- Email: [email protected]
Identified spouse abuse as a risk factor for child abuse1
Pergamon
Child Abuse & Neglect, Vol. 24, No. 11, pp. 1375–1381, 2000 Copyright © 2000 Elsevier Science Ltd. Printed in the USA. All rights reserved 0145-2134/00/$–see front matter
PII S0145-2134(00)00192-7
IDENTIFIED SPOUSE ABUSE AS A RISK FACTOR FOR CHILD ABUSE PETER D. RUMM Division of Chronic Disease and Health Promotion, Department of Public Health, State of Wisconsin, Madison, WI, USA
PETER CUMMINGS Harborview Injury Prevention and Research Center; School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
MARGOT R. KRAUSS Epidemiology Department, Walter Reed Army Institute of Research, Washington, DC, USA
MICHELLE A. BELL School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
FREDERICK P. RIVARA Harborview Injury Prevention and Research Center; School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
ABSTRACT Context: There are limited data on the extent to which spouse abuse in a family is a risk factor for child abuse. Objective: To estimate the subsequent relative risk of child abuse in families with a report of spouse abuse compared with other families. Design: Cohort study. Setting: Analysis of a centralized US Army database Participants: Married couples with children with at least one spouse on active duty in the US Army during 1989 –95. Main Outcome Measures: The US Army Family Advocacy Program’s Central Database was used to identify child and spouse abuse. The exposure was an episode of identified spouse abuse and the main outcome was a substantiated episode of subsequent child abuse.
Doctors Rumm and Krauss were assigned previously to the Department of Preventive Medicine, Madigan Army Medical Center, Tacoma, WA. The opinions in this article are those of the authors and do not reflect the positions or policies of the US Army Medical Department, the US Army Family Advocacy Program, or the US Government. Received for publication May 25, 1999; final revision received September 13, 1999; accepted September 15, 1999. Requests for reprints should be sent to Frederick P. Rivara, MD, MPH, Harborview Injury Prevention and Research Center, 325 Ninth Avenue, Box 359960, Seattle, WA 98104-2499. 1375
1376
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara
Results: During the study period of an estimated 2,019,949 person years, 14,270 incident child abuse cases were substantiated. Families with an incident case of spouse abuse identified during the study period were twice as likely to have a substantiated report of child abuse compared with other military families, rate ratio, 2.0, (95% confidence interval [CI] 1.9 –2.1). Young parental age had the highest rate ratio, 4.9 (95% CI 4.5–5.3) in the subgroup analysis controlling for rank. Identified spouse abuse was associated with physical abuse of a child, rate ratio 2.4 (95% CI 2.2–2.5), and with sexual abuse of a child, rate ratio 1.5 (95% CI 1.3–1.7). Identified spouse abuse was not associated with child neglect or maltreatment, rate ratio, 1.0 (95% CI 0.9 –1.1). Conclusion: An identified episode of spouse abuse in a family appears to be associated with an increased risk of subsequent child abuse and serves as an independent risk factor. Therefore, care providers should consider the potential risk to children when dealing with spouse abuse. © 2000 Elsevier Science Ltd. Key Words—Spouse abuse, Child abuse, Domestic violence, Military.
INTRODUCTION IN 1994, THE National Center on Child Abuse and Neglect received 1,001,628 reports that were either considered to be substantiated or suspected according to the laws of the reporting state (Fingerhut & Warner, 1997). Child abuse has been associated with lower socioeconomic status, less parental education, large families, and young parental age (Straus, Gelles, & Steinmetz, 1980, 1990). Several authors have suggested that an additional risk factor for child abuse is belonging to a family with a history of spouse abuse (Gayford, 1975; Giles-Sims, 1985; McKibben, De Vos, & Newberger, 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Two national surveys have suggested that families with spousal violence are more likely to report behaviors consistent with child abuse (Ross, 1996; Straus et al., 1990). Moreover, several case series of battered woman have reported a family prevalence of child abuse as high as 54% (Gayford, 1975; Giles-Sims, 1985). While these studies have brought attention to the problem of child abuse in families with spouse abuse, they have relied solely on self-reported data or retrospective record reviews. The case series of battered women may have selected women with an atypically high prevalence of child abuse in their families and the studies lacked comparison groups. The temporal sequence of spouse and child abuse episodes was unclear in all studies (Gayford, 1975; Giles-Sims, 1985; McKibben et al., 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Since 1975, the US military has had a federally mandated initiative, known as the Family Advocacy Program, to reduce and track abuse. This branch of service has kept a joint database of identified spouse and child abuse episodes since the late 1980s. We analyzed these data to compare the incidence of substantiated child abuse in families with and without identified spouse abuse. METHODS Data Sources and Study Sample This cohort study analyzed spouse and child abuse data reported from the US Army Medical Command (MEDCOM) Central Registry, during 1989 –95. This study was approved by the US Army Medical Command, Madigan Army Medical Center, and by the University of Washington. Abuse Identification and Substantiation Reports of spouse or child abuse in the military are made usually by primary care physicians or emergency room personnel and does not differ in this respect from what happens in the civilian sector. Child abuse is reported to the local children’s protective service office as well as to the military authorities. Medical personnel are urged by their commanders to report all cases of suspected spouse or child abuse. By federal law, the medical commander at each military
Spouse abuse
1377
