Introduction

In 2011, an estimated 676,569 (9.1 per 1,000) children were determined to be victims of abuse or neglect. Of those victims, 17.6 % (118,825) suffered physical abuse and another 9 % (60,839) experienced psychological maltreatment (USDHHS 2011). While these figures reflect a substantial number of children who fall victim to abuse at the hands of their caregivers, the actual numbers are much higher. Only a fraction of abuse cases are ever reported and, ultimately, substantiated. In fact, in a general population survey by Straus et al. (1998), 49 per 1,000 parents reported perpetrating severe physical assault toward their children.

Child maltreatment has been studied for several decades and various individual and environmental indicators have been identified as risk factors for abuse and neglect. This information has led to the development of prevention programs that contributed to a reduction in maltreatment substantiations (Finkelhor and Jones 2006; Jones et al. 2006). Since the early 1990s, substantiated child abuse cases have decreased from a rate of 15.3 to the current rate of 9.1 (USDHHS 2011). This decrease can be attributed to a greater awareness and enhanced knowledge of the effects of the various indicators and the resultant increase in the number of prevention programs and services. Understanding child maltreatment indicators allows professionals to identify those who are at risk so the appropriate and necessary services or assistance can be provided to prevent future abuse.

One indicator consistently shown to increase the risk of child abuse is intimate partner violence (IPV) (Herrenkohl et al. 2008; Jouriles et al. 2008). While some studies report gender symmetry in IPV perpetration, the type and intensity of the violence tend to be more gender specific (Johnson 2006; Swan et al. 2008). Johnson’s (2008) review of data from general population surveys and shelter samples distinguishes patterns of violence that account for the gender differences. Situational couple violence, the most commonly reported form of IPV in general population surveys, occurs when couple conflicts accelerate to the point of violence. Males and females perpetrate situational couple violence fairly equally. On the other hand, intimate terrorism occurs when one partner uses coercive control with violence, or the threat of violence, as reinforcement. A third type, violent resistance, occurs when victims of IPV fight back against intimate terrorism. Both intimate terrorism and violent resistance are more often reported in agency/shelter samples and intimate terrorism is almost entirely perpetrated by men (Johnson 2011). Johnson’s typologies (Johnson 2008) will be used to examine IPV victimization and its effect on changes in family structure and subsequent child maltreatment.

Archer (2000) reviewed 82 studies using a variety of samples. Women were found slightly more likely than men to use physical aggression and to use it more frequently, but to be more likely than men to be injured from partner violence. The two studies on refuge (shelter) samples showed large effect sizes in the direction of men being more violent. The seven studies of couples undergoing treatment (for husband being assaultive, husband being alcoholic, and marital violence) had effects sizes in the direction of men being more violent, but effect size values were lower than for the shelter sample. Because of these differences, we chose to study only female victims of male perpetrators.

IPV Victimization and Child Abuse

Many previous studies that examined co-occurrence of IPV and child abuse either focused on abuse by the IPV perpetrator or failed to identify the abuser. In these studies, IPV was directly related to abuse of the child, particularly physical abuse (Berger 2005). Few studies have examined child abuse perpetrated by the victim; however, the findings in studies that do exist suggest the need for further examination. For example, while a majority of these studies support that female victims of domestic abuse are more likely to maltreat their children, Guterman et al. (2009) failed to find significance between maternal abuse and paternal coercion toward the mother.

The effect of IPV on women is no less devastating to a woman’s psychological health than abuse that occurs during childhood (Levendosky and Graham-Bermann 2001). Similar to child abuse, IPV assaults the victim’s psyche and is associated with, among others, increased depression (Campbell 2002; Nixon et al. 2004), anxiety (Pico-Alfonso et al. 2006), and substance abuse (Herrenkohl et al. 2008). These psychological manifestations have also been identified as child maltreatment predictors. Therefore, it is logical to assume that IPV promotes negative psychological responses in the victim that, ultimately, increase the risk of child abuse.

