The impact of childhood trauma on substance use trajectories from adolescence to adulthood: Findings from a longitudinal Hispanic cohort study.

Abstract

Background:

Adverse Childhood Experiences (ACE) are associated with substance use in adolescence and adulthood. However, there is a lack of longitudinal research examining the effect of ACE on substance use trajectories from adolescence through emerging adulthood.

Objective:

This study examined the role of ACE in substance use trajectories among Hispanic emerging adults.

Participants:

We surveyed a cohort of Hispanic adolescents (n=1,399) in Southern California across eight survey waves (beginning in 9^th grade and continuing through emerging adulthood).

Methods:

Growth curve models were used to examine the effect of ACE on past 30-day cigarette, marijuana, and alcohol use over seven time points, and an interaction term of ACE*time was included to investigate the cross-level effect of ACE.

Results:

ACE was a significant predictor at 9^th grade across all substances. Every additional ACE was associated with significantly higher past 30-day cigarette use (β=0.05, 95%CI=0.01, 0.10), marijuana use, (β=0.15, 95%CI=0.06, 0.25) and alcohol use (β=0.14, 95%CI=0.06, 0.21). Across all models, cross level interactions between ACE and time indicated that young adults exposed to more ACE experience significantly steeper inclining trajectories of 30-day cigarette use (β=0.05, 95%CI=0.02, 0.68), marijuana use (β=0.07, 95%CI= 0.03, 0.11), and alcohol use (β=0.02, 95%CI=0.02, 0.68) than young adults with fewer ACE.

Conclusion:

ACE continue to have an impact on substance use trends through emerging adulthood. Results highlight the graded effect of ACE on substance use during and beyond adolescence and illustrate that ACE exposure is linked to an escalation of substance use frequency.

Introduction

Substance use in adolescence

Although adolescent substance use prevalence has decreased over the past decade, by the time youth graduate from high school 60% will have used alcohol in the past year, 36% will have used marijuana, and 24% will have tried a cigarette (Johnston et al., 2019). The initiation of substance use during adolescence is of critical concern given the impact of substances on neurological development and social adjustment, the long-term negative consequences for health and wellbeing, and increased risk for substance use disorders in later life (Bauman & Phongsavan, 1999; National Institute on Drug Abuse, 2007; Sussman et al., 2008). Two of the leading causes of preventable disease and mortality, cigarette smoking and alcohol dependence, are typically initiated during adolescence (Gilpin et al., 1999; Reidpath et al., 2014; World Health Organization, 2017) due to the confluence of the cognitive, physiological, and psychological changes that occur during this developmental period along with the growing interest in risk taking behaviors and susceptibility to peer influence (Bauman & Phongsavan, 1999; National Institute on Drug Abuse, 2007).

Substance use in emerging adulthood

The aforementioned risks extend into early adulthood as the brain continues to develop well into young adulthood and young people experience greater independence. Emerging adulthood, a transitional developmental period between ages 18 and 25, is characterized as a time of identity exploration, transition, and development (Arnett, 2005) when individuals seek out novel experiences and grapple with adult value systems and responsibilities (Arnett, 2005). Compared to other age categories, emerging adults have the highest prevalence of substance use, behaviors that jeopardize ongoing development and challenge the adoption of adult roles and obligations (Andrews & Westling, 2016; Arnett, 2005). The combination of developmental stressors and newly acquired freedom, peer pressure, decision-making skills, and high disinhibition likely account for a proportion of the increases in substance use during this period (Andrews & Westling, 2016). The risk behaviors adopted during emerging adulthood can carry forward and set the stage for life-long patterns of behaviors that threaten work force participation, educational achievement, and family functioning (Brener et al., 1999; Forster et al., 2018; Nelson et al., 2008).

