Latent Class Patterns of Adverse Childhood Experiences and Mental Health Among National Guard Recruits

Abstract

Background:

Most studies have relied on a cumulative risk approach when examining ACEs. This approach assumes equal weighting of adversities and fails to consider how the nature, severity, and combination of ACEs may differentially impact outcomes.

Objective:

Employing a person-centered approach, we identified distinct patterns of ACEs in a sample of Army National Guard recruits and investigated how these patterns relate to internalizing symptoms (i.e., self-reported mental health) and externalizing problems (i.e., substance use and rule-breaking behaviors).

Participants and setting:

Participants were 1,201 Army National Guard recruits from the Advancing Research on Mechanisms of Resilience (ARMOR) study.

Methods:

Latent class analysis (LCA) was performed to identify distinct latent classes of recruits with similar ACEs patterns. Associations between classes and mental health and behavioral outcomes of interest were examined using the Bolck-Croon-Hagenaars (BCH) method.

Results:

LCA revealed four latent classes of ACEs: (1) low adversity (51.1%), (2) emotional and physical maltreatment (18.3%), (3) poly-adversity (16.1%), and (4) parental separation (14.5%). There were no significant interclass differences found in internalizing symptoms or substance use. However, the emotional/physical maltreatment and low adversity classes showed significantly higher counts of rule-breaking behaviors than the other classes.

Conclusions:

Findings suggest recruits enter military service reporting distinct patterns of ACEs, which are differentially associated with rule-breaking behaviors but not mental health outcomes or substance use.

Introduction

Adverse childhood experiences (ACEs) refer to a collection of stressful and/or potentially traumatic events that individuals may experience before the age of 18 years and include abuse (physical, emotional, and sexual), neglect (emotional and physical), and household dysfunction (household mental illness, substance use, and witnessing intimate partner violence). The landmark Adverse Childhood Experiences (ACE) study conducted by Felitti and colleagues (1998) found that over half of participants in a large community sample reported at least one ACE. Since the original ACE study, extensive research has documented the profound impact ACEs can have on physical, mental, and behavioral health outcomes later in life (Petruccelli et al., 2019). Studies have shown that ACEs have strong cumulative associations with negative outcomes (Hughes et al., 2017). Recently, researchers have suggested that ACEs often occur in clusters allowing for the identification of subgroups of individuals who share similar patterns of ACEs exposure, and that analyzing these patterns is important to understand how ACEs affect well-being later in life.

There is also evidence that ACEs are more common in military than civilian populations (Blosnich et al., 2014; Blosnich et al., 2021; Katon et al., 2015; McCauley et al., 2015), and that exposure to early childhood adversity may interact with subsequent military-related stressors to amplify risks of poor mental health outcomes following deployment (Cabrera et al., 2007; Crede et al., 2023; LeardMann et al., 2010; Sareen et al., 2013). However, much of the research on ACEs in military populations has focused on veterans, and there is limited data available regarding patterns of ACEs reported by new recruits before their exposure to military-related stressors. Identifying specific ACE profiles and their links to mental and behavioral health outcomes in new military personnel can help design targeted early interventions for recruits at high risk of mental health and behavioral issues.

Most extant ACEs literature has employed the cumulative risk, or “variable-centered” approach which entails totaling the cumulative number of ACEs endorsed by an individual to prioritize the relationship between variables. Notably, ACEs research using the cumulative risk approach has demonstrated that reporting four or more total ACEs significantly increases an individual’s risk for negative outcomes (Campbell et al., 2016; Felitti et al., 1998; Hughes et al., 2017). However, the cumulative risk approach assumes equal weighting of all adversities and fails to consider how the nature and severity of co-occurring ACEs may impact outcomes (Barboza, 2018; Evans et al., 2013; Lian et al., 2022). The categorical risk, “or person-centered” approach considers the information given by looking at the events from a typological rather than numerical lens, which may allow researchers to better gauge the impact of ACEs on health outcomes (Font & Maguire-Jack, 2016; Schilling et al., 2016).

Latent class analysis (LCA) has recently become a popular statistical method to identify unobserved subgroups or latent classes within a population. These classes “characterize heterogeneity in a population for a given phenomenon” based on response patterns in observed categorical variables (Nylund-Gibson & Choi, 2018, p. 441). LCA is a data driven approach that allows researchers to determine the optimal number of latent classes (subgroups of people) that best fit the data. In recent years, there has been “explosive growth” in studies using LCA to discover distinct subgroups of people reporting similar patterns of ACEs exposure (Wang et al., 2024, p. 1701). However, the number and types of classes identified vary across studies depending on the sample (e.g., clinical or at-risk samples) and methods (e.g., types of ACEs included as indicators in LCA) used. Therefore, to guide our hypotheses, we limited our review of LCA studies to those most like the current study. Specifically, we reviewed research conducted with young adults from the general population as well as studies of ACEs in military populations. Because findings of LCA studies vary based on the number and types of ACEs assessed, we focused on those studies that included ACEs indicators closely aligned with those used in the original ACE study (Felitti et al., 1998).

