Predicting Child Abuse Potential from the Mmpi-2-rf Higher Order Scales and the Aseba within a Sample of Care Givers Referred for Evaluation

PREDICTING CHILD ABUSE POTENTIAL FROM THE MMPI-2-RF HIGHER ORDER SCALES AND THE ASEBA WITHIN A SAMPLE OF CARE GIVERS REFERRED FOR EVALUATION Valerie J. Russell, M. A. Western Carolina University (March, 2012) Director: Dr. Kia A. Asberg The purpose of the current study is to examine the association between the higher-order scales of the Minnesota Multiphasic Personality Inventory, Second Edition, Restructured Form (MMPI-2-RF) and the DSM-oriented scales of the Achenbach System of Empirically Based Assessment (ASEBA), Adult Self-Report with physical child abuse potential, as measured by the Child Abuse Potential Inventory (CAP). Abuse and neglect has been shown to have serious and long-lasting negative effects on children’s mental health. Current research has identified a variety of predictors of child abuse potential. However, no previous studies could be found that have examined the correlation between scores on the MMPI-2-RF and the ASEBA with child abuse potential. The participants were 177 parents and caregivers who were court-ordered by the Georgia Division of Family and Children Services to receive a psychiatric evaluation in north Atlanta. Results show that males and females significantly differed on several of the predictor variables, and CAP scores were significantly correlated with all of the variables except the Drug and Alcohol Use Scales. The results of a multiple regression analysis indicate that the overall model did significantly predict CAP score [F(10,61) = 27.50, p < .001] with an R of .844. However, only Depressive Problems (Beta = .300, p < .05) and Emotional/Internalizing Dysfunction (Beta = .620, p < .001) were significantly predictive of child abuse potential. Findings of this study suggest that depression and other emotional disorders could be used as a “red flag” when assessing child abuse potential. In addition, it may be that internalizing disorders such as depression may be more predictive of child abuse potential than externalizing disorders (e.g., substance use). 1 CHAPTER ONE: INTRODUCTION Although statistics show that the incidence of child maltreatment, especially physical and sexual abuse, has been declining since the 1990s (Jones, Finkelhor, & Halter, 2006), abuse continues to be a reality for a significant number of children. For example, in the United States, the Department of Health and Human Services (2010) reported that, in 2008, there were 3.3 million reports of children being abused or neglected and 355,500 children were determined through investigations or assessments of these reports to be confirmed victims of child abuse or neglect. This number is most likely an underestimation as many cases of child abuse go unreported. Also, the secretive nature of child abuse can make it difficult to identify. In addition, many individuals are unaware of the extent of child abuse (i.e., do not know the prevalence of child abuse) or do not frequently think about child abuse (Polnay, 2001). For example, a recent study found that 31% of professionals, including counselors, psychologists, and teachers, endorsed that they have suspected child abuse, but did not report the incident (Owhonda, 2010). In addition, an Ohio study utilizing vignettes found that teachers underreported 33% of abuse cases, whereas overreporting occurred in 4% of cases and correct identification (along with reports) were made in just over 60% of scenarios (Webster, O’Toole, O’Toole, & Lucal, 2005). Overall, teachers in this sample were more likely to underreport than overreport child abuse. Further, Webster et al. (2005) noted that variables associated with decreases in underreporting included, among others, greater perceived knowledge of child abuse and “positive evaluation of the police in dealing with reports of child abuse” (p. 1291). Thus, 2 it is important to increase the understanding among professionals of what constitutes abuse, and what the risk factors for perpetration are, as well as improve the way that governmental agencies (including the police) are dealing with these types of reports. Also, it has been shown that many people are less willing to report suspected abuse if the family lives in a neighborhood with a high degree of perceived “social disorder” (Gracia & Herrero, 2006). For example, feelings of mistrust and powerlessness and fear of retaliation in a neighborhood reduces the willingness of residents to report suspected child abuse. Furthermore, although each state determines their legal definition of child abuse, many individuals are not familiar with the exact terms. Thus, the “definition” or “schema” (mental representation or template) of child abuse may vary from person to person. As a result, what may be considered child abuse to one person may seem like normative discipline to another. For example, a study of 199 university students and nonstudent adults found that perceptions of abuse were affected by several factors, including the relationship between victim and perpetrator, abuse type, and victim and perpetrator gender (Bornstein, Kaplan, & Perry, 2007). Similarly, Sherrill, Renk, Sims, and Culp (2011) found that undergraduate student raters’ attributions of abuse (depicted in vignettes) were significantly impacted by the age of the perpetrator in the vignette, and the gender role adherence and sexual attitudes of the rater. Also, the definition of “reasonable suspicion” (i.e., what is needed to report abuse) concerning child abuse also varies by state, which