installation must appoint a team of medical and legal professionals, social workers, and counselors to investigate reports of spouse or child abuse (Mollerstrom, Milner, & Patchner, 1992). In a formal committee format, these professionals evaluate an episode based on the abuse history, physical examination findings, and the results of military and civilian criminal investigations, to decide whether it is a substantiated episode. Findings on all reported episodes are sent to the Central Registry Database in San Antonio, Texas. Episodes of identified spouse abuse and child abuse in this study met the following criteria: (1) At least one parent in the family was an active duty soldier; (2) the parents were married; (3) there was at least one child in the family; and (4) the episode was considered substantiated by medical evidence, a confession of guilt, or a legal determination of culpability. The cohort was divided into exposed and unexposed groups by the presence or absence of identified spouse abuse. The Outcome—Child Abuse The outcome of interest was the first substantiated episode of child abuse occurring in the family. These child abuse episodes were linked to the presence or absence of prior identified spouse abuse in the family using a coded family identifier (SPSS, 1995). Child abuse was subdivided into three categories: physical abuse, sexual abuse, and neglect (including a small number of cases coded for maltreatment by emotional abuse or deprivation) as determined by the local child protection team. Follow-up Time The time during which the cohort was followed (exposure time) was measured in family-years, the number of years each married family with children accumulated in the Army during the study period. Family-years were stratified for married Army families with children by the age of the soldier (⬍ 25 years, 25–30 years, and 31 years or more); the number of children in the family (1, 2–3, 4 or more); and military rank. Rank was selected as a proxy variable for income and educational status and was divided into three categories: (1) junior enlisted soldiers, (2) noncommissioned officers, and (3) all other officers. Family years of exposure were only known for the entire military population; the US Army did not agree to release individual level data on number of years of service. We assumed that families with identified spouse abuse had an average time in the Army equal to the average length of service for all Army families within the same stratum of rank; these periods were 4 years for junior enlisted soldiers, and 6 years for both senior noncommissioned officers and other officers. We also assumed that, on average, an identified episode of spouse abuse would occur mid-way through this period of service and therefore each family with identified spouse abuse was assigned two or three family-years of follow-up time according to the rank of the soldier. To determine how sensitive the main conclusion was to these assumptions, the data were re-analyzed by subtracting 1 year from the follow-up time in service for each family with an identified episode of spouse abuse. Statistical Analysis Incidence rates were calculated for child abuse episodes per 1000 family-years of exposure for the families who had identified spouse abuse, and all other Army families. Rates were estimated to compare Army families and for subgroups according to rank, age of the military soldier, and the number of children. Rate ratios and 95% confidence intervals (CI) were estimated to compare the identified spouse abusing families and all other Army families. Soldier’s age, number of children in the family, and rank were examined as potential confounders, effect modifiers, and independent risk factors using stratified methods (Centers for Disease Control, 1996; Mantel, & Haenszel, 1952; Rothman, 1986). Tests of heterogeneity were performed to determine the significance of any difference in rates between separate strata (Stata Corporation, 1997).
1378
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara Table 1. Characteristics of Families with Children in the U.S. Army, 1989 –95 Category
Families with Identified Spouse Abuse (%)a
Other Army Families (%)
50.9 44.1 5.0
24.1 59.9 16.0
27.5 40.9 31.6
15.9 28.0 56.0
43.8 51.2 4.9
39.5 54.1 6.3
Rank Junior enlisted Non-commissioned Officer Parental Age ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more a
Percentages are based on the number of family-years during which Army families were at risk for child abuse during 1989 –95.
RESULTS A total of 21,643 Army families with children had identified episodes of spouse abuse during the study period and were at risk for subsequent child abuse during an estimated 53,959 family-years. The rest of the Army averaged 280,856 families per year during the 7-year period for a total of 1,965,990 family-years. In general, families with identified spouse abuse had active duty parents who tended to be younger and of lower rank compared to other military families (Table 1). However, the families with identified spouse abuse and other Army families did not differ in the number of children they had. The child abuse rate among families with identified spouse abuse was 32.3 episodes/1000 familyyears (Table 2). The median time interval from the identified spouse abuse episode to a subsequent child
Table 2. Number and Rate of Child Abuse Episodes Among Army Families with Children During 1989 –95 Families with Identified Spouse Abuse
Rank of Soldier Junior enlisted Non-commissioned Officer Age of Soldier ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more Type of Abuse Physical Neglectb Sexual All types a b
Other Army Families
Number
Ratea
95% CI
Number
Rate
95% CI
931 756 58
42.2 26.4 17.6
39.6, 45.0 24.6, 28.3 13.4, 22.7
4876 7134 965
10.3 6.1 3.1
10.0, 10.6 6.0, 6.2 2.9, 3.3
758 685 302
55.7 30.5 16.9
52.0, 59.8 29.8, 31.4 15.1, 18.9
1986 6187 4802
6.4 11.3 4.4
6.1, 6.6 10.0, 11.5 4.3, 4.5
477 974 294
21.1 34.2 101.1
19.3, 23.1 29.4, 38.7 68.4, 131.1
2625 6732 3618
3.4 6.3 29.1
3.2, 3.5 6.1, 6.5 28.3, 30.1
1212 370 163
22.5 6.9 3.0
21.3, 23.8 6.2, 7.6 2.6, 3.5
6563 4963 1449
3.3 2.5 0.7
3.2, 3.4 2.4, 2.6 0.5, 0.9
1745
32.3
30.8, 33.9
12975
6.6
6.5, 6.7
Rates are expressed as the number of substantiated reports per 1000 family-years of exposure. Neglect includes maltreatment by emotional deprivation or emotional abuse.