Substance use in Hispanic populations

Generally, the prevalence of substance misuse among Hispanic populations is similar to non-Hispanic White counterparts with the exception of higher rates of alcohol dependence (Alvarez et al., 2007; Substance Abuse and Mental Health Services Administration, 2019). However, Hispanic populations in the U.S. experience more substance use related negative consequences, including poorer treatment outcomes and lower utilization of these resources compared to non-Hispanic populations, leading the Substance Abuse and Mental Health Services Administration to identify Hispanics as a priority population for prevention and treatment (Alvarez et al., 2007; Substance Abuse and Mental Health Services Administration, 2012). Research exploring substance use behaviors from adolescence through emerging adulthood in Hispanic populations has been growing, with much of the research focusing on the impact of role transitions, the protective effects of ethnic identity, risks and protections involved in the acculturative process, and negative effects of discrimination (Allem et al., 2013; Rogers et al., 2020; Unger et al., 2014).

Adverse Childhood Experiences

A critical intersection between substance use behaviors and transitions to adulthood is the impact of childhood stressors. The seminal study by Felitti and colleagues (1998) highlighted the influence of child maltreatment (physical, sexual, and psychological) and household dysfunction (paternal interpersonal violence, household incarceration, household substance misuse, household mental illness, and divorce) across an array of health outcomes, including substance use (Felitti et al., 1998; Forster et al., 2018; Forster, Rogers, et al., 2019; Shin et al., 2018). Since then, a substantial body of work has focused on the detrimental effects of adverse childhood experiences (ACE) on social, economic, and health status. Retrospective studies have identified a graded and predictive impact of ACE on substance use among adults (Anda et al., 2006; Felitti et al., 1998; Shonkoff et al., 2012) and adolescents (Anda et al., 1999).

The biological mechanisms driving poorer health outcomes are the trauma related changes in the brain that affect morphological and cognitive processes, as well as gene expression that can contribute to the formation of habits that are implicated in the development of addiction (Enoch, 2011; Stephens & Wand, 2012). Essentially, early life adversity can alter the brain structures and physiological processes involved in mood and behavior (Perry & Pollard, 1998; Teicher et al., 2002), in turn promoting later life disability and disease (Danese & McEwen, 2012). Maladaptive coping behaviors, including substance use, have also been attributed to deficits in cognitive and emotional processing (Gilbert, 2009; Pollak et al., 2000). According to diathesis-stress models, early exposure to chronic stressors increases vulnerability for pathology and psychopathology (Albott et al., 2018; Spielman et al., 1987). In addition to changes in cognitive and emotional processes, ACE can impede healthy attachment and emotional self-regulation which compromise protective factors such as social support and the adoption of health promoting behaviors (Felitti & Anda, 2010; Forster et al., 2017; Grigsby et al., 2020). Considering the heightened risk for substance use during emerging adulthood and that ACE have been associated with a higher probability of tobacco, marijuana, and alcohol use among Hispanic populations (Allem et al., 2015; Forster, Vetrone, et al., 2019; Rogers et al., 2020), there is need for research assessing the longitudinal effects of traumatic stressors across adolescence and emerging adulthood among Hispanic populations—a gap filled by the present study.

Using data from a cohort of Hispanic adolescents followed from early adolescence through emerging adulthood, we hypothesized that on average, (1) for every additional ACE there will be a higher level at baseline [9^th grade] of 1.a) past 30-day cigarette use, 1.b.) past 30-day marijuana use, and 1.c.) past 30-day alcohol use, while controlling for sex, school, nativity, and socioeconomic status (SES). We also hypothesized about the cross-level effects of ACE, specifically: (2) that the slope of substance use over time will differ across levels of ACE with steeper increases of substance use among emerging adults with higher levels of ACE. This analytic approach builds upon prior research by examining the shape of individual growth trajectories as well as the impact of ACE on the intercept and slope. This enables identifying the role of ACE in the initial level of substance use and whether this exposure influences a pattern of behavior over time that cannot be detected with a single time point or two-time points.