Studies using LCA to examine patterns of ACEs in general community samples have identified multiple latent classes, typically ranging from three to five (Barboza, 2018; Cavanaugh et al., 2015; Merians et al., 2018; Romm & Berg, 2024). Most studies have found a low adversity class, a childhood maltreatment class, a household dysfunction class characterized by parental substance abuse, poor family health, or divorce, and a fourth class characterized by high adversity. Across these studies, the child maltreatment and high adversity classes have typically demonstrated worse mental health and physical health outcomes. Barboza (2018) highlighted the importance of categorizing the types of ACEs people report—such as sexual abuse, parental separation, or living with someone experiencing mental illness—rather than treating all ACEs as having the same weight or impact. They found that compared to the low adversity class, their high-risk classes were differentially associated with psychosocial outcomes.

Several studies have used data from the National Epidemiological Survey on Alcohol and Related Conditions (NESARC) to examine distinct patterns of ACEs (McLafferty et al., 2019; Ross et al., 2018; Tynan et al., 2022). Tynan et al. (2022) conducted LCA with a sample of over 36,000 NESARC participants, including 3,111 U.S. military veterans. Consistent with LCA studies using community samples, those authors identified four distinct patterns of ACEs (low adversity, maltreatment, household dysfunction, and multiple adversity). However, veterans were more likely to be members of the maltreatment, household dysfunction, and multiple adversity classes compared to the “low adversity” class. Two additional studies have examined patterns of ACEs reported in subsamples of U.S. military veterans using the NESARC dataset (McLafferty et al., 2019; Ross et al., 2018) and found evidence of the same four classes of ACEs patterns. McLafferty and colleagues (2019) found that individuals in the adversity classes had higher rates of psychopathology and were less likely to have supportive networks. After controlling for sociodemographic characteristics and alcohol and drug use, Ross et al. (2018) found that all three adversity classes had significantly elevated odds of adult incarceration compared to the low adversity class.

To our knowledge, only one study has used LCA to identify distinct patterns of ACEs and examined their relationship to mental health outcomes in military recruits (Reed-Fitzke et al., 2023). Self-reported ACEs were collected from over 3,600 male Marines who responded to the Recruitment Assessment Program (RAP) survey at the Marine Corps Recruit Depot before basic training. However, those ACEs data were not linked to current mental health or behavior problems present before basic training. Rather, the study focused on whether ACEs subgroups differed on demographic, military deployment experiences, and follow-up measures of mental health (depression, PTSD symptoms) and social support collected up to 4 years later from the Millenium Cohort study. Those authors identified five latent classes: low adversity, low adversity – parental separation, elevated adversity, moderate adversity – violence/safety, and moderate adversity – parental loss. Compared to the low adversity class, the parental loss class was associated with later PTSD and depression. The elevated adversity class was associated with PTSD and social support problems, and the parental separation class was associated with only social support problems. Research is needed to understand how distinct ACEs profiles may differentially impact recruits’ mental health and behavior at the outset of their military before basic training.

The dimensional model of adversity and psychopathology (DMAP; McLaughlin et al., 2014; McLaughlin & Sheridan, 2016) provides a framework for conceptualizing differential impacts of ACEs. DMAP suggests that two different dimensions of adversity (i.e., threat and deprivation) lead to differential impacts on individuals through neuropsychological mechanisms. McLaughlin and colleagues (2014) proposed that adversities related to threat (i.e., potential to suffer harm and/or violence) and adversities related to deprivation (i.e., lack of vital care and resources such as neglect) generally lead to different outcomes. Their research suggests that ACEs related to threat such as abuse are more likely to impact internalizing problems like depression, anxiety, and emotion regulation while deprivation ACEs are more closely linked to cognitive impairments, which can influence poor decision-making. Other studies have aligned with this model in finding that ACEs of threat are more closely associated with internalizing symptoms (Henry et al., 2021; Milojevich et al., 2019).

The present study expands upon this previous research by employing a person-centered approach (i.e., latent class analysis) to: (1) identify distinct patterns of ACEs in a sample of National Guard recruits and (2) examine associations between these ACEs patterns and self-reported internalizing symptoms (i.e., anxiety, depression, anger) and externalizing problems (i.e., substance use and rule-breaking behaviors). We hypothesized that three to five distinct profiles of ACEs would emerge within the sample, consisting of a low adversity class, a poly-adversity class, and classes based on maltreatment and household dysfunction. We also expected that self-reported internalizing symptoms and externalizing behaviors would differ as a function of adversity profiles. Specifically, we expected that recruits exposed to childhood maltreatment (i.e., the poly-adversity and maltreatment classes) would be associated with both internalizing symptoms and externalizing problems. We expected that recruits who are members of a household dysfunction class would be associated with externalizing symptoms. We also expected that recruits in the low adversity class would report lower symptoms across all outcomes compared to recruits in the other classes. The current study expands on previous findings in several ways. First, most extant literature examining risk factors and outcomes in military samples fails to examine risk factors at baseline, prior to any military-related stressors. Second, most LCA studies of ACEs in military populations have been limited to veteran samples. To our knowledge, this is the first study to use LCA to uncover distinct patterns of ACES among new recruits and examine whether these ACEs typologies may be associated with internalizing symptoms (i.e., self-reported mental health) and externalizing problems (i.e., substance use and rule-breaking behaviors). Through early identification of heterogeneous subgroups with distinct problems, unique mental health needs can be recognized which could support the argument for more tailored resilience interventions.