Spouse abuse
1379
Table 3. Rate Ratios for Child Abuse among U.S. Army Families with and without Identified Spouse Abuse during 1989 –95 Category All types Rank of Soldierb Enlisted soldier Non-commissioned Officer Age of Soldierc ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more Type of Abuse Physical Neglect Sexual
Rate Ratiosa
95% CI
2.01
1.92, 2.11
2.08 2.92 0.30
1.95, 2.22 2.70, 3.14 0.24, 0.35
0.04
4.88 1.98 0.81
4.54, 5.26 1.84, 2.13 0.72, 0.91
⬍0.001
3.14 1.59 1.19
2.87, 3.43 1.50, 1.69 1.06, 1.34
⬍0.001
2.36 0.96 1.46
2.23, 2.50 0.87, 1.06 1.25, 1.69
NA
p*
* p value for a test of heterogeneity. Adjusted for the rank and age of the military parent except as indicated. Adjusted for the age of the military parent. c Adjusted for the rank of the military parent. a
b
abuse episode was 21.2 months, with 74.8% occurring within 2 years. The rate was higher among families with lower rank, younger parental age, and with more children in the family. Army families without an identified episode of spouse abuse had a lower rate of child abuse, 6.6 episodes/1000 family-years, (95% CI 6.5– 6.7). Among Army families, the most common type of child abuse was physical. Neglect was intermediate in incidence, and sexual abuse was least common. Families with identified spouse abuse were 4.9 times as likely as families without identified episode of spouse abuse to have a subsequent substantiated case of child abuse. After adjusting for rank and parental age, families with identified spouse abuse were twice as likely to have a report of child abuse as other military families: (adjusted rate ratio 2.0, 95% CI ⫽ 1.9 –2.1) (Table 3). Among families of enlisted men, those with identified spouse abuse were two to three times more likely to have subsequent reported child abuse compared with other families. Among families of officers, however, those with identified spouse abuse were only one-third as likely to have reported child abuse. Families with identified spouse abuse were at higher risk for subsequent child abuse, compared with other Army families, if the family member in the Army was either younger than age 25 years or 25–30 years, (adjusted rate ratios 4.9, [95% CI 4.5–5.3] and 2.0 [95% CI 1.8 –2.1] respectively). However, if the soldier was older than 30 years, families with identified spouse abuse were somewhat less likely to have later child abuse, adjusted rate ratio: 0.8, (95% CI .7–.9). For each grouping of family size by the number of children, the families with identified spouse abuse were at a greater risk for subsequent child abuse, and the rate ratio was greatest for families with only one child: (rate ratio 3.1, 95% CI 2.9 –3.4). Identified spouse abuse was associated with subsequent physical child abuse and sexual child abuse, but not with child neglect; adjusted rate ratios: 2.4, (95% CI 2.2–2.5); 1.5, (95% CI 1.3–1.7); and 1.0, (95% CI .9 –1.1), respectively. Sensitivity Analysis Changing the estimated time in military service for families with identified spouse abuse by subtracting 1 year from the estimated time in service increased the rate ratio to 3.5 (95% CI 3.4 –3.7).