Methods

Procedures

Data are from Project Reteniendo y Entendiendo Diversidad para Salud (RED), a longitudinal cohort study designed to assess acculturation and substance use patterns among Hispanic/Latino adolescents enrolled in eight high schools in Southern California (Unger, 2018; Unger et al., 2009). Investigators visited classrooms of all eight high schools and distributed consent forms for students to complete with their parents. The first survey wave occurred in 2005 with 9^th grade students and then repeated in 10^th and 11^th grade. In order to extend the study beyond high school into emerging adulthood, the cohort was re-contacted in 2011. This led to five post-high school waves of data collection that occurred in 2011, 2013, 2014, 2016, and 2018. The Institutional Review Board (IRB) approved all study procedures.

The original cohort included (N=3,218) Hispanic adolescents from seven Hispanic serving schools in the Los Angeles Area who were enrolled in 9th grade. Of those invited, 75% provided parental consent and student assent and 92% of those who consented completed the survey. Before the second wave, one of the school districts transferred participants to a new school, so an additional school was included in the study at the beginning of 10^th grade, resulting in 704 additional participants. The high school sample was 2,969 students who provided data during at least one of the first three collection waves, which was restricted to only those who identified as Hispanic/Latino (n=2,722). The analytic sample for the current paper (n=1,399) was limited to participants who provided data on ACE as well as substance use for at least one of the waves. For the analytic sample, substance use response rates across waves were similar, with 71% providing complete data on substance use at seven or more waves, 24% across 5 to 6, and 5% across 2 to 4 waves.

Measures

Self-reported Adverse Childhood Experiences were measured with nine items in the second wave of post-high school data collection. Items included child maltreatment (e.g., physical, sexual, and psychological abuse) and household dysfunction (e.g., parental partner violence, incarceration, alcohol misuse, illicit substance use, mental illness, and divorce). The maltreatment questions were prefaced with, “While you were growing up, that is your first 18 years of life, how often did a parent, step-parent, or adult living in your home…” with responses coded in line with the original ACE measure (Felitti et al., 1998). Response options ranged from “never,” “once or twice,” “sometimes,” “often,” to “I prefer not to answer.” Household disfunction items asked participants if, before they turned 18 years old, they lived with anyone who was mentally ill, misused substances, was incarcerated, or was physically violent with their spouse/partner. Response options were “yes,” “no,” or “I prefer not to answer.” ACE items were dichotomized (0=no and 1=yes) and then summed to create an index of childhood adversity (range 0-9).

Past 30-day cigarette use and alcohol use were assessed by asking respondents “During the past 30 days how many days did you use…” Response options included days from 0 to 30. Past 30-day marijuana use was captured by asking respondents to report, “In the last 30 days how many times have you used marijuana.” Response options included times from 0 to 40.

Demographic covariates included sex, nativity, and socioeconomic status. Sex was assessed by asking, “What is your sex?” Response options including female and male with female as the reference group. Nativity was measured with one question; “In what country were you born?” Response options were “U.S.” and “other” with other as the reference group. Socioeconomic status (SES) is a complex construct to assess in adolescents because many cannot provide reliable reports of the indicators most commonly used to quantify SES (Bradley & Corwyn, 2002; Unger et al., 2009). In order to capture SES this study created a standardized SES variable previously validated in this population (Unger et al., 2009; Unger et al., 2014). Several proxy measures were used to build the SES index and included parents’ education rated on a 6-point scale ranging from “8th grade or less” to “advanced degree,” a ratio of the number of rooms per person in the home captured by dividing the number of people in the house by the number of rooms in the house, and the U.S. census median household income in the respondent’s provided zip code provided. The index also included dichotomous measures of eligibility for free/reduced price lunch at school (1 = no, 0 = yes), homeownership (1 = family owns its home, 0 = family rents home from a landlord), presence of a computer in the home (1 = yes, 0 = no), presence of a gaming console in the home (1 = yes, 0 = no), and availability of the Internet at home (1 = yes, 0 = no). In order to weight each indicator equally, items were standardized to a mean of 0 and a standard deviation of 1 and summed to create the SES index (Unger et al., 2009; Unger et al., 2014).