Methods

Participants and procedures

The present study is a secondary analysis of data from the Advancing Research on Mechanisms of Resilience (ARMOR) study conducted by Polusny and colleagues (2023). The ARMOR study is a longitudinal investigation of neurobiological, cognitive, and social mechanisms underlying adaptive functioning following military stressor exposure among National Guard recruits (Polusny & Erbes, 2023). Self-reported ACEs and mental health data were collected at baseline, before recruits were shipping off to BCT. Subsequently, participants were followed up at four time points after BCT. Analyses for the present study were based on baseline survey data only. The analytic sample (N = 1154) included participants who responded to ACEs-related questionnaire items. The analytic sample was comprised of mostly males (n = 796, 69.2%) with a mean age of 18.97 (SD = 3.03); 85% of the sample was under age 21. Most soldiers were White (n = 698, 60.6%) and single (n = 656, 57.0%). The ARMOR study was approved by institutional review boards of the University of Minnesota and Minneapolis VA Health Care System and the relevant National Guard command.

ACEs Measurement

To represent the landmark ten ACEs indicators (e.g., emotional, physical, and sexual abuse, emotional and physical neglect, parental separation, witnessing violence between caregivers, and living with someone with a substance use problem, mental illness, or who is incarcerated; Felitti et al., 1998), the present study utilized survey items to correspond with each indicator, which can be found in Table 1. After a recent scoping review (Wang et al., 2024), it was determined that most existing ACEs literature using LCA based their ACEs indicators on the original ten ACEs. The ACEs questionnaire used in the present study included nine of the ten ACEs indicators, and items were prefaced with “While you were growing up, during your first 18 years of life…”. The indicator for witnessing violence between caregivers was captured using the Deployment Risk and Resilience Inventory-2 Prior Stressors scale (DRRI-2; Vogt et al., 2013), where items were prefaced with “In my life…”. On average, the analytic sample reported experiencing 2.34 ACEs (SD = 2.33).

Table 1.

Adverse Childhood Experiences Indicators and Items Endorsed

Indicators/Items	n	%
Emotional abuse (either or both items)	473	39.40%
Adult in home swear at, insult, put down, or humiliate you?	442	36.80%
Adult in home made you afraid you’d be physically hurt?	306	25.50%
Physical abuse
Adult in the home often push, grab, slap, or throw at you?	278	23.10%
Sexual abuse (either or both items)	116	9.70%
Person in home 5+ years older ever touch or fondle you or have you touch their body in a sexual way?	105	8.70%
Person in home 5+ years older try to have oral, anal, or vaginal sex with you?	83	6.90%
Emotional neglect (either or both items)	405	33.70%
Often felt no one in family loved you or thought you were important?	253	21.10%
Often felt family did not look out for or support each other?	334	27.80%
Physical neglect
Often did not have enough to eat or had to wear dirty clothes?	121	10.10%
Parental separation
Were your parents ever separated or divorced?	549	45.70%
Household substance use
Anyone in home who misused alcohol or used street drugs?	299	24.90%
Household mental illness
Anyone in home mentally ill or attempt suicide?	318	26.50%
Household incarceration
Anyone in home go to prison?	141	11.70%
Witnessed domestic violence
Saw or heard physical fighting between my parents or caregivers?	314	26.10%

Emotional abuse.

Two ACEs items were used to determine whether recruits had been exposed to emotional abuse during childhood. The first item asked: “Did a parent or other adult in the household often swear at you, insult you, put you down, or humiliate you?” The second item asked: “Did a parent or other adult in the household often act in a way that made you afraid that you might be physically hurt?” Emotional abuse was coded as “1” if at least one item was answered “yes” and was coded as “0” if responses to both items were “no”. Nearly 40% of the analytic sample reported experiencing emotional abuse (n = 473, 39.4%).

Physical abuse.

The physical abuse item asked: “Did a parent or other adult in the household often push, grab, slap, or throw something at you?” Physical abuse was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Over 23% of the analytic sample endorsed being physically abused (n = 278, 23.10%).

Sexual abuse.

