Beyond risk: resilience portfolios and youth mental health

Abstract

Background

Given concerns about the mental health of young people, there has been a growing body of research aimed at understanding strengths, resilience, and protective factors. The current study utilizes latent profile analysis to examine different combinations of strengths. Group differences on demographics, social determinants of health (SDoH), and mental health indices are examined across the identified latent classes.

Methods

Project Lift Up is a longitudinal, national study of youth and young adults designed to understand bystander behaviors for self-directed violence. 4,981 adolescents and young adults aged 13–22 years were recruited online between June 13, 2022 - October 30, 2023. Three additional waves of data were collected, approximately 6 months apart: Wave 2 (collected between January 13, 2023 - February 26, 2024); Wave 3 (June 13, 2023 – November 19, 2024); and Wave 4 (December 15, 2023 – May 8, 2025). The qualifying longitudinal cohort, defined as completing baseline and either the 6- or 12-month surveys, was 3,360. A total of 2,049 participants completed all four survey waves and are the analytic sample for the current paper (61% of the qualifying cohort).

Results

Latent profile analysis identified seven strength profiles: Despondent (5.5%), Isolated (14.2%), Pessimistic (14.3%), Support-Focused (26.8%), Self-Reliant (8.9%), Multi-faceted Strengths (23.8%), and Resilient Activists (6.3%). Profiles differed in both overall strength levels and configurations across regulatory, interpersonal, meaning making, and environmental domains. Depressive symptoms and self-directed violence were highest in the Despondent and Pessimistic profiles and lowest in the Self-Reliant and Multi-faceted Strengths profiles, with Resilient Activists showing intermediate risk. Despite high strengths, Resilient Activists experienced uniquely elevated SDoH adversity, including discrimination, food insecurity, and structural hardship. In contrast, Self-Reliant and Multi-faceted Strengths profiles showed the lowest SDoH risk across domains. Females, sexual-minority, and gender-minority participants were overrepresented in profiles with poorer mental health and greater adversity and underrepresented in the most robust strength profiles.

Conclusion

Practitioners should incorporate assessment of strengths portfolios when working with young people. Programs to prevent mental health problems, including suicide and self-injury, should include specific strengths-based components across the four domains of the Resilience Portfolio Model – regulatory, interpersonal, meaning making and environmental.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28573-w.

Background

Resilience is the process of achieving wellbeing after adversity and trauma and it began with the study of human development among samples of young children [1, 2]. The goal was to understand how some groups of children achieved outcomes associated with positive functioning even after experiencing a variety of adverse childhood experiences. While initial studies examined factors like dimensions of personality or intelligence that are believed to be relatively stable, current research uses multidimensional models that emphasize that a variety of resources and assets across areas of the social ecological model, within the individual; family; peer and other social networks; communities and community structures like workplaces, faith based organizations, and neighborhoods; and within societies [3], may be present in different combinations and contribute to positive indicators of wellbeing [2, 4]. To date, much of this work has focused on one strength or asset at a time (e.g. growth mindset, grit) despite multidimensional resilience theories that emphasize both that the array of strengths, and the combined dose of strengths are what best predict thriving after trauma and adversity [4, 5]. The current study conducted a latent profile analysis of a large national sample of adolescents and young adults who answered self-report questions about an array of assets and resources identified in previous studies. The classes represent different combinations of strengths. Group differences on demographics, social determinants of health (SDoH), and mental health indices were examined across the latent classes identified.

Multidimensional models of resilience

Given continued concerns about the mental health of young people [6], there has been a growing body of research aimed at understanding strengths, resilience, and protective factors. Models of these strengths emphasize multiple and multidimensional protective factors and focus on modifiable variables rather than personality traits [7]. For example, a well cited model for understanding strengths in adolescence and designing prevention strategies that promote well-being is Positive Youth Development which emphasizes five key developmental strengths for this age group – caring, connection, confidence, competence, and character [8]. While there is empirical support for the model, it focuses mainly on intraindividual variables. Ungar’s multisystemic resilience reminds us to look across the social ecology and to take into account global cultural differences [9]. Importantly, these models help us understand how the presence of strengths explains unique, and often highly significant amounts of variation in outcomes above and beyond the well-documented and well-researched impact of adversities like adverse childhood experiences. Studies using a dual factor approach support this [10, 11]. Further, strengths are measured as the presence of competencies rather than simply low levels of risk factors.

The current study is based on one of these multidimensional models - the Resilience Portfolio Model (RPM). It has been used across different community samples and age groups globally. It notes the importance of multiple strengths across multiple domains and emphasizes both the overall quantity of high scores on different strengths variables and also that different specific strengths are linked to different indicators of well-being [7]. The RPM organizes assets (within the individual) and resources (outside of the individual) across four different domains. The regulatory domain includes strengths like emotion regulation, optimism, and hope that help individuals cope with difficult feelings and sooth agitation and disequilibrium so as to modulate experiences of stress and emotion overload. The interpersonal domain encompasses social supports and relationship skills (i.e. communication, conflict management) that help keep connections to social networks healthy and strong. The meaning making domain includes strengths like sense of purpose and also behaviors like community volunteering or activism that help individuals get outside themselves and connect to larger movements and experiences. Finally, a fourth and recently added domain focuses on the environment. This recognizes that exposure to features of the built and natural environments as well as broader policy contexts also affect well-being.

