BACKGROUND. Children exposed to intimate partner violence (IPV) are at increased risk for adverse mental and behavioral health sequelae, as has been documented by both systematic reviews and meta-analyses. Studies addressing the physical health impact of childhood IPV exposure, however, have not been summarized in a manner that might facilitate additional hypothesis-driven research and accelerate the development of targeted interventions.
METHODS. To identify a comprehensive set of articles examining the association between childhood IPV exposure and physical health, we searched online bibliographic databases including Medline, CINAHL, PsychInfo, and Sociological Abstracts using the keywords “domestic” or “intimate partner violence” and “infant,” “child,” or “pediatric.” From >2000 articles retrieved in the initial search, we used online abstract and bibliographic information to identify 94 articles potentially meeting the inclusion criteria of studies that (1) examined a postnatal physical health outcome related to IPV exposure and (2) had a contemporaneous control group. Thorough review of these 94 published studies yielded 22 that met these inclusion criteria. The data then were abstracted independently by 2 of the authors, and differences were settled with the assistance of a third author.
RESULTS. Childhood exposure to IPV increases the likelihood of risk-taking behaviors during adolescence and adulthood and is likely associated with underimmunization. Minimal data and study limitations preclude establishing a clear connection between IPV exposure and general health and use of health services, breastfeeding, or weight gain.
CONCLUSIONS. The impact on physical health from exposure to IPV during childhood is still uncertain. Future studies should be grounded in a theoretical model that specifies how IPV exposure can affect child health, should adjust for confounders adequately, should include a community-based sample, and should be of larger scale.
Each year in the United States, between 3.3 and 10 million children are estimated to be exposed to intimate partner violence (IPV).1,2 Although some of these children are “caught in the crossfire” and directly injured3 (leading potentially to intervention by child protective services), many more children who reside in homes mired by conflict and aggression are affected in a less direct manner by a multicausal pathway linking a “risky family environment” to ultimate poor mental and physical health.4
Evidence linking poor mental health outcomes and IPV exposure is mounting as several recent review articles and meta-analyses have aggregated existing studies.5–7 In particular, IPV-exposed children seem to have higher rates of both internalizing and externalizing problems than non–IPV-exposed children, leading to maladaptive emotional and behavioral development.5,7 Specifically observed problems include anxiety, depression, posttraumatic stress, and aggressiveness.8 On the basis of these observations, intervention programs have been designed and tested to ameliorate these untoward mental health outcomes.9,10
In comparison, evidence linking physical health outcomes and IPV exposure remains unsummarized. Consequently, we may underestimate the full adverse impact of IPV on children and underconceptualize the myriad ways by which IPV exposure may be harming children's physical health, thereby hampering our capacity to develop hypothesis-driven research that ultimately would guide interventions to improve outcomes.
We therefore conducted a comprehensive systematic review of the published literature that addresses the association between childhood IPV exposure and physical health in an effort to benchmark current knowledge and accelerate the development of effective, specific interventions to mitigate these problems.
Data collection and organization were divided into 2 phases. During the first phase, described below, we broadly searched the medical literature with the goal of obtaining a comprehensive collection of existing studies to answer the question, “Does childhood exposure to IPV affect physical health?” After assembling relevant studies, 4 categories were established to organize and frame our results: general health and use of health services, breastfeeding, failure to thrive (FTT), and risk-taking behaviors.
We searched the published medical literature from 1966 to 2003, including Medline via Ovid and the Institute of Science, Cumulative Index to Nursing and Allied Health, PsychInfo, and Sociological Abstracts. Articles written in all languages were included in the search, and the assistance of a fluent speaker for translation was obtained when needed.
These databases were searched by using key words including “domestic violence,” “intimate partner violence,” “battered women,” “partner abuse,” or “spouse abuse” and “infant,” “child,” “adolescent,” or “pediatric.” We augmented this search with a review of the bibliographies of related articles.