Two items assessed sexual abuse: 1) “Did an adult or person at least 5 years older than you ever touch or fondle you or have you touch their body in a sexual way?” and 2) “Did an adult or person at least 5 years older than you ever try to actually have oral, anal, or vaginal sex with you?” Sexual abuse was coded as “1” if at least one item was answered “yes” and was coded as “0” if responses to both items were “no”. Approximately 10% of the sample reported a history of sexual abuse (n = 116, 9.7%).

Emotional neglect.

Two items were used to measure emotional neglect. In the first item, recruits were asked: “Did you often feel that no one in your family loved you or thought you were important or special?” The second item asked: “Did you often feel that your family didn’t look out for each other, feel close to each other, or support each other?” Emotional neglect was coded as “1” if at least one item was answered “yes” and was coded as “0” if responses to both items were “no”. Over a third of the analytic sample endorsed emotional neglect (n = 405, 33.7%).

Physical neglect.

The ACEs questionnaire included the following item to capture whether a recruit had experienced physical neglect: “Did you often feel that you didn’t have enough to eat, had to wear dirty clothes, and had no one to protect you?” Physical neglect was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Approximately 10% of the sample reported physical neglect (n = 121, 10.1%).

Parental separation.

The ACEs questionnaire also asked recruits: “Were your parents ever separated or divorced?” Parental separation was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Almost half of the analytic sample reported parental separation (n = 549, 45.7%).

Witnessed IPV.

The DRRI-2 Prior Stressors scale included the following item for recruits to indicate whether they had witnessed intimate partner violence in their home: “I saw or heard physical fighting between my parents or caregivers.” Witnessed IPV was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Over a quarter of the analytic sample reported witnessing IPV (n = 314, 26.1%).

Household mental illness.

One ACEs item captured whether respondents were exposed to living with an individual who experienced mental illness. The item asked: “Was a household member depressed or mentally ill or did a household member attempt suicide?” Household mental illness was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Over a quarter of the sample endorsed having lived with someone with a mental illness (n = 318, 26.50%).

Household substance use.

One ACEs item captured whether respondents were exposed to living with an individual with a substance use problem. This item asked: “Did you live with anyone who was a problem drinker or alcoholic or who used street drugs?” Household substance use was coded as “1” if respondents indicated “yes” and coded “0” if respondents indicated “no”. Approximately one quarter of the sample reported having lived with someone with a substance use problem (n = 299, 24.90%).

Household incarceration.

One ACEs item captured whether respondents were exposed to living with an individual who was incarcerated. The item asked: “Did a household member go to prison?” Household incarceration was coded as “1” if respondents endorsed the item and coded “0” if respondents indicated “no” to the item. Over a tenth of the sample endorsed living with someone who was incarcerated (n = 141, 11.7%).

Internalizing Symptoms

Internalizing symptoms were assessed using the Patient-Reported Outcomes Measurement Information System (PROMIS) anger, anxiety, and depression short forms. All three measures were constructed using a psychometrically sound item bank from the National Institutes of Health (Ader, 2007; Cella et al., 2010). PROMIS-Anger (Pilkonis et al., 2011) is a five-item scale measuring angry mood, negative social cognitions, and efforts to control anger over the past 7 days. Sample items include “I felt angry” and “I felt like I was ready to explode”. Items are scored on a Likert scale ranging from 1 (never) to 5 (always). Total scores ranged from 0 to 25. PROMIS-Anxiety (Pilkonis et al., 2011) is a four-item scale measuring self-reported fear, anxious misery, hyperarousal, and somatic symptoms related to arousal over the past 7 days. Sample items included “My worries overwhelmed me” and “I felt uneasy”. Items are scored on a Likert scale ranging from 1 (never) to 5 (always). Total scores ranged from 0 to 20. PROMIS-Depression (Pilkonis et al., 2011) is 4-item scale measuring depression symptoms, including feeling hopeless, worthless, helpless, or depressed over the past 7 days. Sample items included “I felt depressed” and “I felt hopeless”. Total scores ranged from 0 to 20. All three measures have demonstrated strong validity when compared to other instruments. T-scores of 50 (SD = 10) demonstrate average levels of distress (Wilford et al., 2018) and were used to interpret the clinical significance of internalizing symptoms.

Externalizing Problems

Externalizing behaviors were assessed using the Alcohol Use Disorder Identification Test (AUDIT; Saunders et al., 1993), Drug Abuse Screening Test (DAST-10; Skinner, 1982), items adapted from the Behavioral Report on Rule-Breaking (BHR; Nelson et al., 2011), and two items from the Revised Conflict Tactics Scale (CTS; Straus et al., 1996).

Substance use.

