Child Abuse & Neglect, Vol. 20, No. 3, pp. 191-203, 1996 Copyright © 1996 Elsevier Science Ltd Printed in the USA. All rights reserved 0145-2134/96 $15.00 + .00
Pergamon
SSDI 0145-2134(95)00144-1
ONSET OF PHYSICAL ABUSE AND NEGLECT: PSYCHIATRIC, SUBSTANCE ABUSE, AND SOCIAL RISK FACTORS FROM PROSPECTIVE COMMUNITY DATA MARK CHAFFIN Department of Pediatrics, University of Arkansas for Medical Sciences, Little Rock, AR, USA
KELLY
KELLEHER
Departments of Psychiatry and Pediatrics, University of Pittsburgh, Pittsburgh, PA, USA
JAN HOLLENBERG Department of Psychiatry, University of Arkansas for Medical Sciences, Little Rock, AR, USA
Abstract---Studies of psychiatric and social risk factors for child maltreatment have been limited by retrospective methodologies and reliance on officially reported or identified samples. Using data from both Waves I and lI of the National Institute for Mental Health's Epidemioiogic Catchment Area survey, 7,103 parents from a probabilistic community sample who did not self-report physical abuse or neglect of their children at Wave I were followed to determine the risk factors associated with the onset of self-reported physical abuse or neglect identified at Wave II. Social factors considered included age, socioeconomic status, social support, education, household size, and gender. In addition, several psychiatric disorders, including substance abuse disorders and depression were examined. Risk models were developed using hierarchical logistic regression. Physical abuse and neglect were found to have distinct sets of risk factors, with minimal overlap between the groups. Social and demographic variables were found to be limited predictors of maltreatment, while substance abuse disorders were strongly associated with the onset of both abuse and neglect (relative risks = 2.90 and 3.24 respectively). Depression was found to be a strong risk factor for physical abuse (relative risk = 3.45). Implications of the findings are discussed in terms of major causal models of maltreatment.
Key Words--Physical abuse, Abuse, Neglect, Substance abuse, Depression, Risk factors.
INTRODUCTION EARLY RISK MODELS for child physical abuse and neglect focused on psychiatric or mental health conditions among abusive parents, emphasizing a range of disorders and personality characteristics (Steele, 1987). Subsequently, both risk and causal models have increasingly emphasized sociological factors that combine through multiple pathways eventuating in physical abuse or neglect. Failer and Ziefert (1981) divided all of these factors into three categories: This paper was supported in part by a grant from the National Center on Child Abuse and Neglect, and funds from the Stannton Farm Foundation. Received for publication May l, 1995; final revision received August 14, 1995; accepted August 16, 1995. Reprint requests should be addressed to Mark Chaffin, Ph.D., UAMS Department of Pediatrics, 1120 Marshall St., Suite C-401, Little Rock, AR 72202. 191
192
M. Chaffin,K. Kelleher,and J. Hollenberg
individual parent factors (such as depressive or substance use disorders), family related factors (such as scapegoating, child temperament, or single parenthood), and environmental factors (such as social stress, poverty, or cultural beliefs). Other models, not considered in detail here, have emphasized more proximal and intermediary variables such as specific interactional patterns, perceptions, cognitions, belief systems, attributions, and complex transactional pathways that may lead to abuse (Azar, 1991; Crittenden, 1993; Milner & Chilamkurti, 1991), or cultural factors such as approval of violence (Strauss, Gelles, & Steinmetz, 1980). Belsky (1993) has noted that there are likely multiple pathways to abuse and neglect, involving multiple factors at different levels of analysis within a broad ecological conception of parentchild interactions, with probably no single group of necessary or sufficient etiologic conditions. However, for purposes of identifying parental risk factors, which is the level of analysis with which this study is concerned, the two main categories of predictors have remained social (e.g., poverty, isolation, and stressors) and psychological or psychiatric. Some sociological theorists have proposed that individual mental health variables play, at best, an artifactual or peripheral role in the etiology of physical abuse or neglect. Instead, they emphasize social variables, most notably socioeconomic status (SES), as central (Gelles, 1993; Gelles & Cornell, 1990). This position is supported by findings that there is no distinct set of personality characteristics or profile that has been found to distinguish abusers from nonabusers (Wolfe, 1985). In contrast, others have noted that abuse cuts across a wide range of social and economic classes, and in some limited studies, there appears to be a dose-effect relationship between parental psychopathology and the severity of violence against children (O'Leary, 1993).
