Introduction
Rates of mental health problems are increasing dramatically among adolescents and young adults (Lipson et al., Reference Lipson, Lattie and Eisenberg2019; Mojtabai & Olfson, Reference Mojtabai and Olfson2020; Theriault et al., Reference Theriault, Rosenheck and Rhee2020). Mental health problems can profoundly affect young people by eliciting psychological distress, poor physical health, and interpersonal trouble (e.g., Goldstein et al., Reference Goldstein, Dawson, Smith and Grant2012; Rohde et al., Reference Rohde, Lewinsohn, Seeley, Klein, Andrews and Small2007; Thornicroft, Reference Thornicroft2011). Rates of mental health problems are particularly high among youth who have contact with the legal system, with some studies estimating that 50–75% of youth with legal system contact have a diagnosable mental health disorder (Underwood & Washington, Reference Underwood and Washington2016). Foundational studies that examine the factors influencing mental health problems among young people emphasize the importance of stressful life experiences (e.g., Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards and Marks1998; Kessler et al., Reference Kessler, McLaughlin, Green, Gruber, Sampson, Zaslavsky and Angermeyer2010). Legal system contact itself is a significant life stressor (Sugie & Turney, Reference Sugie and Turney2017) and is often only one of many life stressors that youth with legal system involvement experience (Becker & Kerig, Reference Becker and Kerig2011; Charak et al., Reference Charak, Ford, Modrowski and Kerig2019). Two common life stressors that legal system-involved youth experience, that also are associated with the developmental of mental health problems, are parental harshness (PH) and exposure to community violence (ETV) (Abram et al., Reference Abram, Teplin, Charles, Longworth, McClelland and Dulcan2004; Becker & Kerig, Reference Becker and Kerig2011; Dierkhising et al., Reference Dierkhising, Ko, Woods-Jaeger, Briggs, Lee and Pynoos2013; Pane Seifert et al., Reference Pane Seifert, Tunno, Briggs, Hill, Grasso, Adams and Ford2021).
PH is defined by critical, demeaning, hostile, and maltreating (e.g., physical/emotional abuse) parenting that includes verbally and physically aggressive behaviors (Kim et al., Reference Kim, Pears, Fisher, Connelly and Landsverk2010). In general population and at-risk samples, PH predicts mental health problems, including frequent antisocial behavior (Doom et al., Reference Doom, Peckins, Hein, Dotterer, Mitchell, Lopez-Duran and Abelson2022; Kingsbury et al., Reference Kingsbury, Sucha, Manion, Gilman and Colman2020), high levels of depression (Calhoun et al., Reference Calhoun, Ridenour and Fishbein2019), and suicidal ideation (Doom et al., Reference Doom, Peckins, Hein, Dotterer, Mitchell, Lopez-Duran and Abelson2022). In legal system-involved samples, studies show that PH predicts antisocial behavior (Vaughan et al., Reference Vaughan, Frick, Ray, Robertson, Thornton, Wall Myers, Steinberg and Cauffman2021) as well as elevated rates of substance use disorders (Bosk et al., Reference Bosk, Anthony, Folk and Williams-Butler2021). Furthermore, the trajectory of exposure to PH, not just whether the exposure occurred, impacts mental health. For example, Wiggins et al. (Reference Wiggins, Mitchell, Hyde and Monk2015) examined trajectories of PH across early/middle childhood in at-risk youth and found that a trajectory characterized by higher and steeper increases in PH associated with consistently high externalizing and internalizing symptoms whereas a trajectory characterized by high but stable levels of PH associated with high but decreasing externalizing and internalizing symptoms. These results suggest that changes over time in PH have implications for how youth experience mental health problems.
ETV is defined by witnessing or being a victim of violent acts (e.g., shootings, stabbings, robberies) in one’s community, outside the home (DeCou & Lynch, Reference DeCou and Lynch2017). In both general population and legal system-involved samples, ETV relates to greater mental health problems, such as posttraumatic stress disorder (PTSD; Ford et al., Reference Ford, Hartman, Hawke and Chapman2008; Fowler et al., Reference Fowler, Tompsett, Braciszewski, Jacques-Tiura and Baltes2009), externalizing symptoms (Affrunti et al., Reference Affrunti, Suárez and Simpson2018; Phan & Gaylord-Harden, Reference Phan and Gaylord-Harden2022), and frequent and problematic substance use (Löfving-Gupta et al., Reference Löfving-Gupta, Willebrand, Koposov, Blatný, Hrdlička, Schwab-Stone and Ruchkin2018; Udell et al., Reference Udell, Hotton, Emerson and Donenberg2017). Further, ETV is a well-documented risk factor for future violent/aggressive behavior in youth with legal system involvement (Baskin & Sommers, Reference Baskin and Sommers2014; Myers et al., Reference Myers, Salcedo, Frick, Ray, Thornton, Steinberg and Cauffman2018). For youth with legal system involvement, trajectories of ETV, estimated across late adolescence/emerging adulthood, that were characterized by high and increasing or high and stable ETV associated with more mental health problems compared to trajectories with low and stable or decreasing ETV (Baskin & Sommers, Reference Baskin and Sommers2015). Thus, levels of ETV are not static over time, and changes in these levels over time differentially predict mental health problems.
