Positive Childhood Experiences are Associated with Alcohol Use in Adolescent and Emerging Adult Females by ACEs Dimension

Abstract

Purpose

Experiencing multiple adverse childhood experiences (ACEs) is associated with alcohol use in female adolescents and emerging adults. Protective experiences during childhood (PACEs) have been theorized to off-set the health and behavioral consequences from the accumulation of ACEs throughout childhood. This study examines the association between protective experiences and subsequent alcohol and binge alcohol use frequency over one month among female adolescent and emerging adults reporting high and low levels of two ACE dimensions (household dysfunction and emotional abuse/neglect).

Methods

One hundred forty-three females between the ages of 15-24 who indicated at least one binge episode in the past two weeks completed the 6-item ACEs scale, the PACEs scale, and demographics at baseline. Alcohol consumption was measured prospectively over the next month during weekly appointments using the timeline follow back approach (TLFB).

Results

Two PACEs factors had significant direct associations, a source of unconditional love was associated with less frequent alcohol use (β = −.437, 95% CI −0.744, −0.131, exp(β) = 0.65, p = .005) in the context of high household dysfunction; and having a trusted adult to count on for help and advice (β = −1.373, 95% CI −2.283, −0.464, exp(β) = 0.25, p = 0.003) predicted fewer binge occasions in the context of high emotional abuse/neglect. Regardless of ACE dimension exposure, non-sport social group membership was associated more frequent alcohol use over the month across all ACE dimensions (β = 0.11-0.74, 95% CI −0.11, 0.74, exp(β) = 1.37 – 1.62, p ≤ 0.002); and having a trusted adult to count on for help and advice was associated with a 5.7 times more frequent of alcohol use among those with low household dysfunction (β = 1.74, 95% CI 0.83, 2.65, exp(β) = 5.70, p ≤ 0.001).

Discussion

Few PACE items are associated with direct reductions in alcohol outcomes. Indeed, there is consistently heightened risk associated with non-sport group membership for alcohol use frequency, regardless of experiences of childhood adversity. Future research should identify which protective factors have the most potential to off-set alcohol use by ACE dimension.

Adverse childhood experiences (ACEs), or the occurrence of abuse, and/or household dysfunction before the age of 18 are common, occurring in approximately 61% of adults. ACEs are widely recognized as important predictors of later health and behavioral outcomes in a dose-dependent manner.¹ Indeed, exposure to multiple ACEs is associated with alcohol use in adolescence, evident by accumulated ACEs increasing the likelihood of alcohol use initiation in adolescence by 20% to 70%.² Exposure to multiple ACEs has also been associated with poly use of alcohol and cannabis in adolescent and emerging adults.³ Meta-analytic evidence supports that moderate associations exist between multiple ACEs and heavy alcohol use, and strong associations exist between multiple ACEs and mental illness and problematic alcohol use.⁴ ACEs are more prevalent in those with substance use disorders than in the general population, and ACEs are positively associated with both the development and severity of substance use disorders in adolescence and adulthood.⁵ Importantly, gender differences have also been observed, wherein multiple ACEs are associated with at-risk alcohol use in females, but not males.⁶ There is evidence that adversity has differential impact by gender: males who experience household mental illness have a greater likelihood of heavy/binge drinking than females, and females who experience emotional abuse have a greater likelihood of binge/heavy drinking and alcohol use problems than males.⁷ Therefore, female adolescents and emerging adults with exposure to multiple ACEs should be considered as a target population for alcohol use prevention efforts. It is critical to understand the developmental and environmental contexts through which alcohol use behavior develops from exposure to childhood stress for interventions to be successful at preventing alcohol use in those with cumulative ACEs exposure.

Alcohol is the most commonly used substance among adolescents in the U.S. Nationally collected Youth Risk Behavior Surveillance System (YRBSS) data suggest that 73% of adolescents report initiating alcohol use by the 12^th grade and 30% report current use over the past 30 days.⁸ Adolescence through young adulthood is a critical developmental period and alcohol use during this time may impact developing systems. Risks associated with adolescent alcohol use include deficits in overall cognitive and executive functioning and interference with brain development.⁹ Importantly, the onset of regular drinking before age 15 is problematic, whereas the incidence of alcohol dependence increases when adolescents initiate alcohol use before the age of 15, and also increases when females initiate alcohol use between the ages of 15-17.¹⁰ Therefore, female adolescents and emerging adults with exposure to multiple ACEs who report alcohol use should be considered as a target population for alcohol use prevention efforts. It is critical to understand the developmental and environmental contexts which alcohol use behavior develops from exposure to childhood stress for interventions to be successful at preventing alcohol use in those with cumulative ACEs exposure.

ACEs and Alcohol Use Mechanisms

The exact mechanisms by which ACEs increase vulnerability to alcohol use are not entirely understood, with timing, frequency, duration, and type of adversity influencing the effect. Severe and chronic stress caused by childhood adversity can propel dysregulation of the hypothalamic pituitary adrenal (HPA) axis leading to hypo- or hyper-secretion of cortisol.¹¹ In children without exposure to adversity, morning basal cortisol levels increase from childhood through adolescence. However, in females who have experienced sexual abuse, there is evidence of a transition from hypercortisolism to hypocortisolism, where attenuation of cortisol activity can be detected in adolescence and significantly lower levels of cortisol by early adulthood.¹² Further, the structure and functions of brain regions with numerous cortisol receptors are affected, namely the corticolimbic system (prefrontal cortex, anterior cingulate cortex, amygdala, and hippocampus).¹³ ACEs can lead to neurocognitive adaptations that facilitate rapid identification of threat, yet result in impaired emotional regulation and executive control.¹⁴ Resulting alterations to meso-corticolimbic brain regions are implicated in reduced endogenous reward sensitivity and elevated ventral striatal dopamine response to substances.¹⁵ Psychological pathways of increased vulnerability are also implicated, with substance use viewed as an attempt to cope with and lessen negative affect produced by ACEs per a self-medication hypothesis.¹⁶ On the contrary, those with dysregulated stress response systems may feel the calming effects of alcohol more intensely, whereas GABAergic systems have been shown to mediate the self-administration of alcohol in animal models by stimulating the dopaminergic reward system.¹⁷ By understanding which protective and compensatory activities do the most work to change childhood stress-alcohol use trajectories, we can more reliably intervene to prevent alcohol use in high risk adolescents.

ACE Protective Factors

Importantly, not all individuals that are exposed to ACEs go on to develop later behavioral and health issues. To explain the multiple pathways through which ACEs impact social and behavioral outcomes throughout the lifespan, the role of compensatory and protective childhood experiences (PACEs) must be accounted for, as suggested by the Intergenerational and Cumulative Adverse and Resilient Experiences (ICARE) model (Figure 1).¹⁸ The ICARE model accounts for the co-occurrence of ACEs and PACEs during childhood to explain why inconsistencies in development of consequences from ACEs. PACEs include ten specific experiences needed for optimal development that can be categorized into two domains: supportive relationships (unconditional love from a caregiver (PACE 1), having a close friend (PACE 2), volunteering in the community (PACE 3), being part of a group (PACE 5), and having a mentor (PACE 7)), and enriching resources (engaging in physical activity (PACE 4), having a hobby (PACE 6), living in a safe, clean home with enough food (PACE 8), receiving a good education (PACE 9), and having rules and routine (PACE 10)). Theoretically, PACEs are associated with resilience in the face of ACEs, and therefore may mitigate the behavioral and health effects of adversity. For instance, support from a trusted adult mitigated the association between ACEs and alcohol consumption in adulthood in a sample of UK adults.¹⁹ As noted above, ACEs are associated with disruptions in cortisol production patterns and reactivity. One mechanism that may explain how PACEs mitigate negative impacts of adversity is through mitigating this influence. For example, PACE 4 accounts for childhood engagement in physical activity, and meta-analytic evidence has shown that physical activity is associated with more regulated diurnal cortisol patterns.²⁰

Figure 1.