Studies of Social Demographics: SES, Parental Age, Gender, Ethnicity, and Family Size Two National Incidence Studies (MS-I, MS-H) of abuse and the social characteristics of abusive families were commissioned in 1980 and 1988 by the National Center on Child Abuse and Neglect. NIS-II surveyed community professionals and child protection agencies in a national probability sample of 29 counties. Low family income (< $15,000/year) and larger family size (> four children) were identified as risk factors for physical abuse and neglect (MS-U, 1988). However, only identified cases and no controls were used. Other studies have noted a high number of female parents among identified cases of physical abuse or neglect. However, when the design controls for differential at-risk status across gender (e.g., by comparing male vs. female single parents, or male vs. female baby-sitters), it appears that males are more at-risk for physically abusing children (Margolin, 1992). General population studies using well-operationalized definitions of violence toward children have been conducted under the First and Second National Family Violence Surveys (Gelles, 1980; Wolfner & Gelles, 1993). The second study examined a national probability sample of 3,232 households with children using the Conflict Tactics Scale (Straus, 1979). The findings supported the contribution of a number of social variables to violence against children, including parental age (younger), economics (poorer), ethnicity (African American), and number and ages of children. Women were at greater risk for minor violence towards children (e.g., spanking, grabbing), but there was no gender difference for abusive violence (e.g., beating up). However, because of the cross-sectional methodology, it was not clear whether the risk profiles identified would be predictive of the onset of abuse. The survey also made cursory inquiry into drug and alcohol use, but did not systematically assess diagnosable substance use disorders or other psychiatric diagnoses.
Studies of Mental Health Risk Factors: Depressive and Substance Abuse Disorders Studies examining mental health characteristics of abusive or neglectful parents have noted the prevalence of two main disorders: depression and substance abuse. Other Axis I disorders have generally not been found to be associated with risk for violent behaviors unless they are also complicated by a substance abuse disorder (Swanson, Holzer, Ganju, & Jono, 1990).
Onset of abuse
193
Determination of relative risk, however, has been compromised by the use of cross-sectional and retrospective methodologies in most of these studies. For example, while identified physically abusive and neglectful parents have been found to exhibit greater levels of depression relative to controls (Culp, Culp, Soulis, & Letts, 1989), it is not clear what role identification bias or retrospective response biases might play in these findings. In addition, it is possible that either becoming or being identified as a maltreating parent may predispose to depression rather than vice versa. Nonetheless, odds ratios among severely depressed mothers compared to those without depression had been estimated to be 3.95 for abuse and 1.87 for neglect (Zuravin, 1988). Similarly, substance abuse has been strongly implicated in child maltreatment, both from studies examining rates of substance abuse among identified maltreating parents as well as studies of child maltreatment among identified substance abusers. Murphy, Jellinek, Quinn, Smith, Poitrast, and Goshko (1991) examined substance abuse in a sample of 206 cases of serious child abuse or neglect before a Boston juvenile court. In 43% of the cases, at least one of the parents had a documented problem with either alcohol or drugs. Black and Myer (1980) interviewed 200 alcohol or opiate addicted Boston families at an addiction treatment center, inquiring about child care and child injuries. Forty-one percent (41%) of the children met criteria for serious neglect, abuse, or both. All were felt to be at least mildly neglected. General population case/control studies using well-operationalized definitions of substance abuse and other psychiatric disorders have been conducted using data from the National Institute of Mental Health's Epidemiologic Catchment Area (ECA) Wave I data, and have supported the association between some psychiatric disorders and interpersonal violence in general (Dinwiddie & Bucholz, 