Across studies and samples, PH and ETV appear to separately confer risk for a host of mental health problems at single points in time and across time. However, exposure to multiple stressful life experiences is relatively common (Green et al., Reference Green, Goodman, Krupnick, Corcoran, Petty, Stockton and Stern2000; Kilpatrick et al., Reference Kilpatrick, Resnick, Milanak, Miller, Keyes and Friedman2013; Pane Seifert et al., Reference Pane Seifert, Tunno, Briggs, Hill, Grasso, Adams and Ford2021). Therefore, the co-occurrence of PH and ETV elevates risk for mental health problems beyond any one stressor alone. In particular, youth with legal system contact are likely to experience multiple life stressors that increase the likelihood of future mental health problems across externalizing and internalizing spectrums. Repeated experience of multiple life stressors across late adolescence may represent snares that trap youth in legal system involvement and promote poor mental health in emerging adulthood (McGee et al., Reference McGee, Hayatbakhsh, Bor, Aird, Dean and Najman2015; Moffitt, Reference Moffitt1993). However, it is unclear how these youth experience multiple life stressors simultaneously across late adolescence and how their experience of multiple stressors impacts future mental health functioning.
In both general population and legal system-involved samples, existing research on the co-occurrence of PH and ETV predominately estimates stressful life experiences by collapsing across subtypes. Commonly, researchers create a binary exposure metric (e.g., exposed versus not exposed) for each experience that indicates whether individuals have some form of exposure to each subtype (Lynch & Cicchetti, Reference Lynch and Cicchetti1998; Stevens & Mennen, Reference Stevens, Mennen and Negriff2018; Valentino et al., Reference Valentino, Nuttall, Comas, Borkowski and Akai2012). However, this binary approach does not account for any features of the experience (e.g., frequency, change over time; see Lacey & Minnis, Reference Lacey and Minnis2020 for a review). Second, some researchers measure cumulative frequency of exposure to stressful life experiences and collapse across experiences (Huesmann et al., Reference Huesmann, Dubow, Boxer, Bushman, Smith, Docherty and O’Brien2021; McGinnis et al., Reference McGinnis, Sheridan and Copeland2022; Schafer et al., Reference Schafer, Ferraro and Mustillo2011). Cumulative life stressor metrics provide more detailed assessment in that they reflect the number of exposures. However, these cumulative scores frequently are not disaggregated by subtype, making it unclear how specific subtypes combine to increase mental health problems. Finally, in some instances, researchers examine change in one stressor while controlling for another. For example, one recent longitudinal study of trajectories of ETV in at-risk youth also measured PH and examined whether PH was a risk factor for membership in a given trajectory of ETV (Zhao et al., Reference Zhao, Ettekal, Nickerson, Schuetze, Shisler, Godleski and Eiden2022). Thus, the extant literature shows that individuals are often exposed to both PH and ETV and, when measured separately, each experience impacts mental health. However, methodological limitations leave questions regarding 1) how PH and ETV simultaneously change over time in youth with legal system involvement and 2) how simultaneous change over time impacts mental health.
To characterize joint trajectories of PH and ETV and clarify their associations with mental health problems in youth with legal system involvement, first we examined simultaneous fluctuations in the co-occurrence of PH and ETV across middle and late adolescence in a sample of legal system-involved youth. Then, we determined how trajectories of co-occurrence predicted externalizing and internalizing mental health problems three years later during emerging adulthood. For all analyses, we examined the robustness of any relationships by considering additional factors (i.e., demographics, broader environmental factors, individual difference factors) that also relate to PH, ETV, and/or mental health. Based on prior work (Cecil et al., Reference Cecil, Viding, Barker, Guiney and McCrory2014; Estrada et al., Reference Estrada, Gee, Bozic, Cinguina, Joormann and Baskin-Sommers2021; Stevens & Mennen, Reference Stevens, Mennen and Negriff2018), we hypothesized that trajectories characterized by low levels of both subtypes, high levels of one subtype, and high levels of both subtypes would emerge. We further hypothesized that trajectories primarily characterized by higher levels of PH relative to ETV would predict greater dysfunction in internalizing outcomes; trajectories primarily characterized by higher levels of ETV relative to PH would predict more antisocial behavior; and trajectories characterized by high levels of one or both subtypes would promote dysfunction across outcomes.
Method
Participants and procedure
The present study utilized data from the Pathways to Desistance study,Footnote 1 a longitudinal study of serious juvenile offenders located in Phoenix, AZ and Philadelphia, PA (Mulvey et al., Reference Mulvey, Steinberg, Fagan, Cauffman, Piquero, Chassin and Hecker2004; Schubert et al., Reference Schubert, Mulvey, Steinberg, Cauffman, Losoya, Hecker and Knight2004; see also: https://www.pathwaysstudy.pitt.edu/index.html). Participants completed a 4-hour baseline assessment during which they provided information on a wide range of individual and social background factors. Six follow-up assessments were conducted every six months for the three years following the baseline interview; after three years, participants were reinterviewed annually for four years.