Problems with alcohol use were measured with the AUDIT (Saunders et al., 1993). The AUDIT, a ten-item screening tool, was originally developed to identify items distinguishing between hazardous and non-hazardous alcohol use over the past 12 months. Sample questions include: “How many drinks containing alcohol do you have on a typical day when you are drinking?” and “How often have you found that you were not able to stop drinking once you started?” Items are rated by respondents on Likert scales ranging from 0 to 4, and total scores can range from 0 to 40. The AUDIT holds excellent sensitivity and specificity at a cut-score of 8, with 92% of hazardous drinkers scoring 8 or more and 94% of non-hazardous drinkers scoring less than 8. The DAST-10 is a ten-item screening tool used to assess nonalcoholic, non-medicinal drug use and abuse problems over the past 12 months. Sample items include “Are you unable to stop abusing drugs when you want to?” and “Have you engaged in illegal activities in order to obtain drugs?” Response options are “yes” coded as “1” and “no” coded as “0”, and total scores range from 0 to 10. The DAST-10 has shown strong psychometric properties in clinical and non-clinical populations including outpatients and undergraduates (McCabe et al., 2006; Yudko et al., 2007).

Rule-breaking behaviors.

The BHR is a 33-item measure of adolescent and adult antisocial behavior using items from several other published measures (Clark & Tifft, 1966; Hindelang et al., 1979; Nye & Short, 1957). To reduce participant burden, only 18 items from the BHR were administered for the present study. Participants were asked to report engagement in a range of physically aggressive and nonaggressive rule-breaking behaviors within the last six months. Sample items included “How often did you go onto someone’s property when they did not want anyone there?” and “Not counting a fight with your brother or sister, how often did you get into fistfights to defend yourself?” Items were scored using a Likert scale ranging from 0 (never) to 3 (very often) with total scores ranging from 0 to 54. The BHR has demonstrated good internal consistency with undergraduate students (α = 0.92; Nelson et al., 2011). Two items from the widely used Revised Conflict Tactics Scale (CTS2; Straus et al., 1996) were used to screen for psychological and physical aggression towards an intimate partner over the past 6 months. Participants were asked to indicate on a Likert scale ranging from 0 (never) to 3 (very often) the following: “How often did you push, shove, or slap a romantic partner?” and “How often did you insult, swear, shout, or yell at a romantic partner?” A combined frequency count was used for the BHR and CTS items to assess rule-breaking and aggressive behavior across classes.

Demographics.

Participants self-reported on their sex, race/ethnicity, relationship status, current living arrangement, current employment, and education level. This information was used to provide descriptive statistics about the sample.

Data Analysis Plan

Descriptive statistics were analyzed using the Statistical Package for the Social Sciences (SPSS; IBM Corp., 2023) Version 29. Mplus statistical software (Muthén and Muthén, 1998–2017) Version 8 was used for all subsequent analyses. Data analysis was conducted in two steps. In the first step, latent class analysis (LCA) was used to detect distinct profiles (or latent classes) of childhood adversity in the sample based on their response patterns related to each of ten ACEs indicators. Conditional probabilities, ranging from 0 to 1, represent the likelihood of reporting having experienced an ACE indicator given membership in a certain class. LCA was conducted following recommendations by Sinha and colleagues (2021). We fit models specifying one to five latent classes (1 class, then 2 classes, then 3 classes, etc.) and compared the fit of these models to determine the optimal number of classes. Model selection was based on multiple fit indices, parsimony, and interpretability. Fit indices analyzed included the Bayesian Information Criterion (BIC; Schwarz, 1978), Sample-size Adjusted BIC (SABIC; Sclove, 1987), Consistent Akaike Information Criterion (CAIC; Bozdogan, 1987), Vuong-Lo-MendellRubin likelihood ratio test (VLMRLRT; Lo, Mendell, & Rubin, 2001), Bootstrap likelihood ratio test (BLRT; McLachlan, 2000), and entropy (Golino et al., 2021). BIC, SABIC, and CAIC balance the fit of the model against the complexity of the model, while BIC rewards parsimony with lower values indicating better model fit. The VLMRLRT and BLRT are procedures for testing between two nested models (e.g., comparing a three-class model to a four-class model), and these statistics provide a p-value to determine if adding a class leads to a statistically significant improvement in model fit. If the p-value is not significant, then we conclude there is not a meaningful difference between the two models, and the simpler model is retained. Finally, we considered entropy, which is a measure of how accurately the model defines classes ranging from 0 to 1, with 1 indicating perfect accuracy.

In the second step, we examined associations between patterns of ACEs (latent class membership) and internalizing symptoms and externalizing problems. In line with recommendations from Bakk and Vermunt (2016) and Nylund-Gibson and Choi (2018), the BCH method (Bolck, Croon, & Hagenaars, 2004) was used to explore significant mean differences in internalizing and externalizing outcomes across latent classes. The BCH method has demonstrated minimal bias and performed well in terms of estimation error when used with skewed continuous distal outcomes (Bakk & Vermunt, 2016; Dziak et al., 2016). Full Information Maximum Likelihood estimation (FIML; Enders & Bandalos, 2001) was used to analyze data with partially missing data by incorporating all available information.