may compromise identification and reporting of abuse. Specifically, professionals living in states with clear definitions of reasonable suspicion have been found to be more confident in reporting child abuse 3 than professionals living in states with vague definitions (Flieger, 1999). Further, different forms of abuse are also more difficult to identify. Physical abuse may leave marks whereas the signs from sexual and emotional often are hidden or not visible. Thus, it is imperative that state departments of family and child services have adequate measures of predicting child abuse (e.g., child abuse potential) in order to prevent further abuse from occurring in families who have been reported. Furthermore, children who are maltreated are at a greater risk of having poorer psychological adjustment relative to their non-abused counterparts (e.g., McGloin & Widom, 2001), which suggests more research is needed to better understand what factors contribute to abuse. To that end, the present study aimed to provide a review of the literature on potential outcomes in children who have experienced abuse, the key individual and family characteristics of perpetrators that contribute to abuse, and address the need for investigations of child abuse potential (Munz, Wilson, & D’Enbeau, 2010) within groups of care takers that have been identified by Child Protective Services after an incident of abuse or neglect. 4 CHAPTER TWO: LITERATURE REVIEW Outcomes Following Child Abuse To better understand the importance of predicting child abuse potential among perpetrators, one must also understand and acknowledge that the effects of abuse on children are wide-ranging and often severe. The abuse of infants and toddlers has been shown to result in a heightened risk for developmental delays in adaptive behavior, cognition, and communication (Scarborough, Lloyd, & Barth, 2009). In addition, child abuse has been linked to a 50% to 60% chance of developing some form of psychopathology in adulthood (McGloin & Widom, 2001), with substance abuse being one of the most prevalent problems following maltreatment, especially physical abuse (Lo & Cheng, 2007). Another study of parents and their college-age children indicated that boys and girls who had been physically abused by their mothers or fathers, and especially by both parents, displayed more aggressive behavior (Muller & Diamond, 1999). In addition, severity of child sexual abuse predicts also the risk of involvement with the criminal justice system (i.e., arrests, incarceration; Asberg & Renk, 2012). In addition to externalizing symptoms, child abuse has also been shown to predict internalizing symptoms, including posttraumatic symptoms, depression, and anxiety (Naar-King, Silvern, Ryan, & Sebring, 2002). For example, data from the World Mental Health Survey suggested that childhood sexual abuse is associated with lifetime mood and anxiety disorders, while childhood physical abuse is associated with lifetime anxiety disorders, and any type of abuse or maltreatment is associated with both lifetime moodand 12-month anxiety disorders (Gal, Levav, & Gross, 2011). 5 Furthermore, the results of a comparison between sexually or physically abused mental-health-center clients and non-abused clients indicate that clients who were physically abused as children were significantly more likely to have auditory and tactile hallucinations, and clients who experienced any type of childhood abuse or partner aggression had significantly higher rates of hallucinations, delusions, and thought disorder relative to their non-abused counterparts (Read, Agar, Argyle, & Aderhold, 2003). Sexual, verbal, physical, and fear of physical abuse have also been correlated with obesity (Williamson, Thompson, Anda, Dietz, & Felitti, 2002). Obesity is also a mediating factor in the association between early child abuse and risk of type 2 diabetes among adult women (Rich-Edwards et al., 2010), suggesting that child abuse has far reaching negative outcomes that involve both mental and physical health. Perhaps one of the most detrimental effects of child abuse is the greater likelihood of the abused to become abusive or violent in adolescence and adulthood (Gómez, 2011). A cycle of violence may develop that is difficult to interrupt (Parkes, 2008). Specifically, an analysis of data from the National Longitudinal Study of Adolescent Health found that adolescents who experienced violence in the form of child abuse or adolescent dating violence were 97% more likely to become perpetrators of intimate partner violence in young adulthood (Gómez, 2011). Previous experiences of childhood abuse have also been shown to be predictive of child abuse perpetration (Medley & Sachs-Ericsson, 2009). It should be noted, however, that only a small fraction of children who suffer maltreatment develop into perpetrators of abuse (Heyman & Sleps, 2002). Overall, children who experience abuse and neglect are at a higher risk of a wide range of psychopathology and poor adjustment (Thornberry, Henry, Ireland, & Smith, 6 2010), but outcomes vary (Mullen, Martin, Anderson, Romans, & Herbison, 1994) depending on a variety of ameliorating circumstances (e.g., support) and individual characteristics (Banyard & Williams, 2007). Also, despite intervention by Child Protective Services, a majority of high risk families will end up back in the system, suggesting that abuse is a perpetual problem (DePanfilis & Zuravin, 1998). Thus, there is a need to better understand predictors of child abuse, and the usefulness of assessment tools that are currently available, in order to