Some empirical studies using the RPM focus on children and young people and support its utility for understanding youth resilience [12–16]. This research highlights key strengths such as optimism, social connections, self-esteem, and sense of purpose and underscore that it is a collection of strengths that helps promote positive youth outcomes. Research to date has established that strengths across the RPM domains explain variance in well-being and mental health outcomes and that some specific strengths emerge as strong correlates. Few studies, however, have created empirical resilience profiles and whether such profiles may be related to social locations young people occupy as defined by social identity communities and contextual SDoH risk.

Current study

To date much research on the RPM has been explored among adults and more research on young people is needed. We sought to take a novel approach to understanding multidimensional resilience by using person-centered latent profile analysis. We hypothesized that latent classes that emerged would represent distinct strength profiles which are further compared to examine demographic, SDoH, and mental health symptom differences in group membership. The main aims were exploratory and descriptive: Are there a set of combinations of strengths, assets, and deficits that can be described in a group of 13–22 year olds? How do these groups differ on exposure to adversities and deficits in SDoH? How do youth with different strength profiles differ on indices of mental health (depression, anxiety, suicide thoughts and behaviors, non-suicidal self-injury), that is – are there certain combinations of strengths that are related to mental health?

Methods

Study population

Project Lift Up is a longitudinal, national study of youth and young adults designed to understand bystander behaviors for self-directed violence. Detailed cohort design and recruitment procedures have been described previously [17–20]. The current manuscript reports a cross-sectional, person-centered analysis using measures collected across the longitudinal study. 4,981 adolescents and young adults aged 13–22 years were recruited online between June 13, 2022 - October 30, 2023. Three additional waves of data were collected, approximately 6 months apart: Wave 2 (collected between January 13, 2023 - February 26, 2024); Wave 3 (June 13, 2023 – November 19, 2024); and Wave 4 (December 15, 2023 – May 8, 2025). The qualifying longitudinal cohort, defined as completing baseline and either the 6- or 12-month surveys, was 3,360. A total of 2,049 participants completed all four survey waves and are the analytic sample for the current paper (61% of the qualifying cohort). Comparisons between those who did and did not complete all four waves among the qualifying cohort are provided in Table 1. The two samples were statistically similar with the exception of sex at birth with more females completing all four waves compared to males. The current analyses are not longitudinal but do use variables from across each wave as different strengths were measured at different times in order to capture a wide range of strengths while reducing survey burden at any one wave. Accordingly, the present study should be interpreted as a cross-sectional latent profile analysis conducted within a longitudinal cohort rather than a longitudinal analysis of change over time.

Table 1.

Demographic characteristics of sub-sample who completed all four survey waves versus those who did not among the qualifying cohort

Characteristic	All participants in the qualifying cohort (N = 3,360)	Did not complete all 4 waves (n = 1,311)	Completed all 4 waves (n = 2,049)	P value
	n (%)	n (%)	n (%)
Age
Adolescent	1867 (55.6)	742 (56.6)	1125 (54.9)	0.33
Young adult	1493 (44.4)	569 (43.4)	924 (45.1)
Sex at birth
Male	1185 (35.3)	540 (41.2)	645 (31.5)	< 0.001
Female	2068 (61.5)	733 (55.9)	1335 (65.1)
Intersex	25 (0.7)	8 (0.6)	17 (0.8)
Prefer not to answer	82 (2.4)	30 (2.3)	52 (2.5)
Sexual identity
Gay	247 (7.3)	92 (7.0)	155 (7.6)	0.55
Lesbian	211 (6.3)	81 (6.2)	130 (6.3)	0.85
Bisexual	777 (23.1)	301 (23.0)	476 (23.2)	0.86
Heterosexual	1386 (41.3)	558 (42.6)	828 (40.4)	0.22
Queer	259 (7.7)	92 (7.0)	167 (8.1)	0.23
Polysexual, omnisexual, sapiosexual or pansexual	255 (7.6)	98 (7.5)	157 (7.7)	0.84
Demisexual	112 (3.3)	45 (3.4)	67 (3.3)	0.80
Asexual	262 (7.8)	97 (7.4)	165 (8.1)	0.49
Questioning/unsure	128 (3.8)	42 (3.2)	86 (4.2)	0.14
Any sexual minority	1947 (57.9)	744 (56.7)	1203 (58.7)	0.26
Gender identity
Cisgender boy	894 (26.6)	402 (30.7)	492 (24.0)	< 0.001
Cisgender girl	1252 (37.3)	438 (33.4)	814 (39.7)	< 0.001
Non-binary	600 (17.9)	233 (17.8)	367 (17.9)	0.92
Transgender girl	147 (4.4)	67 (5.1)	80 (3.9)	0.09
Transgender boy	352 (10.5)	110 (8.4)	242 (11.8)	0.002
Genderqueer	252 (7.5)	98 (7.5)	154 (7.5)	0.97
Gender fluid	242 (7.2)	104 (7.9)	138 (6.7)	0.19
Agender	138 (4.1)	51 (3.9)	87 (4.3)	0.61
Questioning/unsure	204 (6.1)	77 (5.9)	127 (6.2)	0.70
Any gender minority	1206 (35.9)	464 (35.4)	742 (36.2)	0.63
Race and ethnicity
White	2474 (73.6)	972 (74.1)	1502 (73.3)	0.59
Black or African American	399 (11.9)	150 (11.4)	249 (12.1)	0.53
Asian	420 (12.5)	149 (11.4)	271 (13.2)	0.11
Two or more races	310 (9.2)	130 (9.9)	180 (8.8)	0.27
Hispanic or Latino	719 (21.4)	302 (23.0)	417 (20.3)	0.06
Race or ethnic minority	1495 (44.5)	584 (44.5)	911 (44.5)	0.96