Results

Latent Classes of ACEs

Table 2 presents model fit indices for the class enumeration process in LCA, ranging from 1 to 5 latent class solutions. The 4-class solution demonstrated better (lower) BIC and SABIC values than the 5-class solution and provided better (higher) class separation, which also assisted in overall interpretability. After a review of the conditional probabilities and interpretability, we selected the 4-class solution. The 4-class model is not only supported by fit indices but also demonstrates 4 distinct and interpretable profiles of ACEs that were crucial to understanding our research questions.

Table 2.

Fit Indices for Latent Class Models

No. classes	LL	BIC	SABIC	CAIC	BLRTp	VLMR-LRTp	Entropy
1-Class	−6161.14	12392.77	12361.01	-	-	-	-
2-Class	−5142.25	10432.52	10365.81	10326.5	0.00	0.00	0.86
3-Class	−5048.52	10322.59	10220.95	10161.05	0.00	0.0001	0.82
4-Class	−4965.93	10234.93	10098.35	10017.85	0.00	0.0000	0.78
5-Class	−4944.92	10270.45	10098.92	9997.83	0.00	0.0096	0.76

Note. LL = Log likelihood, BIC = Bayesian Information Criteron, SABIC = Sample-size Adjusted Bayesian Information Criterion, Consistent Akaike Information Criterion, BLRT = Bootstrapped Likelihood Ratio Test, VLMR-LRT = Vuong-Lo- Mendell-Rubin adjusted likelihood ratio test

Based on conditional probabilities for experiencing ACEs (Figure 1), we interpreted and labeled four classes based on their distinct patterns of ACEs. The largest class was labeled Low Adversity, which comprised 51.1% of the sample. The Low Adversity class was characterized by a low probability of reporting all ACEs. The second largest class was labeled Emotional/Physical Maltreatment and comprised 18.3% of the sample. The maltreatment class was characterized by a high probability of experiencing emotional abuse, physical abuse, and emotional neglect. The next largest class, making up 16.1% of the sample, was labeled the Poly-Adversity class. This class was characterized by high probabilities of maltreatment and household dysfunction such as emotional and physical abuse, emotional neglect, parental separation, someone in their home with a substance use or mental illness problem or witnessing intimate partner violence. The smallest class, comprising 14.5% of the sample, was labeled the Parental Separation class, which was characterized by a high probability of a recruit’s parents being separated or divorced but low probabilities in all other ACEs. A breakdown of demographics by analytic sample and latent class is available in Table 3.

Figure 1.

Conditional probabilities and group membership prevalence for latent classes of adverse childhood experiences (N = 1201). Note. IPV = Intimate partner violence. Conditional probabilities represent the prevalence of individuals in a given class selecting the more adverse level of the variable.

Table 3.

Demographics overall and by latent class

	Analytic Sample N (%)	Class 1: Emotional/Physical Maltreatment N (%)	Class 2: Parental Separation N (%)	Class 3: Poly-Adversity N (%)	Class 4: Low Adversity N (%)
Variable
Age
17–20	973 (84.32%)	170 (79.81%)	123 (88.49%)	156 (85.71%)	524 (84.93%)
21+	178 (15.42%)	43 (20.19%)	16 (11.51%)	26 (14.29%)	93 (15.07%)
Gender
Male	827 (68.90%)	148 (69.48%)	95 (68.34%)	123 (67.58%)	430 (69.69%)
Female	320 (26.60%)	57 (26.76%)	39 (28.06%)	53 (29.12%)	158 (25.61%)
Race
White	725 (60.40%)	124 (58.22%)	99 (71.22%)	111 (60.99%)	364 (58.99%)
Black	102 (8.50%)	22 (10.33%)	9 (6.47%)	18 (9.89%)	50 (8.10%)
Asian	128 (10.70%)	23 (10.80%)	13 (9.35%)	18 (9.89%)	70 (11.35%)
Multiracial	116 (9.65%)	23 (10.80%)	9 (6.47%)	18 (9.89%)	60 (9.72%)
Other	52 (4.32%)	9 (4.23%)	3 (2.16%)	7 (3.85%)	31 (5.02%)

Association Between Latent Classes and Internalizing and Externalizing Outcomes

Table 4 presents means and standard deviations in outcomes for both the overall analytic sample and across latent classes to illustrate their respective associations with each of our outcomes. Overall, our sample of recruits endorsed low levels of anger (M = 45.99, SD = 10.72), anxiety (M = 50.11, SD = 9.58), or depression (M = 48.66, SD = 9.37), all of which fell below the cutoff for clinical significance. Substance use across the sample was very low as reflected by overall AUDIT (M = 1.93, SD = 3.70) and DAST-10 (M = 0.12, SD = 0.33) scores. However, the sample reported a noticeable level of engagement in various rule-breaking behaviors (M = 3.18, SD = 4.59). There were no significant differences in levels of anxiety, depression, anger, alcohol use, or illicit drug use across classes. However, there were small but statistically significant differences in rule-breaking behaviors, as demonstrated by a significant overall chi-square test (p < .05). In the BCH method, chi-square tests are used to compare expected to observed values related to latent class membership. Soldiers in the emotional/physical maltreatment class (M = 4.05) reported significantly higher levels of rule-breaking and deviant behaviors than those in the parental separation (M = 2.27) and poly-adversity classes (M = 2.63), but there were no differences between the maltreatment and low-adversity classes. We also found that soldiers in the parental separation class (M = 2.27) and the poly-adversity classes (M = 2.63) reported lower rule-breaking and deviant behavior than those in the low adversity class (M = 3.32).