The Intergenerational and Cumulative Adversity and Resilience Experiences Model (Adapted from Hays-Grudo & Morris, 2020). Copywrite American Psychological Association.

Importantly, PACEs may have direct effects and moderating effects that reduce the probability of harm as risk rises (such as in those with ACEs).²¹ Overall, a review of extant literature suggests mixed findings of PACEs as a protective factor for alcohol use in the context of childhood adversity (please see Han et al, 2023 for a review of this literature).²² This evidence suggests that positive childhood experiences may promote less drinking in adults without childhood adversity, but may not necessarily be protective for drinking outcomes in the context of childhood adversity. The heterogeneity in findings suggests that a more nuanced approach may be necessary to understand the contribution of PACEs to impact various behavioral and health outcomes in those with and without childhood adversity.

Cumulative vs. Dimensional Risk and Resilience

As evidenced above, much of the research on both ACEs and PACEs focuses on a cumulative risk/resilience approach, whereas regardless of the dimensional nature of childhood adversity (e.g., neglect, abuse, household dysfunction; or physical, emotional, environmental) or positive childhood experiences (e.g., supportive relationships, resources), ACEs and PACEs are summed to provide cumulative risk and protection scores. However, the cumulative risk approach assumes that all adverse experiences influence development through the same underlying mechanism; the stress response system. ²³ While many issues have been raised with this approach,²⁴ one of the most important issues is that it limits ACEs interventions to targeting one universal underlying mechanism (allostatic load). The cumulative protection approach also assumes that all protective factors are equally protective in all contexts. Few effective intervention approaches for children exposed to adversity have been developed as a result of cumulative risk models, and approaches that have been utilized for stress regulation have varying effects in children and adolescents.²⁵ In order to develop more targeted, effective interventions, some scholars have utilized a dimensional approach to ACEs, focusing on cumulative risks in each dimension to allow exploration of developmental mechanisms in each dimension beyond dysregulated stress response systems.²⁴

Present Study

While cumulative risk approaches have identified moderate associations between multiple ACEs heavy alcohol use and strong associations between multiple ACEs and problematic alcohol use across existent literature,⁴ this study adopts a cumulative dimensional approach to examine unique associations in two ACE dimensions; household dysfunction and emotional abuse/neglect, to begin to disentangle whether individual PACEs factors have direct effects on subsequent alcohol use among individuals with high and low exposure to these adversity dimensions. By accounting for the experience of individual PACEs factors in a Poisson regression model, this study is expected to highlight which PACEs have the strongest associations with prospective alcohol use in those with high and low levels of household dysfunction, as well as high and low levels of emotional abuse/neglect. We hypothesize that associations between each PACE and alcohol use (drinking occasions per month and frequency of consuming 4 or more drinks per occasion per month) will be negative; regardless of whether the individual has high or low levels of household dysfunction and emotional abuse/neglect, as these factors have been theoretically associated with resilience among individuals with experiences of childhood adversity,²⁶ and in the absence of childhood adversity, ²⁷ and therefore would be expected to predict lower quantity and frequency of alcohol use. We will explore which individual PACEs have the strongest associations with alcohol use in each dimension of ACEs to identify which factors may be the most important in female adolescent alcohol use in those with high levels of childhood household dysfunction and emotional abuse/neglect.

Method

Participants

This study used data collected as part of a larger study examining the effect of alcohol use on nutritional status in adolescent and young adult females. Recruitment efforts yielded 150 females between the ages of 15-24 to participate; this study included data from 143 participants who completed all measures of interest. Participants that were omitted from the current analysis dropped out after the baseline assessment (n = 1); 6 participants did not report data on PACEs or maternal educational status (less than 5%). Case-wise deletion was deemed the best option to address these cases. We performed a t-test to examine differences in age, race/ethnicity, ACEs sum score, and PACEs sum score to determine whether there were significant differences in these variables between participants who were included in this analysis and those who were not. Independent t-tests suggested that those who were omitted from this analysis were more likely to be younger than 21 years old (equal variances not assumed; t(148) = 9.48, p < .001) than those who were not omitted for this analysis, however, there were no other significant differences between these groups of participants. Eligibility criteria included residing 45 minutes from the research office, being between the ages of 14-24, and indicating a binge drinking episode in the two weeks prior to screening (defined as 4 or more drinks during a drinking event). Recruitment occurred through a combination of convenience sampling and respondent-driven sampling. Recruitment events were conducted at local schools; flyers were shared with community partners and posted locally; and participants were encouraged to identify other individuals who may be interested in participating in the study. Data were collected between February 2017 and November 2018. All procedures were approved by Oklahoma State University’s Institutional Review Board. Written informed consent and parental assent for minors (participants <18 years of age) was provided by all participants prior to study participation.

Procedure

At baseline, participants responded to several questionnaires including a 6-item adverse childhood experiences (ACEs) scale (Appendix A), the protective and compensatory experiences (PACEs) scale (Appendix A), and demographic questions. Over the subsequent month, participants attended weekly appointments and completed timeline follow back (TLFB) assessments of their daily alcohol use. At each visit, participants were asked to self-report their alcohol use using a TLFB approach.

Measures

Adverse Childhood Experiences

A brief version of the 10-Item ACEs scale ²⁸ was administered with six items in total. The ACEs scale is widely used and original versions of the ACEs scale have demonstrated good reliability and validity.²⁹ While the ACEs scale is usually summed to produce an overall cumulative score, more recent approaches examining ACEs by dimensionality allow for a better understanding of treatment approaches for individuals who have experienced specific ACEs.³⁰ Reflecting empirically-derived dimensions from confirmatory factor analysis on the ACEs scale,³¹ four items represent household dysfunction (parental separation/divorce, substance use, mental health issues, and legal issues) and two items represent emotional abuse/neglect (verbal abuse and feeling unloved; see Appendix A for the full scale and overarching dimensions). Physical and sexual abuse were not measured due to the study methodology and inability to quickly and appropriately respond to reports of abuse from minors in the sample. Each of the 6 items were coded dichotomously (1 = “yes” and 0 = “no”). Scores for each ACE dimension were summed, ranging from 0-4 for household dysfunction and 0-2 for emotional abuse/neglect, with a higher score indicating more exposure to each dimension of childhood adversity, respectfully. Then, each ACE dimension was split into high/low categories, with those who reported 0-1 items (household dysfunction) and 0 items (emotional abuse/neglect) categorized as low exposure, whereas those who reported 2-4 items (household dysfunction) and 1-2 items (emotional abuse/neglect) considered to have a high exposure in each dimension. In the household dysfunction dimension, we allowed one incident of household dysfunction to be considered low because the household dysfunction dimension contains both divorce/separation and household incarceration, and both experiences may remove conflict from the home and do not always result in negative outcomes for offspring. For emotional abuse/neglect, evidence is more consistent showing that both of these experiences lead to negative outcomes in children, therefore, we considered 0 experiences of emotional abuse/neglect to be low, and any indication of emotional abuse/neglect (1-2) to be high.

