Differences in childhood physical abuse reporting and the association between CPA and alcohol use disorder in European American and African American women

Abstract

The goal of the current study was to examine whether the magnitude of the association between childhood physical abuse (CPA) and alcohol use disorder (AUD) varies by type of CPA assessment and race of the respondents. Data are from the Missouri adolescent female twins study and the Missouri family study (N = 4508) where 21.2% identified as African American (AA) and 78.8% as European American (EA); mean age = 23.8. Data were collected using a structured comprehensive interview which assessed CPA experiences using behavioral questions about specific abusive behaviors and trauma checklist items. Cox proportional hazards regression analyses were conducted, adjusting for additional risk factors associated with AUD, including co-occurring psychiatric disorders (defined as time-varying) and parental alcohol misuse. Overall, CPA reporting patterns were highly correlated (tetrachoric rho = 0.73); although, only 25.8% of women who endorsed behaviorally defined CPA also endorsed checklist items whereas 72.2% of women who endorsed checklist items also endorsed behavioral questions. Racial disparities were evident, with behaviorally defined CPA increasing the hazard for AUD in EA but not AA women. Additional racial disparities in the risk for AUD were observed: increased hazard for AUD were associated with major depressive disorder in AA, and cannabis dependence and paternal alcohol problems in EA, women. Results demonstrate the relevance of the type of CPA measure in assessing CPA in studies of alcohol-related problems – behavioral items may be more inclusive of CPA exposure and more predictive of AUD– and highlight racial distinctions of AUD etiology in women.

CPA and Alcohol-Related Problems

The relationship between childhood physical abuse (CPA) – a physical trauma or physical injury caused by slapping, beating, hitting, or otherwise harming a child – and problem drinking is bidirectional and intergenerational. Over ten percent (10.9%) of adolescent girls report experiencing CPA (Hanson et al., 2006), a well-documented risk factor for alcohol-related problems in adolescence and young adulthood (Kilpatrick et al., 2000). Although the impact of CPA on later substance involvement has received much less attention than other childhood adversities [e.g., childhood sexual abuse (CSA), early substance initiation], CPA has been linked with a host of psychiatric and substance related pathologies in adolescence including early substance initiation, alcohol and other substance use disorders, major depression, and conduct disorder (Goldstein et al., 2013; Kaplan et al., 1998; Lo & Cheng, 2007). Research has also shown that those who experience CPA are at increased risk for psychiatric dysfunction persisting into adulthood (Fergusson, Boden, & Horwood, 2008; Springer, Sheridan, Kuo, & Carnes, 2007). Furthermore, adolescents whose parents misuse alcohol are at elevated risk for both CPA and alcohol-related problems (Kaplan, Sunday, Labruna, Pelcovitz, & Salzinger, 2009; Kelleher, Chaffin, Hollenberg, & Fischer, 1994; Kumpfer, 1999; Li, Pentz, & Chou, 2002). As women are increasingly using alcohol and experiencing alcohol-related problems (Johnston et al., 2014) and CPA is an exposure also linked to a host of psychiatric problems in women (Briere & Elliott, 2003), understanding the relationship between alcohol problems and disorder and CPA in this population is of high importance.

Although the majority of research supports the link between CPA and alcohol-related problems, methodology for the assessment of CPA has not been consistent across studies, possibly accounting for discrepancies in the magnitude of the associations between CPA and alcohol-related constructs reported in the literature. The primary focus of the current investigation was to examine the relationship between varying assessments of CPA and alcohol use disorder (AUD), to determine if questions that ask about specific physically harsh/aggressive behaviors on the part of others (e.g., parental/others’ harsh disciplinary practices) – what we will term here “behaviorally defined CPA” – yield different estimates than those including abuse terminology such as “physical assault” commonly presented with definitions in a trauma checklist. This is an important question as many large scale epidemiologic studies are limited in their ability to include numerous questions about trauma exposure. While single item trauma questions are typically included in these studies, additional behavioral assessment may allow for a more comprehensive query of specific occurrences that may have been experienced during childhood (e.g. harsh punishment) and have been shown to detect an increased number of CPA cases (Carlin et al., 1994; Silvern, Waelde, Baughan, Karyl, & Kaersvang, 2000). As behaviors operationally defined as physical abuse by investigators may not be categorized as such by the respondents, discrepancies between behavioral assessment and trauma checklist responses may exist. Conversely, the trauma checklist may serve as a more anonymous or generic form of assessment allowing individuals to endorse physical abuse without directly describing the event or implicating a perpetrator.

Previous research has shown variability in prevalence estimates between differentially defined CPA reports. Carlin et al. (1994) reported of the 28.4% of women who were found to meet criteria for CPA, defined as endorsing any major assault question on the Emotional and Physical Abuse Questionnaire before the age of 18, only 11.4% considered themselves as physically abused. Conversely, only 1% of respondents endorsed physical abuse without endorsing any behaviorally assessed items. Silvern and colleagues (2000) compared individual endorsement of CPA as measured by the Conflict Tactics Scale (research-defined abuse) and self-defined abuse. Results again suggested higher rates for research-defined CPA (19.6%) with only 8.8% of these women self-defining as CPA survivors; only 1% of women self-defined as CPA survivors without meeting criteria for research-defined physical abuse. However, whether the CPA endorsement patterns also differ in their associations with substance use or other psychiatric disorders has not yet been studied. Research on CSA has revealed different associations with alcohol-related constructs for trauma checklist versus specific behavior questions. Sartor and colleagues (Sartor et al., 2012) found that women who endorsed CSA on behaviorally based questions only were significantly more likely to report a history of AUD compared to those who endorsed CSA on the trauma checklist only or on both the behavioral and trauma checklist assessment. In the present report, we will investigate whether similar associations exist with CPA reporting patterns.