Participants were included in the present studyFootnote 2 if they completed baseline and at least four of six 6-month follow-up assessments of PH and ETV (n = 1027; see Supplemental Method Sections 3a and 3b for information on missingness). The six 6-month follow-up assessments were chosen for analysis because a) PH was no longer assessed when participants turned 20 years old and b) uneven amounts of time between assessments would yield uneven opportunities for ETV. All outcomes were assessed three years following the final timepoint that was included in trajectory analyses (i.e., third annual follow-up, six years after baseline). On average, participants were 15.91 years old at baseline (SD = 1.15) and 21.92 years old at the three-year follow-up timepoint (SD = 1.16). Most of our sample identified as Black (42.2%) or Hispanic (32.6%), with the remaining participants identifying as white (21.3%) or another racial group (3.9%; see Table 1 for sample characteristics and descriptive statistics [mean, standard deviation, range]). Cronbach’s alphas for all measures can be accessed at https://www.pathwaysstudy.pitt.edu/codebook/measures.html.
Sample characteristics and descriptive statistics for key study variables
Note. MDD = major depressive disorder. PTSD = posttraumatic stress disorder. a Socioeconomic status was assessed using the Index of Social Position (Hollingshead, Reference Hollingshead1957), where higher values indicate lower socioeconomic status. b Values refer to the mean number of crimes committed by participants in the sample. c Substance dependency terminology is used because participants were assessed using DSM-IV substance use disorder criteria. Min and Max refer to observed scores. Range refers to total possible scores.
Measures
Trajectories
PH (Conger, et al., Reference Conger, Ge, Elder, Lorenz and Simons1994). The Maternal and Paternal Hostility subscales contained 12 items each that documented the frequency of harsh behavior perpetrated by each parent/primary caregiver (e.g., “How often did your mother threaten to hurt you physically?”; “How often did your father push, grab, hit, or shove you?”). Participants rated each item on a 1–4 Likert-type scale (1 = “Always”, 4 = “Never”), where higher scores indicated harsher behaviorFootnote 3 . A PH score (PH score) was calculated using the average (i.e., mean) of the Maternal and Paternal Hostility subscale scores. If participants only had one subscale score, that one score was used as the PH score. Rates of PH in this sample were comparable to those found in other legal system-involved youth samples (e.g., Vaughan et al., Reference Vaughan, Frick, Ray, Robertson, Thornton, Wall Myers, Steinberg and Cauffman2021).
ETV (Selner-O’Hagan et al., Reference Selner-O’Hagan, Kindlon, Buka, Raudenbush and Earls1998). Items documented seven witnessed (e.g., “Have you ever seen someone else get killed as a result of violence, like being shot, stabbed, or beaten to death?”), six victimization (e.g., “Have you ever been beaten up, mugged, or seriously threatened by another person?”), and four exposure to death (e.g., “Have you found a dead body?”) experiences. Participants responded to each item based on a dichotomous choice (yes/no), and a total score was calculated using the sum of all items (ETV score). Higher scores indicated a greater number of exposures to community violence. Rates of ETV in this sample were comparable to rates of ETV in other legal system-involved and psychiatrically hospitalized youth samples (Muller et al., Reference Muller, Goebel-Fabbri, Diamond and Dinklage2000; Myers et al., Reference Myers, Salcedo, Frick, Ray, Thornton, Steinberg and Cauffman2018).
Trajectory covariates. Baseline scores for PH, ETV, and biological sex (male/female) were included as covariates in trajectory analyses to account for exposure to PH and ETV prior to trajectory estimation (VanderWeele et al., Reference VanderWeele, Mathur and Chen2020; see Supplemental Method Section 8). Biological sex was included to account for the disproportionate number of male participants in the study (Table 1).
Mental health outcomes
Externalizing: antisocial behavior. The Self-Reported OffendingFootnote 4 (Huizinga et al., Reference Huizinga, Esbensen and Weiher1991) scale was used to measure 13 nonviolent (e.g., destroying property, entering buildings to steal, selling drugs) and nine violent (e.g., carjacking someone, shooting someone, being in a fight) crimes. Participants were asked if they had committed any of these crimes and, if yes, reported how many times they engaged in the crime during in the past year. Total frequency scores for nonviolent and violent crimes were calculated based on how many times participants engaged in each crime. Higher scores indicated more frequent nonviolent/violent crime.