Protective and Compensatory Experiences

The Protective and Compensatory Experiences scale (PACEs) was utilized to assess the occurrence of positive childhood experiences before the age of 18 at baseline.¹⁸ The 10 items were coded dichotomously to represent endorsement of the experience (1 = “yes” and 0 = “no”). While the PACEs scale is frequently summed to produce a cumulative score, we assessed each PACE item independently as a predictor in the regression models. Preliminary evidence suggests the PACEs scale is a valid and reliable assessment of positive childhood experiences.³²

Alcohol Use

With the TLFB approach, alcohol use was measured over the month following baseline at weekly in-office visits. Participants were asked if they consumed alcohol on each day of the previous week, and if so, how many drinks were consumed each day. From this, scores were summed to create variables reflecting 1) alcohol use frequency over the subsequent month and 2) the frequency of 4 or more drinks per drinking occasion over the month. TLFB is a retrospective calendar-based method for gathering retrospective estimates of substance use. The interview-based TLFB for alcohol use exhibits sound reliability and validity, and has been shown to be a reliable measure to assess adolescent alcohol use.³³

Covariates

Participants reported their age, race and ethnicity, and maternal education status (a proxy for socioeconomic position in childhood).³⁴ Race was dichotomized as White (1) and non-White (0), and ethnicity was dichotomized as Hispanic (1) and non-Hispanic (2). Maternal education was dichotomized as “No college” (0) or “Some college or more” (1). All sociodemographic factors were included as covariates in each adjusted model, as these factors have been shown to correlate with alcohol use. Childhood adversity is socially patterned, such that children with low socioeconomic status are at increased risk for experiencing childhood adversity. Studies suggest an inverse relationship between socioeconomic status and PACEs, with higher income individuals more likely to experience PACEs. Religious participation, a proxy for overall religiosity, was assessed with the question “how often do you go to church and other religious activities?” reported on a 7-point Likert scale, where higher numbers indicated less attendance over the year. Religiosity was controlled for in post-hoc analyses to test whether the variance in PACE 5 was due to religion, as religious service attendance has been associated with risk and protection from alcohol use among adolescents.

Analytic Approach

Statistical analyses were conducted in IBM SPSS v. 29. Bivariate correlation analyses were conducted to examine the correlation between all study variables (Appendix C). Multicollinearity between PACE items was assessed with the variance inflation factor (VIF), whereas a VIF >= 2.5 is indicative of significant multicollinearity. Examination of the distribution of both the frequency of alcohol use and binge alcohol use suggested zero-inflation, and the Shapiro-Wilk test for normality confirmed that both the frequency of alcohol use (p < .001) and the frequency of binge alcohol use (p < .001) were not normally distributed. Model fit for both Poisson and Negative Binomial models were considered for each outcome for those with high/low household dysfunction and high/low emotional abuse/neglect. Model fit statistics are detailed in Appendix B. In all comparisons, Akaike’s information criterion (AIC) and Bayesian information criterion (BIC) differed slightly. When examining model fit in the alcohol frequency outcome, both the AIC and BIC were slightly lower than the AIC and BIC in the negative binomial output in those with high household dysfunction, yet the AIC and BIC were lower than the AIC and BIC in the Poisson output for those with low household dysfunction, high emotional abuse/neglect, and low emotional abuse/neglect, indicating a better fit. In the binge alcohol use frequency outcome, the AIC and BIC were slightly lower than the AIC and BIC in the negative binomial output in those with high household dysfunction, low household dysfunction, and low emotional abuse/neglect, yet the AIC and BIC were lower than the AIC and BIC in the Poisson output for those with high emotional abuse and neglect. However, the chi-square likelihood ratio for each of the Poisson analyses were significant. This suggested that the current model outperformed the null model, whereas the chi-square likelihood ratios were not significant for any of the tested negative binomial models. Therefore, Poisson regression analyses were interpreted. A priori power analyses were conducted with G*Power 3.1.9.6 for a Poisson regression. The analysis indicated that there is a 99% chance of correctly rejecting the null hypothesis that all the regression coefficients are simultaneously zero by following 49 participants over 28 days, providing ample power to detect effects in each ACE dimension (see Appendix C for full output). Separate Poisson regression analyses were conducted to examine which PACE items were significantly associated with the number of drinking occasions per month and the number of the frequency of binge drinking episodes per month (4+ per drinking occasion) in samples stratified by level (high/low) of each ACE dimension accounting for age (continuous), race, ethnicity (coded as binary variables), and maternal educational status (coded as a binary variable). The McFadden pseudo-R² was computed to examine the amount of variance accounted for in each outcome. Exploratory post-hoc Poisson regression analyses were conducted controlling for frequency of participation in religious activities and sensitivity analyses were conducted in the high and low household dysfunction groups with the ACE item measuring substance and alcohol use in the home omitted. To correct for multiple comparisons, Bonferroni corrections were applied to all a priori analyses (a = .006 for 8 models) and to all post hoc analyses (a = .003 for 14 total models).

Results

Participants

Sociodemographic characteristics are detailed in Table 1. Mean age of participants was 19.19 (2.65), and over half identified as White (61.5%), while 17.5% identified as American Indian/ Alaskan Native and 16.8% identified as Hispanic. Ninety-six participants (67.2%) indicated that their mother had attended some college or beyond, while 47 (32.7%) indicated that their mother did not attend any college. Only 4 participants did not drink alcohol during the month-long study duration. On average, participants drank 6.92 days over the month, and had 1.73 binge drinking episodes (four or more drinks in one drinking occasion). Ninety-four participants were less than 21 years old, and 49 were of legal drinking age. There was a significant difference in the frequency of alcohol use per month between those who were of legal drinking age (21+) and those who were not (equal variances not assumed; t(141) = −5.01, p < .001), but there was not a significant difference in the frequency of binge drinking episodes between those who were of legal drinking age (21+) and those who were not (equal variances not assumed; t(141) = −.029, p = .977). Figures 2 and 3 detail the frequency of alcohol use and binge alcohol use per month by age, respectively. Forty-eight participants (33.6%) indicated having high levels of household dysfunction, while only 50 (35%) indicated experiencing high levels of emotional abuse/neglect. Thirty-six participants (25.2%) reported experiencing all 10 PACE items, and the lowest number of PACEs experienced by any one participant was 4, indicating a high level of PACEs in this sample.

Table 1.