Eligible studies were limited to trials that investigated postneonatal physical health outcomes related to exposure to IPV. Studies without a contemporaneous control group were excluded. The primary outcome of a “physical health problem” was defined as any phenomenon that impacts the bodies' structure, function, or form. We elected to include intermediate outcomes such as physician visits, immunizations, and risk behaviors that, through omission or commission, directly impact physical health.4
Developmental delay and IQ were deemed to be related more closely to mental and behavioral health and were not considered for the purposes of this review. In addition, direct trauma from a parent either as the result of child abuse or involvement in a dispute between parents was not considered. Finally, studies that did not clearly examine violence between caregiving adults, such as works that included violence between siblings in the definition of family violence, were excluded.
The initial literature search yielded >2000 articles. Titles of articles were reviewed to screen for eligibility. If insufficient data were available from the online databases to determine if a study met criteria, the abstract was obtained and reviewed. Furthermore, citations included in articles reporting eligible studies were reviewed and pertinent abstracts retrieved.
After reviewing this information, 94 articles remained. Two authors (M.H.B.-M. and M.B.) independently read the text of these selected articles to determine inclusion. Several articles that were clearly ineligible, such as review articles, were excluded based on the abstract. The full text, however, of the majority of the 94 articles was reviewed to determine relevant studies. Disagreements were resolved by discussion and consensus mediated by the third author (C.F.).
The full text of all eligible articles was abstracted according to a prespecified abstraction form by 2 authors (M.H.B.-M. and M.B.). Abstracted data included information regarding the study population, design, outcome measures, statistical analyses, and study limitations. Differences in abstracted data were resolved by review of the study under question, discussion, and consensus mediated by the third author (C.F.). Final review of the abstracted data winnowed the initial set of articles that met inclusion criteria down to 22.11–32 Reasons for exclusion are detailed in Fig 1.8,33–103
We present our results organized around the 4 most common categories of outcome reported in these studies: (1) general health and use of health services11–18; (2) breastfeeding19; (3) FTT20,21; and (4) risk-taking behaviors.22–32
General Health and Use of Health Services
Do patterns of illness and illness-related visits differ between children in homes with IPV and those in homes without IPV? Are children in homes in which IPV occurs less likely to receive routine health maintenance such as immunizations or well-child visits?
Evidence likely supports that children exposed to IPV are at risk for underimmunization; the evidence is inconclusive regarding overall health status and use of health services.
Eight studies addressed these issues (Table 1).11–18 In 2 related articles, Attala et al11,12 compared a composite physical health index, immunization rate, hemoglobin, lead, general physical examination, and weight of children of women in 3 groups: nonabused women in the community, abused women in the community, and women residing in an IPV shelter. They concluded that IPV-exposed community children were more likely to be underimmunized than non–IPV-exposed community children and that IPV-exposed, sheltered children were less likely to have a normal physical examination than IPV-exposed community children. Additionally, sheltered children had lower physical health index scores than community, non–IPV-exposed children. None of the children involved were known victims of child abuse.
Although inclusion of a community sample of IPV-exposed children strengthened these studies, the small number of children in each group potentially precluded Attala et al from detecting additional differences between the groups. In addition, Attala et al commented that the 3 groups were different with regards to factors such as race and age but did not control for confounding in the analysis.
Also accessing a shelter population, Brooks et al13 used information from a nationally representative database in Wales to compare actual immunization rates and well-child visits in a convenience sample of children who were residing in an IPV shelter with the expected rates based on the current local child population and child patients from 1 socioeconomically depressed general practice. Children in the shelter were significantly less likely than their counterparts in both communities to be properly immunized and also were less likely to have received appropriate well-child care. The authors did not discuss measurement of, or adjustment for, potential confounders.
Kernic et al14 investigated the link between childhood IPV exposure and academic and school health concerns by comparing the children of women who initiated police or court reports of IPV and children who were enrolled in a local public school. The children of abused women were more likely than their peers to have a nursing visit that resulted in being sent home from school and were much more likely to receive a speech pathology referral. IPV-exposed children also were 4 times as likely to visit the school nurse secondary to alcohol or drug use, but this did not reach statistical significance. In this work, the authors considered potential confounders and effect modifiers and accounted for them in their description of the analysis. Additional strengths of this work include the use of a community-based sample and separate analyses for children with comorbid IPV and child abuse.