The path from IPV victimization to child abuse perpetration can be complex. Several explanations have been proposed, including: female IPV victims may project their experiences onto their children, assigning negative, aggressive attributes to the child’s behavior (Casanueva et al. 2009; Lieberman 2007; Lieberman and Van Horn 2005, 2008); the child may be seen as an extension of the abuser, especially in cases where the child resembles the abusive partner; and, IPV may interfere with the mother–child relationship by impeding the mother’s emotional and psychological availability, her responsiveness, and the warmth she is able to provide to her child (Cummings 1998; Levendosky and Graham-Bermann 2000; Osofsky 1998).

Early studies of child maltreatment concentrated on direct effects of individual and community-level factors with minimal consideration of any mediating or moderating effects of external variables. More recent research has identified several variables that influence the impact of child maltreatment risk factors. For example, belief in the use of corporal punishment has been identified as a mediating factor in the relationship between educational attainment and child abuse (Juby 2009). In addition to assessing the impact of IPV victimization on child maltreatment, this study also examines indirect effects of changes in family structure.

Change in Family Structure

IPV does not exist in a void; rather, it resonates throughout the home environment and affects the entire family system. It can lead to separation, child removal, or even death (Campbell et al. 2012; Centers for Disease Control and Prevention, n.d.). These disruptions may cause additional stress and tension, both of which have been conceptually and empirically associated with child maltreatment (Stith et al. 2009). The National Research Council (USDHHS 1999) reports that maladaptive parenting that arises from a parent’s behavioral characteristics such as an inability to control anger, impulsivity, background of abuse, or poor coping skills, is exacerbated by life events that cause stress. Similarly, Burrell et al. (1994) suggest that “stress is the most noteworthy correlate of child abuse potential” (p. 1046) and is also an important correlate of other variables that are associated with child abuse potential. The authors of the current study propose that IPV contributes to changes in the family structure that, ultimately, are associated with risk of child maltreatment.

In summary, IPV victimization has been associated with increased risk of child maltreatment. It is the purpose of this study to examine the source of the abuse and the direct and indirect effects of changes in the family structure. Our general hypothesis is that IPV victimization leads to changes in family structure that increase the risk of child maltreatment. The data allow us to test four areas: (1) female IPV victims, especially victims of intimate terrorism, are more likely to abuse their children, (2) IPV is associated with changes in family structure, and 3) changes in family structure increase the risk of child maltreatment.

Methods

Data for this study were obtained from an existing dataset. The data were initially collected for a study of IPV and substance abuse in women (Downs et al. 2006; Downs and Rindels 2004). Retrospective data on childhood experiences were obtained to examine the associations among these experiences, IPV, mental health and substance abuse.

Participants

Participants for this study were recruited from seven domestic violence programs/shelters and five substance use disorder treatment programs in a Midwestern state. The initial sample consisted of 447 participants. Some of the respondents failed to identify a father figure and were excluded from the analysis. The final sample included 399 females (domestic violence program/shelter, 49.7 %; substance abuse treatment, 50.3 %) who were either in substance abuse treatment or were receiving assistance from a domestic violence program or residing in a shelter. The majority of respondents were White, non-Hispanic (75.4 %), while Blacks made up 16.3 % of the sample. Only 32 % of the women were employed at the time of the interview and the women’s personal median income was $6,000.00 (with a range from $0 to $145,600). The women’s ages ranged from 18 to 68 (\(\bar{x} = 24.17\)) and 28.6 % of them were single and had never been married.

Procedures

With approval from the University’s Institutional Review Board, women from the substance use disorder treatment program were recruited through group meetings. To protect the women’s confidentiality, staff and male clients were asked to leave prior to the recruitment. At that time, one of the interviewers briefly described the study to the women, informed the women that volunteering for the study did not mean they had to answer all of the questions, informed the women that those who volunteered to be interviewed would be paid $20, answered the women’s questions, and asked women interested in being interviewed to sign up on a schedule sheet.

Recruitment at the domestic violence agencies took place during support groups. Meeting women in the groups had the advantage of efficiency; however, women who were in the shelter only a few days missed the opportunity to be asked to volunteer for the study. The effect on recruitment was that women who stayed in the shelter longer had a greater chance of being in the sample. Women who were in a short-term crisis may have left the shelter after a few days time and missed the opportunity to be in the study. Thus, an additional method of recruitment was used. Flyers describing the study were posted in prominent places in the shelter with a toll-free number to call to have an interview scheduled. Flyers were eventually placed in five of the seven domestic violence programs.