Participant sociodemographic characteristics

Variable	Total N = 149
Age (years)
M (SD)	19.09 (2.65)
(% missing)	(0%)
Race: N (%)
White	92 (61.7%)
Black/ African American	7 (4.7%)
American Indian/ Alaskan Native	25 (16.8%)
Asian	1 (.7%)
Native Hawaiian/Pacific Islander	0 (0%)
Other	24 (16.1%)
(% missing)	(0%)
Ethnicity: N (%)
Hispanic	26 (18.3%)
Non-Hispanic	116 (81.7%)
(% missing)	(4.7%)
Maternal Education Level: N (%)
Less than high school	12 (8.1%)
High School/ GED	35 (24.3%)
Some college/Technical School	30 (20.8%)
Community college/Technical school	11 (7.4%)
College graduate	35 (24.3%)
Graduate/professional degree	21 (14.6%)
(% missing)	(3.4%)
ACEs Dimensions N (%)
Household Dysfunction
High	51 (34.5%)
Low	97 (65.5%)
Emotional Abuse
High	53 (35.8%)
Low	95 (64.2%)
PACEs: N (yes), (% of endorsement)
PACE 1	139 (93.3%)
PACE 2	147 (98.7%)
PACE 3	101 (67.8%)
PACE 4	104 (69.8%)
PACE 5	87 (58.4%)
PACE 6	105 (70.9%)
PACE 7	138 (92.6%)
PACE 8	140 (94%)
PACE 9	133 (89.9%)
PACE 10	128 (85.9%)
(% missing)	(0%)
Alcohol Consumption Frequency: Days/Month
M (SD)	6.84 (4.81)
(% missing)	(0%)
Heavy Alcohol Consumption Frequency/Month
M (SD)	2.82 (2.62)
(% missing)	(0%)

Figure 2.

Number of Days of Alcohol Use Over the Month by Age

Figure 3.

Number of Prospective Binge Drinking Episodes Over the Month by Age

PACEs and the Frequency of Drinking Events Per Month

Household Dysfunction

Results are detailed in Tables 2 and 3. Bivariate correlations are detailed in Appendix C. The VIF did not indicate significant multicollinearity between PACEs in any analyses. In the model for those with high levels of household dysfunction, older age (b = .128, 95% CI .080, 1.75, p < 0.001) and membership in a non-sport related social group (e.g., scouts, youth group) before the age of 18 significantly predicted more alcohol use occasions per month (PACE 5; b = .482, 95% CI .225, 7.39, p < 0.001). This suggests that each additional year of age was associated with a 13.6% increase in drinking days (IRR = 1.136), and membership in a non-sport related social group was associated with a 61.9% increase in drinking days (IRR = 1.619). Only having unconditional love (PACE 1; b = −.437, 95% CI −.744, −.131, p = .005) significantly predicted less frequent alcohol use over the month, wherein those who endorsed this experience drank 64.6% less days than those who did not (IRR = .646). For those with low levels of household dysfunction, older age (b = .125, 95% CI .095, .156, p < 0.001), membership in a non-sports social group (PACE 5; b = .331, 95% CI .135, .527, p < 0.001), and having a trusted adult to count on for help and advice (PACE 7; b = .125, 95% CI .095, .156, p < 0.001) significantly predicted more frequent alcohol use over the prospective month. For each additional year of age, there was a 13.4% increase in drinking days (IRR = 1.134). Compared to those who did not endorse membership in a non-sports related social group, there was a 39.2% increase in drinking days (IRR = 1.392), and compared to those who did not endorse having a trusted adult to count on for help and advice, there was a 69.8% increase in drinking days (IRR = 5.698). The predictors accounted for a significant amount of variance in the days of alcohol use in the high household dysfunction group (likelihood ratio c²(3.54) = 59.32, p < .001, McFadden pseudo-R² = .215), representing approximately 21.5% of the variance. In the low household dysfunction group, the predictors also accounted for a significant amount of variance in the days of alcohol use (likelihood ratio c²(1.91) = 146.90, p < .001, McFadden pseudo-R² = .254), representing approximately 25.4% of the variance in alcohol use frequency.

Table 2.

Multivariate regression analyses of the association between childhood protective factors and the frequency of alcohol use over the subsequent month in each ACE dimension subgroup

	Model 1a						Model 2
			95% CI
			95% CI
High Household Dysfunction	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	−.147	.399	−.852	−6.717	2.732	−.271	.130	−1.556	−8.436	1.129
PACEs 2	NC
PACEs 3	−.024	.902	−.125	−4.958	4.382	−.046	.807	−.247	−5.224	4.096
PACEs 4	−.025	.888	−.142	−4.174	3.626	.056	.741	.333	−3.236	4.502
PACEs 5	.238	.148	1.474	−.921	5.888	.343	.048	2.056	.034	7.299
PACEs 6	.089	.602	.525	−3.615	6.156	.062	.710	.375	−4.348	6.312
PACEs 7	−.050	.775	−.288	−5.667	4.253	.006	.976	.031	−5.438	5.605
PACEs 8	−.026	.878	−.155	−5.575	4.781	−.055	.752	−.319	−6.045	4.408
PACEs 9	−.011	.947	−.067	−5.129	4.801	−.077	.644	−.467	−6.177	3.873
PACEs 10	−.160	.332	−.981	−5.482	1.899	−.031	.858	−.180	−4.246	3.556
Age						.428	.010	2.727	.240	1.659
Maternal Ed						.019	.916	.106	−3.759	4.172
Race						.251	.214	1.268	−1.630	7.007
Ethnicity						.016	.931	.087	−4.910	5.349
Low Household Dysfunction	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	.275	.055	−.158	−.258	25.205	.193	.120	1.572	−2.291	19.412
PACEs 2	−.252	.074	1.948	−17.058	.802	−.156	.207	−1.272	−12.649	2.791
PACEs 3	.023	.838	−1.810	−1.966	2.419	.002	.988	.015	−1.961	1.991
PACEs 4	.039	.722	.206	−1.840	2.646	.059	.550	.600	−1.425	2.652
PACEs 5	.266	.017	.358	.466	4.672	.263	.009	2.692	.662	4.432
PACEs 6	.016	.875	2.429	−1.808	2.120	−.020	.824	−.223	−1.944	1.553
PACEs 7	.275	.013	.158	1.606	12.985	.271	.006	2.845	2.104	11.933
PACEs 8	−.235	.079	2.550	−16.087	.903	−.244	.040	−2.091	−15.033	−.363
PACEs 9	−.124	.223	−1.777	−5.414	1.282	−.046	.623	−.494	−3.761	2.266
PACEs 10	.176	.171	−1.227	−1.608	8.910	.228	.051	1.979	−.031	9.281
Age						.506	<.001	5.739	.557	1.149
Maternal Edu						.025	.798	.257	−1.734	2.247
Race						−.133	.256	−1.146	−3.585	.967
Ethnicity						.074	.550	.600	−2.077	3.867
High Emotional Abuse	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	.002	.991	.011	−5.064	5.121	−.145	.476	−.721	−7.297	3.477
PACEs 2	.115	.484	.706	−8.218	17.072	.144	.404	.846	−7.553	18.306
PACEs 3	.066	.676	.420	−3.355	5.120	−.003	.984	−.020	−4.295	4.212
PACEs 4	.039	.810	.242	−3.417	4.349	.032	.843	.199	−3.435	4.181
PACEs 5	.266	.087	1.750	−.431	6.052	.359	.061	1.940	−.188	7.930
PACEs 6	−.021	.898	−.128	−5.037	4.434	−.095	.578	−.562	−6.633	3.760
PACEs 7	−.270	.131	−1.541	−14.131	1.892	−.212	.245	−1.185	−12.743	3.365
PACEs 8	−.006	.971	−.037	−5.164	4.978	−.045	.787	−.272	−5.802	4.433
PACEs 9	−.074	.629	−.486	−5.908	3.613	−.032	.833	−.212	−5.153	4.179
PACEs 10	−.155	.331	−.983	−5.523	1.905	−.053	.781	−.280	−5.035	3.815
Age						.418	.015	2.571	.183	1.567
Maternal Edu						−.059	.758	−.311	−4.943	3.634
Race						.088	.695	.396	−3.998	5.930
Ethnicity						−.049	.801	−.254	−5.998	4.664
Low Emotional Abuse	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	NC	-	-	-	-	-	-	-	-	-
PACEs 2	−.164	.117	−1.585	−16.131	1.825	NC	.307	−1.028	−12.196	3.896
PACEs 3	−.057	.650	−.456	−2.806	1.759	−.025	.825	−.222	−2.340	1.870
PACEs 4	.028	.817	.232	−2.073	2.621	.071	.537	.620	−1.557	2.961
PACEs 5	.229	.066	1.865	−.141	4.378	.234	.039	2.106	.117	4.251
PACEs 6	.104	.350	.940	−1.083	3.026	.035	.739	.335	−1.628	2.285
PACEs 7	.196	.082	1.758	−.438	7.117	.240	.021	2.358	.673	8.026
PACEs 8	−.099	.374	−.893	−14.017	5.328	−.025	.806	−.246	−9.958	7.766
PACEs 9	−.165	.131	−1.526	−6.471	.851	−.124	.257	−1.142	−5.741	1.558
PACEs 10	.098	.371	.900	−2.370	6.284	.132	.179	1.356	−1.222	6.427
Age						.484	<.001	4.976	.498	1.163
Maternal Edu						.048	.650	.455	−1.588	2.528
Race						−.058	.639	−.471	−2.869	1.773
Ethnicity						.096	.487	.699	−2.166	4.508

