Pediatrics, Vol. 84, No. 3, September 1989, 531-535.
Victimization of Mothers of Abused Children: A Controlled Study
Linda McKibben, MD, MPH, Edward De Vos, EdD, and Eli H. Newberger, MD
From the Department of Pediatrics, Boston City Hospital and Boston University School of Medicine, the Department of Pediatric Medicine, The Children’s Hospital, and the Department of Pediatrics, Harvard Medical School, Boston, Massachusetts
ABSTRACT. To search for indicators of violence against mothers of child abuse victims by husbands or boyfriends, the women’s medical records were reviewed and compared to records of mothers of a nontraumatized child comparison group. Of the 32 children ascertained in a 6 month interval, the records of mothers of 19 (59.4 %) were diagnostic or highly suggestive of current or previous victimization. Although the prevalence of documented violence against the mothers of children in the comparison group was an unexpectedly high 16%, the case-control difference was highly significant (P < .001). Although differences were found in the (younger) ages and (higher) parity of mothers of abused children, these differences did not predict risk of mothers’ exposures to violence in a multivariate analysis. The rate of violence against single mothers of child abuse victims, however, was four times the rate against mothers who were married (P = .022). These findings suggest a need to broaden the diagnostic conceptualization of child abuse to include maternal victimization and argue for including data concerning maternal risk in formulating diagnoses and disposition plans for abused children. Pediatrics 1989;84:531-535; child abuse, maternal victimization, battered women.
With the publication 27 years ago by Kempe and colleagues1 of the influential paper, “The Battered Child Syndrome,” a diagnostic concept was coined that stipulated a causal connection between the aggressive actions of a perpetrator (a parent or foster parent) and the symptoms of a child victim. In the intervening years, the notion has been formalized in laws obliging physicians and others to report child abuse, and the concept has been expanded in clinical studies and legal definitions of child abuse.2
An important conceptual advance was made in 1980 with the publication of the Straus et al3 national survey of family violence, in which parent to child violence was perceived as one among many modes of physical aggression within families. Violence between parents was estimated to be as frequent as, or more frequent than, violence toward children. Violence between parents and toward children were estimated from interviews to be related. Women who were victims of severe violence, for example, were 150% more likely to use severe violence in resolving conflicts with their children than women who were not.
A recent review of data from studies concerning the posited intergenerational transmission of violence suggests that minimally 30% of children will express toward their own offspring physical or sexual abuse or extreme neglect similar to that which they had suffered.4
Although a literature is developing regarding the psychologic impact of violence between adults on children, there are few reports of physical victimization of both children and mothers.5-11 With a view to exploring the extent of overlap of maternal and child abuse, a retrospective case-control study was designed.
METHODS
Case Selection
Cases were selected through a multitiered review of all child abuse reports filed at Boston City Hospital during the 6-month period, January I, 1986, through July 6, 1986. Given the purpose of this study, only those reports filed by emergency room staff were retained for further review; there were 95 such filings (59% of the total). These reports were evaluated by the following inclusion criteria: (1) physical examination findings positive for signs of physical abuse except where sexual abuse was the chief complaint; (2) presence of biologic mother at the time of the child protective report, and (3) injuries that were not self-inflicted. Reasons for exclusion included injury known to be inflicted by a day-care worker or a case that was deemed “unsubstantiated” after investigation by the Massachusetts Department of Social Services.
On the basis of these inclusion and exclusion criteria, 63 case reports were retained for further review. The mothers’ charts were then obtained through the following process. First, the mother’s name and address, as listed on the child abuse report, were compared with those of the insurance guarantor (usually the mother) on the child’s computerized demographic record. The mother’s social security number was often available from this source. Subsequently, the mother’s name was entered into the hospital’s computerized data base. These records were reviewed and the mother’s identity was verified through a comparison of social security number, address, and consistency of information in the medical chart itself (eg, birth records of the child). Following a definite match, the medical record number was retrieved and the medical chart requested. Sixteen mothers had no medical record numbers at Boston City Hospital and presumably had not received medical care there. Forty-seven charts were requested; 5 of these were not available after two requests from the medical records department. The remaining 42 charts were reviewed. Ten of these medical records were excluded because of insufficient information.
Control Selection
Control mothers were also selected based on initial review of children’s records. The emergency room daily logs were scanned for records of children with nontraumatic complaints who had sought medical attention within the same month as the index case. In addition to time of emergency room visit, the children were also matched for age, sex, race, and socioeconomic status (dependency on Medicaid was used as a proxy for low socioeconomic status). The same process used to locate a case mother’s chart was used to retrieve the control mother’s medical chart.