1380
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara
COMMENT This is the first large reported cohort study to quantify the degree to which identified spousal violence among parents affects the risk for subsequent child abuse. The results suggest that identified spouse abuse is an important predictor of subsequent child abuse. In a large cohort of Army families, we found that the adjusted relative risk for child abuse was twice as great in families with a report of spouse abuse, compared with other families. This association was especially strong among the families of enlisted soldiers, the families of young soldiers, and families with one child. The association appeared to be strongest for physical child abuse, weaker for sexual abuse, and absent for child neglect. The present study used an estimate of the time in service for the families with identified spouse abuse to determine family-years at risk for subsequent child abuse. We hypothesized that families in which spouse abuse was identified might leave the military sooner than those without identified spouse abuse. Therefore, we conducted a sensitivity analysis, examining the effect on the rate ratios by subtracting 1 year from the estimated time in service for families with identified spouse abuse. The rate ratios were even larger, indicating that our main analysis is a conservative estimate of the effect of identified spouse abuse on the risk of later child abuse. It is possible that some families with identified spouse abuse may have had child abuse detected due to increased surveillance by the Family Advocacy Program. This bias, if present, could have caused us to over-estimate the true association between spouse abuse and child abuse. However, the military strongly urges reporting of child abuse whenever it is suspected. The finding that a history of identified spouse abuse among families of officers appeared to protect against subsequent child abuse seems doubtful. We postulate that many soldiers who committed spouse abuse which was then identified as such were forced to cut short their military careers and that this effect was strongest among officers, who are expected to be role models for other soldiers. If this was so, then we over-estimated the family-years of exposure for officer families with identified spouse abuse, biasing the estimate of association toward zero. We lacked the follow-up data on families after they left the Army. It is possible that families might leave the military sooner or later depending on their likelihood of committing or being reported for committing child abuse. If the outcome was related to the time when families left the military, this could have biased our risk ratio estimates; the direction of any bias is unknown. It is also possible that officers were treated differently in terms of reporting spouse or child abuse, although we have no way of knowing this. As noted above, once they were identified as being abusers, officers may have been more likely to leave the service than enlisted personnel. Although we were not able to determine the rates of child abuse by the gender of the military parent, the US Army is predominantly male. In 1998, there were 483,880 active duty soldiers in the military, of whom 9% were female. Thus, the data were predominantly about males. Finally, as in any study of abuse, some spouse and child abuse episodes undoubtedly went undetected, although if anything this is likely to be less in the military than in the civilian sector (Centers for Disease Control, 1996; Straus et al., 1980, 1990). We examined identified spouse abuse as a risk factor for later child abuse. Regardless of how many episodes of spouse abuse were unknown to the military, our analysis suggests that children in families where spouse abuse was discovered were at increased risk of abuse, compared with all other children. Our study has several strengths. The study involved a very large cohort for which data on both types of abuse were collected and recorded prospectively in a systematic manner. In addition, unlike many civilian studies, subjects could be tracked across geographic boundaries as long as they remained in the cohort of Army families. Our cohort study is in general agreement with findings reported by previous authors. McKibben conducted a paired case-control study of 64 families, which was reported in 1989 (McKibben et al., 1989). Professionals who reviewed medical charts found that 59% of the mothers of abused
Spouse abuse
1381
children were themselves victims of violence compared to 12.5% of controls. Straus and colleagues used data from a large national survey in the US and reported that families with episodes of spouse abuse were more likely to report parental behaviors consistent with child abuse (Straus et al., 1980). Previous studies on family abuse did not examine the categories of child abuse (Gayford, 1975; Giles-Sims, 1985; McKibben et al., 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Our study suggests that children in a family with identified spouse abuse are a greater risk for subsequent sexual abuse. This is a finding that has not been specifically discussed in the previous literature, although a few authors have suggested that marital problems and family turbulence increase the risk for sexual abuse in children (Gelles, & Stares, 1989; Kempe, 1962; Van Hasslet, Morrison, Bellack, & Hersen, 1988). We found that children in Army families with identified spouse abuse were at increased risk for subsequent child abuse. Care providers should consider the risk to children when dealing with families with spouse abuse. REFERENCES Centers for Disease Control. (1996). EpiInfo statistical program (Ver. 6.04b). Atlanta, GA: The Centers for Disease Control. Fingerhut, L. A., & Warner, M. (1997). Health United States 1996 –97 and injury chartbook (48 pp.). Hyattsville, MD: National Center for Health Statistics. Gayford, J. J. (1975). Wife battering: A preliminary survey of 100 episodes. Medical Practice, 1, 194 –197. Gelles, R. J., & Stares, M. A. (1989). Intimate violence. New York: Simon and Schuster. Giles-Sims, J. (1985). A longitudinal study of battered children of battered wives. Family Relationships, 34, 205–210. Kempe, C. H. (1962). The battered child syndrome. Journal of the American Medical Association, 181, 17–24. Mantel, N., & Haenszel, W. (1952). Statistical aspects of the analysis of the data from retrospective studies. Journal of the National Cancer Institute, 22, 719 –748. McKibben, L., De Vos, E., & Newberger E. H. (1989). Victimization of mothers of abused children: A controlled study. Pediatrics, 84, 531–535. Mollerstrom, W. W., Milner J. S., & Patchner, M. A. (1992). Family violence in the Air Force: A look at offenders and the role of the family advocacy program. Military Medicine, 157, 371–375. Ross, S. M. (1996). Risk of physical abuse to children of spouse abusing parents. Child Abuse & Neglect, 7, 589 –598. Rothman, K. J. (1986). Modern epidemiology. Boston: Little Brown and Company. SPSS. (1995). SPSS statistical program (Ver. 6.1.). New York: SPSS Corporation. Stark, E., & Flitcraft, A. H. (1988). Woman and children at risk: A feminist perspective on child abuse. International Journal of Health Services, 10, 97–117. Stata Corporation. (1997). Stata statistical software (Ver. 5.0). College Station, TX: Stata Corporation. Straus, M. A., Gelles, R. J., & Steinmetz, S. K. (1980). Behind closed doors. Garden City, NY: Anchor Press. Straus, M. A., Gelles, R. J., & Steinmetz, S. K. (1990). Physical violence in the American family. Beverly Hills, CA: Sage Publications. Van Hasslet, V. B., Morrison, R. L., Bellack, A. S., & Hersen, M. (1988). Handbook of family violence. New York: Plenum Press.