Note. Separate multivariate regression analyses were run for each ACE subgroup. ^a Model 1 for high household dysfunction and high emotional abuse have no predictive value (indicated by a negative R² statistic); NC = standardized beta not computed; PACE 2 cannot be estimated in high household dysfunction and PACE 1 cannot be estimated in low emotional abuse due to being constants or having missing correlations; β = standardized beta; p = p- value, t = t statistic, 95% CI = 95% confidence interval; LLCI = Lower level confidence interval; ULCI = upper level confidence interval; age = continuous; Maternal Edu = maternal educational attainment (0 = no college, 1 = college); Race (0 = not white, 1 = white); Ethnicity (0 = not Hispanic, 1 = Hispanic).

Table 3.

Multivariate regression analyses of the association between childhood protective factors and the frequency of binge alcohol use over the subsequent month in each ACE dimension subgroup

	Model 1					Model 2
			95% CI					95% CI
High Household Dysfunction	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	.073	.672	.427	−1.93	2.96	.046	.814	.237	−2.403	3.036
PACEs 2	-	-	-	-	-	-	-	-	-	-
PACEs 3	−.206	.294	−1.064	−3.688	1.145	−.190	.371	−.908	−3.831	1.469
PACEs 4	.092	.603	.524	−1.494	2.541	.122	.521	.649	−1.500	2.901
PACEs 5	.010	.952	.061	−1.708	1.814	.013	.946	.069	−1.996	2.135
PACEs 6	.183	.286	1.081	−1.175	3.880	.098	.596	.536	−2.233	3.829
PACEs 7	.016	.925	.094	−2.447	2.686	−.014	.946	−.068	−3.245	3.035
PACEs 8	−.248	.153	−1.456	−4.609	.749	−.224	.251	−1.169	−4.678	1.267
PACEs 9	−.162	.327	−.992	−3.829	1.308	−.208	.267	−1.130	−4.443	1.272
PACEs 10	.074	.652	.454	−1.480	2.338	.150	.435	.790	−1.358	3.079
Age						.243	.173	1.395	−.127	.680
Maternal Edu						.055	.784	.277	−1.949	2.562
Race						.070	.750	.321	−2.069	2.843
Ethnicity						.119	.553	.599	−2.059	3.775
Low Household Dysfunction	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	.097	.518	.650	−5.069	9.991	.037	.803	.251	−6.335	8.160
PACEs 2	−.060	.684	−.409	−6.368	4.195	.036	.807	.245	−4.522	5.791
PACEs 3	−.094	.437	−.781	−1.806	.787	−.126	.309	−1.024	−1.998	.641
PACEs 4	.153	.193	1.311	−.452	2.202	.156	.191	1.321	−.459	2.264
PACEs 5	.219	.063	1.885	−.064	2.423	.202	.089	1.721	−.172	2.346
PACEs 6	−.018	.871	−.163	−1.257	1.066	−.047	.664	−.437	−1.424	.912
PACEs 7	.160	.165	1.400	−.995	5.735	.156	.177	1.363	−1.037	5.528
PACEs 8	−.118	.400	−.846	−7.162	2.887	−.095	.499	−.679	−6.569	3.229
PACEs 9	.076	.476	.715	−1.268	2.693	.165	.142	1.484	−.514	3.512
PACEs 10	.148	.275	1.099	−1.391	4.830	.139	.316	1.010	−1.533	4.686
Age						.334	.002	3.159	.116	.511
Maternal Edu						.185	.124	1.555	−.292	2.367
Race						.008	.954	.058	−1.476	1.564
Ethnicity						−.070	.636	−.475	−2.458	1.512
High Emotional Abuse	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1	.272	.118	1.597	−.466	4.006	.273	.109	1.646	−.409	3.880
PACEs 2	−.007	.961	−.049	−5.688	5.417	−.051	.718	−.364	−6.068	4.227
PACEs 3	−.122	.393	−.864	−2.657	1.064	−.195	.147	−1.487	−2.931	.456
PACEs 4	.092	.532	.630	−1.173	2.237	.112	.400	.853	−.880	2.152
PACEs 5	.191	.171	1.393	−.441	2.405	.330	.038	2.163	.102	3.334
PACEs 6	−.071	.624	−.493	−2.587	1.571	−.136	.335	−.979	−3.064	1.074
PACEs 7	−.498	.003	−3.147	−9.003	−1.968	−.443	.005	−3.001	−7.937	−1.524
PACEs 8	−.091	.539	−.620	−2.910	1.543	−.027	.844	−.199	−2.236	1.838
PACEs 9	−.402	.006	−2.916	−5.110	−.930	−.360	.007	−2.881	−4.488	−.772
PACEs 10	.101	.484	.706	−1.061	2.201	.028	.856	.183	−1.603	1.920
Age						.381	.008	2.842	.109	.660
Maternal Edu						.166	.292	1.070	−.809	2.605
Race						−.225	.229	−1.226	−3.167	.786
Ethnicity						−.071	.657	−.448	−2.590	1.655
Low Emotional Abuse	β	p	t	LL CI	UL CI	β	p	t	LL CI	UL CI
PACEs 1						-	-	-	-	-
PACEs 2	−.005	.966	−.043	−5.577	5.340	.048	.672	.425	−4.426	6.824
PACEs 3	−.147	.256	−1.144	−2.186	.590	−.132	.335	−.971	−2.189	.755
PACEs 4	.111	.383	.877	−.798	2.056	.100	.470	.726	−1.004	2.155
PACEs 5	.107	.402	.842	−.792	1.955	.118	.378	.888	−.802	2.089
PACEs 6	.122	.293	1.058	−.585	1.913	.068	.583	.551	−.989	1.746
PACEs 7	.165	.155	1.434	−.640	3.953	.152	.218	1.244	−.966	4.175
PACEs 8	−.151	.196	−1.305	−9.739	2.023	−.092	.460	−.743	−8.507	3.885
PACEs 9	.116	.303	1.036	−1.067	3.385	.132	.314	1.014	−1.254	3.850
PACEs 10	.056	.624	.492	−1.980	3.281	.081	.492	.690	−1.748	3.600
Age						.258	.030	2.207	.025	.490
Maternal Edu						.061	.635	.476	−1.095	1.783
Race						.031	.834	.210	−1.452	1.793
Ethnicity						.080	.631	.483	−1.768	2.898