Procedures

Participants were recruited through study advertisements on social media, including Facebook and Instagram, which linked them to a secure survey website. These data were collected directly by the study investigators as part of Project Lift Up and do not represent a secondary analysis of an existing dataset. Those who were eligible (i.e., 13–22 years of age, living in the United States, proficiency in English), were then asked to read an assent/consent form and to indicate their willingness to participate in the survey before continuing with the main survey. Ineligible youth were directed to a web page that included links to general resources for youth (e.g., https://youngwomenshealth.org). Participants were given a $15 incentive as an Amazon gift code for completing Waves 1–3 (Wave 3 also had a $5 early completer bonus if they finished the survey in the first 24-hours after receiving their invitation), and those who completed Wave 4 received $20.

To promote a diverse sample that included adequate representation of historically underrepresented populations, demographic quotas were identified. Once the targeted number of youth in a particular group had been achieved (e.g., aged 13–17, cisgender girls), subsequent youth in this group who were otherwise eligible were deemed ineligible. This quota-based recruitment strategy was intended to maximize diversity rather than produce a nationally representative sample. Numerous data check procedures, consistent with recommended practices for online survey research [21], were used to deter and detect fraudulent entries to ensure the quality of the online data, including age verification, reverse lookup of IP addresses and phone numbers, review of names and addresses, time to survey completion, and direct participant outreach, as needed.

The protocol was reviewed and approved by the University of New Hampshire Institutional Review Board. A waiver of caregiver permission was requested and granted for minors because requiring caregiver consent could potentially place youth in situations where their sexual experiences and/or sexual attraction could be unintentionally disclosed to their caregivers. Appropriate mechanisms were in place to protect youth participants, such as localized referrals to mental health supports.

Measures

Latent profile constructs: strengths

A suite of strength-based measures was included across different waves of this study to assess protective psychosocial resources. We organize them here based on the RPM domains.

Regulatory strengths

Emotional regulation, captured at Wave 2, was assessed through the Recovery of Positive Affect scale [14, 22], a six-item measure assessing participants’ ability to rebound from negative emotional states (e.g., “I can still laugh at a joke, even when I’m having a bad day”). Responses on a four-point scale provided a measure of psychological flexibility.

Self-care items were created for the Wave 2 survey and examined things young people do to take care of themselves; items were informed by a review of practitioner-based literature and measures like the Self Care Inventory developed for adults and more focused on health and illness related behaviors [23] but our focus was on things that would be appropriate for young people. Participants were prompted to think about 14 different things people do to take care of themselves in a typical month, including exercise, creative expression, mindfulness practices, and interactions with pets, with space provided to specify additional strategies they personally utilize. Response options ranged from 1 (never) to 5 (always).

Self-compassion was measured at Wave 4 with selected items from the Self-Compassion Scale for Youth developed by Neff and colleagues [24]. This measure included items from the self-kindness subscale in its entirety, one item from the common humanity subscale (with minor wording adjustments), and two mindfulness items. Wording across all items was slightly adapted for accessibility while preserving the original construct definitions. A sample item is: “I am kind and supportive to myself when I’m having a hard time.”

Participants’ hope was assessed in Wave 3 using six items adapted from Krafft and colleagues’ validated hope scale [25]. This scale captured both dispositional and situational hope by asking participants to rate their agreement with statements such as “I feel hopeful” and “Even in difficult times, I am able to remain hopeful.” Response options ranged from 1 = strongly disagree to 6 = strongly agree .

Interpersonal strengths

Social support, assessed at Wave 4, was measured through two validated scales developed by Hamby and colleagues:²⁶ the Social Support Received Scale and the Social Support Seeking Scale. For social support behavior, participants were prompted to think about the last time they were upset about something and then answer five questions about howe true each was about them. For example, “Someone went with me to get some help”. Social support seeking more generally assessed participants’ propensity to seek emotional or instrumental support from their social networks. It consisted of six items; a sample item being: “I feel better when I talk to people about what’s going on”. Response options for both scales ranged from 0 (not true about me) to 3 (mostly true about me).

Social group connections assessed feelings of closeness and support from peer groups with seven items [14] captured at Wave 4. Six of the items were modified from “group or team” to “friends.” One items was created for the current study to ask about feeling safe among friends. Response options ranged from 1 (not true about me) to 4 (mostly true about me).

Environmental strengths

Use of the built and natural environment was measured at Wave 4 assessing access to and use of 18 different community resources, including “parks,” “public transportation (bus, train, subway),” “farmer’s market,” and “hiking trails” [27]. For each resource, participants indicated whether it existed in their area, was accessible, and had been used in the past year. This structure was informed by prior work on community assets and environmental support for youth well-being. [28, 29]. The current analyses utilized a count of different resources they had used in their community.