Measures

The data obtained from the participants involved retrospective considerations of events that occurred during childhood. While a majority of participants were able to respond to most of these questions, missing data did exist. One of the methods of analysis chosen for this study did not allow for missing data; therefore, missing values were replaced with variable means in the dataset.

Maternal Physical and Psychological Abuse

The Parent–Child Conflict Tactics Scale (CTSPC) was used to assess experiences of maternal physical and psychological abuse during childhood (Straus et al. 1998). To obtain the data, the participants were asked “how many times that your mother did these things during a typical year of your childhood” for each of the retrospective questions regarding their childhood, ages 7 through 18. The CTSPC has five subscales: nonviolent discipline, psychological aggression, minor physical assault (corporal punishment), severe physical assault (physical abuse), and very severe physical assault (severe physical abuse). For this study, the psychological aggression, severe physical assault and very severe physical assault subscales were used. The severe physical assault subscale and the very severe physical assault subscale were ultimately combined into one physical abuse variable, as the combined physical abuse scale had very high correlations with mother severe assault (.95, p < .001) and mother very severe assault (.82, p < .001) and an internal consistency reliability of .78.

Intimate Partner Violence

Assessing IPV perpetrated by the paternal figure involved retrospective self-report from the participants on the frequency with which their childhood father figure physically harmed their mother figure. Ordinal values for this variable ranged from “1” (Never) to “7” (All the time). To examine differing effects of levels of violence, clustering was used to develop categories of IPV that ranged from low to high. In k-means cluster analysis, subjects are divided into clusters that minimize the sum of squares of distances from each subject to the cluster mean. The analysis revealed three clusters and these were identified as: “minimal to no violence”, “some violence” and “frequent violence.” These recategorizations will allow for the use ANOVA to examine the effects based on Johnson’s typologies. Because intimate terrorism is associated with increased frequency (Johnson and Leone 2005), we expect that the “frequent violence” category will include more incidents of intimate terrorism and the “some violence” will include more events related to situational couple violence.

Change in Family Structure

To measure changes in family structure, participants were provided a list of experiences from which to select from (e.g., “Parents separated before you were 18,” “You were sent to a group home or detention home,” “You were placed in a foster home,” etc.). The participants were then asked, “Was there anything else that happened in your childhood that was disruptive or caused major changes in your household?” The number of items selected made up the value for this variable; however, some participants reported numerous events that resulted in skewed data. For that reason, the coding for this portion of the variable ended at nine events and any events listed beyond that were excluded from the quantitative analysis.

Results

Correlations were significant between IPV victimization and change in family structure (r = .286), maternal physical abuse (r = .114), and maternal psychological abuse (r = .229); however, once these variables were introduced into a path analysis (Fig. 1), significance between IPV and physical abuse disappeared while the impact of IPV on psychological abuse was substantially reduced (β = .159). The path between IPV and change in family structure reflected a stronger Beta value (.286) than any other path. Additionally, the number of changes in family structure was a stronger indicator of both physical abuse (β = .242) and psychological abuse (β = .179) than IPV.

Fig. 1
figure 1

Path analysis delineating the relationship between IPV and child maltreatment

Full size image

Change in family structure fully mediated the relationship between IPV victimization and maternal physical abuse (Δ = .069; Table 1). Significance between these variables was lost subsequent to inclusion of the change in family structure variable. These results imply that it is the change in family structure in IPV households that is associated with increased child physical abuse.

Table 1 Total, indirect, and direct effects
Full size table

Additionally, changes in family structure mediated the relationship between IPV victimization and maternal psychological abuse, however, this effect was partial as significance between IPV victimization and maternal psychological abuse was maintained, although reduced, after inclusion of family disruption (Δ = .051). This suggests that IPV victimization is associated with an increase in maternal psychological abuse both directly and indirectly (through changes in family structure).