Externalizing: substance use. The Substance Use/Abuse Inventory (Chassin et al., Reference Chassin, Rogosch and Barrera1991) was used to measure participants’ substance use and dependence. Data were collected when substance use disorders were classified into substance abuse and substance dependence using DSM-IV criteria (First et al., Reference First, Spitzer, Gibbon and Williams1997). The Substance Use subscale had participants report how many of 10 different types of illegal substances they used in the past year. Higher scores indicated more substances tried. The Substance Dependency subscale assessed past-year dependency symptoms for drug and/or alcohol use (e.g., “Have you had any problems or arguments with family or friends because of your alcohol or drug use?”). Higher scores indicated greater dependency on substances. Substance use and substance dependency were moderately correlated (r = 0.52, p < .001).
Internalizing: symptoms. The Brief Symptom Inventory (Derogatis & Melisaratos, Reference Derogatis and Melisaratos1983) was used to assess depression and anxiety symptoms. The Depression subscale consisted of six items assessing the extent to which participants were bothered by symptoms of depression in the past week (e.g., “Feeling no interest in things”). The Anxiety subscale consisted of six items measuring the extent to which participants were bothered by physiological symptoms of anxiety in the past week (e.g., “Feeling tense or keyed up”). For both subscales, participants responded using a 0–4 Likert-type scale (0 = “Not at all”, 4 = “Extremely”), where higher scores indicated greater depression/anxiety symptoms. For both subscales, scores consisted of the mean of each of the six items in that subscale.
Internalizing: diagnoses. The Composite International Diagnostic Interview (World Health Organization, 1990), a structured clinical interview designed to assess mental disorders based on DSM-IV criteria, was used to assess past-year major depressive disorder (MDD) and PTSD. For PTSD, it is notable that at the time of data collection PTSD was considered an anxiety disorder under DSM-IV, and the anxiety symptoms measure overlapped with the PTSD reexperiencing (e.g., intrusive, distressing thoughts) and hyperarousal symptom (e.g., hypervigilance) clusters.
Mental health outcome model covariates. Baseline scores for age, biological sex, racial group membership, parental socioeconomic status (Index of Social Position; Hollingshead, Reference Hollingshead1957), neighborhood conditions (Neighborhood Conditions Measure; Sampson & Raudenbush, Reference Sampson and Raudenbush1999), early behavioral problems (Early Behavioral Problems before age 11), the baseline level of the mental health problem of interest, and study site location (Philadelphia, PA or Phoenix, AZ) were included as covariates in all regression analyses (see Supplemental Method for more details on covariate measures and selection).
Data analysis
Trajectories
Group-based multi-trajectory modeling (GBTM) was used to identify joint trajectories of PH and ETV. GBTM is a form of latent class growth analysis that identifies subgroups of individuals who show similar patterns of change over time on multiple variables simultaneously (Nagin et al., Reference Nagin, Jones, Passos and Tremblay2018). Unlike the dual trajectory approach, which estimates the association between two variables by measuring probability of a trajectory for one variable given membership in a specific trajectory for the second variable, GBTM examines the association between two variables by defining trajectory groups based on patterns of change for both variables simultaneously. Analyses were conducted using SAS software, version 9.4 using the PROC TRAJ procedure with MULTGROUPS option (Jones et al., Reference Jones, Nagin and Roeder2001; Nagin et al., Reference Nagin, Jones, Passos and Tremblay2018). Models simultaneously considered PH and ETV scores for the six 6-month follow-up assessments. We followed guidelines for reporting on latent trajectory modeling (Supplemental Method).
Following recommendations from Nagin et al. (Reference Nagin, Jones, Passos and Tremblay2018), first we estimated trajectories for each indicator (i.e., PH and ETV scores) separately (see Supplemental Method Section 4 for information on PH and ETV distributions). Model solutions found in the individually estimated trajectories were used to inform subsequent GBTM analyses using both indicators. We evaluated 1- through 6-trajectory models, and intercept, linear, quadratic, and cubic trajectory shapes were considered. Model fit was compared using the Bayesian Information Criterion, where smaller values indicated a greater model fit; log likelihood, where smaller values indicated a greater model fit; average posterior probability, where values above .7 were considered adequate; odds of correct classification, where values above 5 were considered adequate; and the ratio of the probability of trajectory membership to the proportion of participants assigned to that trajectory, where values closer to 1 indicated a better fit (Nagin, Reference Nagin2005). Trajectory size (i.e., greater than 5%) and theoretical coherence of the trajectories also were considered.
Mental health outcomes
Three separate negative binomial regression analysesFootnote 5 were used to determine if trajectory membership predicted nonviolent crime frequency, violent crime frequency, and substance dependency three years after the last assessment used in trajectory analyses. One Poisson regression analysis was used to determine if trajectory membership predicted substance use three years after the last assessment used in trajectory analyses. Two separate linear regression analyses were used to determine if trajectory membership predicted depression and anxiety symptoms three years after the last assessment used in trajectory analyses. Two separate binomial logistic regression analyses were used to determine if trajectory membership predicted MDD and PTSD diagnostic status three years after the last assessment used in trajectory analyses. Casewise deletion was used to address missingness (see Table 1 for sample size by mental health outcome). For all regression analyses, the group with the lowest PH and ETV scores served as the reference group. Then, post hoc simple contrasts were conducted to compare trajectories that significantly differed from the reference group to each other. Analyses were conducted using the MASS and emmeans packages in R (Lenth & Lenth, Reference Lenth and Lenth2018; Ripley et al., Reference Ripley, Venables, Bates, Hornik, Gebhardt, Firth and Ripley2013). Bonferroni correction based on the number of variables in each mental health subdomain (i.e., antisocial behavior, substance use, internalizing symptoms, internalizing diagnoses) was applied.