Meaning making strengths

Participants’ sense of purpose was assessed at Wave 3 using six items [14] such as “My life has a clear sense of purpose” and “My values give my life meaning” were rated on a 4-point scale from 1 = not true about me to 4 = mostly true about me. It is one of the most widely used measures of this domain of the RPM and has most consistently explained significant variance across a number of indicators of wellbeing.

The measure of activism was adapted from the Short Critical Consciousness Scale [30] and was assessed at Wave 2. Items were reduced for space limitation and minor adaptations for other items included simplifying language to better reflect contemporary forms of youth activism. Items captured youths’ engagement in various social justice activities, such as participation in protests, political organizations, and advocacy efforts, over the past year. Sample items include “participated in a civil rights group or organization” and “contacted a public official to tell him/her how you felt about a particular social or political issue.”

Profile comparison constructs

Mental health

Mental health symptomatology was assessed at Wave 4 using eight items from the DSM-5-TR Self-Rated Level 1 Cross-Cutting Symptom Measure for ages 11–17 [31], capturing the frequency of a variety of symptoms (depression and anxiety) during the past two weeks (M = 1.77, SD = 1.0). Suicide ideation and attempt were measured with items from the same measure but at Wave 1, with participants reporting whether they had ever (1) tried to die by suicide (attempt) and (2) had ever thought about killing themselves or dying by suicide (ideation) (response options: yes, no, prefer not to answer). Non-suicidal self-injury (NSSI) was also assessed at Wave 1 by asking participants whether they had ever intentionally engaged in specific forms of self-injury (for example, cutting the skin, hitting oneself). For each behavior reported, participants indicated the number of times it occurred in the past 12 months, and NSSI was operationalized as any lifetime engagement in self-injury [32].

Social determinants of health (Wave 1)

Economic vulnerability was assessed using two items: frequency of not having enough money to pay bills in the past 12 months and frequency of cell phone service being disconnected due to inability to pay in the past 12 months. Participants answered using a 5-point scale ranging from 0 (never) to 4 (always).

Food insecurity was measured using the same scale, assessing how often participants skipped meals or ate less in the past 30 days due to insufficient money for food.

Discrimination was assessed using the Everyday Discrimination Scale, [33] which has been validated in racially and socioeconomically diverse youth samples; participants reported the frequency of unfair or demeaning treatment in daily life (e.g., people act as if they are better than you are), and those endorsing any item identified perceived target identity (e.g., based on race, sexual orientation, disability) via a single mark-all-that-apply follow-up to reduce respondent burden.

Adversity was measured using 12 lifetime items capturing youth- and family-level experiences of economic hardship, family separation, parental incarceration, chronic illness, and substance misuse, adapted from prior work on adverse childhood experiences [34].

Barriers to mental health treatment were assessed using five items adapted from the Barriers to Care Scale [35], simplified for accessibility, with an additional open-ended item allowing participants to report other reasons for not seeking care.

Poor home conditions were measured using eight yes/no items assessing current housing problems (e.g., lead paint or lead pipes, mold, no smoke detectors or they do not work), summed to create a total score [36].

Neighborhood disorder was assessed with 12 youth-adapted items measuring perceived severity of neighborhood conditions (rated as no problem, minor problem, or serious problem), including gun shots at night, homelessness, drugs, and graffiti [37].

Internal consistency for the multi-item measures demonstrated excellent reliability (Cronbach’s α = 0.77–0.93) (Supplemental Table S1).

Demographic characteristics (Wave 1)

Sexual identity was assessed by asking participants which label best represented how they thought of themselves, with response options including gay, lesbian, bisexual, heterosexual, queer, polysexual/omnisexual/sapiosexual/pansexual, demisexual, asexual, two-spirit, not having a sexuality, not using labels, questioning/unsure, or another identity. For analytic purposes, responses were collapsed into five categories: (1) exclusively heterosexual; (2) gay or lesbian; (3) bisexual, queer, polysexual, or demisexual; (4) asexual or no sexuality; and (5) questioning, mostly heterosexual, or not using labels. Gender identity was measured by asking participants to select all terms that described them from a list including cisgender boy/man, transgender boy/man, cisgender girl/woman, transgender girl/woman, nonbinary, genderqueer, genderfluid, pangender, agender, Indigenous or other cultural gender minority identity, questioning or unsure, or another identity. This multiple-response measure was reduced to a single categorical variable for analyses: (1) cisgender, (2) transgender, (3) nonbinary, (4) agender, and (5) questioning or gender variant.

Age was treated as a continuous variable ranging from 13 to 22 years (M = 17.4, SD = 2.2). Sex assigned at birth was reported by the choice of male, female, intersex, or prefer not to answer. Race was assessed using self-identified categories (White, Black or African American, Asian, Hawaiian or Pacific Islander, American Indian or Alaska Native, other, unknown/not reported, or prefer not to answer), and ethnicity was measured separately by self-reported Hispanic or Latino origin (yes/no).