Child Abuse & Neglect, Vol. 24, No. 11, pp. 1375–1381, 2000 Copyright © 2000 Elsevier Science Ltd. Printed in the USA. All rights reserved 0145-2134/00/$–see front matter
PII S0145-2134(00)00192-7
IDENTIFIED SPOUSE ABUSE AS A RISK FACTOR FOR CHILD ABUSE PETER D. RUMM Division of Chronic Disease and Health Promotion, Department of Public Health, State of Wisconsin, Madison, WI, USA
PETER CUMMINGS Harborview Injury Prevention and Research Center; School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
MARGOT R. KRAUSS Epidemiology Department, Walter Reed Army Institute of Research, Washington, DC, USA
MICHELLE A. BELL School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
FREDERICK P. RIVARA Harborview Injury Prevention and Research Center; School of Public Health and Community Medicine, University of Washington, Seattle, WA, USA
ABSTRACT Context: There are limited data on the extent to which spouse abuse in a family is a risk factor for child abuse. Objective: To estimate the subsequent relative risk of child abuse in families with a report of spouse abuse compared with other families. Design: Cohort study. Setting: Analysis of a centralized US Army database Participants: Married couples with children with at least one spouse on active duty in the US Army during 1989 –95. Main Outcome Measures: The US Army Family Advocacy Program’s Central Database was used to identify child and spouse abuse. The exposure was an episode of identified spouse abuse and the main outcome was a substantiated episode of subsequent child abuse.
Doctors Rumm and Krauss were assigned previously to the Department of Preventive Medicine, Madigan Army Medical Center, Tacoma, WA. The opinions in this article are those of the authors and do not reflect the positions or policies of the US Army Medical Department, the US Army Family Advocacy Program, or the US Government. Received for publication May 25, 1999; final revision received September 13, 1999; accepted September 15, 1999. Requests for reprints should be sent to Frederick P. Rivara, MD, MPH, Harborview Injury Prevention and Research Center, 325 Ninth Avenue, Box 359960, Seattle, WA 98104-2499. 1375
1376
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara
Results: During the study period of an estimated 2,019,949 person years, 14,270 incident child abuse cases were substantiated. Families with an incident case of spouse abuse identified during the study period were twice as likely to have a substantiated report of child abuse compared with other military families, rate ratio, 2.0, (95% confidence interval [CI] 1.9 –2.1). Young parental age had the highest rate ratio, 4.9 (95% CI 4.5–5.3) in the subgroup analysis controlling for rank. Identified spouse abuse was associated with physical abuse of a child, rate ratio 2.4 (95% CI 2.2–2.5), and with sexual abuse of a child, rate ratio 1.5 (95% CI 1.3–1.7). Identified spouse abuse was not associated with child neglect or maltreatment, rate ratio, 1.0 (95% CI 0.9 –1.1). Conclusion: An identified episode of spouse abuse in a family appears to be associated with an increased risk of subsequent child abuse and serves as an independent risk factor. Therefore, care providers should consider the potential risk to children when dealing with spouse abuse. © 2000 Elsevier Science Ltd. Key Words—Spouse abuse, Child abuse, Domestic violence, Military.
INTRODUCTION IN 1994, THE National Center on Child Abuse and Neglect received 1,001,628 reports that were either considered to be substantiated or suspected according to the laws of the reporting state (Fingerhut & Warner, 1997). Child abuse has been associated with lower socioeconomic status, less parental education, large families, and young parental age (Straus, Gelles, & Steinmetz, 1980, 1990). Several authors have suggested that an additional risk factor for child abuse is belonging to a family with a history of spouse abuse (Gayford, 1975; Giles-Sims, 1985; McKibben, De Vos, & Newberger, 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Two national surveys have suggested that families with spousal violence are more likely to report behaviors consistent with child abuse (Ross, 1996; Straus et al., 1990). Moreover, several case series of battered woman have reported a family prevalence of child abuse as high as 54% (Gayford, 1975; Giles-Sims, 1985). While these studies have brought attention to the problem of child abuse in families with spouse abuse, they have relied solely on self-reported data or retrospective record reviews. The case series of battered women may have selected women with an atypically high prevalence of child abuse in their families and the studies lacked comparison groups. The temporal sequence of spouse and child abuse episodes was unclear in all studies (Gayford, 1975; Giles-Sims, 1985; McKibben et al., 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Since 1975, the US military has had a federally mandated initiative, known as the Family Advocacy Program, to reduce and track abuse. This branch of service has kept a joint database of identified spouse and child abuse episodes since the late 1980s. We analyzed these data to compare the incidence of substantiated child abuse in families with and without identified spouse abuse. METHODS Data Sources and Study Sample This cohort study analyzed spouse and child abuse data reported from the US Army Medical Command (MEDCOM) Central Registry, during 1989 –95. This study was approved by the US Army Medical Command, Madigan Army Medical Center, and by the University of Washington. Abuse Identification and Substantiation Reports of spouse or child abuse in the military are made usually by primary care physicians or emergency room personnel and does not differ in this respect from what happens in the civilian sector. Child abuse is reported to the local children’s protective service office as well as to the military authorities. Medical personnel are urged by their commanders to report all cases of suspected spouse or child abuse. By federal law, the medical commander at each military
Spouse abuse
1377