prevent the occurrence of such abuse or, in cases where abuse has occurred, to prevent re-victimization. Predictors of Child Abuse Potential Numerous attempts have been made by researchers to identify predictors of child abuse. On a family level, husbands’ and wives’ partner aggression have been found to be strongly connected with mothers’ and fathers’ parent aggression (Slep & O’Leary, 2005). Specifically, partner aggression has been found to be correlated with parent aggression toward children, with 45% of families reporting both parent and partner aggression, including 5% reporting severe parent and partner aggression. Similarly, Appel and Holden (1998) found that 40% of violent families experienced co-occurring partner and child abuse, suggesting that violence within families take many forms that can be detrimental to children. Moreover, a study of 62 women and their children at a domestic violence shelter found that level of partner-child aggression prior to entering the shelter, level of partnermother intimate partner violence after leaving the shelter, and frequency of contact between the children and the partners after departure each significantly predicted post 7 shelter partner-child aggression (McDonald, Jouriles, Rosenfield, & Corbitt-Shindler, 2011). These findings are in line with the cycle of violence described above. Further, poor relationship quality, marital violence, and low marital satisfaction have been shown to be predictive of child abuse (Agathonos-Georgopoulou & Brown, 1997; O’Keefe, 1995), possibly by ways of increasing stress (Guterman, Lee, Taylor, & Rathouz, 2009). Also, parental happiness with the parent-child relationship, as measured by the Parent Satisfaction with Youth Survey, has been correlated with child abuse potential after controlling for social desirability (Bradshaw, Donohue, Cross, Urgelles, & Allen, 2011). On an individual level, maternal characteristics have been shown to predict child abuse potential (Hien, Cohen, Caldeira, Flom, & Wasserman, 2010). Specifically, Hien and colleagues (2010) found that a non-clinical sample of urban mothers (N = 152) who reported high levels of anger arousal and reactivity, as indicated by responses to the Novaco Anger Inventory, were more likely to have a high abuse potential. The authors of the study used the definition provided by Cloitre, Koenen, Cohen, and Han (2002) that describes reactivity as “affect dysregulation,” which is characterized by “the tendency to have low threshold, high intensity emotional reactions followed by slow return to baseline” (p. 1067). A highly reactive individual becomes upset easily, is unable to calm down and self-soothe, and allows their emotions, especially anger, to control his or her behavior. In addition, parents with a high level of reactivity do not reason logically and have little control of their anger or behavior. Also, cognitive processes such as stress, avoidant coping, irritability, and an external locus of control (LOC) have been found to be predictive of abuse potential and 8 disciplinary style among care takers (Rodriguez, 2010). For example, Rodriguez (2010) describes avoidant coping as a style of problem-solving that is characterized by avoidance of the problem, resignation, seeking alternative rewards, and lashing out at others, whereas approach coping involves logically analyzing the problem, seeking support and information, and taking action to evaluate different solutions. In other words, parents who use the avoidant coping style do not take positive steps to solve problematic parent-child relationships and are more likely to lash out at their children. Similarly, parental LOC, along with ability to empathize with the child and level of frustration tolerance, has also been correlated with child physical abuse risk in a sample of mothers of children with externalizing behavior problems (McElroy & Rodriguez, 2008). LOC refers to the perceptions an individual holds regarding the cause of events that affect him or her. Parents with an external LOC believe they are not in control of parent-child interactions, whereas an internal LOC indicates that the parent feels they are in control. It may be that mothers with external LOC feel less responsible for what happens to them or believe that the child is in control of his or her misbehavior, resulting in a perceived detachment from the consequences of child maltreatment. Overall, avoidant coping and external LOC may predict a care giver’s risk of engaging in child maltreatment. In addition, it has been illustrated that lower perceived social support (which is related to higher perceived stress) and a childhood history of physical abuse are significantly related to adult child abuse potential (Crouch, Milner, & Caliso, 1995). Also, a recent study of home-based family support and child maltreatment prevention services found that intimate partner psychological aggression, depression, and substance use were risk factors for attrition in such programs (Damashek, Doughty, 9 Ware, & Silovsky, 2011). Such attrition is problematic because it increases the risk of reabuse. Moreover, studies find that depressive symptoms have a direct, negative impact on effective parenting; however, trauma often co-occurs along with other factors, such as substance use and mental disorders, that have been shown to be predictive of child abuse potential (Rinehart et al., 2005). Emotional problems and insecure attachment styles have also been significantly and positively correlated with child abuse potential in a sample of domestic violence victims, with depression and anxiety as the