1993; Kelleher, Chaffin, Hollenberg, & Fischer, 1994; Swanson, Holzer, Ganju, & Jono, 1990). Kelleher, Chaffin, Hollenberg, and Fischer 0994), using ECA data from a communitybased sample of over 11,000 parents, compared those self-reporting physical abuse or neglect to separate control groups matched on age, race, site, gender, and SES. Maltreating parents were found to have lifetime prevalence rates for a DSM-IH substance abuse disorder of 43% for parents who physically abused their children and 51% for parents who neglected their children. The association remained robust after controlling for measures of social support, depression, and antisocial personality. Nonetheless, because the study used retrospective lifetime prevalence estimates for the disorders in question, it was not possible to establish the temporal relationship of substance abuse disorders and child maltreatment or determine relative risks. Two other recent community-based sample retrospective studies have found associations between depression or substance abuse and violence, including violence toward children (Dinwiddie & Bucholz, 1993; Swanson ct al., 1990). Dinwiddie and Bucholz (1993) also found an association between anxiety disorders, such as panic disorder or obsessive compulsive disorder (OCD), and physical abuse, especially in alcohol involved populations. Swanson and colleagues (1990) found that associations between many psychiatric disorders, especially anxiety disorders, and interpersonal violence appear to be mediated to some extent by comorbid alcohol and drug abuse disorders. In other words, the risk of violence associated with many psychiatric disorders (including major mental illnesses) is not markedly increased unless a substance abuse disorder is also present. Along with psychiatric disorders, a number of social variables (male gender, young age, lower SES) were found to be highly associated with violence in general, although no analysis of child abuse in particular was undertaken. Neither study examined child neglect.
Summary of Methodological Issues in Assessing Risk Factors Much of the empirical research into etiologic factors, both sociological and psychological or psychiatric, has been conducted by retrospectively examining the characteristics of officially reported or identified cases. These data are vulnerable to a number of biases. First, official identification and referral patterns in child maltreatment may be biased. For example, low-
194
M. Chaffin,K. Kelleher,and J. Hollenberg
income African American children seen in public hospitals are more likely to be both correctly and incorrectly reported for abuse relative to middle-class White children (Newberger, Reed, Daniel, Hyde, & Kotelchuck, 1977). Where general population studies have been conducted, most have been retrospective and have used lifetime prevalence data, making it difficult to adequately sequence presumed risk or causal factors and the onset of abuse or neglect. Utilization of prospective methodologies has been largely limited to "at risk" populations because of the large numbers of subjects and/or long follow-up periods required to obtain an adequate number of cases. In addition, many studies have not utilized control groups, distinguished between abuse and neglect, or used well-operationalized definitions of abuse, neglect, social variables, or psychiatric disorders (Mash & Wolfe, 1991; Widom, 1993). Finally, most studies have relied on univariate approaches and not included both key social and mental health variables in the same predictive model in order to determine the relative predictive power of the social and mental health factors (Belsky, 1993).
Research Question The present study extends the Kelleher, Chaffin, Hollenberg, and Fischer (1994) findings by using both ECA Wave I and Wave II data to prospectively examine whether substance abuse disorders and depression, as well as other psychiatric disorders, are risk factors for physical abuse or neglect. The main question concerns the relative risk first becoming abusive or neglectful, by self-report, at Wave II assessment 1-year later. Rather than using a matched control procedure in order to mitigate the predictive contribution of social factors, this study includes both social and psychiatric factors in the predictive analysis in order to determine the most powerful predictors of the onset of new abuse or neglect in a representative, community-based, nonreferred population.