IPV victimization and change in family structure accounted for a larger portion of the variance in maternal psychological abuse (R 2 = .074) than in maternal physical abuse (R 2 = .058). The presented model resulted in goodness of fit with the data (TLI = 1.002; CFI = 1.000; RMSEA = .000) and the χ2 was non-significant at .950 (p = .330). The absence of statistical significance for the Chi square attests to the model’s validity as this measure compares the degree to which the proposed model reflects that of the actual data. Standard acceptable values for the fit indices are: TLI > .95, CFI > .95, RMSEA < .06 (Hu and Bentler 1999; Schreiber et al. 2006).

Clustering of the IPV variable resulted in three groups. Mean scores for violence ranged from 1.23 (minimal to no violence) to 6.46 (very frequent violence) (Table 2). A great deal of variation existed between groups on the changes in family structure variable (2.12, minimal to no violence; 2.52, some violence; 3.97, frequent violence) and on the psychological abuse variable (2.94, minimal to no violence; 3.52, some violence; 3.73, frequent violence).

Table 2 Mean values across clusters
Full size table

As in the path analysis, physical abuse did not vary significantly across groups in the analysis of variance; however, the minimal to no violence group was significantly less likely to psychologically abuse than the more violent groups (some violence, p = .004; frequent violence, p = .001; Table 3). The frequent violence group reported a significantly larger number of changes in family structure than the minimal to no violence and the some violence groups (p = .000 and p = .000, respectively).

Table 3 Analysis of variance results comparing level of violence with maternal child maltreatment and changes in family structure
Full size table

Discussion

This study presents a multivariate illustration of the influence of IPV perpetration on child maltreatment. Our hypothesis that maternal victims of IPV were more likely to psychologically and physically abuse their children was supported; however, the paths to these outcomes varied. While maternal IPV victimization was directly, as well as indirectly, related to increased child psychological abuse, the path to child physical abuse was indirect—fully mediated through family disruption. Maternal figures of IPV were more likely to physically abuse their children when multiple structural changes occurred in the household. On the other hand, psychological abuse of children in IPV households existed apart from increased changes in structure.

While substantial research supports the direct effect of IPV and child maltreatment, few studies have examined the indirect influence of IPV on physical and psychological abuse of children. Application of the ecological model of child maltreatment can help explain how IPV victimization may increase the risk of abuse. This framework incorporates multiple interrelated levels of systems that contain influences that contribute to, or protect against, abuse. According to this model, it is the lower level systems, or primary relationships, that exert the most immediate effect on parenting (Belsky 1980; Cicchetti and Toth 2005). For example, paternal supportiveness toward mothers plays a protective role in maternal maltreatment risk (Zelenko et al. 2001), as well as decreases maternal rejecting and punitive behaviors (Brunelli et al. 1995), and increases maternal responsivity (Guterman et al. 2009). Alternatively, abusive primary relationships increase stress and tension within the family system, elevating the risk for child maltreatment. IPV victimization may co-occur with child abuse when the adult victim over-disciplines the child in an attempt to avoid conflict with an abusive partner, or from the adult victim’s diminished tolerance for or ability to manage parenting stresses (Coohey 2004). In her practice with IPV victims, one of the authors (Rindels, B., personal communication, September, 12, 2011) reports that some of her clients acknowledge the use of physical force against their children as a means of controlling behavior that may provoke the perpetrator.

While an ecological framework may explain the increased psychological abuse found in our study, it does not adequately explain why maternal IPV victimization was not directly associated with an increase in physical abuse. Taylor et al. (2009) similarly found that, when examining direct effects, IPV victimization was significantly associated with psychological abuse but not physical abuse. What was interesting in our study, however, was the path from IPV victimization to family disruption was the strongest of all the relationships, followed by the path from family disruption to maternal physical abuse. The indirect impact of IPV victimization on maternal physical abuse was also stronger than the indirect effect on maternal psychological abuse. Additionally, in the path model, a change in family structure eliminated the significant correlation between IPV victimization and maternal physical abuse. These findings suggest that IPV households with multiple changes in family structure are at high-risk for physical abuse. While the impact of IPV victimization is not direct, its ultimate effect on physical abuse (through disruption) is nonetheless significant. Because of the differing paths of IPV victimization on these two types of maltreatment, it is important that future studies distinguish between the major forms of abuse and neglect and thoroughly examine external influences.