Using data obtained from the Longitudinal Studies of Child Abuse and Neglect, English et al15 sought to determine if, in a group of children reported to child protective services, a global physical health score was poorer for children exposed to IPV than for children not exposed to IPV. They concluded that “the influence of DV [domestic violence] [on global health score] … drops out when controlling for caregiver depression, conflict tactics, family health and competence, and the number of CPS [child protective services] referrals,” stating that the effect of IPV is likely indirect.
Arcos et al16 studied the infants of abused and nonabused women by comparing the 2 groups for well-child care, anthropometric measurements, and sickness-related physician visits. Both groups sought appropriate well-child care, although the abused women were less likely to complete visits at the normally scheduled times, and no difference existed in anthropometric measurements. In both the neonatal and postneonatal period, children exposed to IPV were significantly more likely to have sickness-related physician visits. Although the majority of the children of these women were followed for 6 months (90.8% in the DV group; 91.3% in the control group), by 11 months only 47.4% of the children originally in the IPV-exposed group and 58.7% of children in the non–IPV-exposed group remained enrolled in the study. Although the authors mentioned that the 2 groups were similar with regards to biopsychosocial variables, the specifics were not presented, and no mention was made with regards to adjustment for confounders or consideration of effect modification.
Using the North Carolina Pregnancy Risk Assessment Monitoring System, Martin et al17 examined, among other objectives, the impact of abuse before, during, and after pregnancy on both the presence of well-child care after delivery and the mean number of well-child visits during infancy. Using bivariate analyses, they found no differences in either outcome and emphasized the importance of pediatric IPV screening based on their finding that abused women “do take their infants to well-infant care visits.”17
Interested in the impact of childhood IPV exposure on adult women's health, Bensley et al18 analyzed data from the Washington State Behavioral Risk Factor Surveillance System. Using a logistic-regression model with adjustment for factors such as age, education, and income, witnessing IPV without concurrent child abuse was not associated with poor self-reported health status. The authors did not comment on examining for effect modification. The response rate for the survey was 57%, and the exposure and outcome variables were assessed via self-report.
Are women who are abused by their partners less likely to breastfeed than their nonabused peers?
The evidence is insufficient to draw a conclusion.
One study examined this question (Table 2).19 In a trial of women from 2 Supplemental Nutrition Programs for Women, Infants, and Children (WIC) clinics, Bullock et al19 compared initiation and duration of breastfeeding between abused and nonabused women and found no significant difference in either outcome. Breastfeeding initiation was investigated by using a χ2 test for difference of proportions. The authors provided some power calculations and noted that they had 80% power to find a 28% difference in breastfeeding initiation by using a 1-sided test; what constituted a clinically meaningful difference in breastfeeding between the 2 groups was not addressed. Using survival analyses to investigate breastfeeding duration, the authors considered covariates including lifetime abuse, age, marital status, and ethnicity.
Are infants who are born to abused mothers more likely to have poor weight gain than infants who are born to mothers who are not abused?
The evidence is insufficient to draw a conclusion.
Two studies addressed this issue (Table 3).20,21 McFarlane and Soeken20 studied infants from 0 to 12 months born to women who were known to be in relationships with IPV to determine if changing abuse status led to differences in relative weight gain at 6 and 12 months. Although weight gain between 0 and 6 months was not affected by abuse status at either 6 or 12 months, the rate of weight gain was significantly greater (P = .019) between 6 and 12 months for infants whose mothers stated that the abuse ended by 6 months and did not resume by 12 months than for infants in families with continuing abuse. In addition to comparing weights by using t tests, McFarlane and Soeken analyzed the data by using a linear-regression model that considered ethnicity and parity as potential covariates.
In a second study by Bithoney and Newberger,21 hospitalized children with nonorganic FTT were matched with hospitalized children with normal growth to determine, in part, familial characteristics associated with FTT. Using a stepwise-regression model to determine statistically significant variables, the authors determined that children without FTT were more likely to live in homes in which frequent violent disagreements occurred between parents (P < .07); this statistical method, however, did not mention accounting for the matched pairs, and thus the statistical significance of this finding is potentially overstated.
Are children who are exposed to IPV more likely to engage in risk-taking behaviors as adolescents or adults?