Dummy coded variables for school were included to control for any school level differences. Random effects for schools were not included because of low intraclass correlation coefficients (ICC) of students nested within schools (<0.03) and too few schools to accurately model random effects. Age was not included as a covariate because of limited variability since all participants began as freshmen. Race/ethnicity was limited to only those identifying as Hispanic.

Data Analysis

Univariate and bivariate analyses were conducted to describe the sample and patterns in missing data. Means plots were calculated to graph the substance use trends across time. Growth curves were estimated using SAS PROC MIXED. Models were iteratively assessed to determine appropriateness of model parameters beginning with empty models and subsequently adding fixed and random effects of time and differing covariance patterns (See Table 2). The random linear model with random slopes and random intercepts proved to be the most appropriate model with the best fit, based on the lowest Akaike’s Information Criteria (AIC) and Bayesian Information Criteria (BIC) (See Table 2). The final random linear models included the random intercepts and slopes for time as well as the fixed effects of time invariant independent variables (ACE) and the time invariant baseline model covariates (SES, sex, nativity, and school). A second set of models included an interaction term to determine the cross-level effect of ACE with time on substance use. To visualize the interaction effects, the final model’s predicted values were stratified by levels of ACE and plotted with 95% confidence intervals using the SAS PROC PLM procedure. To address missing data across waves, maximum likelihood estimation of mixed models allowed for data that are missing at random on the dependent variable (Allison, 2012; Molenberghs & Kenward, 2007). Therefore, the sample was restricted to complete data on ACE but allowed for some missing data on dependent variables. To contrast participants who did and did not provide ACE or substance use items, we compared covariate differences. Compared to the full sample, the analytic sample had fewer males (χ²=40.497, p<0.001) and lower SES (t=−3.820, p<0.001), however, there were no significant differences on first versus later generation participants (χ²=3.501, p=0.061).

Table 2:

Model Testing.

		Past 30-Day Cigarette Use	Past 30-Day Marijuana Use	Past 30-Day Alcohol Use
Empty means Model	AIC	54851.4	66885.1	50295.2
	BIC	54867.2	66900.8	50310.9
Fixed linear random intercept Model	AIC	54650.6	66559.1	49679.6
	BIC	54671.6	66580.0	49700.6
Random linear Model	AIC	52312.2	64780.3	48861.0
	BIC	52327.9	64801.3	48882.0
Fixed Quadratic random linear Model	AIC	54619.7	66555.8	49680.0
	BIC	54646.0	66582.0	49706.2

Notes: The Bolded items denote the lowest model fit statistics.

Results

The final analytic sample was comprised of 1,399 Hispanic participants who provided ACE and substance use data on at least one or more time points. At baseline, the average age was 14.49 (SD=0.36) and 26.14 (SD=0.38) at the final emerging adult wave (See Table 1). Among the analytic sample, 59.26% were female, 87.78% were US-born and 59.9% were second generation (born in the U.S. but neither parent was born in the US). The average ACE was 2.75 (2.19) with 58% of the sample reporting verbal abuse, 51% physical abuse, 37% parental divorce, 30% household alcohol use, 25% parental intimate partner violence, 22% household mental illness, 22% household incarceration, 17% household drug use, and 16% sexual abuse. Across all substances, average use was less than one day at baseline and increased over time. On average, past 30-day substance use increased from baseline to the final post-high school collection wave for cigarettes ($\stackrel{\_}{\text{X}}=0.30$ SD=2.20 to $\stackrel{\_}{\text{X}}=1.26$ SD=5.06), marijuana ($\stackrel{\_}{\text{X}}=0.15$ SD=0.58 to $\stackrel{\_}{\text{X}}=0.67$ SD=1.35), and alcohol ($\stackrel{\_}{\text{X}}=0.98$ SD=3.29 to $\stackrel{\_}{\text{X}}=3.66$ SD=5.37). Trends in average product use were assessed with means plots stratified by ACE. On average, those with higher levels of ACE had higher levels of substance use at baseline with averages increasing and remaining consistently higher than students with no ACE, up through the final wave (See Panel Figure 1).