Data analysis

Latent profile analysis (LPA) was used to identify distinct profiles using cross-sectional measures drawn from multiple waves of the longitudinal cohort. All indicator variables were standardized (z-scored) to place constructs on a comparable scale; original means and standard deviations are reported in Supplemental Table 1. Model selection was guided by multiple fit indices (Table 2), including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Consistent Akaike Information Criterion (CAIC), with lower values indicating better fit. Likelihood-based tests were also examined using the Vuong–Lo–Mendell–Rubin adjusted likelihood ratio test (VLMR-LRT), which assesses whether the addition of a profile results in a statistically significant improvement in model fit. Final model selection additionally considered theoretical interpretability and profile size, with a minimum class prevalence of 5%.

Table 2.

Fit statistics for varying size latent class models of strengths

Class	Log likelihood	AIC	BIC	AICc	CAIC	Entropy	VLMR	p> VLMR
1	-28,901.47	57,842.94	57,955.44	57,843.35	57,975.44	---	---	---
2	-27,459.60	54,981.20	55,155.58	54,982.18	55,186.58	0.8117	2,883.74	< 0.001
3	-27,060.72	54,205.44	54,441.69	54,207.24	54,483.69	0.8027	797.76	< 0.001
4	-26,884.54	53,875.08	54,173.21	53,877.95	54,226.21	0.7703	352.35	< 0.001
5	-26,687.51	53,503.01	53,863.02	53,507.21	53,927.02	0.7780	394.07	< 0.001
6	-26,515.36	53,180.71	53,602.60	53,186.49	53,677.60	0.7949	344.30	< 0.001
7	-26,417.05	53,006.11	53,489.87	53,013.74	53,575.87	0.7924	196.60	< 0.001
8	-26,343.55	52,881.11	53,426.75	52,890.85	53,523.75	0.7890	147.00	< 0.001
9	-26,296.39	52,808.78	53,416.29	52,820.91	53,524.29	0.7952	94.33	0.014
10	-26,192.40	52,622.80	53,292.18	52,637.60	53,411.18	0.7644	207.98	< 0.001

AIC Akaike Information Criterion, BIC Bayesian Information Criterion, AICc Corrected Akaike Information Criterion, CAIC Consistent Akaike Information Criterion, VLMR Vuong-Lo-Mendall-Rubin test

Bold values reflect the best fit using these indices

Following profile enumeration, differences across profiles in lifetime and recent mental health symptomatology, SDoH, and demographic characteristics were examined using chi-square tests with post hoc pairwise comparisons. Bonferroni corrections were applied to all pairwise tests to account for multiple comparisons. Missing data were minimal (< 5%) and were conservatively handled by coding missing values as “no” for dichotomous variables and imputing the mean for continuous variables.

Results

Latent profiles of strengths

After consideration of all the indices detailed in Table 2, theoretical significance, and profile sizes, we decided on a 7-profile solution (Table 3, Supplemental Table 2). Profiles consisted of “Despondent” (5.5%, n = 109), “Isolated” (14.2%, n = 299), “Pessimistic” (14.3%, n = 288), “Support Focused” (26.8%, n = 561), “Self-reliant” (8.9%, n = 179), “Multi-faceted Strengths” (23.8%, n = 490), and “Resilient Activists” (6.3%, n = 123). Participants in the “Despondent” profile had very low levels of all strengths measured across RPM domains. Those in the “Isolated” profile had low to moderate strength levels overall but particularly so for social support seeking and support receiving while demonstrating moderate regulatory strengths. Participants in the “Pessimistic” profile had low levels of support seeking but moderate support receipt and social connectedness alongside very low levels of hope, purpose in life and self-compassion. Participants in the “Support Focused” group had moderate levels of most strengths with the exception of support seeking which was high. Participants in the “Self-Reliant” profile had high levels of regulatory and meaning making strengths (hope, purpose in life and emotional regulation) with low/very low levels of support seeking and receipt. Those in the “Multi-faceted Strengths” profile had high levels of all strengths with the exceptions of activism, self-care, and accessing healthy built environment (moderate). Finally, participants in the “Resilient Activist” profile had high levels of all strengths with an extremely high level for activism.

Table 3.

Latent profile construct levels by class

Key for latent marginal mean scores: Extremely high (> 1.4); Very high(1.0 to 1.39); High (0.4 to 0.99); Moderate + (0.11 to 0.39); Moderate (-0.10to 0.10); Moderate – 0.11 to -.39); Low (-0.4 to -0.99); Very low (-1.0 to –1.39); Extremely low (< -1.4)

Differences in mental health symptomatology, SDoH and demographic characteristics across profiles

Several differences in recent depressive symptoms and lifetime self-directed violence were found across profiles (Table 4). Participants with the lowest rates of suicidal thoughts and behaviors, NSSI, and depressive symptoms were in the Self-Reliant and Multi-faceted Strengths profiles while the highest were in the Despondent and Pessimistic profiles. Participants in the Resilient Activist profile fell in between with lower depression scores than the Despondent and Pessimistic groups but higher than the Self-Reliant and Multi-faceted Strengths groups. Personal suicidal thoughts and behaviors followed a similar pattern.

Table 4.