Evidence exists to support the link between IPV exposure and adolescent and adult risk-taking behaviors.
Seven publications conducted within the greater context of the Adverse Childhood Experiences (ACEs) Study met criteria for inclusion.22–28 Using a large population of adults enrolled in the Kaiser Health system, the authors of these articles conducted a large, survey-based study that examined the strength of relationship between ACE and adult health behaviors and health outcomes. The study was conducted in 2 waves (ie, 2 distinct time periods). During wave 1, 70% of the people returned surveys. In the second wave of the survey, which contained several additional questions, 65% of the people returned surveys. Previously developed and validated questionnaires were used in the construction of the larger survey. The ACE investigation included exposure to a battered mother; childhood emotional, physical, or sexual abuse; household substance abuse; mental illness in or incarceration of a member of the household; and parental separation or divorce. In a series of articles, the authors separately analyzed the contributions of individual and summed total of ACEs to outcomes including, but not limited to, smoking, alcohol abuse, drug abuse, sexually transmitted diseases, and teen pregnancy/unintended adult pregnancy. This group considered the causal pathway connecting ACEs and poor health outcomes and was careful in their statistical adjustment for confounders. The results of these studies are summarized in Table 4.
Four studies, in addition to the ACE studies, investigated the link between childhood IPV exposure and later risk-taking behaviors (Table 5).29–32 Using a nationwide sample, Caetano et al29 interviewed adults about childhood exposure to IPV and recent alcohol problems. Stratifying by race and gender in a χ2 analysis, they concluded that exposure to IPV was related to alcohol problems in black men but not in Hispanic or white men or in women of any ethnicity. In a logistic-regression model that controlled for age, income, marital status, employment, and, interestingly, alcohol consumption, Caetano et al found that both men and women who had childhood IPV exposure were more likely than those who had not to have adult alcohol problems. Finally, they stated that, within this regression model, men who witnessed IPV and had been abused themselves were less likely to have alcohol problems.
Although one of the objectives of Caetano et al was to look at ethnic differences in the impact of childhood exposure to IPV on alcohol problems, they did not discuss looking for effect modification in the stratified analyses or describe testing for interaction between ethnicity and childhood exposure in the regression model.
Carlson30 recruited subjects from residential drug-treatment centers and from a shelter for runaway adolescents to explore the relationship between child abuse and/or childhood IPV exposure and, among other outcomes, substance abuse. Using a 2 × 2 analysis of variance to examine potential interaction between child abuse and witnessing IPV, Carlson commented that substance abuse was unrelated to witnessing IPV but did not provide these statistics. In addition, power calculations were not presented.
Fergusson and Horwood31 investigated the relationship between childhood IPV exposure and poor psychosocial adjustment, including substance abuse/dependence, in adolescence. Using a χ2 test, they determined that a relationship existed between IPV exposure and all types of substance abuse/dependence. With the exception of alcohol abuse/dependence after witnessing mother-perpetrated IPV, the association between IPV and substance abuse/dependence disappeared after adjusting for covariates related to “family context.”
Conducting a secondary data analysis, Elliott et al32 examined a sample of adolescent girls to investigate the link between experiencing or witnessing family violence and sexually risky behavior. Using a logistic-regression model that considered multiple covariates and interaction between child abuse and IPV, either witnessing IPV or experiencing violence increased the odds of engaging in sexually risky behavior, but the occurrence of both together decreased this association.
Exposure to IPV during childhood seems to increase the likelihood of engaging in health-adverse behaviors later in life and likely increases risk for underimmunization. Because of limitations of the published studies and a relative paucity of data addressing certain key questions, we cannot determine if a connection exists between childhood IPV exposure and general health and use of health services, breastfeeding, or weight gain. Specifically, our ability to draw strong conclusions were hampered by 4 major threats to drawing valid scientific inferences that we found among the included studies:
Poorly specified theoretical model of cause and effect: Childhood IPV exposure potentially leads to poor physical health indirectly through a series of interrelated intermediaries that function along a causal pathway, an example of which (designed by Repetti et al4) is depicted in Fig 2. In investigating the diverse repercussions of IPV exposure, factors along this causal pathway (such as maternal depression or substance abuse, chaotic home environment, or poor role modeling) should be considered as cascading forces that lead to poor outcomes rather than as confounders. Studies that treated these intermediary influences as confounding influences and adjusted for them in the analyses may well have inadvertently attenuated the association between IPV exposure and physical health outcomes.15,18,29,31
Failure to consider confounding and effect modification: The complex, intricate relationship between childhood IPV exposure and health outcomes is certainly entangled by common confounding influences and likely modified by both protective and accentuating factors. In addition to specifying the hypothesized chain or web of cause and effect, investigators should specify a priori which additional factors are thought to confound or modify any relationship between IPV exposure and physical health. Each of these variables then should be measured appropriately and considered in an adjusted analysis.