Chart Review
Thirty-two case and 32 control mothers’ charts were evaluated and assigned to one of four categories of maternal victimization: (1) diagnostic, (2) suggestive, (3) notable for relationship problems and psychosomatic complaints, and (4) nonsuggestive. These categories were operationalized using the following criteria: diagnostic – a chart with explicit data both confirming intentional physical trauma and identifying the perpetrator as an intimate man; (2) suggestive – a chart with documented intentional physical abuse but with no specification of the perpetrator’s identity, either as man or woman, acquaintance or stranger (3) relationship problem – a chart without specific indication of physical abuse but rather of significant marital/relationship problems with such associated psychosomatic complaints as hyperventilation syndrome, depression, and tension headaches; and (4) non-suggestive – a chart with no evidence of physical abuse and/or relationship problems. (It should be recalled that charts with insufficient information for evaluation were excluded before this review.) The following additional information was extracted from the medical record: mother’s date of birth, race, employment status, type of health insurance coverage, and number of other children.
The data were gathered from the medical records by one of us, who was unaware of the hypothesis and the classification status of case and control mothers. Because the first eligible control mother was used, however, there was little opportunity for bias in the selection of the comparison group. The possible intrusion of bias in the coding of data from the records was addressed in meetings among the coauthors as the study proceeded.
The study design was reviewed and approved by the chief of the Boston City Hospital’s Child Protection Program, the chairman of the Department of Pediatrics, and the Institutional Review Board before the initiation of the research.
Statistical Methods
Data analysis was done with statistical methods appropriate both to the level of measurement of the variables and to the sampling design of the study. The appropriate unit of analysis for the central questions in this retrospective study was the matched pairs that constituted the sample rather than individual subjects. Thus, for example, the major question regarding differences in the presence or absence of violence against mothers of case children and control children was addressed using McNemar’s test with a continuity correction.12 Where the variable being considered was at least interval level, as in the comparison between cases, and controls regarding maternal age, a paired t test was used.
RESULTS
The child’s record served as the basis for selection of control children’s records. They were matched for sex, age, race, socioeconomic status, and time of emergency room visit. The matching variables are compared in Table 1.
All control children were seen in the pediatric emergency room within 1 month of their target case child’s visit. The criteria for matching were adequately met. Characteristics of case mothers and control mothers are compared in Table 2.
Although case mothers were somewhat younger than control mothers, the difference was not statistically significant (paired t = -1.83, P = .08). Similarly, although case mothers were twice as likely to be employed than control mothers, the difference again was not significant (McNemar’s test, x2 = 3.125). Case and control mothers did differ significantly, however, on marital status and parity. Fewer than one of five case mothers were married (19.4%) in comparison to nearly half of the control mothers (45.2%) (McNemar’s test, x2 = 4.083, P < .05). Similarly, case mothers had more children (1.28 vs 0.56 children); the difference was significant (paired t = 2.70, P = .011).
| TABLE 1. Case and Control Children Variables That Served as the Basis for Matching | ||
|
Cases Children |
Control Children |
|
|
Sex (No. [%]) |
15 (46.9) |
15 (46.9) |
|
Age |
7.04 y |
6.89 y |
|
Race/ethnicity (No. [%]) |
23 (71.9) |
|
|
Health coverage (No. [%]) |
25 (78.1) |
25 (80.6) |
|
TABLE 2. Case and Control Mother Variables |
||
|
|
Case Mothers |
Control Mothers |
|
Employment status (No.[%]) |
22 (68.7) |
28 (87.5) |
|
Age (y) |
28.81 |
31.25 |
|
Marital status (No. [%]) |
6 (19.4) |
14 (45.2) |
|
Other children (No.) |
1.28 |
0.56 |
The central focus of this study was the relationship between child abuse and violence against mothers in mother-child pairs where a protective report was filed on behalf of the child at an emergency room visit. The cross-tabulation of maternal victimization categories for case and control pairs is given in Table 3. The marginal totals indicate that case pairs were far more likely to be classified in the suggestive and diagnostic categories (37.5% and 21.8%, respectively) than were the control pairs (9.4% and 3.1%, respectively).
|
TABLE 3. Cross-Tabulation of Victimization Categories for Case and Control Mothers (Full Categorization) |
|||||
|
Case Category |
Control Category |
||||
|
|
Non- |
Relationsbip |
Suggestive |
Diagnostic |
Total |
|
Nonsuggestive |
6 |
1 |
0 |
1 |
8 (25.0) |
|
Relationship |
5 |
0 |
0 |
0 |
0 (15.6) |
|
Suggestive |
9 |
1 |
2 |
0 |
12 (37.5) |
|
Diagnostic |
5 |
1 |
1 |
0 |
7 (21.8) |
This scale of violence against mothers, however, is not really ordinal. The amount of violence within the suggestive category may have been greater than within the diagnostic category; the difference was that a perpetrator was identified in the latter and not in the former. Thus, there were alternative methods for collapsing the table, depending upon emphasis.