Youth characteristics across classes

Characteristic	Class 1: Despondent 5.5% (n = 109)	Class 2: Isolated 14.2% (n = 299)	Class 3: Pessimistic 14.3% (n = 288)	Class 4: Support Focused 26.8% (n = 561)	Class 5: Self-Reliant 8.9% (n = 179)	Class 6: Multi-faceted Strengths 23.8% (n = 490)	Class 7: Resilient Activists 6.3% (n = 123)	p value
Construct	n (%)/M (SD)	n (%)/M (SD)	n (%)/M (SD)	n (%)/M (SD)	n (%)/M (SD)	n (%)/M (SD)	n (%)/M (SD)
Mental health (lifetime) [n, %]
NSSI	97 (89.0)	239 (79.9)	262 (91.0) b	459 (81.8) c	123 (68.7) a, c, d	298 (60.8) a, b, c, d	95 (77.2) c, f	< 0.001
Suicidal ideation	101 (92.7)	253 (84.6)	266 (92.4)	469 (83.6) c	116 (64.8) a, b, c, d	297 (60.6) a, b, c, d	98 (79.7) e, f	< 0.001
Suicide attempt	40 (36.7)	82 (27.4)	118 (41.0) b	153 (27.3) c	26 (14.5) a, b, c, d	73 (14.9) a, b, c, d	50 (40.7) d, e, f	< 0.001
Mental health (past 30 days) [M, SD]	0.75 (1.0)	0.15 (0.9) a	0.55 (0.9) b	0.10 (0.9) a, c	-0.42 (1.0) a, b, c, d	-0.56 (0.85) a, b, c, d	0.02 (0.8) a, c, e, f	< 0.001
SDoH (lifetime) [M, SD]
Adversity	0.21 (1.0)	-0.001 (1.0)	0.06 (0.9)	-0.01 (1.0)	-0.30 (0.9) a, b, c, d	-0.34 (0.9) a, b, c, d	0.15 (1.0) e, f	< 0.001
Barriers to mental health care	0.34 (1.0)	0.18 (0.9)	0.26 (0.9)	0.07 (0.9)	-0.04 (1.0) a, c	-0.33 (1.0) a, b, c, d, e	0.13 (1.0) f	< 0.001
Discrimination	0.14 (1.0)	0.03 (1.0)	0.16 (0.9)	0.06 (0.9)	-0.14 (1.0) c	-0.28 (1.0) a, b, c, d	0.45 (0.8) b, d, e, f	< 0.001
Neighborhood disorder	-0.06 (0.9)	0.02 (0.9)	0.03 (0.9)	-0.05 (1.0)	-0.10 (0.9)	-0.22 (0.9) b, c, d	0.10 (1.0) f	< 0.001
Poor home conditions	0.12 (1.1)	0.13 (1.1)	0.10 (1.0)	-0.02 (1.0)	-0.21 (0.8) b, c, d	-0.21 (0.8) a, b, c	0.13 (1.1) f	< 0.001
Food insecurity	0.13 (1.1)	0.11 (1.1)	0.06 (1.1)	0.004 (1.0)	-0.25 (0.7) a, b, c	-0.23 (0.8) a, b, c, d	0.28 (1.2) e, f	< 0.001
Phone turned off	0.04 (1.0)	0.02 (1.0)	-0.06 (0.9)	-0.08 (0.8))	-0.08 (0.8)	-0.19 (0.7) b	-0.03 (0.9)	0.03
Economic instability	0.25 (1.1)	-0.02 (0.9)	0.02 (1.0)	-0.08 (0.9) a	-0.17 (0.9) a	-0.24 (0.9) a, b, c	0.25 (1.1) d, e, f	< 0.001
Demographics [n, %]
Female	75 (68.8)	169 (56.5)	211 (73.3) b	381 (67.9) b	109 (60.9)	310 (63.3)	80 (65.0)	0.001
Hispanic	17 (15.6)	80 (26.8)	61 (21.2)	108 (19.3)	43 (24.0)	85 (17.3) b	23 (18.7)	0.03
American Indian or Alaska Native	7 (6.4)	12 (4.0)	9 (3.1)	21 (3.7)	5 (2.8)	11 (2.2)	7 (5.7)	0.287
Asian	12 (11.0)	37 (12.4)	28 (9.7)	81 (14.4)	20 (11.2)	79 (16.1)	14 (11.4)	0.17
Black	15 (13.8)	50 (16.7)	21 (7.3) b	64 (11.4)	31 (17.3) c	47 (9.6)	21 (17.1)	0.001
White	80 (73.4)	205 (68.6)	231 (80.2)	416 (74.1)	125 (69.8)	355 (72.5)	90 (73.2)	0.06
Gender minority	60 (55.1)	108 (36.1) a	178 (61.8) b	216 (38.5) a, c	32 (17.9) a, b, c, d	91 (18.6) a, b, c, d	57 (46.3) c, e, f	< 0.001
Sexual minority	82 (75.2)	187 (62.5)	222 (77.1) b	356 (63.5) c	75 (41.9) a, b, c, d	191 (39.0) a, b, c, d	90 (73.2) e, f	< 0.001
Age [M, SD]	17.4 (2.1)	17.3 (2.1)	17.8 (2.3)	17.5 (2.3)	17.3 (2.4)	17.5 (2.2)	17.9 (2.1)	0.04

^a significantly different from Class 1; ^b significantly different from Class 2; ^c significantly different from Class 3; ^d significantly different from Class 4; ^e significantly different from Class 5; ^f significantly different from Class 6. Pairwise differences use a Bonferroni adjustment for multiple comparisons

The Resilient Activist group had particularly and uniquely high levels of SDoH deficits, including discrimination, food insecurity, and structural (neighborhood disorder and poor home conditions). Barriers to mental health care and adversity were most notable in the Despondent and Pessimistic profiles. Participants in the Self-Reliant and Multi-faceted Strengths groups had the lowest scores on SDoH across all domains with no significant differences between those two groups. Interestingly, the first four groups, all of which have less robust strength profiles (as defined by many assets with scores below average) do not significantly differ on most of the SDoH variables, though these groups do differ on mental health outcomes.