strongest predictors (Rodriguez, 2006). Additionally, insecure attachment style in childhood has been correlated with child abuse potential in adulthood, while controlling for abuse history, in an at-risk sample of mothers raising children with behavioral problems (Rodriguez & Tucker, 2011). Furthermore, depression and other trauma symptomology, such as PTSD, anxiety, and anger/irritability, as well as intravenous drug use, have also been found to account for significant variance in scores on the Child Abuse Potential Inventory (CAP; Milner & Wimberley, 1979) among pregnant alcohol and other drug abusing women (Erickson & Tonigan, 2008). Hien et al. (2010) speculate that substance use is the mediating factor between distressing emotional states and high child abuse potential. Specifically, parents who cope with their negative emotions by using alcohol and other drugs are less likely to utilize the decision making process necessary for effective parenting. Substance use in response to stress can also be conceptualized as a form of avoidance coping (e.g., Banyard & Williams, 2007). In one of the few studies to utilize logistic regression to examine substantiation of child maltreatment, Wekerle and colleagues (2007) found that although “the total number of caregiver vulnerabilities [depression, history of trauma] 10 was a far more robust predictor of maltreatment substantiation than any specific vulnerability”(p. 438), substance abuse was the strongest individual predictor. Also, parental substance abuse has been linked to neglect recidivism (see Wekerle et al., 2007, for a review) and physical abuse perpetration (Walsh, MacMillan, & Jamieson, 2003). Overall, substance use may be an important variable to consider in the context of child abuse potential, and it is also important to consider from an intervention standpoint, as substance using parents who abuse their children often are excluded from receiving services specifically tailored to their co-occurring problems (Donohue, Romero, & Hill, 2005). Other predictors of child abuse that are personal characteristics of parents include parental stress and anger expression (Rodriguez & Green, 1997) as well as mental health problems, adverse life experiences, and neglect of the child’s hygiene (AgathonosGeorgopoulou & Browne, 1997). Moreover, one study that examined the scores of physically abusive parents on the Minnesota Multiphasic Personality Inventory – Second Edition (MMPI-2) was conducted by Stredny, Archer, and Mason (2006). The highest elevations (relative to other scales) were found on the psychopathic deviate and paranoia scales, but the mean scores on all scales were within normal limits. In addition, an examination of the characteristics of domestic violence perpetrators found that men who were attending court-mandated domestic violence treatment programs had no clinical elevations on any scale (Scott, Flowers, Bulnes, Olmsted, & Carbajal-Madrid, 2009), but were significantly different from the control group on scales pertaining to antisocial/psychopathic tendencies and symptoms associated with serious disturbances or “faking bad” (endorsing answers to several test items that were infrequently endorsed by 11 the sample used to standardize the MMPI-2). One interpretation of these findings, or the lack of clinical elevation, may be that it is not the clinical elevation per se that determines the utility of the violence predictor, but whether or not the predictor can differentiate between confirmed perpetrators of violence and non-violent individuals, as well as distinguish between high and low risk individuals. No published study, however, could be found that explored child abuse predictors from the MMPI – 2 – Restructured Form (MMPI-2-RF) (Ben-Porath & Tellegen, 2008). Statement of the Problem Given that millions of children are abused each year (Trickett, Negriff, Ji, & Peckins, 2011) and the probability of re-abuse following intervention by child protective services is high (around 85% for high-risk families; DePanfilis & Zuravin, 1998), it is of the utmost importance to examine the variables that may predict re-abuse. Also, just as there has been a recently emerging movement to identify risk and resilience variables within samples of abuse survivors (e.g., Asberg & Renk, 2012; Banyard & Williams, 2007), there is a call for examining such variables among perpetrators as well. For example, studies have examined substantiation of abuse reports (Wekerle, Wall, Leung, & Trocmé, 2007) and perpetration leading to fatalities (Yampolskaya, Greenbaum, & Berson, 2009) among caregivers referred or investigated for child maltreatment, but more research is needed to illuminate key variables to target for intervention. Also, relative to studies on differences between perpetrators and non-perpetrators, far fewer studies have investigated heightened child abuse potential among caregivers involved with, and referred for evaluation by, Child Protective Services. 