Supplemental analyses
To understand how sociodemographic, environmental, and individual difference factors impacted the likelihood of trajectory membership, we used multinomial logistic regression analyses to determine how baseline age, racial group membership, biological sex, socioeconomic status, neighborhood conditions, and early behavioral problems predicted trajectory membership. Full analytic methods and results are presented in the Supplemental Material.
Results
Trajectories
Model fit indices suggested that a four-trajectory solution optimally characterized the sample (Figure 1, Table 2). Further supporting selection of this solution, the four-trajectory solution yielded trajectories that were theoretically coherent and conceptually distinct (see Supplemental Figure 2). The first trajectory was characterized by low and stable levels of PH and ETV (trajectory 1 [Low]; 26.2% of the sample). For the first trajectory, both PH and ETV were best fit using an intercept model, reflecting stability over time.Footnote 6 The second trajectory was characterized by moderate and decreasing levels of PH and ETV (trajectory 2 [Moderate and Decreasing]; 44.1% of the sample). Both PH and ETV were best fit using linear models, reflecting linear change over time. The third trajectory was marked by moderate and stable levels of PH but high and stable levels of ETV (trajectory 3 [Moderate PH/High ETV]; 16.6% of the sample), and both PH and ETV were best fit with an intercept model. The fourth trajectory reflected high and stable levels of PH but moderate and stable levels of ETV (trajectory 4 [High PH/Moderate ETV]; 13.1% of the sample), with both experiences best characterized by an intercept model. See Supplemental Table 5 for model parameter estimates.
Group trajectory estimates for the optimal model solution. A four-trajectory model optimally characterized the data. The first trajectory (green) represented low and stable levels of both parental harshness and exposure to community violence. The second trajectory (orange) displayed moderate and decreasing parental harshness and exposure to community violence. The third (purple) showed moderate and stable levels of parental harshness but high and stable exposure to community violence. The fourth trajectory (red) was characterized by high and stable parental harshness but moderate and stable exposure to community violence. Error bands represent 95% confidence intervals.
Regression analysis results
Note: Italics denote that trajectory membership significantly predicted the outcome of interest compared to trajectory 1 (Low). MDD = major depressive disorder. PTSD = posttraumatic stress disorder.
Mental health outcomes
Membership in trajectory 3 (Moderate PH/High ETV) predicted greater nonviolent crime frequency three years later compared to trajectory 1 (Low) and was the only trajectory that differed from the Low reference group. Membership in trajectories 2 (Moderate and Decreasing), 3 (Moderate PH/High ETV), and 4 (High PH/Moderate ETV) predicted greater violent crime frequency, substance use, and substance dependency three years later compared to trajectory 1. No post hoc simple contrast comparisons, which compared trajectories that significantly differed from the reference group to each other, were significant for any variable. Membership in trajectories 3 (Moderate PH/High ETV) and 4 (High PH/Moderate ETV) each predicted greater depression symptoms. Only membership in trajectory 4 (High PH/Moderate ETV) predicted a higher likelihood of having an MDD diagnosis compared to trajectory 1 and was the only trajectory that differed from the Low reference group. Membership in trajectory 3 (Moderate PH/High ETV) predicted greater anxiety symptoms compared to trajectory 1 and was the only trajectory that differed from the Low reference group. PTSD diagnostic status was not significantly predicted by membership in any trajectory with PH/ETV elevations relative to trajectory 1 (Table 2, Figure 2).Footnote 7 , Footnote 8
Mental health problems three years later by trajectory. Membership in trajectories with PH/ETV elevations significantly predicted (A) externalizing and (B) internalizing mental health problems three years later compared to the Low reference group. Trajectory 2 (Moderate and Decreasing; orange) membership predicted violent crime, substance use, and substance dependency. Trajectory 3 (Moderate PH/High ETV; purple) membership predicted nonviolent and violent crime, substance use, substance dependency, depression symptoms, and anxiety symptoms. Trajectory 4 (High PH/Moderate ETV; red) membership predicted violent crime, substance use, substance dependency, depression symptoms, and MDD diagnosis. Error bars represent standard errors. Asterisk denotes that trajectory membership significantly predicted the outcome compared to trajectory 1 (Low).