Racial Disparities in CPA and Alcohol-Related Problems

As racial disparities in exposure to CPA, alcohol use, and alcohol use disorder (AUD) have been found between EA and AA cohorts (Grant et al., 2012; Hawkins et al., 2010; Sedlak et al., 2010), our second goal is to investigate whether the associations between CPA endorsement and AUD differ by race/ethnicity. In the majority of epidemiological investigations, AA children and adolescents experienced significantly higher rates of physical abuse than those in other racial groups, but lower rates of early alcohol use and AUD (Deater-Deckard, Pettit, Lansford, Dodge, & Bates, 2003; Hawkins et al., 2010; Sedlak et al., 2010; Zolotor, Theodore, Chang, Berkoff, & Runyan, 2008). Not all study findings have been consistent, with at least one report of significantly higher rates of CPA in EA than AA adolescents (Finkelhor, Ormrod, Turner, & Hamby, 2005). Methodological differences such as question type (i.e. trauma checklist vs behavioral assessment), time period assessed, nature of the sample (e.g. general population, adult, or college) and inclusion of spanking as physical abuse may account for these differences. Furthermore, trauma exposure is associated with elevated risk for early initiation of alcohol use and AUD in adolescent girls and women (Sartor et al., 2011; Wu et al., 2010) yet trauma exposure is higher and rates of problem drinking are lower in AA than EA women (Duncan et al., 2014; Grant et al., 2012; Grucza, Bucholz, Rice, & Bierut, 2008; Heath et al., 1999; Sartor et al., 2013). This seemingly paradoxical relationship between trauma exposure and alcohol outcomes in EA and AA women suggests existing environmental etiological models of AUD development do not fit as well for AA as EAs.

One possible explanation for this discrepancy, as it pertains to the association of CPA and AUD, is that AA and EA populations may have differing acculturation of harsh physical punishment, with AA adolescents viewing the use of physical discipline as more normative than their EA counterparts (Deater-Deckard et al., 2003). Therefore, AA respondents may be less likely to categorize harsh physical punishment as a trauma (on checklist assessment) but may still endorse the behavior when queried directly in a behavioral assessment. However, the impacts of race and source of CPA assessment on the association with AUD have yet to be addressed.

Thus, the aims of the study were to examine a) the consistency of reporting of CPA on behaviorally worded questions and items using abuse terms (trauma checklist); b) whether behavioral and trauma checklist assessments of CPA are equally robust predictors of AUD; and c) whether the association between CPA reporting and AUD holds in both AA and EA women. Based on the current literature, we predicted 1) there will be a different pattern of CPA endorsement in AA vs. EA women with AA women less likely than their EA counterparts to report physical abuse or assault on the checklist but more likely to endorse behavioral questions; and 2) for both EA and AA women, those who endorse behaviorally-defined CPA (with or without endorsement of trauma checklist items) would be at higher risk for AUD than those who endorse CPA on checklist items only.

Methods

Detailed descriptions of the Missouri Adolescent Female Twin Study (MOAFTS) and the Missouri Family Study (MOFAM) methods have been previously reported (Calvert, Bucholz, & Steger-May, 2010; Heath et al., 2002; Waldron, Bucholz, Lynskey, Madden, & Heath, 2013) with summaries pertaining to the current investigation provided below. Procedures for both MOFAM and MOAFTS were approved by the Human Research Protections Office at Washington University. MOFAM was also approved by the Ethics Board of the State Department of Health and Senior Services (which was not required at the time that MOAFTS began).

Participants

Of a possible 4522 female participants assessed in the MOFAM and MOAFTS studies, 14 individuals were dropped from the analyses for missing history of CPA. Therefore, the total sample utilized in the current study (N = 4508) consisted of 721 female participants from MOFAM and 3787 female twins who completed the fourth wave of data collection for MOAFTS where trauma checklist and behaviorally based measures of CPA were obtained; 21.2% (n = 954) identified as African American. Additional demographic variables are reported in Supplement 1.

MOAFTS

MOAFTS is a longitudinal investigation of alcohol-related problems and associated psychopathology in female adolescents and young adults. Twins born to Missouri-resident parents between July 1, 1975 and June 30, 1985 were identified through birth records and recruited from 1995 to 1999. Cohorts of 13-, 15-, 17-, and 19-year-old twin pairs and their families were ascertained in the first two years, with new cohorts of 13-year-old twins and their families added in the subsequent two years. Parental diagnostic interviews were collected at baseline in 78% of the families recruited and eligible for participation. Baseline (Wave 1) interviews were completed with 3,258 twins. Wave 3 retest interviews were conducted with a subset of Wave 1 participants (n = 1,370) two years after Wave 1 assessments. (Data were not drawn from Wave 2 interviews, as they did not cover all domains of interest.) Wave 4 assessments were conducted from 2002 to 2005 with 3,787 twins (80% of twins identified from birth records) and contained detailed trauma history assessment. Wave 5 interviews were conducted between 2005 and 2008 with 3,428 twins. Data from Waves 1, 3, and/or 5 were available for 95.5% of Wave 4 participants. MOAFTS participants (n = 3787) included 2176 monozygotic (57.5%) and 1611 di-zygotic (42.5%) female twins. Ages of the female participants at the time of Wave 4 interview ranged from 18 to 29 years (M = 21.7; SD = 2.76). 14.6% percent of participants identified as AA and 85.4% as EA. MOAFTS Wave 4 was chosen as the CPA assessment time-point as it included the largest number of respondents, most intensive trauma assessment, and respondents were of similar age – emerging adulthood – as the baseline MOFAM respondents.