A cross-tabulation of dichotomized maternal victimization categories that identify the mother as victim, disregarding the identity of the perpetrator and his relationship to the mother, is presented in Table 4. Thus, the diagnostic and suggestive categories were pooled, and relationship problems and nonsuggestive categories were pooled.
|
TABLE 4. Cross-Tabulation of Victimization Categories for |
|||
|
Case Category |
Control Category |
||
|
|
Victim |
Nonvictim |
Total No. (%) |
|
Victim |
3 |
16 |
19 (59.4) |
|
*The percentages for victims and nonvictims in the control category are 12.5 and 87.5, respectively. |
|||
The case pairs were far more likely to be victims of violence than were the control pairs (59.4% vs 12.5%, respectively). One of eight control mothers were victims of violence, an impressive prevalence, and nearly three of five case mothers were victimized. The difference was highly significant (McNemar’s test, x2 = 11.53, P < .001). The difference between case and control pairs with respect to the victim of violence dichotomy as shown in Table 4 was striking. The result requires no elaboration with respect to the child-focused variables that served as the basis for case and control selection. Nevertheless, given the differences between case and control pairs on mother-level variables (viz, number of other children, maternal age, and marital status), this difference needs to be examined in more detail.
A significant difference was noted earlier between case mothers and control mothers with respect to the number of other children. A one-way analysis of variance was performed on the case data to evaluate the relationship between this variable and maternal victimization. The analysis revealed no overall significance when using the full four-category victimization scale. However, case mothers categorized as suggestive and diagnostic victims of violence had more children than did mothers in the other two categories. The dichotomized violence scheme was employed and a t test was used to compare victims of violence (mean 1.47 children) with nonvictims (mean = 1.00 children); the difference was not significant (t = 1.158).
A significant difference was also noted earlier in the ages of case and control mothers. An analysis was conducted of case data to determine whether within that group any relationship existed between maternal victimization and maternal age. Again, the overall differences using all four violence categories were not statistically significant. However, victimized case mothers appeared somewhat older than nonvictimized mothers. A t test on the dichotomized grouping indicated that victimized case mothers (mean age 30.84 years) were significantly older than non victimized case mothers (mean age 25.85 years) (t = 2.213, P = .034). For control mothers, no differences in age were found.
Marital status, as well, demonstrated a significant relationship to case-control status. The case data were analyzed to test whether or not a relationship existed between victimization and marital status. The results appear in Table 5. Whereas 16.7% of married case mothers were victims of violence, 68.0% of single case mothers were victims, more than four times the rate (x2 = 5.236, P = .022).
|
TABLE 5. Victimization and Marital Status of |
||
|
Child |
Mother’s Marital Status |
|
|
Married |
Single |
|
|
Victim (No. [%]) |
1 (16.7) |
17 (68) |
A similar analysis was conducted for control mothers. Although the number of control mothers who were victims of violence was small, the results were consistent with the case data. Fewer than one in four single control mothers was a victim a violence, although all four victims were single; none of the married control mothers were victims (Fisher’s exact test [one-tail], P = .076).
Given the original focus of this study and the relationship between marital status and victimization, an analysis was performed limited to case control pairs of the same marital status. In this subsample, case pairs were again far more likely to be victims of violence than were control pairs (66.7% vs 22.2%, respectively). Six of nine mothers of case children were victimized, in comparison to two of nine mothers of control children. The difference was again significant (McNemar’s test, x2 4.90, P < .05).
DISCUSSION
The 59.4% concurrence of maternal victimization and child abuse in this study is impressive and worrisome. The findings exceed previous estimates in the literature.
Stark et al10 reviewed the medical records of mothers of children referred to a child protective team at Yale-New Haven Hospital during 1 year without restriction to emergency room visits. Their “positive” and “probable” groups were defined similarly to our categories of diagnostic and suggestive, respectively. In their study, these two groups accounted for 41% of the cases of victimization. Their positive group, in a manner similar to our own diagnostic group, accounted for approximately 25% of the cases; in our own, the result was 23%. However, our suggestive group was more than double their probable group (40% vs 16%, respectively). Our finding of three times the amount of violent marital conflict may signify greater risk for the Boston mothers, assuming that health care providers in both locations are similarly documenting these issues. This comparison suggests that traumatized children visiting inner-city emergency rooms are at high risk for exposure to violence against their mothers.