Note. Separate multivariate regression analyses were run for each ACE subgroup. ^a Models 1 and 2 for high household dysfunction have no predictive value (indicated by a negative R² statistic); NC = standardized beta not computed; PACE 2 cannot be estimated in high household dysfunction and PACE 1 cannot be estimated in low emotional abuse due to being constants or having missing correlations; β = standardized beta; p = p-value, t = t statistic, 95% CI = 95% confidence interval; LL CI = Lower level confidence interval; UL CI = upper level confidence interval; age = continuous; Maternal Edu = maternal educational attainment (0 = no college, 1 = college); Race (0 = not white, 1 = white); Ethnicity (0 = not Hispanic, 1 = Hispanic).

Emotional Abuse

For high emotional abuse/neglect, both age (b = .107, 95% CI .060, 1.53, p < 0.001) and non-sports group membership (PACE 5; b = .438, 95% CI .162, .714, p = .002) significantly predicted prospective alcohol use frequency. For each additional year in age, there was a 11.3% increase in drinking frequency (IRR = 1.113), and in those who endorsed membership in a non-sports group, there was a 55% increase in alcohol use frequency (IRR = 1.550). In those with low emotional abuse/neglect, older age (b = .132, 95% CI .100, .164, p < 0.001), membership in a non-sports social group (PACE 5; b = .313, 95% CI .113, .512, p = .002), and having a trusted adult to count on for help and advice (PACE 7; b = .946, 95% CI .443, 1.488, p < 0.001) significantly predicted more frequent alcohol use over the prospective month. For each additional year of age, there was a 14.1% increase in drinking days (IRR = 1.141). Compared to those who did not endorse membership in a non-sports related social group, there was a 36.7% increase in drinking days (IRR = 1.367), and in those who endorsed having a trusted adult to count on for help and advice, there was a 57.5% increase in drinking days (IRR = 2.575). The predictors accounted for a significant amount of variance in the days of alcohol use in the high emotional abuse/neglect group (likelihood ratio c²(3.16) = 69.47, p < .001, McFadden pseudo-R² = .243), representing approximately 24.3% of the variance. In the low emotional abuse/neglect group, the predictors also accounted for a significant amount of variance in the days of alcohol use (likelihood ratio c²(2.29) = 124.25, p < .001, McFadden pseudo-R² = .223), representing approximately 22.3% of the variance in alcohol use frequency.

PACEs and the Frequency of Binge Drinking Events Per Month

Household Dysfunction

In high household dysfunction, none of the variables were significantly associated with binge drinking frequency. In those with low household dysfunction, the only predictor of binge drinking frequency was maternal educational status (b = .641, 95% CI .203, 1.079, p = 0.004), whereas in participants who endorsed that their mothers did not attend college, there was an 89.9% increase in binge drinking events (IRR = 1.898). The predictors accounted for a significant amount of variance in the frequency of binge alcohol use in the high household dysfunction group (likelihood ratio c²(2.75) = 27.39, p < .05, McFadden pseudo-R² = .174), representing approximately 17.4% of the variance in binge drinking frequency. In the low household dysfunction group, the predictors also accounted for a significant amount of variance in the days of alcohol use (likelihood ratio c²(2.18) = 27.92, p < .05, McFadden pseudo-R² = .160), representing approximately 16% of the variance in binge alcohol use frequency.

Emotional Abuse

In those with high emotional abuse, having a trusted adult to count on for help and advice (PACE 7; b = −1.373, 95% CI −2.283, −.464, p = 0.003) predicted less binge drinking occasions over the subsequent month than those who did not endorse this item, such that in participants who endorsed this experience, there was a 25.3% decrease in binge drinking events (IRR = .253). In those with low emotional abuse/neglect, there were no significant predictors of binge drinking frequency in this sample. The tested PACEs and covariates accounted for a significant amount of variance in the frequency of binge alcohol use in the high emotional abuse/neglect group (likelihood ratio c²(1.88) = 58.65, p < .001, McFadden pseudo-R² = .357), representing approximately 35.7% of the variance. In the low emotional abuse/neglect group, the predictors also accounted for a significant amount of variance in the days of alcohol use (likelihood ratio c²(2.10) = 24.36, p < .05, McFadden pseudo-R² = .094), representing approximately 9.4% of the variance in binge alcohol use frequency over the subsequent month.

Post-Hoc Analyses

Exploring PACE 5

Post-hoc analyses were conducted to explore whether religion may explain the variance in PACE 5; membership in a non-sport related social group (e.g., scouts, youth group). Adding religiosity did not account for the variance in PACE 5 (social group membership), which remained significant in the adjusted models for the frequency of alcohol use among those with high household dysfunction (PACE 5; b = .471, 95% CI .212, .729, p < .001), high emotional abuse/neglect (PACE 5; b = .452, 95% CI .177, .727, p = .001), low household dysfunction (PACE 5; b = .343, 95% CI .146, .540, p < .001), and low emotional abuse/neglect (PACE 5; b = .307, 95% CI .106, .508, p = .003).

Sensitivity Analysis: Household Dysfunction

Both environmental (norms, modeling behaviors, availability of alcohol, alcohol use expectancies)^41,42 and genetic⁴³ familial transmission effects influence adolescent alcohol use behaviors. While the household dysfunction dimension in the current analysis includes the item “did you live with anyone who was a problem-drinker or alcoholic or who used street drugs or prescription drugs not as prescribed”, environmental and genetic transmission effects may be present, and are important to account for in this set of analyses. To account for the contribution of environmental and genetic risk, a sensitivity analysis was conducted with this item omitted to examine the impact on associations of each PACE item and alcohol use frequency and binge alcohol use frequency in those with high and low household dysfunction. The household dysfunction dimension was re-coded to reflect whether the participant experienced 0-1 instances of household dysfunction representing mental illness in the household, household incarceration, and parental divorce/separation (0, n = 112) or 2-3 instances of household dysfunction (1, n = 31).