Table 4.

Descriptives of outcome variables overall and by latent class

Outcomes	Analytic Sample M (SE)	MaltreatmentMT Class M (SE)	Parental SeparationPS Class M (SE)	Poly-AdversityP Class M (SE)	Low AdversityL Class M (SE)	Pairwise Comparisons
Anger (T-score)	45.99 (10.72)	46.93 (0.98)	44.20 (1.01)	46.85 (0.87)	45.88 (0.49)	-
Anxiety (T-score)	50.11 (9.58)	50.53 (0.85)	49.81 (1.02)	50.39 (0.81)	49.95 (0.43)	-
Depression (T-score)	48.66 (9.37)	49.14 (0.82)	48.50 (0.97)	48.18 (0.76)	48.67 (0.43)	-
Alcohol Use	1.93 (3.70)	2.30 (0.38)	1.23 (0.25)	1.82 (0.29)	2.02 (0.17)	-
Illicit Drug Use	0.12 (0.33)	0.14 (0.03)	0.08 (0.03)	0.09 (0.03)	0.14 (0.02)	-
Rule-breaking Behaviors	3.18 (4.59)	4.05 (0.51)	2.27 (0.33)	2.63 (0.27)	3.32 (0.21)	MT > PS**; MT > P*; L > PS*; L > P*

Note. All significant comparisons are included in the Pairwise Comparisons column. Pairwise comparisons based on BCH method.

^*p < .05.

^**p < .01.

Discussion

The present study addressed two research questions: 1) Are there distinct subgroups of National Guard recruits based on their reports of childhood adversities? and 2) If so, how do these subgroups differ concerning current mental health symptoms, substance use, and rule-breaking behaviors? Four profile groups emerged from our analyses, characterized by distinct patterns of ACEs, which is consistent with extant literature across both civilian and military samples. While approximately half of recruits were classified as having experienced Low Adversity, over one third were classified as having experienced Emotional/Physical Maltreatment or Poly-adversity, and about 1 in 6 recruits were classified as having experienced Parental Separation but no other childhood adversities. Broadly, this is consistent with extant literature which has examined ACEs using LCA (Wang et al., 2024) and has demonstrated that most general population samples consist of low adversity individuals while the rest of the sample consists of classes related to specific household dysfunction and maltreatment ACEs and a combination of both. However, the limited literature using LCA to examine ACEs in military samples has consisted of substantial variation in ACEs indicators used; thus, our resultant latent classes do not completely align with these previous studies.

ACEs are widely known to pose significant risks for negative physical, emotional, and behavioral health problems in adulthood. Despite endorsing multiple ACEs, most recruits in this study reported low levels of internalizing symptoms and substance use. Unexpectedly, we found no differences in mental health outcomes, including anger, anxiety, and depression symptoms, alcohol use and drug use, across distinct ACEs classes. Recruits with ACEs profiles characterized by poly-adversity and maltreatment reported similar levels of internalizing symptoms and substance use as did recruits in the low-adversity and parental separation groups. Although these findings are inconsistent with previous research which has frequently found that high/multiple ACEs and maltreatment groups are usually associated with the worst study outcomes (Wang et al 2024), our findings underscore the apparent resilience of new recruits. Our findings do not demonstrate or suggest that certain groups of people are not suitable for military service.

Previous research has documented associations between ACEs and internalizing symptoms in military populations (Armour, 2021; Gaska and Kimerling 2018; McLafferty et al., 2019; Reed-Fitzke et al., 2023; Turgoose and Murphy, 2024). However, these associations did not emerge in our sample of young recruits at the outset of their military careers. Most previous studies documenting these links have examined ACEs in samples of deployed military service members and veterans. Thus, the lack of associations between ACEs groups and internalizing symptoms and substance use problems in the present study may be related to the fact that we assessed recruits before exposure to military stressors. When considering possible explanations for the lack of significant differences in mental health symptoms, it is important to point out that most recruits in our predominately male sample reported minimal internalizing symptoms and low levels of externalizing problems. It is possible that recruits with more severe psychopathology (and perhaps greater exposure to multiple ACEs) were screened out during the military’s recruitment process and thus did not have the opportunity to participate in the present study. This may have resulted in a military sample “enriched” with resilient recruits (i.e., those demonstrating positive adaptation in terms of low internalizing and externalizing symptoms following exposure to childhood adversity).