MOFAM

MOFAM is a high risk alcoholism family study oversampled for AA families. From 2003 to 2009, Missouri state birth records were used to identify families with at least one child aged 13, 15, 17 or 19 years (the same age range targeted in MOAFTS) and 1 or 2 additional full siblings. Biological mothers completed brief telephone screening interviews to determine familial risk level for alcoholism. Families where biological fathers had a history of excessive drinking were classified as “high risk.” All others were classified as “low risk.” An additional group of families was selected from men identified as having 2 or more drunk-driving convictions (“very high risk”). In all participating families, mothers were interviewed first. Biological fathers were then solicited for interview and permission was sought from mothers to recruit offspring. Three hundred seventeen families of non-AA (primarily EA) descent and 450 AA families (92% of targeted families) were enrolled in the study. Sample enrollment occurred over 6 years. Three of the intake years had 4 waves of data, collected at two-year intervals; the remainder had 1 to 2 waves. A total of 1,461 (735 female) offspring completed at least one interview. Ages of the female participants included in the current analyses ranged from 13 to 33 years (M = 20.2; SD = 3.78); 55.6% percent identified as AA.

Procedure and Assessment Battery for MOAFTS and MOFAM

MOFAM and MOAFTS assessments were nearly identical by design to facilitate integration of the data across samples (with adjustments for differences in ascertainment strategies). In both studies, data were collected by trained interviewers using an interview modified for telephone administration from the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA; (Bucholz et al., 1994) The SSAGA was designed to assess the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [DSM-IV;(American Psychiatric, 1994)] substance use and other psychiatric disorders and related psychosocial domains. With the exception of the Wave 5 MOAFTS interview, which covered the 2 years between Wave 4 and 5 assessments, interviews queried lifetime psychiatric and psychosocial history. Consent was obtained verbally prior to the start of the interview.

Predictors and Outcomes

Behaviorally defined CPA (MOAFTS Wave 4, MOFAM baseline assessment)

Questions assessing CPA using behavioral descriptions of experiences consistent with physical abuse were asked by the interviewer using the questions from the SSAGA (see Table 1). These questions do not require the respondent to label their experience as CPA or traumatic. For this study, CPA criteria were met if the respondent endorsed experiencing these “often” for either parent.

Table 1.

Items used to define childhood physical abuse in MOAFTS and MOFAM

Behavioral Items

Endorsement of ≥1 item (at the level of “often”)

1. When you were 6 to 13, when you did something wrong, how often were you slapped by your (mother figure/father figure)?

2. When you were 6 to 13, when you did something wrong, how often were you hit with a belt or a stick or something like that by you (mother figure/father figure)?

3. When you were 6 to 13, how often did your (mother figure/father figure) physically punish you so hard that you hurt the next day?

4. When you were 6 to 13, were you injured or hurt on purpose by any adult? Examples of such injuries would include: broken bones, burns, being hit so hard you developed bruises, or any other physical injuries.

Trauma Checklist Events

Endorsement of ≥1 event, with reported age at first experience 13 years or younger

1. Physical Abuse “You were physically abused as a child”

2. Physical Assault “You were seriously physically attacked or assaulted”

MOAFTS = Missouri Adolescent Female Twin sample; MOFAM = Missouri Family Study sample

Trauma checklist defined CPA (MOAFTS Wave4, MOFAM baseline assessment)

The trauma checklist items were broader than the behavioral questions, explicitly worded using the terms abuse and assault. Adapted from the National Comorbidity Survey trauma checklist (Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995), traumatic events (including physical abuse and physical assault) were included in the respondent booklets sent to participants before interview. Brief definitions of events were given in parentheses (see Table 1) but not read to respondents. Participants referred to this list as the interviewer read, “Did event [number] ever happen to you?” Age at first occurrence was queried for those events that were endorsed. For this study, CPA criteria were met if the respondent reported age of CPA was 13 or below for any of the criteria (to be consistent with the age cutoff in the behavioral questions). In the MOAFTS sample, trauma questions were queried prior to behavioral questions; in the MOFAM sample, behavioral questions were queried prior to trauma items.

Alcohol Use Disorder

A Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5;(Association, 2013) AUD proxy was calculated by identifying individuals who endorsed two or more of the 10 symptoms – 7 dependence and 3 abuse (legal problems are excluded from DSM-5) within the same 12-month period at some point during their life [The symptom craving was not assessed; however, the psychometric benefit of adding craving to AUD is unclear (Hasin et al., 2013), as the addition of craving to the DSM-IV alcohol dependence has not been found to add significant information (Cherpitel et al., 2010; Hasin, Fenton, Beseler, Park, & Wall, 2012).] Age of onset was defined as the first age at which DSM-5 AUD proxy was met.

Risk factors for alcohol use disorder

Participant reports of maternal and paternal history of alcohol-related problems/excessive drinking, respondent own regular smoking, cannabis dependence, history of CSA, conduct disorder (CD) and major depressive disorder (MDD) were assessed within the SSAGA interview. Maternal and paternal alcohol use problems were also independently assessed using the SSAGA, by querying: (a) Has drinking ever caused your biological mother/father to have problems with health, family, job or police, or other problems, and (b) Have you ever felt that your biological mother/father was an excessive drinker? Additionally, the mother’s report on her own and paternal alcohol use disorder symptoms was used to assess parental problem alcohol use. Endorsement of a) one or more alcohol problems by maternal report or b) either alcohol question by either twin/sibling was coded as a positive response for an alcohol-related misuse in that parent. Regular smoking was defined as smoking at least 20 cigarettes (lifetime) and smoking at least once per week for 2 months. CSA was queried through both trauma checklist and behavioral questions. Trauma checklist: a) Had ever been raped or sexually molested (trauma checklist); Behavioral: b) “Has anyone ever forced you to have sexual intercourse?”; c) “Before you turned 16, was there any forced sexual contact between you and any family member like a parent or step-parent, grandfather, etc.?”; or d) “Before you turned 16, was there any forced sexual contact between you and anyone who was 5 or more years older than you?” and coded positive if any of these questions was endorsed. Consistent with CPA methods, age of onset for CSA was defined as age of first reported CSA experience. Diagnoses of lifetime DSM-IV CD and MDD were assessed based on independent self-report. When more than one report was available, age at onset was defined as the first time onset was reported (not necessarily the youngest onset age).