Results are detailed in Appendix E. In those with high household dysfunction, unconditional love (PACE 1; b = −.876, 95% CI −1.346, −.405, p < .001) and attending a school that provided the resources and academic experiences the individual needed to learn (PACE 9; b = −.979, 95% CI −1.472, −.487, p < .001) were associated with less frequent alcohol use. This suggests that in those who endorsed unconditional love there was a 41.7% decrease in drinking days (IRR = .417), and in those who endorsed going to a school with resources that facilitated learning there was a 37.6% decrease in days of alcohol use over the subsequent month (IRR = .376). On the contrary, regular involvement in sports group or other organized forms of physical activity (PACE 4; b = .679, 95% CI .234, 1.160, p = .003), older age (b = .172, 95% CI .081, .262, p < .001), and endorsing a race category other than White (b = 1.271, 95% CI .509, 2.232, p = .002) were significantly associated with more frequent drinking over the month. In those with low household dysfunction both non-sports group membership (b = .401, 95% CI .227, .575, p < .001) and older age (b = .115, 95% CI .087, .143, p < .001) were predictive of more frequent drinking over the month. For those with high household dysfunction, involvement in sports group or other organized forms of physical activity (PACE 4; b = 1.481, 95% CI .574, 2.387, p = .001) was the only significant predictor for binge alcohol use frequency, and there were no significant predictors of binge alcohol use frequency in those with low household dysfunction.

Discussion

This study provides evidence that PACEs have varying strengths of association with alcohol use frequency in adolescent and emerging adult females exposed to varying levels of (low/high) of ACE dimensions (household dysfunction and emotional abuse/neglect). While we hypothesized that PACEs would be inversely related to alcohol use, only 2 PACEs were inconsistently inversely related to alcohol use, 2 PACEs were consistently positively associated with alcohol use frequency. Critically, findings did not identify any significant PACEs that were inversely related to alcohol use for those who endorsed low adversity dimensions. Indeed, contrary to our hypotheses in all high and low dimensions of adversity, alcohol use frequency was positively associated with membership in a non-sports group (PACE 5). In those with low household dysfunction and low emotional abuse/neglect, having a trusted adult to count on for help and advice (PACE 7) was positively associated with alcohol use frequency. Having unconditional love from an adult (PACE 1) was inversely associated with alcohol use frequency only for those with high household dysfunction. Among those who endorsed emotional abuse/neglect, having a trusted adult to count on for help and advice (PACE 7) was significantly inversely associated with binge drinking. These results demonstrate that outcomes of interest are critical for determining protective and compensatory childhood experiences in the context of alcohol use behavior; indeed, participating in a non-sport group was associated with more frequent drinking regardless of household dysfunction and emotional abuse/neglect.

Age and Drinking Frequency

Across models, alcohol use frequency was associated with older age. Further, after omitting familial transmission effects, older age remained a significant predictor of alcohol use frequency in those with both high and low household dysfunction. While the current sample included those both below and above legal drinking age in the U.S. (21+), the consistent findings for older age may be related to the participant’s ability to access and purchase alcohol. These results may also be explained by the initiation of college, whereas those who attend college have higher rates of drinking than same age peers who do not attend college.⁴⁴ Further, emerging adults may experience important role transitions that influence drinking behavior such as employment transitions, marital transitions, and transitioning into a parental role. Previous literature has shown that the acquisition of a spousal role and a parental role was associated with a decrease in alcohol consumption or heavy drinking, but this was not true for employment transitions.⁴⁵ Further, divorce among women was associated with an increase in heavy drinking.⁴⁵ As such, as age increases, alcohol availability, transitioning to a college environment, and transitioning into employment may all impact drinking frequency in emerging adults as we observed in this sample.

Unconditional love (PACE 1) and Alcohol Use Frequency

Only in the high household dysfunction group was an individual PACE associated with less frequent alcohol use– wherein having a source of unconditional love was negatively associated with alcohol consumption frequency (PACE 1). Further, after omitting parental substance and alcohol use from those household dysfunction dimension, unconditional love remained a significant predictor of less frequent alcohol use in those with high household dysfunction. Having a loving adult is associated with less frequent alcohol use, but only in the context of high household dysfunction (parental divorce/separation, parental alcohol and substance use, household mental illness, and household incarceration). Household dysfunction may affect the parent’s involvement with and sensitivity to the child in various ways, such that a parent may be removed from the home environment completely (parental divorce/separation, household incarceration), or may be less sensitive to child needs due to internal factors (household mental illness, parental alcohol and substance use). The level of involvement and parental sensitivity may be one moderating factor in the context of household dysfunction. For example, previous literature has shown that maternal sensitivity moderated divorce effects on both internalizing and externalizing behavioral problems in children between the ages of 11-15, whereas when the post-divorce home environment was less supportive and their mother was less sensitive, child outcomes tended to be more negative.⁴⁶ As such, despite high levels of household dysfunction that occur, having unconditional love from an adult is negatively associated with the frequency of alcohol use in the current sample.

Social group membership (PACE 5 and PACE 4) and Alcohol Use

Across both high and low ACE dimensions, non-sport group membership (PACE 5) was consistently associated with a higher frequency of alcohol consumption. Since the item states “when you were growing up, prior to your 18th birthday: - were you an active member of at least one civic group or a non-sport social group such as scouts, church, or youth group?”, post-hoc analyses were conducted controlling for frequency of religious service attendance (a proxy of religiosity) to disentangle variance contributed through social group membership and religiosity. Religious attendance did not account for the variance for non-sport membership in any post hoc analyses performed. Further, after omitting familial transmission effects from the household dysfunction dimension, regular involvement in sports groups or other organized forms of physical activity (PACE 4) was significantly associated with more frequent alcohol use and binge alcohol use in those with high household dysfunction. From a developmental perspective, during adolescence less time is spent with parents, family involvement decreases, adolescents gain autonomy from parents, and the quality of parent-child communication is reduced.⁴⁷ In parallel, there is an increase in time spent with peers, heightened sensitivity to social reward, and engagement in novel experiences that emphasize socializing with peers and conforming to perceived peer group standards.⁴⁸ Importantly, peers become more influential than parents during the transition from childhood to adolescence.⁴⁹ Indeed, PACEs related to social networks may increase access and opportunity to use alcohol, and therefore represent a risk based on the dimensional experiences of cumulative ACEs. Time with friends has been associated with higher alcohol use in young adult females,⁵⁰ so it may be that more time with friends results in a greater peer influence to use alcohol in some groups, including religious groups among those with higher levels of childhood trauma. During adolescence and emerging adulthood, peer influence becomes increasingly important in terms of drinking behaviors. In emerging adults (18-29) negative peer pressure has been associated with more binge drinking and lifetime alcohol use, whereas positive peer pressure made emerging adults less likely to engage in lifetime alcohol use than those with no peer pressure.⁵¹ Further, females are influenced by peer drinking behaviors more than males.⁵² As such, female adolescents may be particularly vulnerable to peer influences in terms of drinking behaviors, which may explain the positive association between alcohol use frequency and membership in a non-sports related social group across all ACE dimensions in the current study. This finding should be explored in future studies as social group membership is a consistent risk factor for frequency of alcohol use across adversity contexts.