Pascoe et al11 reviewed records of children who were referred to a child-protection team in North Carolina during a 2-year period, where a majority of families were rural. Their results revealed a higher prevalence of wife abuse in women with multiple pregnancies and in alcoholic families. Overall, they found a 40% prevalence of wife abuse when the child’s records were supplemented with information from social workers involved with the cases. The authors suggested that this rate was high partly on the basis that the social workers were attuned to family violence issues. As a corollary, Flitcraft and Stark’s probable and our suggestive groups may also be composed of women abused by male partners; but their medical records may not document the abuse sufficiently. Many health care providers have insufficient knowledge and understanding of family violence13 and many victims seek medical care with nontraumatic complaints.14
The finding of an association between maternal victimization and single marital status corroborates the findings of Goldberg and Tomlanovich.14 They demonstrated an increased risk of spouse abuse in unmarried women.
More research concerning the convergence of child abuse and spouse abuse would appear urgently to be needed. From so few published data, nearly all composed principally of families in poverty, little can be confidently generalized. Nor are any of the reports able sufficiently to probe the kinds of information about family life that would be most useful to practitioners: information concerning the child: medical and developmental history and present status, handicapping conditions, quality of relationships to care givers; information concerning the mother: medical and psychologic history and present status, family and social history including victimization, perceptions of people and institutions who may have tried to help her, history of the spousal relationship and the priority and meaning of the abused child; information concerning the abuser(s): medical and psychologic history and present status, family and social history including victimization, history of previous relationships with women and children, educational and employment status, use and abuse of psychoactive substances.
To address the problem of child and spouse abuse in clinical practice, we recommend the wider use of domestic violence protocols in adult and pediatric medical care. The systematic collection of data and engagement of protective interventions should include both children and their mothers.
Emergency room staff should be equipped with basic knowledge of the legal rights of domestic violence victims and their roles as legally mandated reporters of child abuse and neglect. In addition, they should be advised to document their professional interactions fully.
CONCLUSION
The striking overlap in this study between the victimization of children and their mothers suggests a need for a serious redefinition of both problems, focusing on violence in the family. Such a family level conceptualization may be difficult to bring about in today’s environment of specialized services and smaller human service budgets. But we believe that certain steps can be taken. Just as the staff of pediatric emergency rooms can take the time, pursue the training, and advocate for increased access to services to contend with the realities of mothers’ lives, so may the professionals who hear mothers’ complaints, from physicians, nurses, to battered women’s shelter workers, begin systematically to consider the safety and welfare of their offspring.
ACKNOWLEDGMENTS
This work was supported, in part, by grants from the National Center on Child Abuse and Neglect (90-CA1184) and the National Institute of Mental Health (5 T32 MHI8265).
We thank Robert M. Reece, MD, for helpful support in the formulation and conduct of the research.
REFERENCES
1. Kempe CH, Silverman F, Steele B, Droegemueller W, Silver H. The battered child syndrome. JAMA. 1962;181:17-24
2. Newberger EH, Bourne R. The medicalization and legalization of child abuse. Am J Orthopsychiatry. 1978;48:593-607
3. Straus MA, Gelles RJ, Steinmetz SK. Behind Closed Doors: Violence in the American Family. New York, NY: Anchor Press; 1980
4. Kaufman J, Zigler E. Do abused children become abusive parents?Am J Orthopsychiatry. 1980;57:186-192
5. Nelson KG. The innocent bystander: the child as unintended victim of domestic violence involving deadly weapons. Pediatrics. 1954;73:251-252
6. Jaffe P, Wolfe D, Wilson S, Zak L. Similarities in behavioral and social maladjustment among child victims and witnesses to family violence. Am J Orthopsychiatry. 1988;55:142-146
7. Jaffe P, Wolfe D, Wilson S, Zak L. Children of battered women: the relation of child behavior to family violence and matemal stress. J Consul Clin Psychol. 1985;53:657-665
8. Hershorn M. Rosenbaum A. Children of marital violence: a closer look at the unintended victims. Am JOrthopsychiatry. 1965;55:260-266
9. Elbow M. Children of violent marriages: the forgotten victims. Soc Casework. 1982;63:465-471
10. Stark E, Flitcraft A. Women and children at risk: a feminist perspective on child abuse. Int J Health Services. 1988; 18:97-118
11. Pascoe J, Hebbert V, Perl T, Loda F. Violence in North Carolina families referred to a child protection team. NC Med J. 1981;42:35-37
12. Fleiss JL. Statistical Methods for Rates and Proportions. New York, NY: John Wiley & Sons; 1981
13. Snyder JC, Newberger EH. Consensus and differences among hospital professionals in evaluating child maltreatment. Violence Victims 1986;1:125-139
14. Goldberg WG, Tomlanovich MC. Domestic violence victims in the emergency department: new findings. JAMA. 1984;251:3259-3254
Received for publication Jul 14, 1988; accepted Nov 4, 1988.