Discussion
The present study characterized joint trajectories of PH and ETV in a legal system-involved sample across late adolescence and used trajectory membership to predict mental health outcomes in emerging adulthood. Four trajectories with varying combinations of PH and ETV were identified. Most trajectories were characterized primarily by stability of exposure across time, with one trajectory showing decreases in both PH and ETV across late adolescence. Further, trajectory membership predicted some common and some distinct mental health problems three years later in emerging adulthood. For externalizing outcomes, all trajectories with elevations in PH or ETV predicted more violent crime, substance use, and substance dependency relative to the Low reference group. The Moderate PH/High ETV trajectory predicted more nonviolent crime compared to, and was the only trajectory to significantly differ from, the Low reference group. For internalizing outcomes, distinct associations between trajectories and specific symptom clusters were identified. Both the Moderate PH/High ETV and High PH/Moderate ETV groups predicted greater depression symptoms. However, only the High PH/Moderate ETV group predicted greater likelihood of having MDD and significantly differed from the Low group, and only the Moderate PH/High ETV group predicted higher anxiety symptoms and significantly differed from the Low group. These results suggest that meaningful variation in trajectories of PH and ETV exists and that this variation designates greater risk for some, but not all, mental health problems.
Characteristics of joint trajectories
Our results indicate that features of exposure to stressful life experiences capture important variability in the lived experience. Specifically, the amount of exposure to PH and ETV in late adolescence differed across trajectories. Differences in the amount of exposure across “exposed” groups reinforces the importance of assessment methods that account for amount of exposure.
Another relevant feature of exposure to stressful life experiences includes the pattern of exposure over time. Most trajectories were characterized by stability in one or both experiences across late adolescence. It is possible that the stability of stressful life experiences in this study reflects the inequitable social structures in the United States that limit upward mobility and prevent people from leaving environments that are rife with family and community violence (Chetty et al., Reference Chetty, Friedman, Hendren, Jones and Porter2018). On average, youth in the Pathways to Desistence study resided in environments characterized by moderate to high levels of disadvantage. People who live in environments characterized by disadvantage have limited opportunities to relocate to places where resources may be higher and community violence less prevalent (Warner & Fowler, Reference Warner and Fowler2003). Remaining in a disadvantaged neighborhood likely results in continued, chronic ETV. Moreover, the hardship of living in these neighborhoods may strain parents’ ability to engage skillfully with their children (Barajas-Gonzalez & Brooks-Gunn, Reference Barajas-Gonzalez and Brooks-Gunn2014), contributing to the maintenance of PH. Ultimately, limited abilities to change one’s context may account for the stability in both experiences of stressful life experiences across trajectories.
Further, certain individuals, namely those who identify as Black, are disproportionately burdened by inequitable social structures (e.g., neighborhood disadvantage, legal system contact) that limit social mobility (Brunson & Weitzer, Reference Brunson and Weitzer2009; Swisher et al., Reference Swisher, Kuhl and Chavez2013). This is reflected in the present sample such that Black (and Hispanic) youth are overrepresented in the sample as compared to the general population. However, racial group membership did not differentially predict trajectory membership. More research is needed to explore the ways in which structural and cultural inequalities influence the amount and pattern of change over time in exposure to stressful life experiences (Neblett, Reference Neblett2019; Rucker & Richeson, Reference Rucker and Richeson2021).
That said, one group, the Moderate and Decreasing trajectory, was characterized by decreasing levels of both subtypes of stressful life experiences. Compared to trajectories characterized by stability, this trajectory demonstrated significantly lower amounts of initial ETV. Though youth in this trajectory also are likely to reside in disadvantaged neighborhoods, their qualitative experience of that neighborhood may differ in terms of their ETV. Lower community violence exposure may allow youth greater cognitive capacity to process their current living situation (Sharkey, Reference Sharkey2010). Alternatively, lower community violence exposure may reflect the presence of neighborhood-level protective factors that buffer against violence (e.g., higher collective efficacy, more neighborhood organizations; Gardner & Brooks-Gunn, Reference Gardner and Brooks-Gunn2009; Sampson et al., Reference Sampson, Raudenbush and Earls1997) and increase ability to mitigate exposure to stressful life experiences over time.
Mental health outcomes predicted by joint trajectory membership
Investigation into relative amounts and patterns of change over time in stressful life experiences also revealed that how youth experienced this stress meaningfully predicted future mental health problems. Our results add to a small body of literature suggesting that specific combinations of stressful life experiences are especially potent predictors of negative mental health outcomes (e.g., Briggs et al., Reference Briggs, Amaya-Jackson, Putnam and Putnam2021). Further, the present study contributes to evidence suggesting that nuance in the relationships between combinations of stressful life experiences and mental health emerges when the relative amount of each stressful life experience is considered. Though some studies suggest that there is a dose-response relationship between the experience of multiple stressful life events and negative mental health outcomes (e.g., Chapman et al., Reference Chapman, Whitfield, Felitti, Dube, Edwards and Anda2004), results from the present study indicate that subgroups characterized by high levels of exposure to one stressful life event, in the context of moderate exposure to another, often show the strongest relationships to severe mental health outcomes (Estrada et al., Reference Estrada, Gee, Bozic, Cinguina, Joormann and Baskin-Sommers2021). Our results underscore the importance of understanding the subtype(s) and amount of exposure to multiple stressful life experiences when characterizing the relationships among life stressors and mental health outcomes.