installation must appoint a team of medical and legal professionals, social workers, and counselors to investigate reports of spouse or child abuse (Mollerstrom, Milner, & Patchner, 1992). In a formal committee format, these professionals evaluate an episode based on the abuse history, physical examination findings, and the results of military and civilian criminal investigations, to decide whether it is a substantiated episode. Findings on all reported episodes are sent to the Central Registry Database in San Antonio, Texas. Episodes of identified spouse abuse and child abuse in this study met the following criteria: (1) At least one parent in the family was an active duty soldier; (2) the parents were married; (3) there was at least one child in the family; and (4) the episode was considered substantiated by medical evidence, a confession of guilt, or a legal determination of culpability. The cohort was divided into exposed and unexposed groups by the presence or absence of identified spouse abuse. The Outcome—Child Abuse The outcome of interest was the first substantiated episode of child abuse occurring in the family. These child abuse episodes were linked to the presence or absence of prior identified spouse abuse in the family using a coded family identifier (SPSS, 1995). Child abuse was subdivided into three categories: physical abuse, sexual abuse, and neglect (including a small number of cases coded for maltreatment by emotional abuse or deprivation) as determined by the local child protection team. Follow-up Time The time during which the cohort was followed (exposure time) was measured in family-years, the number of years each married family with children accumulated in the Army during the study period. Family-years were stratified for married Army families with children by the age of the soldier (⬍ 25 years, 25–30 years, and 31 years or more); the number of children in the family (1, 2–3, 4 or more); and military rank. Rank was selected as a proxy variable for income and educational status and was divided into three categories: (1) junior enlisted soldiers, (2) noncommissioned officers, and (3) all other officers. Family years of exposure were only known for the entire military population; the US Army did not agree to release individual level data on number of years of service. We assumed that families with identified spouse abuse had an average time in the Army equal to the average length of service for all Army families within the same stratum of rank; these periods were 4 years for junior enlisted soldiers, and 6 years for both senior noncommissioned officers and other officers. We also assumed that, on average, an identified episode of spouse abuse would occur mid-way through this period of service and therefore each family with identified spouse abuse was assigned two or three family-years of follow-up time according to the rank of the soldier. To determine how sensitive the main conclusion was to these assumptions, the data were re-analyzed by subtracting 1 year from the follow-up time in service for each family with an identified episode of spouse abuse. Statistical Analysis Incidence rates were calculated for child abuse episodes per 1000 family-years of exposure for the families who had identified spouse abuse, and all other Army families. Rates were estimated to compare Army families and for subgroups according to rank, age of the military soldier, and the number of children. Rate ratios and 95% confidence intervals (CI) were estimated to compare the identified spouse abusing families and all other Army families. Soldier’s age, number of children in the family, and rank were examined as potential confounders, effect modifiers, and independent risk factors using stratified methods (Centers for Disease Control, 1996; Mantel, & Haenszel, 1952; Rothman, 1986). Tests of heterogeneity were performed to determine the significance of any difference in rates between separate strata (Stata Corporation, 1997).
1378
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara Table 1. Characteristics of Families with Children in the U.S. Army, 1989 –95 Category
Families with Identified Spouse Abuse (%)a
Other Army Families (%)
50.9 44.1 5.0
24.1 59.9 16.0
27.5 40.9 31.6
15.9 28.0 56.0
43.8 51.2 4.9
39.5 54.1 6.3
Rank Junior enlisted Non-commissioned Officer Parental Age ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more a
Percentages are based on the number of family-years during which Army families were at risk for child abuse during 1989 –95.
RESULTS A total of 21,643 Army families with children had identified episodes of spouse abuse during the study period and were at risk for subsequent child abuse during an estimated 53,959 family-years. The rest of the Army averaged 280,856 families per year during the 7-year period for a total of 1,965,990 family-years. In general, families with identified spouse abuse had active duty parents who tended to be younger and of lower rank compared to other military families (Table 1). However, the families with identified spouse abuse and other Army families did not differ in the number of children they had. The child abuse rate among families with identified spouse abuse was 32.3 episodes/1000 familyyears (Table 2). The median time interval from the identified spouse abuse episode to a subsequent child
Table 2. Number and Rate of Child Abuse Episodes Among Army Families with Children During 1989 –95 Families with Identified Spouse Abuse
Rank of Soldier Junior enlisted Non-commissioned Officer Age of Soldier ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more Type of Abuse Physical Neglectb Sexual All types a b
Other Army Families
Number
Ratea
95% CI
Number
Rate
95% CI
931 756 58
42.2 26.4 17.6
39.6, 45.0 24.6, 28.3 13.4, 22.7
4876 7134 965
10.3 6.1 3.1
10.0, 10.6 6.0, 6.2 2.9, 3.3
758 685 302
55.7 30.5 16.9
52.0, 59.8 29.8, 31.4 15.1, 18.9
1986 6187 4802
6.4 11.3 4.4
6.1, 6.6 10.0, 11.5 4.3, 4.5
477 974 294
21.1 34.2 101.1
19.3, 23.1 29.4, 38.7 68.4, 131.1
2625 6732 3618
3.4 6.3 29.1
3.2, 3.5 6.1, 6.5 28.3, 30.1
1212 370 163
22.5 6.9 3.0
21.3, 23.8 6.2, 7.6 2.6, 3.5
6563 4963 1449
3.3 2.5 0.7
3.2, 3.4 2.4, 2.6 0.5, 0.9
1745
32.3
30.8, 33.9
12975
6.6
6.5, 6.7
Rates are expressed as the number of substantiated reports per 1000 family-years of exposure. Neglect includes maltreatment by emotional deprivation or emotional abuse.