The only study outcome associated with ACEs class membership was rule-breaking behaviors. Although there were no differences in rule-breaking behaviors between recruits in the Emotional/Physical Maltreatment and Low Adversity groups, both these groups reported higher counts of rule-breaking and deviant behavior than those in the Poly-adversity and Parental Separation classes. Our finding linking maltreatment and rule-breaking behaviors was not surprising and is consistent with previous literature (Ongah et al., 2021; Vachon et al., 2015). However, elevated levels of rule-breaking observed in the Low Adversity group suggests that other factors not assessed in the present study may have contributed to externalizing behaviors.

Previous research has established that the development of externalizing problems may be influenced by individual differences (e.g., self-esteem, personality), family processes (e.g., low supervision, absence of affection, strict parenting), socio-environmental factors (e.g., community violence, peer pressure, low-quality education), and cultural factors (e.g., social barriers such as inequality or discrimination) (Cordova et al., 2011; Perrotta et al., 2023). Therefore, it is possible that recruits coming from “good” homes are experiencing some of these other factors which have been linked to rule-breaking behaviors. Despite research suggesting a singular mechanism underlying psychopathology, deemed the ‘p’ factor, there are theories suggesting that response to stress and adversity is based on multiple levels of systems. Ecological systems theory (EST; Bronfenbrenner, 2000) suggests that development across the lifespan is based on an individual’s surrounding layers of systems. Differential susceptibility theory (Ellis et al., 2011) suggests there are individual differences in sensitivity to both positive and negative characteristics in their environments.

Contrary to our hypotheses, recruits in the Poly-Adversity class exhibited significantly lower rule-breaking behaviors than those in the Low Adversity and Emotional/Physical Maltreatment classes. The potential for “parentification” of children in dysfunctional homes (Boszormenyi-Nagy, 1965) provides one possible explanation for our findings related to lower rule-breaking and aggressive behaviors in the Poly-Adversity group. Parentification is described as children taking on parental roles in the home, often related to their caregivers’ inability to do so due to substance use, mental illness, incarceration, or other factors. This often leads to “overfunctioning” in the child, which manifests itself in excessive responsibility taking, people-pleasing, and perfectionism (Bowen, 1966). Reasons behind these findings may be explained by other factors at play in recruits’ homes that need further examination.

Limitations, Implications, and Future Directions

One limitation of the present study is that all the ACEs and outcomes data are self-reported, and we did not have a way to corroborate the self-reporting of these events using outside data sources. Additionally, ACEs reports were based on retrospective recall which may be biased by current levels of distress (Martin-Wagar et al., 2024). By sticking to Felitti et al.’s original ten ACEs, the overall scope of ACEs examined was limited, and our operationalization of ACEs did not include bullying, community violence, or poverty experiences that might correspond with important outcomes for military personnel. Another limitation of the study is that data regarding developmental timing, duration, and chronicity of ACEs was not collected. Life course theory (Elder, 1998) emphasizes the importance of these factors in stress response and would be important to consider in future research.

Additional research is needed to examine the relationship between ACEs, rule-breaking behaviors, and indicators of optimal military job performance (e.g., successful completion of BCT vs early separation or attrition). Such investigations can determine just how disruptive rule-breaking is during the training period, which can provide valuable insights into targets for intervention and implications for early intervention. Longitudinal research is needed to further examine the trajectories of the military career life cycle between servicemembers with various ACEs exposure. Investigating the interaction between ACEs and exposure to subsequent military stressors (i.e., during deployments) and their impact on outcomes beyond the training period will be crucial to identifying specific targets for intervention. Risk for negative outcomes could extend across various domains, including mental health, physical health, and reintegration/transition success. Research is also needed to examine other factors predictive of rule-breaking in the military. Better understanding of factors contributing to rule-breaking behaviors in the military could inform screening and discipline-related training processes.

Conclusion

Findings from the present study demonstrate that newly enlisted National Guard recruits display heterogeneous profiles of childhood adversities. Although recruits endorsed a notable amount of childhood adversities, most appeared resilient following their exposure to childhood adversity as demonstrated by low levels of anger, anxiety, and depression symptoms and substance use. However, the present study shows that rule-breaking behaviors are significantly associated with two ACEs profiles. Our finding that recruits in the Low Adversity and Emotional/Physical Maltreatment classes reported engaging in significantly higher rule-breaking behaviors than those in the Parental Separation and Poly-adversity classes also suggests that multiple factors may be at play related to rule-breaking behaviors among military recruits. While most recruits in our sample denied engaging in rule-breaking behaviors, such behaviors are worth paying specific attention to as discipline and obedience are some of the most important responsibilities of servicemembers, and a lack thereof can have catastrophic consequences including serious illness/injury or death.