METHOD
Subjects The ECA Program consisted of probability sample surveys at five research sites: New Haven, Connecticut; Baltimore, Maryland; St. Louis, Missouri; Durham, North Carolina; and Los Angeles, California. The surveys were coordinated by the National Institute of Mental Health program staff with 4,000 noninstitutionalized individuals sampled at each site. Rather than using simple random sampling, the ECA study used a complex multistage sampling process in which all respondents did not have an equal probability of selection. Sample weights have been derived for the ECA to compensate for any differences in selection probability among the various respondents. In addition to sampling differences, these weights have been adjusted in order to match respondent frequencies from the U.S. Census data on age, sex, and race (Eaton & Kessler, 1985, pp. 44-46). Once an individual was selected for the sample, repeated attempts as necessary were made to interview the individual. The survey consisted of a face-to-face structured interview during which information was obtained on sociodemographics, health service utilization, psychiatric symptoms, and functioning using the Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981). Of the individuals sampled, 65% to 80% were successfully interviewed at the sites with the majority of nonrespondents resulting from refusal to participate. Wave II data was similarly collected from the same subjects l-year after Wave I data collection. For this study, subjects were selected from the total subject pool as shown in Figure 1, resulting in a sample of 7,103 parents. It should be noted that the DIS does not inquire into current parenting status, only lifetime parenting status, so doubtless some number of parents included in the study were not currently parenting and thus were not at risk for abusing or neglecting.
Onset of abuse
195
Combined with minimization biases in self-report data, the incidence of abuse and neglect was probably substantially underestimated by this method. Although the ECA was not specifically designed to examine child maltreatment, the size and composition of the ECA sample and its two wave data collection strategy does offer a unique opportunity to test risk factors and causal assumptions about child physical abuse and neglect. Measures
Child physical abuse and neglect for this study were defined as all adult respondents who answered " y e s " to any one of five questions on abusive or neglectful behaviors included in the antisocial personality section of the DIS. More specifically, a single question in the DIS inquires about serious physical abuse of a child resulting in bruises, bed days, or medical care. Acknowledging respondents were coded as "abusive." Four additional questions inquired about neglectful behaviors engaged in by the respondent including leaving very young children unattended for extended periods, inadequately feeding or caring for children, or having a health-care professional suggest that children were being neglected. Any of the respondents who acknowledged at least one of these behaviors was coded as "neglectful." DSM-III psychiatric disorders were coded consistent with DIS diagnostic algorithms employing DSM-III diagnostic criteria, including exclusionary requirements. Subject parents were coded as having a disorder at Wave I if they met the symptom requirements at any time in the year prior to Wave I data collection. Substance abuse disorders (e.g., alcohol abuse, alcohol dependence, drug abuse, drug dependence, etc.) and depressive disorders (i.e., major depression and dysthymia) were each collapsed into a single present/absent variable. Age in years, marital status, gender, completion of high school, and number of persons in the household were recorded from the demographic section of the DIS. Due to small cell sizes for races other than White, race was coded dichotomously as " W h i t e " and "Other." A regression model of race and SES was undertaken and the residual value of race corrected for
Original E C A Sample 20,861 Removesubjectsat site wher~ questions not asked
v
|
5,372 ] Removesubjectsnot followed at WaveII 4,210 Removesubjectswithno childrenat Wave I or II 3,964 Removesubjectsadmitting lifetimeprevalenceabuseor neglectat Wave I 212
Remainin SubJects 7,103 New AbuseCases 63
,
New NeglectCases 84
l
Both 4
Figure 1. Selection of study subjects from ECA sample.
Controls 6,952
196
M. Chaffin, K. Kelleher,and J. Hollenberg
SES was used in all analyses to correct for racially related socioeconomic disparities. The resulting variable was included for demographic purposes only, and should not be presumed to be reflective of any ethnic group or culture. The ECA variable for household SES, which is derived from occupational status, education, and household income, was used. In addition, presence of social support to deal with problems was estimated by coding if the subject endorsed any of a group of items reflecting availability of a confidante, including friends, family, clergy, or other informal helpers with whom they had discussed personal problems within the past year.