Table 1.

Descriptive results at baseline and final collection waves.

	Baseline First High School Wave	Final Post-High School Wave
	% (Frequency)	% (Frequency)
Sex
Female	59.26% (829)	--
Male	40.74% (570)	--
Nativity
U.S. Born	87.78 (1228)	--
Non-U.S. Born	12.22% (171)	--
	Mean (Standard Deviation)	Mean (Standard Deviation)
SES	0.017 (4.49)	--
Age	14.49 (0.36)	26.14 (0.38)
Past 30-Day Substance Use (mean days in past month)
Cigarette use	0.30 (2.20)	1.26 (5.06)
Marijuana use	0.15 (0.58)	0.67 (1.35)
Alcohol use	0.98 (3.29)	3.66 (5.37)

Note: --Time invariant items were not reported at both time points.

Figure 1.

Average Past 30-day substance use across survey wave, stratified by ACE.

The growth curve analysis was used to estimate participants’ trajectories of substance use from 9^th grade through emerging adulthood; 8 waves of data collection. The models used retrospective ACE as well as covariates to first assess the fixed linear effect of ACE on substance use. The final model chosen was the random linear model that included all theoretically relevant covariates, the main hypothesized predictors of time and ACE, and an interaction term to establish if the growth of substance use over time differs across levels of ACE. Across all models there was significant variance in random slopes; however, there was only significant variance in random intercepts for alcohol use. When assessing the fixed linear effect of ACE on substance use at baseline (hypothesis 1.a-1.c), on average, ACE was a significant predictor across all substances. For every additional ACE there was a significant increase in past 30-day cigarette use (β=0.05, 95%CI=0.01, 0.10), past 30-day marijuana use (β=0.15, 95%CI=0.06, 0.25), and past 30-day alcohol use (β=0.14, 95%CI=0.06, 0.21). When assessing the fixed linear effect of time (survey wave) on substance use, a similar association was seen with steady significant increases in substance use across survey waves. For every additional time point there was a significant (β=0.14, 95%CI=0.06, 0.22) increase in past 30-day cigarette use, (β=0.40, 95%CI=0.25, 0.54) past 30-day marijuana use, and past 30-day alcohol use (β=0.45, 95%CI=0.36, 0.53).

Sex was also significant across all substances with a significant increase for males compared to females (β=0.44, 95%CI=0.25, 0.63) for past 30-day cigarette use (β=0.94, 95%CI=0.55, 1.32), for past 30-day marijuana use, and (β=0.80, 95%CI=0.54, 1.06) for past 30-day alcohol use. Regarding nativity, this was only a significant predictor for past 30-day alcohol use. Compared to those born outside of the U.S., respondents born in the U.S. saw an increase (β=0.41, 95%CI=0.07, 0.80) in past 30-day alcohol use (See table 3).

Table 3:

Growth Curve Models.

	Past 30-Day Cigarette Use	Past 30-Day Marijuana Use	Past 30-Day Alcohol Use
Variance Components	Parameter Estimate (SE)	Parameter Estimate (SE)	Parameter Estimate (SE)
UN1,1	--	0.77 (0.70)	3.16*** (0.42)
UN2,1	−0.04 (0.05)	0.35 (0.21)	−0.58*** (0.11)
UN2,2	0.70*** (0.04)	1.83*** (0.11)	0.56*** (0.04)
Time	8.54*** (0.13)	30.84*** (0.53)	11.25*** (0.21)
Fixed Effects	Parameter Estimate (SE)	Parameter Estimate (SE)	Parameter Estimate (SE)
Intercept	0.15 (0.20)	0.16 (0.38)	0.25 (0.26)
Time	0.14*** (0.04)	0.40*** (0.07)	0.45*** (0.04)
ACE	0.05* (0.03)	0.15** (0.05)	0.14*** (0.04)
Nativity	0.17 (0.15)	−0.24 (0.30)	0.41* (0.20)
Sex	0.44*** (0.10)	0.94*** (0.19)	0.80*** (0.13)
Time* ACE	0.05*** (0.01)	0.07*** (0.02)	0.02* (0.01)