12 The present study is an attempt to expand upon the existing literature by identifying the most important variables that predict child abuse (as measured by the CAP) from the higher order scales of the MMPI-2-RF and the Adult Self-Report (ASR) form of the Achenbach System of Empirically Based Assessment (ASEBA) among parents who were referred by child protective services for a parental fitness evaluation after their children were removed from the home. Previous research has examined the correlations between the MMPI-2 and child abuse and interpersonal violence (Stredny et al., 2006; Scott et al., 2009), and interpersonal violence has been shown to correlate with child abuse (Slep & O’Leary, 2005; Appel & Holden, 1998; McDonald et al., 2011), but to date no published study has examined the correlations between the MMPI-2-RF and child abuse potential. Furthermore, previous studies have identified predictor variables that differentiate between abusers and non-abusers, whereas the present study explored and identified the variables that differentiated between highand low-risk individuals within a clinical sample that have already been determined by state child and protective services (CPS) to have abused or neglected their children. Such predictor variables may have important practical implications, including the use of more serious intervention for high-risk individuals, the distribution of resources by CPS, and reduction of the potential for reabuse through education of families and those in charge of providing interventions. It should be noted that although several characteristics of the children themselves may predict their risk of being abused (e.g., delinquency, sociopathy, internalizing problems; Todd & Gesten, 1999), the present study focused on parental characteristics that predict elevations on a well-established measure of child abuse potential. 13 Furthermore, some variables may be correlated with child abuse potential without being “red flags” in and of themselves (e.g., poverty). Such variables were not directly assessed in this study, which may present a limitation. It is unlikely, however, that they would be directly related to the probability of child abuse and, therefore, are perhaps insufficient in the prediction of child abuse potential. Instead, variables related to stress, maladaptive coping, and psychopathology, which are often seen to a higher degree among impoverished groups (see Wekerle et al., 2007, for a review) were assessed. To address the overall goal of the study, the present analysis compared relevant variables to determine which were the most important predictors of child abuse potential within a clinically referred/identified care-giver sample. Although it was hypothesized a priori that select MMPI-2-RF scales (e.g., Behavioral/Externalizing Dysfunction, Emotional/Internalizing Dysfunction, and Thought Dysfunction) and ASEBA scales (e.g., Depressive Problems, Anxiety Problems, Avoidant Personality Problems, Antisocial Personality Problems, and Alcohol and Drug Substance Use) would correlate significantly with participants’ CAP scores (higher risk vs. lower risk), a multiple regression identified the most robust predictors of child abuse potential. For more specific details on the analyses, please see the method section. 14 CHAPTER THREE: METHOD Participants The participants that comprised the overall sample used in this study were 177 parents and primary caretakers who were court-ordered by Georgia’s Division of Family and Children Services to receive a psychiatric evaluation. The de-identified data was provided by a private practice group in Atlanta, Georgia, where the evaluations were conducted. About 70% (124) of the participants were female and about 30% (53) were male. The overwhelming majority of the sample (148/83%) was White, whereas 14 participants (8%) were Black, two (1%) were Filipino, and one was Latina (0.5%). Data on race were missing for 12 participants. In terms of marital status, 61 (34.5%) of the participants endorsed that they were married, 44 (25%) were single, 30 (17%) were separated, 23 (13%) were divorced, and four (2.3%) were widowed. Data on marital status were missing for 15 participants. Ages of participants ranged from 18 to 59-years (M = 33.39, SD = 9.03). Data on age were missing for eight participants. The specific types of abuse perpetrated by these individuals are unknown; however, different forms of child maltreatment, such as physical, emotional, sexual, and psychological abuse, as well as neglect, often co-occur (Dong et al., 2004), thus the examination of child abuse potential, regardless of abuse type, may still be relevant. We examined demographic variables for four subdivisions of participants. The first group represents the entire sample. The second group represents only the participants who had exact scores and excludes participants with categorical data (discussed below). The third group had exact scores and elevated CAP scores (i.e., 15 higher than 129). The last group had exact scores and CAP scores that were nonelevated. Demographic information for the overall sample (n = 177), the subsample with exact scores (n = 62), participants with elevated scores (n = 26), and participants with non-elevated scores (n = 36) are shown in Table 1. Table 1 – Sample Demographics Demographic Variable Overall Sample Subsample with Exact Scores Subsample with Elevated Scores Subsample with Non-Elevated Scores Mean Age 33.4 33.2 30.8 34.9 Gender (n) Female Male 124 (70.1%) 53 (29.9%) 46 (74.2%) 16 (25.8%) 25 (96.2%) 1 (3.8%) 21 (58.3%) 15 (41.7%) Race (n) White Black Filipino Latina 148 (89.7%) 14 (8.5%) 2 (1.2%) 1 (0.6%) 54 (90.0%) 5 (8.3%) 1 (1.7%) 0 (0.0%) 20 (83.3%) 3 (12.5%) 1 (4.2%) 0 (0.0%) 34 (94.4%) 2 (5.6%) 0 (0.0%) 0 (0.0%) Marital Status (n) Married Single Separated Divorced Widowed 61 (37.7%) 44 (27.2%) 30 (18.5%) 23 (14.2%) 4 (2.5%) 20 (34.5%) 16 (27.6%) 12 (20.7%) 8 (13.8%) 2 (3.5%) 9 (40.9%) 5 (22.7%) 3 (13.6%) 5 (22.7%) 0 (0.0%) 11 (30.6%) 11 (30.6%) 9 (25.0%) 3 (8.3%) 2 (5.6%) Measures The measures used in the court-ordered evaluations included the Minnesota Multiphasic Personality Inventory, Second Edition (MMPI–2; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989), the