Change in family structure completely mediated the relationship between maternal IPV victimization and physical child abuse, and partially mediated the relationship between IPV victimization and psychological child abuse. The context in which IPV occurs is often fraught with dysfunction and repeat victimization. It is known that adverse childhood experiences often preclude both IPV perpetration and victimization. Therefore, the instability that occurs in IPV households may not only be attributed to the current violence but may result from partners who bring a history of disorder to the relationship.

Recent research (Schilling et al. 2007; Turner and Butler 2003; Turner et al. 2006) on victimization in childhood can be applied to adulthood, since many of the problems experienced by adults are a carryover from childhood. These studies have focused on repeat adverse experiences and suggest that victims are more likely to experience successive victimizations. This may help to explain some of the findings in this study. IPV often leads to divorce, remarriage, and splitting of the parental infrastructure. Splitting of the family structure is associated with increases in child maltreatment and in child behavioral problems (Baldridge 2010), which can lead to out-of-home placements in foster care, groups homes, etc. It appears that being a victim creates vulnerabilities in individuals that make them more susceptible to adverse experiences in the future. This could explain why the change in family structure was so prevalent in our study population. A change in structure contributes to stress within the household and this provides a plausible explanation for the path between the number of structural changes and increased child maltreatment we observed (Appleyard et al. 2005; Guterman et al. 2009).

This study illuminates the detrimental effects of IPV victimization and its subsequent association with child maltreatment. Households wherein IPV occurs are not safe for children. These environments are not only risky because of potential abuse by the perpetrators, but they also create dynamics that increase the risk of child maltreatment by the IPV victim. Additionally, children growing up in these environments are exposed to multiple victimizations which put them at future risk, both as children and as adults.

Limitations

The respondents in this study answered questions regarding their childhood. Retrospective data has its limits since recall may not be entirely accurate; however, Williams (1994) found that abuse was actually under-recalled using this methodology. Alternatively, there is the possibility of over-recall in which those with more mental health problems may be more inclined to view their childhoods as abusive. It is also possible that some women categorized in the nonabusive mother figure group did experience abuse, although this would have a conservative effect on the findings of the study.

The data were collected between 1997 and 2001. While the average age of the participants was 34, they were answering questions about their lives growing up, around the 1970s and into the 1980s. During this period, no victim service programs existed. More support services are now available to IPV victims and this may affect the impact of domestic violence on families. Unfortunately, while programs are more prevalent today, they are often greatly underfunded.

The results of the present study cannot be generalized to women who never received services in adulthood. Future research will need to examine these issues with different samples of women and using different methods.

Implications

IPV in the home increases the risk of child maltreatment, not only by the perpetrator, but also by the victim. It is imperative that comprehensive family evaluations be done in all households where IPV exists to determine child safety.

Child protection agencies and those who work with victims of child abuse, whether they are children or adult victims, can benefit from this research. It is important for social workers in these agencies to recognize that IPV affects the entire family system. For example, substance abuse and child maltreatment by the mother may actually stem from IPV perpetration by the father. Substance abuse counselors and child maltreatment worker who are investigating child abuse cases should always inquire about IPV in the home in conjunction with a victim service advocate to make sure this is done safely and does not increase the danger. Treating only the substance abuse or managing only the child maltreatment may be insufficient if these issues are the direct or indirect result of domestic violence. Also, CPS investigators need routinely to include victim service advocates in child abuse investigations to discover IPV if it exists, and to work with the mother to keep the mother and children together instead of removing the children which constitutes a disruptive change in family structure.

Future research should focus on the effects of IPV on the individual types of maltreatment (physical abuse, psychological abuse, sexual abuse, and neglect) rather than using an inclusive measure. As previously stated, maternal child abuse should be investigated further to discern the differing effect of IPV victimization on psychological abuse and physical abuse.