Limited generalizability because of sampling from shelters: Several studies enrolled as subjects children of women who were in IPV shelters. Women who leave a violent relationship and enter a shelter are a self-selecting population that potentially differs from women who either remain in a violent relationship or seek refuge elsewhere. The shelter population may have experienced more severe violence or may have fewer financial resources than a nonsheltered group. These differences may distinguish not only the women but their children as well, as suggested by the finding that children residing in the community have a different risk profile compared with sheltered children.11,12 To enhance generalizability, future studies should seek a broader source population than IPV shelters.
Samples of small size: Several articles included in this review had small sample sizes, which raises concern that the ensuing analyses may have been underpowered to find true differences in unadjusted analyses and more certainly underpowered to detect interactions. Yet, few studies reported power calculations. A greater number of studies conducted with larger samples and clear reporting of their power to detect clinically meaningful effects are needed.
A few aspects of our review warrant close consideration. First, we used demanding inclusion criteria, which necessitated, for example, that studies have a contemporaneous control group to be included. Although enforcing these strict criteria decreased the total number of studies, this degree of methodologic rigor is required to determine if an association exists between childhood IPV exposure and physical health sequelae. We also required that studies examine the specific impact of IPV, as opposed to community violence, child abuse, or general family violence. Although violence exposure in general certainly impacts children negatively, better delineating the impact of specific forms of violence (such as IPV) may enable the design of specific interventions that target individuals who are exposed to certain violent interactions. Our restrictive definition of physical health likely further narrowed the spectrum of included articles. Finally, despite our efforts to conduct a thorough search, our search strategy may have inadvertently missed relevant studies.
These caveats notwithstanding, this review emphasizes the need for a better understanding of the physical health outcomes for children exposed to IPV, which in turn will require a clearer conceptualization of the causal pathway linking the two, including, for example, additional exploration of the impact of IPV on children's physiologic regulatory systems (Fig 2).4 Such knowledge ultimately will enhance our ability to intervene to improve the well-being of these individuals as children and, subsequently, as adults.
- Accepted August 3, 2005.
- Address correspondence to Megan H. Bair-Merritt, MD, General Pediatrics, 3535 Market St, 15th Floor, Room 1550, Children's Hospital of Philadelphia-North, Philadelphia, PA 19104. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Fantuzzo J, Mohr W. Prevalence and effects of child exposure to domestic violence. Future Child.1999;9(3) :21– 32
- ↵Carlson B. Children's observations of interparental violence. In: AR R, ed. Battered Women and their Families. New York, NY: Springer;1984
- ↵Christian C, Scribano P, Seidl T, Pinto-Martin J. Pediatric injury resulting from family violence. Pediatrics.1997;99(2) . Available at: www.pediatrics.org/cgi/content/full/99/2/e8
- ↵Kolbo J, Blakely E, Engleman D. Children who witness domestic violence: a review of empirical literature. J Interpers Violence.1996;11 :281– 293
- ↵Edleson J. Children's witnessing of adult domestic violence. J Interpers Violence.1999;14 :839– 870
- ↵Graham-Bermann S. Designing intervention evaluations for children exposed to domestic violence: applications of research and theory. In: Graham-Bermann S, Edleson J, eds. Domestic Violence in the Lives of Children: The Future of Research, Intervention, & Social Policy. Washington, DC: American Psychological Association;2001