Child Abuse Potential Inventory (CAP; Milner & Wimberley, 1979), and the Adult Self-Report (ASR) form of the Achenbach System of Empirically Based Assessment (ASEBA; Achenbach & Rescorla, 2003). It should be noted, however, that the dataset for the present study consisted of MMPI-216 Restructured Form scores (MMPI-2-RF; Ben-Porath & Tellegen, 2008), which had been extracted previously from the MMPI-2 protocols. Elevated scores on the validity scales for the MMPI-2-RF resulted in the removal of 23 participants from the analysis. This number of invalid responses (13% of the 177 total), is similar to that of Scott et al. (2009), who found invalid MMPI-2 profiles (due to high rates of endorsing infrequent responses) among 16.7% of their sample of domestic violence perpetrators. The following were criteria for removal in the present study: CNS > 18, Fp-r > 100, VRIN-r > 80, or TRIN-r > 80, with meeting any of these criteria resulting in exclusion. After the elimination of 23 participants with invalid profiles, 9 participants without CAP scores, and 83 participants without exact CAP scores (only a designation of elevated or nonelevated scores), analyses were conducted on a remaining sample of 62 participants. The decision to use only those participants for which exact scores on the outcome measure (CAP) was available is based, in part, on the assumption that a) data can more easily be examined for outliers, b) results will be easier to interpret if actual scores are used, and c) it is more appropriate for our methodology and use of multiple regression (where the dependent variable must be continuous). Child Abuse Potential. The CAP is a 160-item screening instrument for physical child abuse potential. It is a self-report questionnaire used with individuals who are 18year-old and older. The CAP was constructed on the basis of personality traits reported in the literature to be “characteristic of individuals who abuse and neglect children” and a factor analysis resulted in the predictive dimensions of loneliness, rigidity, problems, and control (Milner & Wimberley, 1979). It contains 10 scales, including a 77-item clinical Abuse scale, six factor scales that go under the Abuse scale (Distress, Rigidity, 17 Unhappiness, Problems with Child and Self, Problems with Family, and Problems with Others), and three validity scales. Respondents answer “agree” or “disagree” for each item and scores range from 0 to 486. Respondents with scores above 166 are considered to be at medium risk for physical child abuse and respondents with scores above 215 are considered to be at high risk for abuse. Many of the items in the Distress, Rigidity, and Unhappiness scales concern mood and anxiety symptoms and the Problems with Child and Self and Problems with Family scales involve interpersonal or interactional problems. Research has been conducted to examine the construct, convergent, discriminant, and predictive validity of the CAP. In a study of undergraduates, the CAP was administered with an abbreviated MMPI and the Marlowe-Crowne Social Desirability Scale. There was a significant inverse relationship between CAP scores and the MMPI measure of ego-strength, which suggests that the CAP has high construct validity (Robertson & Milner, 1983). In addition, a comparison of CAP scores with the Mental Health Index (MHI) resulted in a positive correlation between CAP scores and MHI measures of psychological distress and a negative correlation between CAP scores and MHI measures of psychological well-being, thus supporting the convergent validity of the CAP (Milner, Charlesworth, Gold, Gold, & Friesen, 1988). Another study of the convergent and discriminant validity of the CAP reported positive relationships between abuse and apprehension, tension, and anxiety and a negative relationship between abuse and stability (Robertson & Milner, 1985). However, the CAP has been shown to have limited predictive validity, perhaps due to unaccounted protective factors (Chaffin & Valle, 2003). For example, in a study 18 of female parents who were enrolled in an at-risk parent-child program it was found that all of the parents who were later reported to a suspected child abuse and neglect team had previously scored above the CAP cutoff score for abuse, but the majority of parents with CAP scores above the cutoff did not subsequently abuse (Milner, Gold, Ayoub, & Jacewitz, 1984). Personality. The MMPI-2-RF was derived from the MMPI-2, which is the most widely used measure of personality in the world (Nichols, 2011). It was developed to assess personality in a variety of settings (Ben-Porath & Tellegen, 2008). The MMPI-2RF was designed to identify and separate the common “patienthood” factor found in many clinical disorders, called demoralization, from the clinical scales of the MMPI-2 in order to make them more unidimensional. It also eliminated invalid “subtle” items that were originally intended to identify underlying factors associated with a syndrome. BenPorath and Tellegen used factor analysis to remove demoralization from the clinical scales and standardized the resulting Restructured Clinical (RC) scales using data from 2,276 individuals randomly chosen from the MMPI-2 normative sample. The MMPI-2-RF has 338 true/false items and provides a set of validity scales, three Higher-Order scales, and nine clinical syndrome scales, as well as 23 Specific Problems Scales and two Interest Scales, with a standard score range of 20 to 120 for each scale. Of particular interest to the present study were the Higher Order scales, which consist of Behavioral/Externalizing Dysfunction (BXD), Emotional/Internalizing Dysfunction (EID), and Thought Dysfunction (THD). Sample items for these three scales, respectively, are: “I have never done anything dangerous for the thrill of it” 19 (scored false), “I am a very sociable person” (scored false), and “I believe I am being plotted against” (scored true). Psychological Functioning. The Achenbach System of Empirically Based Assessment – Adult Self Report (ASEBA-ASR; Achenbach & Rescorla, 2003) is a measure of social, emotional, and behavioral function in adults ages 18-59-years. It especially targets problems for the previous six months and includes scales for adaptive functioning, empirically based syndromes, substance use, Internalizing, Externalizing, and Total Problems, DSM-oriented scales, and a Critical Items scale. Responses to items include “Not True,” “Somewhat or Sometimes True,” and “Very True or Often True”, which are rated on a 4-point Likert-type scale. The ASEBA-ASR consists of 138 items and standard scores for each scale range from 50 to 100. Of particular interest to the present study are the substance use and DSM-oriented scales. The substance use scale items inquire about the number of times per day the respondent used tobacco (including smokeless tobacco), was drunk, and used drugs for nonmedical purposes (including marijuana, cocaine, and other drugs, except alcohol and nicotine). The DSM-oriented scales include Depressive Problems, Anxiety Problems, Somatic Problems, Avoidant Personality Problems, Attention Deficit/Hyperactivity Problems, and Antisocial Personality Problems. The DSM-oriented scales include items such as “Cries a lot,” “Worries about his/her future,” “Feels dizzy or lightheaded,” “Doesn’t get along with other people,” “Is too forgetful,” and “Argues a lot.” Achenbach and Rescorla (2003) analyzed numerous studies in order to determine the reliability and validity of the ASEBA adult forms. They found that the 1-week testretest reliability was high for most scales, the internal consistency was high for the ASR 20 empirically based problem scales and the DSM-oriented scales, cross-informant agreement was modest for substance use, the empirically based problem scales, and the DSM-oriented scales, and the scale scores were substantially stable. They also concluded that the problem items had good content validity, the criterion-related validity of scale scores was good, and the construct validity of the scales was supported by predicting ASEBA adult scores from ASEBA child and adolescent scores, associations between the scales and diagnostic assessment, associations with the Beck Depression Inventory, the Beck Anxiety Inventory, the MMPI, and the Symptom Checklist-90-Revised, and associations with a prior intervention and with scores on the Child Depression Inventory completed at age 11 (Achenbach & Rescorla, 2003). Overall, scales were included in the analysis if their predictive value of child abuse potential has been supported in the literature, resulting in the following list of potential predictors: the BXD, EID, and THD scales of the MMPI-2-RF and the Depressive Problems, Anxiety Problems, Avoidant Personality Problems, and Antisocial Personality Problems DSM-Oriented scales as well as the Alcohol Substance Use scale and Drug Substance Use scale of the ASEBA, resulting in a total of nine predictors. Hypotheses Based on the literature, the following hypotheses were generated: 1. Individuals who experience emotional (as measured by scores on the Emotional/Internalizing Dysfunction scale of the MMPI-2-RF) or cognitive (as measured by the Thought Dysfunction scale of the MMPI-2-RF) dysfunction are at higher risk for re-abuse (i.e., higher scores on the CAP). 21 2. Individuals who exhibit more aggressive or antisocial behavior (as measured by the Behavioral/Externalizing Dysfunction scale of the MMPI-2-RF) are at higher risk for re-abuse (i.e., higher scores on the CAP). 3. Individuals who endorse more frequent use and abuse substances (as measured by the Drug Use and Alcohol Use scales of the ASEBA) are at higher risk for re-abuse. 4. Individuals who endorse more depressive (as measured by the Depressive Problems scale of ASEBA) or anxious (as measured by the Anxiety Problems scale of the ASEBA) symptoms are at higher risk for re-abuse (i.e., higher scores on the CAP). 5. Individuals who use avoidance as a coping strategy more frequently (as measured by the Avoidant Personality Problems scale of the ASEBA) are at higher risk for re-abuse (i.e., higher scores on the CAP). 6. Individuals who have antisocial traits (as measured by scores on the Antisocial Personality Problems of the ASEBA) are at higher risk for re-abuse (i.e., higher scores on the CAP). 7. The combination of the aforementioned variables/subscales will predict significantly re-abuse potential. Summary of Scales Used in Analyses To test the aforementioned hypotheses, the Emotional/Internalizing Dysfunction (EID), Thought Dysfunction (THD), and Behavioral/Externalizing Dysfunction (BXD) scales from the MMPI-2-RF were included in the analysis (Hypotheses 1 – 2). In addition, the following ASEBA scales were chosen to be included in the analysis: the 22 Alcohol Use and Drug Use scales and the Depressive Problems, Anxiety Problems, Avoidant Personality Problems, and Antisocial Personality Problems of the DSMOriented scales (Hypotheses 3 – 6). An overall model predicting child abuse potential (Hypothesis 7) was also examined using all of the aforementioned predictors/scales. Overall, the aforementioned analyses served the purpose of exploring 1) bi-variate relationships among