Analytic Plan The previous study of Wave I data revealed that neglect cases were demographically and psychiatrically different from abuse cases, and that there was virtually no overlap between the groups (Kelleher, Chaffin, Hollenberg, & Fischer, 1994). Consequently, it was decided to analyze the groups separately. Newly abusive and neglectful parents (i.e., denying abuse or neglect at Wave I, admitting at Wave II) were initially compared with control parents (i.e., denying abuse or neglect at both Waves I and II) using bivariate statistical procedures. These included Pearson chi-square tests for nominal data, t-tests for continuous measures meeting parametric assumptions and Wilcoxon signed rank tests for the remaining continuous measures. Hierarchial logistic regression models were constructed using maximum likelihood parameter estimates to predict group membership (case vs. control). Social and demographic variables were entered into the model first. Next, substance abuse disorder was added. Then remaining psychiatric disorders were individually added. This plan was undertaken to examine the effects of psychiatric disorders only in the context of the social variables and to control for any comorbid substance abuse disorders. Only those variables found to be significantly associated with group membership in the bivariate analysis or significantly influencing other regression coefficients were included in the final models for either abuse or neglect. All analyses utilized ECA weighted data (Eaton & Kessler, 1985, pp. 4 4 - 4 6 ) for cell sizes and error terms.
RESULTS
Incidence of Physical Abuse and Neglect Sixty-three (.9%) of the parents endorsed the physical abuse item, while 84 (1.2%) endorsed at least one of the neglect items at Wave II. Only four parents endorsed both abuse and neglect, consistent with the low frequency of overlap found in Wave I lifetime prevalence data (Kelleher, Chaffin, Hollenberg, & Fisher, 1994). This group was felt to be too small to meaningfully analyze and was subsequently dropped from further analysis. When cell counts were weighted to correct for sampling irregularities, the estimated incidence rates were .8% for neglect and 1.1% for abuse. There were no significant site differences in incidence rates or bivariate comparisons, so site was dropped from further analysis. Results of the bivariate comparisons for social and demographic variables are presented in Table 1 for physical abuse and Table 2 for neglect. For child physical abuse, two measures, age and number in household, distinguished the newly abusive parents from controls, with abusive parents being younger and coming from larger households. Although statistically significant, the absolute group differences were not large. Four measures were associated with parents who became neglectful: age, number in household, SES, race, and marital status. Parents newly endorsing neglect were, on average, younger, non-White, from larger households, of lower SES, and unmarried. The number and absolute sizes of the group differences were larger for neglect than for abuse. Results of the bivariate analyses for psychiatric disorders are found in Table 3 for abuse
Onset of abuse
197
Table 1. Comparison of New Physical Abuse Cases and Controls on Pre-Abuse Social Variables" Variable
Sex Male Female Race White Other Current Marital Status Married Unmarried Education Less than HS Grad HS Grad Availability of a Confidante to Discuss Problems Yes No Age in Years (SD) Socioeconomic Status (SD) Number in Household (SD)
Abuse
Control
Case-Control Comparison
50% 50%
42% 58%
ns
54% 46%
59% 41%
ns
69% 31%
66% 34%
ns
54% 46%
51% 48%
ns
21% 79% 39.7 (11.7) 43.7 (23.9) 4.1 (1.9)
14% 86% 43.5 (16.7) 47.2 (23.5) 3.6 (1.9)
ns p < 0.05 ns p < 0.01
weighted data.
and Table 4 for neglect. Substance abuse and depression were associated with both forms of maltreatment. Additionally, OCD and antisocial personality (ASP) were associated with neglect. Because the neglect classification was derived from questions embedded in the ASP module of the DIS, the variables are confounded and so ASP was not included in further analyses. Among parents with a Wave I Substance Abuse Disorder, 2.85% had an onset of abuse and 3.04% had an onset of neglect during the 1-year interassessment interval, compared to .99% and .71% respectively of parents without the disorder. Among parents with a Depressive Disorder at Wave I, 4.10% had an onset of abuse and 2.11% had an onset of neglect compared
Table 2. Comparison of New Neglect Cases and Controls on Pre-Neglect Social Variables~ Variable
Sex Male Female Race White Other Current Marital Status Married Unmarried Education Less than HS Grad HS Grad Availability of a Confidante to Discuss Problems Yes No Age in Years (SD) Socioeconomic Status (SD) Number in Household (SD) ~'weighted data.