The most salient differences between profiles for demographic characteristics involved sex at birth, sexual identity and gender identity. Here we saw females, sexual minority and gender minority participants reflected the most in the Despondent and Pessimistic groups, followed by the Resilient Activist group. Gender minority and sexual minority participants were the least likely to be part of the Self-Reliant and Multi-faceted Strengths groups in comparison to each other group.

Discussion

The current study used latent profile analysis to describe a variety of strength profiles in a group of adolescents and young adults. The overall finding was that many different latent strength profiles emerged from the data. This supports ideas within the RPM that resilience is multifaceted but also that there is no one size fits all [7]. Rather than identifying a single “resilient” pathway, these findings suggest that adolescents and young adults may achieve relatively positive mental health through multiple constellations of strengths that differ across regulatory, interpersonal, meaning making, and environmental domains. This person-centered approach extends previous variable-centered research by demonstrating how strengths co-occur within individuals, providing a more ecologically valid understanding of resilience during a developmental period characterized by rapid psychological and social change.

Two groups, the Despondent and Pessimistic groups, showed the greatest mental health concerns. The Despondent group scored low/very low across all strengths and may represent a group who have few resilience-related resources. The Pessimistic group was similar, with low/very low levels of several regulatory strengths but interestingly had moderate levels of social support receipt with low support seeking. These two groups were also notable for their low level of the meaning making strength - purpose in life - a variable that emerges consistently in RPM studies as related to numerous indicators of well-being [4, 7, 16, 38]. Both of these groups were also the lowest groups on regulatory assets like emotion regulation and self-compassion. Slightly better off was the Support Focused group who were moderate across regulatory, environmental and meaning making strengths but particularly high on the interpersonal domain resources of seeking and receiving social support. Interestingly, participants in this group did not fare as well on mental health outcomes as participants in the Self-reliant or Multi-faceted Strengths groups that scored high across regulatory, interpersonal and meaning making domains. This fits with previous research that shows social support variables are complex and may reflect that relationships can also create compassion fatigue and other negative challenges along with protective connections [26]. The findings also reinforce that no single strength domain appears sufficient on its own. Instead, youth who demonstrated strengths spanning multiple RPM domains consistently experienced the most favorable mental health outcomes, suggesting that resilience is best understood as the cumulative interaction of complementary resources rather than isolated protective factors. Indeed, it was the three groups that scored high on strengths across several different domains of the RPM that had the best outcomes.

Those high on regulatory and meaning making strengths, the Self-reliant group, but with low social connections and the Multi-faceted strengths groups fared the best on mental health indicators and also reported the lowest levels of SDoH deficits. These are groups who are similar in having fortunate external resources but also demonstrate a range of resilience, with some having strengths across all domains and the Self-reliant group scoring high on regulatory and meaning making domains but not on interpersonal strengths. This too is consistent with the ways in which different collections of strengths, or portfolios, can produce similar outcomes, a foundation of multi-dimensional models of resilience. Importantly, these findings suggest that interventions need not seek to create identical patterns of strengths across all youth. Rather, multiple pathways toward positive adaptation may exist, allowing prevention and intervention efforts to build upon youths’ existing strengths while addressing areas of relative vulnerability.

One of the most interesting groups was the Resilient Activists. Like the Multi-faceted Strengths group, these participants had high levels of strengths across the four RPM domains but were unique in the extremely high levels of activism, a marker of meaning making actions. This group was most likely to include participants who had at least one marginalized social identity. This group also had high levels of SDoH deficits and high adversity burden. Given these risk factors, it is not surprising that they also reported higher mental health concerns. But they were also engaging in social action. This fits with research that finds youth motivated to participate in social justice activities to support their identities, create interpersonal connections and to address social problems that affect their communities [39]. Indeed, SDoH deficits, including lack of resources in their communities due to systemic and historic oppression and discrimination and exposure to adversities, can be a catalyst for activism [40]. Engaging in activism can both contribute to stress but also promote strengths and resilience [41]. These findings underscore that resilience does not imply the absence of adversity or psychological distress. Rather, resilience may coexist with ongoing structural inequities, highlighting the importance of considering broader social and environmental contexts alongside individual strengths when studying youth development.

The current study extends the empirical work on the RPM, which often refers to meaning making strengths like activism but seldom measures it. The current study is also innovative in its inclusion of a self-report measure of the built environment; while it is an individual assessment of use of community resources including green spaces, it furthers our understanding of this new domain [7, 42]. Interestingly, in the current analysis, most groups scored similarly in the moderate range on this resource. The exceptions were the Despondent group, which had low scores on the built/natural environment and the Resilient Activist group, who had high levels of this resource. It raises questions about how high access to strengths-based aspects of the physical environment may work against deleterious effects of high SDoH deficits also reported by this group. More broadly, these findings support growing public health recognition that resilience is shaped not only by individual characteristics but also by opportunities to access supportive community environments and resources.