For example, in terms of antisocial behavior, we found that only membership in one trajectory, Moderate PH/High ETV, predicted higher nonviolent crime three years later relative to the Low reference group. As hypothesized, our findings are consistent with previous studies documenting that moderate amounts of family stress in the context of high ETV is an especially potent combination for promoting nonviolent crime (Estrada et al., Reference Estrada, Gee, Bozic, Cinguina, Joormann and Baskin-Sommers2021). Nonviolent crime, which consists of behaviors including theft, selling drugs, and destroying property, can be motivated by a host of factors including one’s ability to regulate behavior or a desire to gain resources (Gardner et al., Reference Gardner, Dishion and Connell2008). In fact, PH, even at moderate levels, is linked to lower self-regulation (Hay et al., Reference Hay, Meldrum, Widdowson and Piquero2017), and ETV is common in neighborhoods with the fewest resources (Gibson et al., Reference Gibson, Morris and Beaver2009). Therefore, it is possible that, for youth in the Moderate PH/High ETV trajectory, their experience of PH contributed to issues in self-regulation that, in the context of a chronically disadvantaged environment, promoted nonviolent crime as means of resource acquisition. In contrast, violent crime was common across all three trajectories with elevations in stressful life experiences. Our findings suggest that the experience of violence across multiple contexts, in addition to the amount of exposure to violence within each context, may be especially relevant to the promotion of future violent behavior (Turner et al., Reference Turner, Shattuck, Finkelhor and Hamby2016). It is possible that youth see violence in one context that is replicated and extended to another, and consequently, they reenact these exemplars of violence in their own behavior. Taken together, these findings reinforce the notion that nonviolent and violent crimes may have different etiologies and therefore interventions should be tailored directly to the type of crime (Kalvin & Bierman, Reference Kalvin and Bierman2017).
For substance use, membership in all trajectories with elevations in stressful life experiences predicted future substance use and problematic substance use compared to the Low group. Our results add to a body of literature documenting the robust associations among various subtypes of life stress and substance use (Bosk et al., Reference Bosk, Anthony, Folk and Williams-Butler2021; Leza et al., Reference Leza, Siria, López-Goñi and Fernandez-Montalvo2021). Experience of life stressors, regardless of subtype and/or amount of exposure, may elevate psychological distress (e.g., Wilson & Rosenthal, Reference Wilson and Rosenthal2003), which youth may seek to alleviate via substances. Therefore, substance use may serve as a form of self-medication that temporarily quells psychological distress but, over time, worsens mental health functioning (Garland et al., Reference Garland, Pettus-Davis and Howard2013).
In terms of internalizing, specific outcomes were differentially predicted by trajectory membership. Though trajectories with high levels of PH or ETV predicted greater depression symptoms, only the trajectory predominated by high levels of PH predicted MDD compared to the Low group. However, the trajectory predominated by high levels of ETV predicted anxiety symptoms when compared to the Low reference group. As hypothesized, the family/caregiving environment appears particularly important for promoting depression that is severe enough to reach diagnostic threshold (LeMoult et al., Reference LeMoult, Humphreys, Tracy, Hoffmeister, Ip and Gotlib2020). Youth whose caregivers are chronically harsh, demeaning, and abusive may feel shame/guilt because they think that their caregivers perceive them as worthy of this harsh behavior (Sekowski et al., Reference Sekowski, Gambin, Cudo, Wozniak-Prus, Penner, Fonagy and Sharp2020). Youth who experience chronic PH, especially in the context of moderate amounts of strain resulting from community violence, may internalize shame/guilt that ultimately produces low positive affect and higher rates of MDD (Clark & Watson, Reference Clark and Watson1991).
In contrast, and contrary to hypotheses, the trajectory characterized by the highest levels of community violence predicted greater (physiological) anxiety symptoms compared to the Low group. In the context of high levels of ETV, it is possible that youth are unsure of when they may experience community violence and whether they were the intended target of community violence, and consequently, constantly fear for their own safety and that of loved ones (Rosen et al., Reference Rosen, Handley, Cicchetti and Rogosch2018). Chronic experience of community violence has been linked to physiological hyperarousal (Estrada et al., Reference Estrada, Richards, Gee and Baskin-Sommers2020) and somatic symptoms of anxiety (Lopez-Tamayo et al., Reference Lopez-Tamayo, Suarez, Simpson and Volpe2022). Conversely, other symptoms of anxiety (e.g., worry, heightened self-monitoring, which were not measured in the present study) may be more strongly associated with PH (Brooker & Buss, Reference Brooker and Buss2014; Gallagher & Cartwright-Hatton, Reference Gallagher and Cartwright-Hatton2008). Critical, demeaning, and punitive parenting may cause youth to monitor and worry about their own behavior in order to avoid future harsh parenting (Brooker & Buss, Reference Brooker and Buss2014). Consequently, PH may manifest in greater cognitive symptoms of anxiety. Researchers may consider incorporating measures of anxiety that capture multiple symptom clusters to refine understanding of the relationships among subtypes of stressful life experiences and anxiety. Additionally, given the high levels of overlap between depression and anxiety (e.g., Kaufman & Charney, Reference Kaufman and Charney2000), researchers can consider examining how stressful life experiences contribute to comorbid depression and anxiety symptoms.