Neglect
Control
Case-Control Comparison
42% 58%
42% 58%
ns
42% 58%
59% 41%
p < 0.01
48% 52%
66% 34%
p < 0.01
52% 48%
51% 48%
ns
21% 79% 37.7 (11.7) 39.6 (16.5) 4.1 (1.9)
14% 86% 43.5 (16.7) 47.2 (23.5) 3.6 (1.9)
ns p < 0.01 p < 0.05 p < 0.05
198
M. Chaffln, K. Kelleher, and J. Hollenberg
Table 3. Comparison of New Abuse Cases and Controls on Pre-Abuse Psychiatric Disorders" Variable
Abuse
Control
Case-Control Comparison
Substance Abuse Disorder Depressive Disorder Antisocial Personality Panic Disorder Schizophrenia OCD
15.1% 16.6% 0% 0% 0% 0%
5.7% 4.3% 1.0% 0.5% 0.4% 0.8%
p < 0.001 p < 0.001 ns ns ns ns
weighted data.
to .95% and .78% respectively of parents without the disorder. For parents with OCD at Wave I (weighted N = 60), 5.18% had an onset of neglect compared to .80% of parents without the disorder. For parents with ASP at Wave I, 6.7% had an onset of neglect compared to .78% of parents without the disorder. Results from the hierarchical logistic regressions for abuse and neglect are presented in Table 5 and Table 6. Among the social variable set for abuse, age failed to remain significant, although number in the household did, with an adjusted odds ratio of 1.13, a small effect. For neglect, two of the original five social variables, SES and age, remained significant. Their adjusted effect sizes, however, were very small (.98, .99). Substance abuse, the next variable entered, remained a strong predictor for both abuse and neglect with adjusted odds ratios of 2.90 and 3.24 respectively, controlling for significant social variables. Depression remained a significant predictor for physical abuse, and OCD remained a significant predictor for neglect, both controlling for significant social variables and substance abuse. Depression was no longer a significant predictor of neglect once social variables and substance abuse were controlled.
DISCUSSION These findings, drawn from prospective community data, reinforce the association between certain types of psychiatric disorders and the later development of physical abuse and neglect in a community-based sample of American parents, It should be noted that these disorders predated the onset of any self-reported maltreatment and were classified in accordance with standardized diagnostic criteria. As such, the diagnosis is not colored by current status nor is it reflective of any retrospective justifications for maltreating behavior. Neither do these diagnoses refer to ill-defined conditions (e.g., a "drinking problem"), but rather represent the presence
Table 4. Comparison of New Neglect Cases and Controls on Pre-Abuse Psychiatric Disorders" Variable
Neglect
Control
Case-Control Comparison
Substance Abuse Disorder Depressive Disorder Antisocial Personality" Panic Disorder Schizophrenia OCD
21.0% 11.0% 8.3% 0% 1.22% 5.4%
5.7% 4.3% 1.0% 0.5% 0.4% 0.8%
p < 0.001 p < 0.05 p < 0.001 ns ns p < 0.001
a weighted data. b Measures of neglect and antisocial personality are confounded.