Notes:

^*P<0.05,

^**P<0.01,

^***P<0.001

The models also controlled for SES, Nativity, and School – Random intercepts were not estimated for cigarette use due to limited variance in random intercepts.

In support of our second hypothesis on the cross-level effects of ACE, the interaction term between time and ACE was positive and significant in all models. Results indicated that the linear rate of change over time differed across levels of ACE for past 30-day cigarette use (β=0.05, 95%CI=0.02, 0.68), past 30-day marijuana use (β=0.07, 95%CI= 0.03, 0.11), and past 30-day alcohol use (β=0.02, 95%CI=0.02, 0.68) (See Table 3). At higher levels of ACE there is a steeper trajectory across all substances. This is visually represented by plotting each model’s predicted probabilities across levels of ACE (See Panel Figure 2). At any single time-point participants with higher ACE have higher average predicted values of substance use; however, as time increases the adverse effect of ACE becomes more pronounced.

Figure 2.

Model predicted probabilities across time stratified by ACE.

Discussion

This is one of the first studies to assess ACE-related longitudinal substance use trajectories among a community cohort of Hispanic emerging adults. Our findings highlight that within this U.S. Hispanic sample, the impact of ACE on substance use behaviors begins as early as adolescence and continues to exacerbate risky behavior well into emerging adulthood. Although previous literature has established that ACE increase risk for substance use, the present study demonstrates the enduring, long-term effect of ACE on tobacco, marijuana, and alcohol uses behaviors. The steady increase in past 30-day substance use over time even among low-ACE adolescents, as demonstrated by the significant positive fixed linear effect of time, is consistent with theoretical models suggesting both adolescence and emerging adulthood are periods characterized by increases in risk taking behaviors and exploration that are aggravated by traumatic stressors and carry forward into adulthood. The significant variance components across models suggest that respondents defined by their ACE exposure have significantly different levels of substance use at baseline and divergent substance use trajectories into adulthood. The patterns observed in the means plots and the significance of ACE in the multivariable models provides compelling evidence that the differences at baseline and across time are linked to childhood trauma. Regarding the average between-person disparity in substance use linked to ACE, adolescents who are ACE exposed begin high school with higher levels of substance use, and this disparity persists for the next decade and a half even after controlling for key confounders. A unique feature of this study was the identification of the significant interaction between ACE and time which suggests that the changes in substance use over time are modified by ACE. This cross-level effect of ACE confirms that ACE are a critical risk factor and account for a significant proportion of between-subject differences and within-subject effects of substance use behaviors spanning from adolescence to emerging adulthood. The expected trajectory of Hispanic respondents who are ACE exposed is not simply constant over time, instead higher levels of ACE can exacerbate substance use behaviors. For example, using the model parameters, even though the average use for all substances was less than one day at baseline for the sample, the models suggest that as respondents progressed through emerging adulthood, participants with no ACE used cigarettes an average of less than one day per month while participants who were ACE exposed (ACE=6 or 9) used cigarettes 3 to 4+ days a month. Compared to their non-ACE exposed peers, ACE exposed individuals are projected to increase current cigarette, marijuana, and alcohol use at a significantly faster rate, placing them at even greater risk for life course addiction as they transition into adulthood.

Although these significant effects were seen across all three substance use models, the change over time was more pronounced for cigarette use and marijuana use compared to alcohol use. This addition to the current literature, that among Hispanic individuals greater ACE exposure is longitudinally associated with greater escalation of past 30-day substance use over time, is critical to our understanding of the developmental nature of trauma related behavioral health outcomes.