Data Analysis

All analyses were conducted in Stata Version 14 (StataCorp LP, College Station, TX) and in all multivariate analyses, adjustments were made for ascertainment by including study of origin and dummy variables representing each of the MOFAM high-risk groups. All analyses were conducted taking familial clustering into account using the cluster command as implemented in Stata.

CPA reporting patterns, race and AUD risk factors

Participants were categorized by reporting pattern into one of four groups: (a) endorsed behaviorally defined CPA but not checklist items, (b) endorsed checklist items but not behaviorally defined CPA, (c) endorsed both question types, and (d) did not endorse either. Analyses investigating this relationship were completed two ways. Tetrachoric correlations and differences in prevalence of risk factor by pattern of CPA reporting grouping were compared across EA and AA women separately. Differences in pattern of CPA reporting between EA and AA women in prevalence of risk factors for AUD were examined using factorial logistic regression analyses to test the association between CPA question reporting pattern, race, and AUD risk factors (maternal alcohol problems, paternal alcohol problems, MDD, CD, regular smoking, cannabis dependence, and CSA). Dummy variables were used to represent behavioral questions only, checklist items only, and behavioral-plus-checklist endorsement groups; participants who did not endorse any questions assessing CPA served as the reference group. Chi-square tests of association were conducted to test for the interaction of CPA endorsement and race in the association with AUD risk factors.

Testing the association of CPA reporting patterns and alcohol-related behaviors: differences between EA and AA women

Cox proportional hazard (PH) regression analyses predicting time to onset of AUD were conducted for the total sample considering the interaction between race and CPA endorsement pattern. We observed a significant interaction for race and CPA behavioral endorsement (X² = 2.77, p = .096), which motivated stratification of further analyses by race, – considering models for EA and AA women separately. Cox PH regression analyses based on a) behavioral questions; b) trauma checklist items and finally c) CPA reporting pattern as predictors for DSM-5 AUD diagnosis were considered. In the complete model, predictors included dummy variables representing behavioral endorsement only, checklist endorsement only, and behavioral and checklist endorsement groups as well as maternal and paternal alcohol problems, MDD, CD, regular smoking, cannabis dependence, and history of CSA as covariates using the method previously described. All CPA variables, as well as MDD, regular smoking, cannabis dependence, and history of CSA were considered as time-varying predictors.

Cox PH regression analysis is a type of survival analysis alcohol-related approach, which accounts for the possibility that participants who have not yet experienced the event of interest may do so in the future. Under this approach, data up until the time of censoring (most recent interview) are used in the calculation of hazard ratios. The proportional hazard assumption, that risk remains constant over time, was tested with the Grambsch and Therneau test of the Schoenfeld residuals (Grambsch & Therneau, 1994). To correct for PH violations, age interactions were included to reflect specific risk periods. Therefore, when reporting the models with PH violations for predictor variables, hazard ratios are reported by risk period.

Time varying covariates included all CPA endorsement variables, regular smoking, cannabis dependence, MDD and CSA. Time invariant covariates included maternal and paternal alcohol-related problems and respondent CD. Covariate status was coded as negative in each year prior to the age at first occurrence and positive for that year and each subsequent year for each covariate considered. Huber-White robust standard errors were used to adjust for the non-independence of observations in siblings.

Results

Consistency of CPA report and impact of race

Six hundred fifty-five women (14.5% of the full sample) endorsed one or more of the behavioral questions assessing CPA. Of those women, 25.8% (169 individuals) also endorsed experiencing physical abuse or assault before the age of 13 on the trauma checklist. Two hundred thirty five women (5.2% overall) endorsed trauma checklist CPA items with 72.2% of those women also endorsing one or more behavioral questions. Tetrachoric correlations for behavioral endorsement and trauma checklist items revealed a weaker correlation for AA (tetrachoric rho = 0.51, p < .0001) versus EA (tetrachoric rho = 0.80, p < .0001) women. Adjusting for familial clustering, AA women were nearly three times as likely as EA women to report experiencing CPA using either question type (29.0% vs. 12.5% respectively, F = 122.79, p < .0001) and were significantly more likely to endorse CPA on the behavioral questions (AA: 26.7%, EA 11.3%; F = 115.89, p < .0001) and trauma checklist (AA: 6.9%, EA: 4.8%; F = 5.28, p =.022). Significant differences across race in overall reporting pattern were also found (F = 49.68, p <.0001); post hoc analyses revealed of those reporting CPA, AA women were significantly more likely than EA women to endorse behavioral items only (F = 12.39, p = .0005) or both behavioral and trauma checklist (F = 12.25, p = .0005) with no difference for trauma checklist only (F = 0.38, p = .539).