- ↵Brooks RM, Ferguson T, Webb E. Health services to children resident in domestic violence shelters. Ambul Child Health.1998;4 :369– 374
- ↵Bensley L, Van Eenwyk J, Wynkoop Simmons K. Childhood family violence history and women's risk for intimate partner violence and poor health. Am J Prev Med.2003;25 :38– 44
- ↵Hillis SD, Anda RF, Felitti VJ, Nordenberg D, Marchbanks PA. Adverse childhood experiences and sexually transmitted diseases in men and women: a retrospective study. Pediatrics.2000;106 (1). Available at: www.pediatrics.org/cgi/content/full/106/1/e11
- ↵Hillis SD, Anda RF, Dube SR, Felitti VJ, Marchbanks PA, Marks JS. The association between adverse childhood experiences and adolescent pregnancy, long-term psychosocial consequences, and fetal death. Pediatrics.2004;113 :320– 327
- ↵Caetano R, Field CA, Nelson S. Association between childhood physical abuse, exposure to parental violence, and alcohol problems in adulthood. J Interpers Violence.2003;18 :240– 257
- ↵Carlson B. Outcomes of physical abuse and observation of marital violence in adolescents in placement. J Interpers Violence.1991;6 :526– 534
- Narusk A. Transmission of drinking habits within the family. Contemp Drug Probl.1991;18 :645– 671
- Sanchez Huesca R, Cruz V, Encinas R, Pantoja G. Early detection of risk factors associated with illegal drug use. Salud Mental.2002;25 :1– 11
- Shaw BA, Krause N. Exposure to physical violence during childhood, aging, and health. J Aging Health.2002;14 :467– 494
- Carlson B. Adolescent observers of marital violence. J Fam Violence.1990;5 :282– 299
- Suh E, Abel E. The impact of spousal violence on the children of the abused. J Independent Soc Work.1990;4 :27– 34
- Davis L, Carlson B. Observation of spouse abuse: what happens to the children? J Interpers Violence.1987;2 :278– 291
- Hilberman E, Munson K. Sixty battered women. Victimology.1977;2 :460– 470
- Stagg V, Wills GD, Howell M. Psychopathology in early childhood witnesses of family violence. Top Early Child Spec Educ.1989;9 :73– 87
- Webb E, Shankleman J, Evans MR, Brooks R. The health of children in refuges for women victims of domestic violence: cross sectional descriptive survey. BMJ.2001;323 :210– 213
- Wildin SR, Williamson WD, Wilson GS. Children of battered women: developmental and learning profiles. Clin Pediatr (Phila).1991;30 :299– 304
- Boothroyd L. Domestic violence: implications for children. Midirs Midwifery Dig.2002;12(suppl 2) :S9– S11
- Garcia J, Davidson L. Researching domestic violence and health: National Perinatal Epidemiology Unit. Midirs Midwifery Dig.2002;12(suppl 2) :S25– S29
- Leal C. Children and family violence in Brazil. Caribb J Criminol Soc Psychol.2000;5 :202– 208
- Margolin G. Effects of Domestic Violence on Children. Washington, DC: American Psychological Association;1998
- McIntosh J. Children living with domestic violence: research foundations for early intervention. J Fam Stud.2003;9 :219– 234
- Moore T, Pepler D, Mae R, Michele K. Effects of Family Violence on Children: New Directions for Research and Intervention. Hillsdale, NJ: Lawrence Erlbaum Associates;1989
- Rossman B. Longer Term Effects of Children's Exposure to Domestic Violence. Washington, DC: American Psychological Association;2001
- Wolak J, Finkelhor D. Children Exposed to Partner Violence. Thousand Oaks, CA: Sage;1998
- Anda RF, Felitti VJ, Chapman DP, et al. Abused boys, battered mothers, and male involvement in teen pregnancy. Pediatrics.2001;107 (2). Available at: www.pediatrics.org/cgi/content/full/107/2/e19
- ↵Dube SR, Felitti VJ, Dong M, Chapman DP, Giles WH, Anda RF. Childhood abuse, neglect, and household dysfunction and the risk of illicit drug use: the adverse childhood experiences study. Pediatrics.2003;111 :564– 572
- Copyright © 2006 by the American Academy of Pediatrics