Limitations, future directions, and clinical implications
Several limitations should be noted. First, PH and ETV were only assessed between the ages of 16–19. It is presumed that family systems are more important earlier in life and community systems are more important as youth age into adolescence and emerging adulthood. However, family and community systems are both likely important to some degree across all stages of development (Finan et al., Reference Finan, Ohannessian and Gordon2018; Schaefer et al., Reference Schaefer, Cheng and Dunn2022). Our results reinforce the need for comprehensive measurement of PH and ETV in the same individuals across multiple periods of development to understand not only how fluctuations in stressful life experiences shape future mental health but also how the timing of those fluctuations impacts mental health.
Second, the variables provided did not differentiate in terms of type or severity of PH or ETV. Additionally, our included sample significantly differed from the overall Pathways to Desistance data set such that the included sample showed significantly lower levels of ETV. Trajectories of co-occurrence may differ when accounting for experiences that are more severe in nature. Further, we were unable to examine additional subtypes (e.g., other Adverse Childhood Experiences [ACEs], structural environmental stressors; Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards and Marks1998) or dimensions (e.g., threat/deprivation; McLaughlin et al., Reference McLaughlin, Sheridan and Lambert2014) of stressful life experiences that may be important for understanding trajectories of co-occurrence and their consequences. Moreover, our community violence measure did not specify who the perpetrator was or how frequently exposures occurred. Future work should consider examining multiple types and dimensions of stressful life experiences to clarify which subtypes and/or dimensions of life stressors are most relevant for understanding how life stress promotes future dysfunction.
Third, trajectory analyses inherently simplify the amount of individual variability captured to identify common amounts of change over time on multiple variables. While this approach is beneficial for designating homogenous groups, there can be variability within groups that is unexamined. Future work can consider alternative methods of trajectory specification to model greater individual variability (e.g., latent growth mixture modeling; see Supplemental Method sections 6a and 6b for more information). Future work also can consider alternate ways of modeling externalizing and internalizing problems (e.g., latent variable modeling; Muthén & Asparouhov, Reference Muthén and Asparouhov2002).
Finally, our sample was limited to youth with legal system contact. Because these youth often experience higher rates of some stressful environmental experiences, it is possible that trajectories of stressful life experiences in this sample may differ compared to those found in samples without legal system contact. Relatedly, our sample included a relatively small number of female participants. Researchers may consider examining joint trajectories of multiple stressful life experiences in non-legal system-involved youth and in samples with a greater proportion of female youth to understand how these findings generalize to other individuals.
Despite these limitations, the present findings have implications for clinical practice with youth who have legal system involvement. Clinicians working with this population will likely benefit from comprehensive assessment that includes documentation of 1) the subtypes and amounts of life stressors experienced and 2) contextual factors that may influence clients’ functioning. For clinicians to accurately understand their clients’ life experiences and conceptualize the causes of the distress that bring them into care, an approach that includes robust assessment of multiple subtypes of stressful life experiences is needed in day-to-day practice (Forkey et al., Reference Forkey, Szilagyi, Kelly, Duffee, Springer, Fortin and Ochs2021). Clinical assessments that use binary metrics to code exposure to stressful life experiences may fail to accurately capture nuance in the lived experience that has implications for the types of mental health distress that youth are likely to experience. Further, assessment of youths’ family and neighborhood characteristics will help clinicians situate them in the structures that may influence their functioning and the applicability of certain interventions (Berkes & Ross, Reference Berkes and Ross2013; Naeem, Reference Naeem2019). Ultimately, comprehensive assessment will allow clinicians to tailor interventions to youths’ lived experience (Baskin-Sommers et al., Reference Baskin-Sommers, Chang, Estrada and Chan2022).
In conclusion, we show that how individuals experience multiple forms of life stressors, specifically PH and ETV, varies across adolescence. Further, this variability in trajectories of stressful life experiences in adolescence predicts externalizing and internalizing mental health problems such that not all trajectories with elevations equally predicted dysfunction across mental health domains. Ultimately, measurement of various stressful life experiences simultaneously across multiple developmental time periods is needed to draw accurate conclusions about the driver of mental health problems and promote the creation of precise interventions tailored to an individual’s lived experience with life stressors.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0954579422001420.
Acknowledgements
None.
Funding statement
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Conflicts of interest
The authors declare no conflicts of interest.
Open access