Rates of AUD risk factors by CPA questions endorsement patterns and race

The prevalence of AUD and AUD risk factors by pattern of CPA question endorsement and race are shown in Table 2. One hundred sixty nine women (3.7%) endorsed both CPA behavioral and checklist question types, 486 (10.8%) behavioral only, 66 (1.5%) trauma checklist only, and 3787 (84.0%) did not endorsed either question type. Significant differences between endorsement patterns were found across all risk factors. History of maternal alcohol problems, MDD, CD, regular smoking, and CSA also significantly differed across race, with differences in paternal alcohol problems and cannabis dependence between EA and AA women reaching trend level (see Table 2).

Table 2.

Prevalence of alcohol use disorder and associated risk factors by pattern of childhood physical abuse and race

	Pattern of CPA question Endorsement					Race
	Both (n = 169) %	Checklist only (n = 66) %	Behavioral only (n = 486) %	Neither (n = 3787) %	X2	EA (n =3554) %	AA (n =954) %	X2
Alcohol use disorder-DSM-5 lifetime	46.1	39.1	37.4	29.3	34.18 p <.001	34.3	18.3	88.57 p <.001
Maternal alcohol misuse or problems	54.4	43.9	41.6	34.8	35.29 p <.001	38.1	30.0	21.47 p <.001
Paternal alcohol misuse or problems	75.7	80.3	67.1	52.8	82.16 p <.001	54.6	59.3	6.81 p = .052
Major depressive disorder	34.9	40.9	29.2	13.7	148.52 p <.001	15.2	21.6	22.42 p <.001
Conduct disorder	17.2	7.6	9.9	3.0	121.60 p <.001	3.6	7.1	22.94 p <.001
Regular smoking	53.3	48.5	42.6	34.5	38.97 p <.001	40.4	21.1	121.07 p <.001
Cannabis dependence	11.8	4.6	4.7	2.6	47.99 p <.001	3.0	4.3	4.32 p = 0.051
Childhood sexual assault	46.4	53.0	26.8	9.7	365.12 p <.001	11.6	20.5	50.02 p <.001

Note: CPA: childhood physical abuse; EA: European American; AA: African American; Regular smoking = smoking ≥ 20 cigarettes (lifetime) and at least once per week for 2 months.

Logistic regression analyses (Table 3), using those without CPA as the reference group revealed elevated rates of all risk factors for women endorsing behavioral items only or both question types as compared to endorsing neither CPA question type, with all risk factors other than maternal alcohol problems and cannabis dependence elevated in those who endorsed the trauma checklist questions type. Odds of experiencing CSA were significantly higher in those who endorsed CPA on both question types as compared to either alone – behavioral only (X² = 7.98, p = .005) or trauma checklist only (X² = 6.74, p = .010). Significant interactions (p < 0.1 for interactions) between CPA endorsement pattern and race were revealed for MDD (X² = 7.64, p = .054), regular smoking (X² = 11.06, p = .011) and CSA (X² = 9.08, p = 0.028) and maternal alcohol problems (X² = 6.34, p = 0.096).

Table 3.

Logistic regression analyses: Differences in prevalence of AUD risk factors by CPA endorsement pattern and race.

	Pattern of CPA Endorsement	Both vs none	Behavioral vs none	Checklist vs none	Race	Reporting Pattern* Race
	X2	OR [95%CI]	OR [95%CI]	OR [95%CI]	OR [95%CI]	X2
Maternal alcohol problems	26.48 p <.0001	2.10 (1.39–3.16)	1.71 (1.31–2.25)	1.74 (0.94–3.20)	0.56 (0.43–0.73)	6.34 p = 0.096
Paternal alcohol problems	31.39 p <.0001	2.67 (1.69–4.20)	1.62 (1.22–2.15)	3.09 (1.46–6.55)	0.91 (0.71–1.18)	1.30 p = 0.730
Major depressive disorder	105.47 p <.0001	3.74 (2.55–5.49)	3.11 (2.33–4.14)	3.97 (2.13–7.40)	1.37 (1.07–1.76)	7.64 p = 0.054
Conduct disorder	47.69 p <.0001	5.37 (2.94–9.82)	2.93 (1.77–4.86)	4.39 (1.68–11.49)	1.32 (0.83–2.08)	2.10 p = 0.351
Regular smoking	70.29 p <.0001	2.83 (1.94–4.13)	2.31 (1.79–3.00)	2.10 (1.15–3.84)	0.44 (0.35–0.56)	11.06 p = 0.011
Cannabis dependence	36.77 p <.0001	5.75 (3.18–10.39)	2.26 (1.26–4.06)	1.91 (0.43–8.36)	1.59 (0.93–2.74)	2.94 p = 0.402
Childhood sexual assault	178.38 p <.0001	7.65 (5.10–11.49)	4.01 (2.96–5.42)	9.76 (5.23–18.18)	1.63 (1.21–2.19)	9.08 p = 0.028

^*Note: Analyses control for ascertainment, study of origin, and years between CPA and most recent assessment; Reference levels are neither CPA reported and European American; Bold significant p < .05 (significant level set at p < 0.10 for interactions);

Prevalence of risk factors for AUD using behavioral questions versus trauma checklist items to define CPA, by race

The prevalence of all AUD risk factors in women categorized as CPA-positive based on behavioral questions (n = 655) was slightly lower than among those who endorsed CPA on trauma checklist items (n=235). Notably, of those who were positive for CPA, the prevalence of CSA was 31.9% using behavioral questions but 48.3% using trauma checklist items. Slightly lower rates of maternal (44.9% vs. 51.5%) and paternal (69.3% vs. 77.0%) alcohol problems were found when defining CPA through behavioral questions versus trauma items. Consistently, behaviorally-defined CPA had lower prevalence of AUD risk factors as compared to trauma item defined CPA including MDD (30.7% vs. 36.6%); CD (11.8% vs. 14.5%); regular smoking (45.3% vs. 51.9%); and cannabis dependence (6.6% vs. 9.8%).