study variables (correlations); 2) group differences between upper and lower CAP groups (t-tests); and 3) predicting child abuse potential (multiple regression) within a sample of parents referred for evaluation after CPS involvement. Findings may aid in the identification of those care givers who are at an elevated risk for re-abusing their child following a substantiated instance of child abuse. Findings may also enhance our understanding of which predictor variables are important within a sample of confirmed or substantiated perpetrators of child maltreatment. Primary Statistical Analyses First, means and standard deviations for the overall sample (N = 62) were calculated for all study measures (relevant subscales only). Also, t-tests of sex differences on subscales were conducted for the overall sample (N = 62) in order to determine whether or not sex needed to be included as a predictor variable in the prediction model. Finally, in order to predict higher CAP scores from MMPI-2-RF and ASEBA scores, a multiple regression was used. Multiple regression is appropriate to use when predicting a continuous variable from a variety of continuous (subscale scores) and/or dichotomous (sex, minority status) variables. For the multiple regression, participants’ continuous score on the CAP was the dependent variable. Scores on the 23 MMPI-2-RF higher-order scales and the relevant scales of the ASEBA-ASR were the predictor variables. Secondary Analysis and Group Designation By definition, this sample of parents/care givers that were identified by CPS can be considered at risk for perpetrating abuse against a child, however, the present study sought to also identify “higher risk” parents within the sample. Specifically, group differences between ‘higher’ and ‘lower’ risk were assessed with t-tests (and discussed in the results below). For the t-tests, high risk for child abuse is represented by a score above the sample mean on the CAP, whereas a score below the mean indicates low risk. In this study, the mean CAP score for the 62 participants, i.e., CAP = 129, was used as the cutoff point for an “elevated score”, i.e., scores at or above 129 were considered “higher risk” for abuse and scores below 129 were considered “lower risk” within this sample. Unfortunately, only one individual in the elevated subsample was male, which likely minimized chances of sex being a significant predictor of abuse perpetration in subsequent analyses. Rationale for Group Designation. Although most studies use a less stringent cutoff of either 166 or 215 as recommended by the creators of the scale, our relatively low estimate may be appropriate given our sample of confirmed perpetrators of child maltreatment. For example, Holden, Willis, and Foltz (1989) suggest that the more liberal cutoff noted above could be used to identify parents “at risk for maltreatment before the occurrence of documented abuse rather than after the abuse has occurred” (p. 66). They report further that abuse potential cutoff scores should be interpreted cautiously “when the CAP is administered to samples containing subjects displaying 24 chronic problematic parenting” (p. 66). For example, CAP scores for physical or mixed physical/sexual abuse perpetrators (N = 37) in their sample (M = 145; SD = 85.4) were not significantly different from CAP scores of parents referred for reasons other than child abuse (M = 168, SD = 80.1). In other words, utilizing a more stringent cutoff (increasing the likelihood of participants being deemed “higher risk”) within this sample of confirmed perpetrators of some form of child maltreatment is in line with the recommendation to exercise caution. Consequently, the elevated and non-elevated groups for the present study were comprised of those above and below the CAP sample mean, respectively. Based on this cutoff, 26 participants comprised the high-risk group and 36 participants made up the low-risk group. Seven participants were missing data for the DSM-oriented scales and for the Alcohol and Drug Substance Use scales of the ASEBA, and one participant was missing data for the Alcohol and Drug Substance Use scales only. Mean substitutions using raw scores were used in cases of missing data. Individuals whose scores place them in the high risk range will be labeled “1” and those in the low risk range will be labeled “0” to indicate group belonging and allow for correlational analysis. 25 CHAPTER FOUR: RESULTS Means and standard deviations for all scales are shown in Table 2 below. An independent samples t-test conducted for age, the only continuous demographic variable, indicated that the elevated (mean age = 30.77, SD = 7.79) and non-elevated (mean age = 34.86, SD = 9.30) groups (as indicated by CAP scores) did not significantly differ on this variable, p = 0.446. Table 2 – Descriptive Statistics Scale Mean Score Standard Deviation Age 33.15 8.87 CAP 128.97 102.16 Alcohol Use 51.97 4.38 Drug Use 54.49 8.49 Depressive Problems 57.43 9.35 Anxiety Problems 56.81 7.55 Avoidant Personality Problems 55.59 8.08 Antisocial Personality Problems 55.89 7.17 Emotional/Internalizing Dysfunction 51.38 14.48 Thought Dysfunction 50.17 11.28 Behavioral Externalizing Dysfunction 50.98 11.81 T-tests were conducted to examine the differences between higher and lower risk groups on the nine subscales. The results show that the higher risk group had significantly higher scores on Depressive Problems (F = 16.57, p < .001), Avoidant Personality Problems (F = 11.57, p < .05), and Emotional/Internalizing Dysfunction (F = 5.37, p < .05) than the lower risk group (see Table 3 below).