Spouse abuse
1379
Table 3. Rate Ratios for Child Abuse among U.S. Army Families with and without Identified Spouse Abuse during 1989 –95 Category All types Rank of Soldierb Enlisted soldier Non-commissioned Officer Age of Soldierc ⬍25 Years 25–30 Years 31 OR ⬎ Years Number of Children One Two–three Four or more Type of Abuse Physical Neglect Sexual
Rate Ratiosa
95% CI
2.01
1.92, 2.11
2.08 2.92 0.30
1.95, 2.22 2.70, 3.14 0.24, 0.35
0.04
4.88 1.98 0.81
4.54, 5.26 1.84, 2.13 0.72, 0.91
⬍0.001
3.14 1.59 1.19
2.87, 3.43 1.50, 1.69 1.06, 1.34
⬍0.001
2.36 0.96 1.46
2.23, 2.50 0.87, 1.06 1.25, 1.69
NA
p*
* p value for a test of heterogeneity. Adjusted for the rank and age of the military parent except as indicated. Adjusted for the age of the military parent. c Adjusted for the rank of the military parent. a
b
abuse episode was 21.2 months, with 74.8% occurring within 2 years. The rate was higher among families with lower rank, younger parental age, and with more children in the family. Army families without an identified episode of spouse abuse had a lower rate of child abuse, 6.6 episodes/1000 family-years, (95% CI 6.5– 6.7). Among Army families, the most common type of child abuse was physical. Neglect was intermediate in incidence, and sexual abuse was least common. Families with identified spouse abuse were 4.9 times as likely as families without identified episode of spouse abuse to have a subsequent substantiated case of child abuse. After adjusting for rank and parental age, families with identified spouse abuse were twice as likely to have a report of child abuse as other military families: (adjusted rate ratio 2.0, 95% CI ⫽ 1.9 –2.1) (Table 3). Among families of enlisted men, those with identified spouse abuse were two to three times more likely to have subsequent reported child abuse compared with other families. Among families of officers, however, those with identified spouse abuse were only one-third as likely to have reported child abuse. Families with identified spouse abuse were at higher risk for subsequent child abuse, compared with other Army families, if the family member in the Army was either younger than age 25 years or 25–30 years, (adjusted rate ratios 4.9, [95% CI 4.5–5.3] and 2.0 [95% CI 1.8 –2.1] respectively). However, if the soldier was older than 30 years, families with identified spouse abuse were somewhat less likely to have later child abuse, adjusted rate ratio: 0.8, (95% CI .7–.9). For each grouping of family size by the number of children, the families with identified spouse abuse were at a greater risk for subsequent child abuse, and the rate ratio was greatest for families with only one child: (rate ratio 3.1, 95% CI 2.9 –3.4). Identified spouse abuse was associated with subsequent physical child abuse and sexual child abuse, but not with child neglect; adjusted rate ratios: 2.4, (95% CI 2.2–2.5); 1.5, (95% CI 1.3–1.7); and 1.0, (95% CI .9 –1.1), respectively. Sensitivity Analysis Changing the estimated time in military service for families with identified spouse abuse by subtracting 1 year from the estimated time in service increased the rate ratio to 3.5 (95% CI 3.4 –3.7).