In sum, these findings support the call for intervention and prevention research and support services. Providing adolescents and young adults with the tools and training necessary to manage traumatic stress as well as encouraging the development of positive approaches to coping with trauma is a key component of prevention work for at risk youth. Prior crosssectional research has shown that cultural assets and support systems may in fact mitigate the negative effects of ACE for youth and young adult health and well-being (Brown & Shillington, 2017; Chatterjee et al., 2018; Karatekin & Ahluwalia, 2020; Robertson et al., 2010). A critical component of future research will be investigating whether early access to resources that leverage the key ingredients of resilience in prevention programs can disrupt risk trajectories among Hispanic youth and emerging adults.

Limitations

These findings should be considered in light of the following limitations. First, these data are based on self-reported ACE and substance use and although the inclusion of biomarkers would provide a more definitive report of misuse, self-reported substance use has been found to be highly accurate under confidential survey conditions such as the present study (Harrison & Hughes, 1997). Second, ACE was assessed retrospectively. However, much of the ACE research has been conducted retrospectively and even studies challenging retrospective vs prospective reports note that retrospective reports provide a meaningful addition to the literature and are validly associated with other subjective measures (Reuben et al., 2016). Third, we cannot definitively anchor ACE to a specific time point in childhood; however, the survey specifies events that occurred prior to age 18. Because of this, we are not able to assess whether the timing of ACE occurred prior to any high school trajectory changes or prior to substance use initiation. Fourth, many of the students in the sample qualify for free lunches, suggesting that the sample skews towards a lower SES threshold. Prior research has shown that there may be a strong association between ACE scores and SES. To this end, all models controlled for a proxy measure for SES. Fifth, similar to other longitudinal studies, there was a considerable amount of attrition and percentage of the sample needed to be excluded because of a lack of ACE data. Therefore, the present study used the maximum likelihood estimations built into the SAS PROC MIXED procedure which produces unbiased parameter estimates by accounting for all included data and other model covariates, even with missing data on the dependent variable (Allison, 2012; Molenberghs & Kenward, 2007). Attrition rates in the current study were comparable to other national studies that cross-over a difficult developmental period and extend over 10 years (Schulenberg et al., 2020). We acknowledge that attrition could have the potential to bias results and that the participants that were excluded due to attrition or not providing information on ACE may represent an especially vulnerable subset of the sample, such that our results may only provide a preliminary understanding of these relationships. In an effort to validate the current approach of maximum likelihood estimations, a series of imputed models were also run and produced similar results. Finally, the generalizability of the present research findings is limited to Hispanic youth and young adults living in urban settings similar to Southern California. Future research using ACE to assess substance use trajectories may consider the effects of individual ACE on use patterns. This line of research would help to disaggregate the relative weight of each experience as it relates to later life substance use patterns, and further explain the ACE substance use relationship.

Conclusions

This study highlights the graded effect of ACE on substance use from adolescence through emerging adulthood and builds on current research by demonstrating that ACE may be linked to escalating and continued substance use over time. The fact that these effects were identified among Hispanic individuals further confirms the effect of ACE that has been observed across populations and draws attention to the need for early prevention and intervention efforts that provide adolescents and young adults with the tools and supports necessary to reduce maladaptive coping behaviors. Disrupting these trajectories for adolescents and young adults could be a key factor in reducing the risk for addiction and promoting resilience and health across the life course.

Highlights:

There is a dose-response relationship with ACE and substance use in adolescence and in emerging adulthood.

The cross-level effect of ACE indicates that ACE is a critical risk factor in substance use trajectories.

The substance use frequency (tobacco, marijuana, and alcohol) trajectories in those with higher levels of ACE are not constant over time, rather higher levels of ACE may exacerbate substance use behaviors.

The effects of ACE are ubiquitous across multiple populations, including Hispanic populations, and highlight the need for early ACE screening and early prevention.