Having a Trusted Adult to Count on for Help and Advice (PACE 7) and Alcohol Use

Having a trusted adult to count on for help and advice predicted less frequent binge drinking episodes among individuals with high emotional abuse/neglect; however, in both the low ACE dimensions, having a trusted adult to count on for help and advice predicted more frequent alcohol use. While adult females with childhood emotional abuse and neglect are more likely to report binge/heavy drinking and alcohol problems than males,⁷ this finding is demonstrates that having a trusted adult to count on for help and advice has a direct effect on reducing alcohol binge drinking; however, this finding is particularly interesting because it highlights the nuance around having a trusted adult to count on for help and advice, which is also a risk for increased frequency of alcohol use in low ACE contexts. The varying effects on alcohol use in high and low adversity contexts could be due to wide ranging effects inclusive of social support dimensions which have evidence of protective effects in the context of adversity,⁵³ and/or having an adult mentor.⁵⁴ Future studies should contextualize the relationships with trusted adults to understand the indirect effect that they may have on adolescent and emerging adult drinking behavior in the context of cumulative dimensional risk.

Maternal Educational Status and Binge Alcohol Use

In participants with low household dysfunction who endorsed that their mothers did not attend college, there was an 89.9% increase in binge drinking events across the prospective month. This finding may be related to one of the proposed explanations of the alcohol harm paradox –which posits that individuals with a lower socioeconomic status tend to engage in less frequent alcohol consumption, but engage in heavier alcohol consumption, than individuals with a higher socioeconomic status. However, after omitting the parental alcohol and substance use item from the household dysfunction dimension (see Appendix E), maternal educational status was no longer associated with binge alcohol use in this dimension. Importantly, the sensitivity analysis accounted for high household dysfunction in 17 participants, whereas after omitting this item, only 31 individuals were categorized as having high household dysfunction. This suggests that parental alcohol and substance use may be driving this association. Parental university education has been shown to moderate the association between problematic familial alcohol use and binge drinking in adolescents, whereas in those without a university-educated parent and with one university-educated parent, the relationship between familial alcohol use and binge drinking was stronger than in those with two university-educated parents.⁵⁶ As such, parental educational status may make adolescents more vulnerable to familial transmission effects in terms of binge alcohol use, especially when other forms of household dysfunction are low.

Model Variance

The PACEs and covariates investigated in those with both high and low household dysfunction and emotional abuse/neglect accounted for around 20-25% of the variance in drinking frequency, leaving 75-80% of the variance in the model unaccounted for. Theoretically, we expected PACEs to account for more variability in alcohol use among those with high ACEs, however, PACEs accounted for the same amount of variability in drinking frequency in those with high and low ACEs. Indeed, there are factors related to variance in a high ACE context that are not captured by PACEs experienced in childhood, which, perhaps is due to lack of measurement of current PACE behaviors. Meanwhile, for binge drinking frequency, high and low household dysfunction and low emotional abuse/neglect account for less variance (between 9 – 17% of variance), whereas a sizeable and actionable 35.7% of the variability in heavy drinking frequency was accounted for in the high emotional abuse/neglect model, reflecting a good model. In high and low household dysfunction and low emotional abuse/neglect, these models left much of the variability unaccounted for. These differences in variability highlight the potential for other factors to be investigated and utilized as avenues for helping aid young women and reducing their adolescent and early adulthood exposure to alcohol.

Strengths, Limitations, and Future Directions

Strengths of the current study include using a prospective design to examine the unique effects of individual PACEs on young women’s alcohol use while accounting for high vs low levels of ACEs. This study provides a more in-depth understanding of the complex relationship between ACEs and PACEs, showing that specific PACEs do appear to have varying effects at different levels of ACEs. Another notable strength is that this sample was racially and ethnically diverse, which adds to the generalizability in these findings (Table 1). However, it should be noted that the interpretability of our findings is limited by our small sample size, particularly within the ACEs subgroups. The sample size also limits the degree to which we were able to examine unique effects of sociodemographic characteristics. A larger sample would allow for the investigation of multinomial sociodemographic categories rather than binomial categories, and comparisons of between and within group effects. While the current study was interpreted based on a Bonferroni correction to account for multiple model comparisons to reduce the possibility of Type I error, the Bonferroni correction is conservative and has a high penalty in light of small samples, and may result in a high Type II error rate. The findings should be interpreted with proper caution due to the possibility of Type II error. Therefore, it is imperative that these results are replicated in much larger, more representative samples. Further, our short follow-up period likely captures an accurate understanding of typical drinking behaviors, but it would be useful to understand the effects over a longer period, particularly from the beginning of adolescence to young adulthood. It is also key to note that this study utilized the 6-item ACEs, and these findings may not hold for those with high levels of sexual or physical abuse. Consideration and investigation including the full 10-item ACEs and additional 2 ACE dimensions that were omitted here (physical and sexual abuse) should be conducted. Analysis incorporating these variables would further elucidate the full relationship investigated herein.

Future research should replicate our study over longer periods of time with larger sample sizes to expand on existing knowledge of the interplay between PACEs, ACEs, and alcohol use among young women. Doing so could identify the most important PACEs, especially for those at the highest risk for risky alcohol use, which may help identify key targets for interventions to mitigate the impact of ACEs. Also, testing moderation effects of positive childhood experiences in the context of each ACE dimension is important to determine which factors are protective for drinking outcomes in those with childhood adversity. Future research should examine how specific PACEs impact other behavioral and health outcomes that are related to cumulative ACEs (e.g., the use of other substances, depression) and explore which PACEs have the strongest association for those with physical and sexual abuse; 2 ACE dimensions that were omitted from the current analysis.

Conclusion

This analysis provides an important first step in understanding how the experience of PACEs factors before the age of 18 are associated with drinking frequency and binge drinking frequency in adolescent and emerging adult females with varying degrees of ACEs per dimension. The analysis highlights that some factors might not be universally effective for all outcomes (e.g., membership in a non-sports social group, PACE 5, was associated with more frequent drinking across all ACE dimensions). It also highlights that in those with high household dysfunction, unconditional love (PACE 1) was inversely associated with drinking frequency, and in those with high emotional abuse/neglect, having a trusted adult to count on for help and advice other than a parent (PACE 7) had significant inverse associations with binge alcohol frequency over the month-predicting less frequent alcohol use with endorsement of these positive childhood experiences. While this should be replicated in larger samples, this analysis offers insight into which factors have the strongest associations in terms of alcohol use frequency and binge frequency in this sample of adolescent and young adult females who use alcohol, factors which may be targeted though intervention for children who experience multiple ACEs to prevent later alcohol use. ACEs impact development in a multitude of ways and examining groups by ACE dimension promotes more granular examinations of which factors function to prevent alcohol use later in those with varying degrees of ACEs per dimension. This study demonstrates that PACEs factors are not consistent, universal cures and vary based on the degree of ACEs in each dimension that were experienced in childhood. It is important to identify intervention targets that are modifiable and will provide the most protection in vulnerable groups, such as those with high ACEs.

Implications and Contributions Statement.

The context of childhood adversity is critical to identifying factors that may promote or inhibit direct effects on alcohol use frequency and quantity. Experiencing unconditional love and having a trusted adult to count on were associated with less frequent alcohol use, but the relationship was mediated by household dysfunction.