Testing for differences in risk for AUD by CPA question endorsement pattern across race

Results of the Cox PH regression analyses (conducted separately for EA and AA participants) predicting AUD using behavioral questions and trauma items to define CPA are reported in Table 4. Behaviorally-defined CPA significantly increased the hazard for AUD in the EA (HR=1.29, 95% CI: 1.10–1.52) but not AA (HR=1.09, 95% CI: 0.79–1.49) women, but trauma checklist-defined CPA did not significantly increase the hazard of AUD for either EA (HR=1.09, 95% CI: 0.86–1.39) or AA (HR= 0.83, 95% CI: 0.44–1.59) women after accounting for the additional AUD risk factors.

Table 4.

Cox regression: Predicting alcohol use disorder from childhood physical assault defined behaviorally vs. trauma checklist in European American and African American women

		Behaviorally-defined CPA				Trauma checklist defined CPA
		African American		European American		African American		European American
		HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Behaviorally defined CPA		1.40 (1.04–1.91)	1.09 (0.79–1.49)	1.50 (1.27–1.77)	1.29 (1.10–1.52)	–	–	–	–
Checklist defined CPA		–	–	–	–	1.15 (0.65–2.02)	0.83 (0.44–1.59)	1.28 (1.01–1.63)	1.09 (0.86–1.39)
Maternal alcohol problem		1.87 (1.32–2.65)	1.69 (1.19–2.40)	1.35 (1.19–1.55)	1.21 (1.06–1.38)	1.88 (1.33–2.67)	1.71 (1.20–2.42)	1.37 (1.20–1.56)	1.22 (1.07–1.39)
Paternal alcohol problem		0.75 (0.51–1.10)	0.71 (0.49–1.03)	1.25 (1.09–1.43)	1.13 (0.99–1.29)	0.77 (0.52–1.12)	0.71 (0.49–1.03)	1.26 (1.10–1.45)	1.14 (1.00–1.30)
Childhood sexual assault	(<18)	┬	┬	1.86 (1.47–2.34)	┬	┬	┬	1.96 (1.55–2.47)	┬
		1.01 (0.69–1.48)	0.85 (0.57–1.26)		1.14 (0.96–1.35)	1.01 (0.69–1.50)	0.87 (0.58–1.31)		1.18 (1.00–1.40)
	(≥18)	┴	┴	1.16 (0.93–1.46)	┴	┴	┴	1.23 (0.99–1.54)	┴
Major depressive disorder		–	1.57 (1.03–2.38)	–	1.13 (0.95–1.35)	–	1.57 (1.04–2.37)	–	1.17 (0.98–1.39)
Conduct disorder	(<14)	–	┬	–	18.47 (9.03–37.76)	–	┬	–	18.92 (9.30–38.49)
			3.13 (2.00–4.91)				3.15 (2.05–4.85)
	(≥14)	–	┴	–	1.70 (1.29–2.25)	–	┴	–	1.73 (1.31–2.29)
Regular smoking	(<18)	–	┬	–	4.46 (3.67–5.42)	–	┬	–	4.49 (3.69–5.45)
			2.74 (1.91–3.93)				2.75 (1.92–3.93)
	(≥18)	–	┴	–	2.24 (1.92–2.61)	–	┴	–	2.26 (1.94–2.63)
Cannabis dependence	(<18)	–	0.67 (0.08–5.52)	–	┬	–	0.69 (0.08–5.64)	–	┬
					3.53 (2.49–5.00)				3.53 (2.52–4.96)
	(≥18)	–	1.58 (0.72–3.50)	–	┴	–	1.63 (0.73–3.61)	–	┴

Note: Models adjusted for ascertainment, study of origin, and years between CPA and most recent assessment; CPA (childhood physical assault); Bold p < .05

Two models were run comparing the three CPA reporting patterns as predictors for AUD in EA and AA women separately (using no reported CPA as the reference group; Table 5) with Model 1 including only familial level risk factors and CSA and Model 2 including all individual level risk factors considered. Model 2 results revealed an increased hazard of AUD for EA women who endorsed CPA on the behavioral questions only (HR=1.33, 95% CI: 1.10–1.61) with no significant increase in hazard associated with CPA endorsement on both behavioral and checklist items or trauma checklist only. None of the three CPA endorsement patterns increased the hazard of AUD in AA women. Other AUD risk factors remained consistent, with maternal alcohol problems, conduct disorder, and regular smoking increasing the hazard of AUD for both EA and AA women with early AUD associated especially with conduct disorder and regular smoking in EA women. Cannabis dependence also increased the hazard for AUD in EA women, while MDD significantly increased the hazard in AA women.

Table 5.

Cox regression: Predicting AUD from childhood physical abuse reporting patterns in European American and African American women

