (USA n=18; UK n=9; Australia n=4) published 2009–2023; sample sizes 360–65,818. Most samples were general population (n=37); 9 targeted high-risk groups (e.g., healthcare workers, police, migrants); 11 studies used representative samples. Weighted mean age 48.58 years (range 20.01–78.69); weighted mean proportion women 53.05% (range 13.4–100%). Attrition frequently high (up to 99%). Trajectory modeling: 84% used GMM variants; 50% allowed within-class variability of intercepts/slopes; 34% used more restrictive LCGA-type models; only 14% allowed free slopes; none allowed differing variances between classes. 20% accounted for class assignment uncertainty. Quality: Median NOS quality 66.67% (mean 62.17%, SD 11.23%; range 33.33–94.44%). Highest risks: GMM quality (74% high risk), selection (14%), outcome assessment (8%), comparability (6%). Stressors and outcomes: Stressors examined—pandemics 29/50 (58%), environmental/natural disasters 9/50 (18%), terrorist attacks 7/50 (14%), involuntary displacement 2/50 (4%), economic crises 2/50 (4%), civil unrest 1/50 (2%). Only 24% included pre-stressor data; 3–27 waves (mean 5.71), spanning 2 months to 15 years. Outcomes: PTSD symptoms 16 (32%), depressive 22 (44%), general distress 16 (32%), anxiety 14 (28%), stress 2 (4%); positive mental health 5 (10%). Mean 3.92 trajectories per study (range 2–6). Resilience factors covered: 54 factors total—34 individual, 12 social, 8 societal. Commonly studied: individual—education (28 studies), individual income (10); social—forms of social support (21), having a partner (16), living with family/others (15); societal—rural residence (8), neighborhood environment (3). Evidence synthesis (resilient vs less favorable): 478 effect estimates: 206 +++ (incremental validity beyond sociodemographics and other factors), 5 ++, 6 +; 222 null (o to ooo; 85 trending positive, 74 negative); 39 negative (- to ---), with 34 --- (unfavorable under highest control). Among numerically positive effects, 51.7% very small, 30.4% small, 12.9% medium, 5.0% large. Individual level: Strongest favorable evidence for individual income/low financial stress (12 +++ effect estimates ≈50% of estimates for income; 3 medium-to-large effects). Education showed mixed evidence overall but more consistently favorable for PTSD outcomes. Cognitive emotion regulation and psychological/cognitive flexibility were associated with more resilient responses (mostly very small to small effects). Several psychological dispositional factors (e.g., optimism, self-efficacy, locus of control, sense of coherence) showed mixed or weak evidence; some coping styles (active, social, religious) linked to unfavorable or null effects, suggesting context-dependent adaptiveness. Social level: Perceived social support consistently associated with resilient trajectories (primarily very small to small effects). Received support yielded more null than positive findings; structural support trended favorable but based on few studies. Having a partner and living with family/others showed mixed patterns, including unfavorable effects for PTSD/general distress in some studies. Societal level: Evidence sparse. Rural residence largely null (only 3 significant favorable estimates; 12%); neighborhood environment showed no significant links but most nulls trended favorable (76.9%). Environmental quality and perceived collective efficacy showed favorable effects (collective efficacy with two +++ small-to-medium effects); other societal factors (e.g., temperature, local house value) were rarely studied. Outcome- and context-related differences: Examination by outcome indicated differing emphases (e.g., individual factors often with general distress/PTSD; social factors with anxiety/depression; societal with positive mental health), with limited estimates for some outcomes. Moderators: More variables and simultaneous resilience factors in models related to less favorable evidence per factor (Spearman r≈-0.22 and -0.25, both p<0.001). Higher proportion of women associated with more favorable evidence (r=0.21, p<0.001); older mean age linked to less favorable evidence (r=-0.21, p=0.002). No significant associations with pre-stressor data inclusion, number of waves, number of trajectories, baseline sample size, representativeness, or stressor type. Recovery vs less favorable comparison: Based on 81 estimates, only 12 +++ (14.8%); many nulls and less consistent findings, including for income and perceived social support.

Findings indicate that, across multilevel psychosocial resilience factors, incremental predictive value for resilient mental health trajectories exists but is typically small in magnitude and context-dependent. Economic security (individual and household income, socioeconomic status) emerged as the most robust predictor of resilient outcomes. Perceived social support consistently related to resilient responses, while structural or received support and close-relationship indicators (partner status, living with family) showed heterogeneous or context-sensitive associations. Psychological processes connected to cognitive emotion regulation and flexibility demonstrated favorable associations, supporting frameworks emphasizing regulatory flexibility and strategy-situation fit. Societal-level factors were least studied; preliminary favorable evidence for environmental quality and collective efficacy suggests potential population-level levers, while rurality and neighborhood quality showed largely null results within single-country designs. Methodological and sample moderators mattered: models with more covariates diluted single-factor effects, and effects tended to be stronger in samples with more women and younger participants. Lack of consistent evidence for pre-stressor data or timing effects may reflect heterogeneous designs and outcomes. Overall, results support moving beyond a proliferation of isolated factors toward testing higher-order mechanisms (e.g., positive appraisal processes, regulatory flexibility) and multilevel interactions, and toward designing interventions and surveillance that incorporate individual, social, and societal components.

There was no single resilience factor that outperformed others across contexts. More resilient responses were associated with higher individual and household income (and lower financial stress), higher socioeconomic status, better cognitive emotion regulation, greater psychological/regulatory flexibility, and higher perceived social support. However, most incremental effects were very small to small, and many factors (e.g., education, optimism, self-efficacy) showed mixed evidence, consistent with the importance of fit between resources and situational demands. Future work should employ large, international longitudinal panels with pre-stressor data, expand the set of individual and especially social and societal factors, investigate higher-order mechanisms, and use modeling approaches that capture temporal dynamics and between-factor interactions to inform multilevel resilience promotion and crisis preparedness.

- Stressor typology: The review used one public health disaster list; absence of a definitive stressor classification may bias inclusions and synthesis. - Causality/prediction: Longitudinal associations are correlational; analyses do not establish causal links or predictive models. - GMM constraints: Trajectory modeling choices may inflate prevalence of resilient responses; limited pre-stressor data can misclassify recovery as resilience. - Synthesis approach: Heterogeneous covariate sets precluded meta-analysis; effect size-informed vote counting inherits limits of significance testing; potential publication bias cannot be ruled out. - Sparse evidence per factor/outcome: Many factors were assessed in single studies or specific stressors, limiting generalizability; insufficient estimates prevented robust between-outcome or stressor-specific recommendations. - Timing heterogeneity: Wide variation in onset and follow-up intervals may modulate factor effects. - Sampling/representativeness: Predominant use of convenience samples, high attrition, and limited diversity may bias findings; exposure levels often insufficiently modeled.