Implications

There are several research, practice, and policy implications from the current study. Latent profile analysis may be a helpful tool to capture a diverse and complex array of strength profiles among adolescents and young adults. By identifying naturally occurring combinations of strengths rather than focusing on individual protective factors in isolation, person-centered approaches may help researchers and practitioners better understand heterogeneity in resilience and tailor interventions to youths’ unique strengths profiles. Further studies over time could examine the stability of these groups and what kinds of interventions might help move young people in groups like the Despondent, Isolated, and Pessimistic groups into those with Multi-faceted strengths. The current study did not have this strengths analysis as a primary aim and thus strengths were measured at different waves, precluding full use of the longitudinal design. This is an important area for further study.

Practitioners need to do more to assess for a wide variety of strengths and more research is needed to create multi-dimensional clinical assessments of resources and assets across the four RPM domains given that the participants in the current study who were doing best had high scores on a variety of strengths domains. Routine assessment of strength profiles, alongside assessment of adversity, SDoH, and mental health symptoms, may help identify assets that can be leveraged during treatment and prevention rather than focusing exclusively on deficits or psychopathology. Clinical interventions can include a discussion of an individual’s strengths profiles and reminders that resilience is not about one specific asset or an immutable personality trait that one is born with or not. Resilience is something that can be cultivated over time using an array of evidence-based strategies [43].

The findings also have important public health implications. Prevention strategies should move beyond reducing risk to intentionally cultivating regulatory, interpersonal, meaning making, and environmental strengths that together support positive adaptation [44]. Programs targeting youth suicide, self-harm, and other mental health concerns may benefit from combining individual skill-building with efforts to strengthen mentoring relationships, foster purpose and civic engagement, increase access to supportive community spaces, and reduce structural barriers associated with SDoH. Such an approach is consistent with contemporary public health models that recognize resilience as both an individual and contextual process rather than solely an individual characteristic [42, 45].

Limitations

There are some limitations to acknowledge. Not all strengths were measured at the same time and assets measured more proximally to the outcome measures (Waves 3 and 4) may have carried more weight in relation to self-reported mental health symptoms. In addition, lifetime suicidal ideation, suicide attempt, and NSSI were assessed at Wave 1, whereas depressive symptoms and several strengths were assessed at later waves. Consequently, temporal ordering cannot be established, and it is possible that earlier mental health experiences influenced the development of later strengths rather than strengths preceding mental health outcomes.

Although national in scope, the study intentionally oversampled youth across sexual and gender minority groups and racially and ethnically diverse communities to ensure sufficient representation of historically understudied populations. The use of demographic quotas enhanced sample diversity but means the analytic sample was not intended to be nationally representative. Only negative outcomes were assessed – mental health symptoms, suicidal thoughts and behaviors, NSSI. The resilience literature is clear that the presence of positive wellbeing, like sense of wellbeing or physical health or educational achievements should also be studied - measures that capture the presence of flourishing rather than just the absence of symptoms. Several measures were created for the current study such as the built/natural environment measure, the measure of self-care, and the measure of activism. While these were adapted from other work, further research is needed.

Conclusions

The current study supports the RPM and other multi-dimensional models of resilience among adolescents and young adults. Latent profile analysis empirically derived a set of seven strength profiles in the current sample, adding more complexity to important dual-process models of resilience and mental health. Three profiles represented youth whose strength portfolios contained relatively few assets and resources, and these groups also struggled with higher levels of mental distress. On the other hand, there were a set of other profiles, each emphasizing different groups of strength that contributed to lower mental health distress. Practitioners should incorporate assessment of strengths portfolios when working with young people and programs to prevent mental health problems, including suicide and self-injury, should include specific strengths-based components across the four domains of the RPM – regulatory, interpersonal, meaning making and environmental.

Abbreviations

SDoH

Social determinants of health

RPM

Resilience portfolio model

NSSI

Non-suicidal self-injury

Authors’ contributions

KJM conceptualized and designed the work, acquired and analyzed the data, drafted the manuscript, acquired the funding, and managed the project. VB conceptualized and designed the work, drafted the manuscript, and acquired the funding.

Funding

This research was supported by the National Institute of Mental Health of the National Institutes of Health under Award Number R01MH128269. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

Data will be available through the NIH Data Archive upon study completion (July 2027). Earlier access is possible upon reasonable request through the corresponding author.

Declarations

Ethics approval and consent to participate

All data were collected under the approval of the University of New Hampshire’s (UNH) Institutional Review Board [Protocol # IRB-FY2022-5]. We confirm that all methods were carried out in accordance with relevant guidelines and regulations put forth by the UNH IRB. Informed consent (for 18-22 year olds) or assent (for 13-17 year olds) was obtained from all subjects involved in the study. A waiver of caregiver permission was requested and granted for minors because requiring caregiver consent could potentially place youth in situations where their sexual experiences and/or sexual attraction could be unintentionally disclosed to their caregivers. Use of any animal or human data or tissue - Not applicable

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.