		African American		European American
		Model 1	Model 2	Model 1	Model 2
		HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Behavioral and checklist		1.10 (0.49–2.46)	0.64 (0.24–1.69)	1.50 (1.16–1.95)	1.23 (0.95–1.59)
Behavioral only		1.29 (0.89–1.87)	1.07 (0.74–1.55)	1.51 (1.25–1.83)	1.33 (1.10–1.61)
Checklist only		1.57 (0.74–3.30)	1.39 (0.64–3.01)	1.12 (0.69–1.84)	1.01 (0.61–1.66)
No CPA report (reference)		1.00	1.00	1.00	1.00
Maternal alcohol problem		1.85 (1.31–2.65)	1.69 (1.19–2.39)	1.35 (1.18–1.55)	1.21 (1.06–1.38)
Paternal alcohol problem		0.74 (0.51–1.09)	0.71 (0.49–1.02)	1.25 (1.08–1.43)	1.13 (0.99–1.29)
Childhood sexual assault	(<18)	┬	┬	1.85 (1.46–2.34)	┬
		1.00 (0.68–1.47)	0.89 (0.59–1.32)		1.14 (0.96–1.36)
	(≥18)	┴	┴	1.16 (0.92–1.45)	┴
Major depressive disorder		–	1.57 (1.04–2.37)	–	1.13 (0.95–1.35)
Conduct disorder	(<14)	–	┬	–	18.52 (9.07–37.81)
			3.32 (2.11–5.22)
	(≥14)	–	┴	–	1.71 (1.29–2.25)
Regular smoking	(<18)	–	┬	–	4.46 (3.67–5.42)
			2.68 (1.87–3.86)
	(≥18)	–	┴	–	2.24 (1.92–2.61)
Cannabis dependence	(<18)	–	0.73 (0.09–5.92)	–	┬
					3.56 (2.52–5.04)
	(≥18)	–	1.69 (0.76–3.76)	–	┴

Note: Models adjusted for ascertainment, study of origin, and years between CPA and most recent assessment; CPA (childhood physical assault); Bold p < .05

Discussion

Given that one out of ten adolescent girls endorse experiencing CPA in their lifetime (Hanson et al., 2006) and the associated risk of psychopathology and AUD is high, it is important to ensure assessment of CPA in alcohol-related research does not include measurement bias that may confound results pertaining to the association between CPA and alcohol involvement. The current study examined the association of CPA endorsement – assessed via behavioral questions and trauma checklist items – with alcohol-related risk factors and the association of endorsement pattern and AUD risk factors with AUD diagnosis and the consistency of these effects in AA versus EA women.

Seventy seven percent of women who endorsed CPA via trauma checklist (physical assault or physical abuse) items also endorsed CPA behavioral questions. However, only 25% of those who endorsed behaviorally worded questions about CPA also endorsed physical abuse or assault (during the same age range) on the trauma checklist items. AA women had higher CPA endorsement rates regardless of question type, and racial differences in CPA endorsement pattern were observed. Analyses also revealed racial distinctions in the contribution of CPA – behaviorally only or both behaviorally and checklist endorsed – to risk for AUD, with risk conferred by CPA only in EA women. These findings generally support previous research investigating CSA endorsement patterns and our hypotheses and point to a) inconsistencies in CPA reporting across behavioral and trauma checklist question types and race, b) the impact of CPA experiences on risk for AUD, and c) distinct etiological models of risk for AUD for EA and AA women. Findings point to including behavioral assessment of CPA in future investigations to capture the most CPA exposure and highest conferred risk for AUD, particularly in EA women.

Consistency of CPA report and impact of race

Findings are consistent with previous reports of variability in CPA endorsement patterns (Carlin et al., 1994; Silvern et al., 2000). We observed an almost a three-fold increase in prevalence estimate of CPA when using behaviorally-defined CPA (14.5%) than trauma checklist-defined CPA (5.2%). Additionally, the majority of women who endorsed trauma checklist items for CPA also endorsed behaviorally defined CPA questions while a minority of women who endorsed behavioral CPA questions also affirmed the trauma check list items. This discrepancy has implications for future epidemiologic measurement of CPA and highlights the value of querying exposure to specific behaviors that fall into the category of physical abuse over asking respondents to explicitly define their experiences as abusive.

These current findings are not consistent with findings from an earlier study by our group comparing behavioral vs. trauma checklist definitions of CSA. The previous study found that almost all of the women endorsing CSA on the behavioral questions also reported rape or sexual molestation on the trauma checklist items (Sartor et al., 2012). This discrepancy is not surprising as physically abusive behavior is more prevalent than CSA. it may be perceived as relatively normative harsh physical punishment and therefore not defined as a trauma (abuse or assault) by the participant. For the small percentage that only endorsed trauma checklist CPA (1.5%; n = 66), their experience may not have fit within the behaviorally-defined parameters (i.e. abuse wasn’t endorsed as “often” or the physical abuse could have been perpetrated by someone other than an adult).

In line with previous research, overall higher rates of CPA, collapsing behavioral endorsement and trauma checklist assessments, were reported in AA as compared to EA women (Hanson et al., 2006; Hawkins et al., 2010). Further analyses revealed AAs were more likely than EAs to endorse behavioral questions only as well as both behavioral and trauma items (but not trauma checklist items only). These findings highlight the possibility of differences in cultural acceptance of CPA in AA and EA women in that, physical punishment is more normative in the AA community, and thus, less likely to be labeled physical abuse or assault. Previous research supports variations in cultural norms pertaining to the use of physical discipline (Deater-Deckard et al., 2003) and AA parents have been reported to be more likely to use harsh physical discipline as compared to EA parents (Zolotor et al., 2008).