1380
P. D. Rumm, P. Cummings, M. R. Krauss, M. A. Bell, and F. P. Rivara
COMMENT This is the first large reported cohort study to quantify the degree to which identified spousal violence among parents affects the risk for subsequent child abuse. The results suggest that identified spouse abuse is an important predictor of subsequent child abuse. In a large cohort of Army families, we found that the adjusted relative risk for child abuse was twice as great in families with a report of spouse abuse, compared with other families. This association was especially strong among the families of enlisted soldiers, the families of young soldiers, and families with one child. The association appeared to be strongest for physical child abuse, weaker for sexual abuse, and absent for child neglect. The present study used an estimate of the time in service for the families with identified spouse abuse to determine family-years at risk for subsequent child abuse. We hypothesized that families in which spouse abuse was identified might leave the military sooner than those without identified spouse abuse. Therefore, we conducted a sensitivity analysis, examining the effect on the rate ratios by subtracting 1 year from the estimated time in service for families with identified spouse abuse. The rate ratios were even larger, indicating that our main analysis is a conservative estimate of the effect of identified spouse abuse on the risk of later child abuse. It is possible that some families with identified spouse abuse may have had child abuse detected due to increased surveillance by the Family Advocacy Program. This bias, if present, could have caused us to over-estimate the true association between spouse abuse and child abuse. However, the military strongly urges reporting of child abuse whenever it is suspected. The finding that a history of identified spouse abuse among families of officers appeared to protect against subsequent child abuse seems doubtful. We postulate that many soldiers who committed spouse abuse which was then identified as such were forced to cut short their military careers and that this effect was strongest among officers, who are expected to be role models for other soldiers. If this was so, then we over-estimated the family-years of exposure for officer families with identified spouse abuse, biasing the estimate of association toward zero. We lacked the follow-up data on families after they left the Army. It is possible that families might leave the military sooner or later depending on their likelihood of committing or being reported for committing child abuse. If the outcome was related to the time when families left the military, this could have biased our risk ratio estimates; the direction of any bias is unknown. It is also possible that officers were treated differently in terms of reporting spouse or child abuse, although we have no way of knowing this. As noted above, once they were identified as being abusers, officers may have been more likely to leave the service than enlisted personnel. Although we were not able to determine the rates of child abuse by the gender of the military parent, the US Army is predominantly male. In 1998, there were 483,880 active duty soldiers in the military, of whom 9% were female. Thus, the data were predominantly about males. Finally, as in any study of abuse, some spouse and child abuse episodes undoubtedly went undetected, although if anything this is likely to be less in the military than in the civilian sector (Centers for Disease Control, 1996; Straus et al., 1980, 1990). We examined identified spouse abuse as a risk factor for later child abuse. Regardless of how many episodes of spouse abuse were unknown to the military, our analysis suggests that children in families where spouse abuse was discovered were at increased risk of abuse, compared with all other children. Our study has several strengths. The study involved a very large cohort for which data on both types of abuse were collected and recorded prospectively in a systematic manner. In addition, unlike many civilian studies, subjects could be tracked across geographic boundaries as long as they remained in the cohort of Army families. Our cohort study is in general agreement with findings reported by previous authors. McKibben conducted a paired case-control study of 64 families, which was reported in 1989 (McKibben et al., 1989). Professionals who reviewed medical charts found that 59% of the mothers of abused
Spouse abuse
1381
children were themselves victims of violence compared to 12.5% of controls. Straus and colleagues used data from a large national survey in the US and reported that families with episodes of spouse abuse were more likely to report parental behaviors consistent with child abuse (Straus et al., 1980). Previous studies on family abuse did not examine the categories of child abuse (Gayford, 1975; Giles-Sims, 1985; McKibben et al., 1989; Ross, 1996; Stark & Flitcraft, 1988; Straus et al., 1980, 1990). Our study suggests that children in a family with identified spouse abuse are a greater risk for subsequent sexual abuse. This is a finding that has not been specifically discussed in the previous literature, although a few authors have suggested that marital problems and family turbulence increase the risk for sexual abuse in children (Gelles, & Stares, 1989; Kempe, 1962; Van Hasslet, Morrison, Bellack, & Hersen, 1988). We found that children in Army families with identified spouse abuse were at increased risk for subsequent child abuse. Care providers should consider the risk to children when dealing with families with spouse abuse. REFERENCES Centers for Disease Control. (1996). EpiInfo statistical program (Ver. 6.04b). Atlanta, GA: The Centers for Disease Control. Fingerhut, L. A., & Warner, M. (1997). Health United States 1996 –97 and injury chartbook (48 pp.). Hyattsville, MD: National Center for Health Statistics. Gayford, J. J. (1975). Wife battering: A preliminary survey of 100 episodes. Medical Practice, 1, 194 –197. Gelles, R. J., & Stares, M. A. (1989). Intimate violence. New York: Simon and Schuster. Giles-Sims, J. (1985). A longitudinal study of battered children of battered wives. Family Relationships, 34, 205–210. Kempe, C. H. (1962). The battered child syndrome. Journal of the American Medical Association, 181, 17–24. Mantel, N., & Haenszel, W. (1952). Statistical aspects of the analysis of the data from retrospective studies. Journal of the National Cancer Institute, 22, 719 –748. McKibben, L., De Vos, E., & Newberger E. H. (1989). Victimization of mothers of abused children: A controlled study. Pediatrics, 84, 531–535. Mollerstrom, W. W., Milner J. S., & Patchner, M. A. (1992). Family violence in the Air Force: A look at offenders and the role of the family advocacy program. Military Medicine, 157, 371–375. Ross, S. M. (1996). Risk of physical abuse to children of spouse abusing parents. Child Abuse & Neglect, 7, 589 –598. Rothman, K. J. (1986). Modern epidemiology. Boston: Little Brown and Company. SPSS. (1995). SPSS statistical program (Ver. 6.1.). New York: SPSS Corporation. Stark, E., & Flitcraft, A. H. (1988). Woman and children at risk: A feminist perspective on child abuse. International Journal of Health Services, 10, 97–117. Stata Corporation. (1997). Stata statistical software (Ver. 5.0). College Station, TX: Stata Corporation. Straus, M. A., Gelles, R. J., & Steinmetz, S. K. (1980). Behind closed doors. Garden City, NY: Anchor Press. Straus, M. A., Gelles, R. J., & Steinmetz, S. K. (1990). Physical violence in the American family. Beverly Hills, CA: Sage Publications. Van Hasslet, V. B., Morrison, R. L., Bellack, A. S., & Hersen, M. (1988). Handbook of family violence. New York: Plenum Press.