Risk for AUD by CPA report

When investigating the impact of CPA endorsement patterns on lifetime AUD diagnosis, findings revealed CPA endorsement on behavioral questions only increased the hazard for AUD and only in EA women. For AA women, no significant risk for AUD was associated with any CPA endorsement pattern in either model. First, these findings suggest that using trauma checklist items may result in an underestimate of CPA and reduced power to detect effects above and beyond other AUD risk factors. Furthermore, any hazard associated with CPA in AA women disappeared after adjusting for multiple co-occurring risk factors (Table 4). This reflects the specificity of risk conferred by CPA and suggests behaviorally endorsed CPA may be a marker of a pathological family environment more generally. Secondly, as AA girls and women are more likely to be exposed to potentially traumatic events (McCutcheon et al., 2010; Roberts, Gilman, Breslau, Breslau, & Koenen, 2011), they may not be as severely affected as would individuals in populations with lower overall exposure to trauma. Thus, EA women may be more vulnerable (and AA less vulnerable) to the adverse effects of behaviors that we would define as physical abuse. Additionally, the risk associated with AUD from behaviorally endorsed CPA in EA women may be due to greater prevalence of alcohol use in the EA as compared to AA women. It may be that in AA women, CPA may increase risk for non-alcohol substance involvement, or other types of risky behaviors. Overall, these findings are consistent with the findings regarding CSA reported by Sartor and colleagues (Sartor et al., 2012) who found similar patterns for AUD risk associated with endorsement of behavioral only and both behavioral and trauma items but not trauma checklist only. However, racial disparities were not examined in that study. Further examination of racial disparities in the risk associated with trauma exposure for AUD and the impact of family level environmental effects is warranted.

In the current study, additional racial disparities in risk for AUD were also revealed: MDD significantly increased the hazard in AA women while cannabis dependence and paternal alcohol problems increased the hazard for EA women only. The increased risk associated with MDD in AA women suggests that internalizing pathways to AUD may be more prominent among AA women. Although externalizing disorders are more commonly linked to substance related-behaviors (Zucker, 2008), internalizing behaviors, such as MDD, are also known to elevate risk for AUD (Colder, Chassin, Lee, & Villalta, 2010; Swendsen & Merikangas, 2000). Racial disparities in the impact of cannabis dependence for AUD point to racial difference in substance use and initiation, consistent with previous research showing a higher prevalence of co-occurring of alcohol and cannabis involvement in EA women and greater likelihood of AA than EA women to initiate cannabis before alcohol use (Sartor et al., 2013). Additionally, the finding that paternal alcohol problems trended toward association with AUD in EA, but not AA, women could be an indicator of lower paternal impact on behaviors in the AA versus EA community. Because ascertainment in the MOFAM sample was based on father’s alcohol-related behaviors, which were controlled for in the analyses, we also conducted analyses using MOAFTS data only and found the same pattern that was observed with the combined sample. Lastly, increased hazard for AUD was revealed for maternal alcohol problems, conduct disorder, and regular smoking in both AAs and EAs, highlighting shared risk factors for AUD across race.

Limitations

Certain limitations should be considered when interpreting results of the study. First, prospective assessment of CPA would likely result in more accurate overall estimates of the prevalence of CPA, although it is unlikely that our use of retrospective reports impacted the consistency in CPA reporting across question types, as the two sets of questions were administered in the same interview. Second, our sample was ascertained from Missouri birth records, and, as Missouri is not necessarily representative of the rest of the country, findings may not be consistent with other regions. Additionally, a portion of this sample was selected based on high risk status pertaining to paternal alcohol involvement; however, only a small portion of the total sample (10.2%, n = 458) was classified as high risk and risk status was taken into account in all analyses. Third, the lifetime prevalence estimate of AUD (30.9%) in the current sample was slightly elevated in comparison to nationally representative samples (Grant et al., 2015; Hawkins et al., 2010). Therefore, findings in this population may not be generalizable to a more heterogeneous community; however, CPA and other risk factor prevalence estimates are comparable to larger epidemiologic samples (Hasin & Grant, 2015). Additionally, our sample was limited to African American and European American women only and did not investigate CPA experience, endorsement, and predictive value in ethnic subgroups of AAs and EAs or other minority groups (e.g., Hispanics or mixed race). As Hispanics have been shown to experience even higher prevalence of CPA than African Americans (Finkelhor et al., 2005; Hawkins et al., 2010) and ethnic variation within racial groups in alcohol problems and exposure to CPA have been reported (Brown, Donato, Laske, & Duncan, 2013), examining differential reporting and the impact of CPA on AUD in racial subgroups and other ethnic populations might yield different results. Furthermore, the current findings may not generalize to men, who have been reported to experience CPA at high rates and may have different endorsement patterns as well as different associations with AUD than those reported for the female-only sample in the current study. Lastly, a portion of the population under investigation is not entirely through the age of highest AUD risk and may still go on to develop alcohol related pathology.

Future directions

Further investigation of factors impacting respondent’s endorsement patterns of sensitive topics such as CPA is crucial to developing effective assessment measures that can accurately capture CPA exposure more efficiently and thus improve our ability not only to conduct epidemiologic research with minimum burden on respondents but to identify individuals at risk in the general population, e.g., through screeners in medical settings. As behaviorally defined CPA may be a proxy for a more general pathological family environment in AA women, the exploration of family-level risk factors (e.g. parental support) with CPA endorsement would be useful to disentangle this relationship. Investigation of other problem drinking behaviors as well as problems with other substances, and inclusion of males in future studies are also key to refining the measurement of the relationship between CPA and substance-related problems. AA women who experience CPA might be at increased risk of non-alcohol-related substance use disorders and the etiological impact of different CPA endorsement patterns on AUD in men remains unknown.

Conclusion

Our findings highlight the ability of behavioral questions to capture the majority of individuals (91.0%) exposed to CPA and point to behaviorally assessed CPA as a significant predictor of AUD in EA women above and beyond other AUD risk factors. In addition, AA women were more likely to endorse behaviorally defined CPA than EA women. As such, future studies aiming to assess CPA with minimal burden but maximum coverage should consider including behavioral questions to query CPA. We found this method to be the most robust assessment of CPA cases as trauma checklist items only captured 32.6% of CPA, underestimating exposure to CPA particularly in AA women. Lastly, CPA was only found to increase the hazard for AUD in EA women suggesting racially distinct etiological models of AUD.