Onset of abuse
199
Table 5. Hierarchial Logistic Regression for Physical Abuse" Variable
O d d s Ratio
Social V a r i a b l e s Age N u m b e r in h o u s e h o l d Substance Abuse Other Diagnoses b Depression
Significance
9 5 % C o n f i d e n c e Interval
0.99 1.13 2.90
ns p < 0.05 p < 0.01
0 . 9 8 - 1.01 1.02 - 1.25 1.52 - 5.53
3.45
p < 0.001
1.80 - 6.61
a w e i g h t e d data. b V a r i a b l e s in this s u b g r o u p w e r e individually a d d e d to separate m o d e l s c o n t a i n i n g social variables a n d s u b s t a n c e abuse.
of a coherent set of symptoms associated with a diagnosable disorder. Furthermore, the subsequent onset of new maltreatment in these parents was not necessarily detected or reported, and included parents who likely remained hidden from traditional child protection agencies. Thus, these findings are not subject to whatever biases might be present in the way cases are detected and referred to public child protection agencies. Utilization of a community sample may account for the relatively small impact found for demographic and social variables in predicting maltreatment. For example, although poverty has repeatedly been found to be a significant feature of cases involved with public child protection agencies (NIS-II, 1988), socioeconomics appear to play a very limited role in the community sample risk model. This may support the contention that abuse and neglect actually cut across social class far more evenly than is widely thought, yet it is the poorest parents, or members of minority groups, who are disproportionately reported. Similar arguments could explain the relatively small influence found for age, education, and availability of social support. One social variable noted in previous studies to be a predictor of maltreatment remained statistically significant in this study--household size. However, even here the absolute size of the relative risk was quite small. Although many of the social risk factors studied were highly prevalent in both case and control groups, their relatively low risk ratios raise significant concerns about etiologic models that emphasize socioeconomics, family makeup, and gender to explain violence toward, or neglect of, children. Rather, these results more closely fit etiologic models, which include parental psychiatric status as a prominent factor in the development of abusive or neglectful behavior, such as stress-diatheses models in which social stressors may activate constitutional or psychological dispositions for maltreating behavior (e.g., Wolfner & Gelles, 1993), or in which individual parental disorders severely compromise parental functioning or predispose to abuse. Table 6. ltierarchial Logistic Regression for Neglect" Variable Social Variables Age N u m b e r in h o u s e h o l d Race SES Marital status Substance Abuse Other Diagnoses b Depression OCD
O d d s Ratio
Significance
9 5 % C o n f i d e n c e Interval
0.98 1.05 0.62 0.99 0.67 3.24
ns ns ns p < 0.05 ns p <: 0.001
0.96 0.93 0.35 0.97 0.38 1.63
2.02 7.21
ns p < 0.01
0.854.78 2.18 - 2 4 . 3 8
-
1.00 1.19 1.12 0.99 1.17 6.44
w e i g h t e d data. b Variables in this s u b g r o u p w e r e individually a d d e d to separate m o d e l s c o n t a i n i n g social variables a n d s u b s t a n c e abuse.
200
M. Chaffin, K. Kelleher, and J. Hollenberg
Of the psychiatric disorders studied, substance abuse disorders appear to be the most common and among the most powerfully associated with maltreatment. This supports earlier findings (Kelleher, Chaffin, Hollenberg, & Fischer, 1994) that close to half or more of abusive or neglectful parents have a lifetime prevalence substance abuse disorder. Substance abuse was the only variable that demonstrated a substantive association with both types of maltreatment, approximately tripling the risk of maltreatment when other factors were controlled. In addition, substance abuse disorders are highly prevalent in the population at large. They were the most prevalent disorders reported in both case and control groups. The finding that approximately 3% become newly abusive and 3% become newly neglectful during a relatively short time frame is cause for concern, given the size of the substance abusing parent population and the fact that new incidence numbers can be expected to accumulate over time. It should be noted that all reported incidence rates are probably underestimated because not all parents in the sample were actively caring for children and were thus not actually at risk. Also, it appears that substance abuse may play a mediating role between socioeconomic or other demographic variables and neglect, or may dramatically tip the scales towards neglect more in some populations, given that all demographic variables that were significant in the bivariate analyses with neglect failed to reach statistical significance when substance abuse was controlled. Although SES did remain statistically significant once substance abuse was controlled, in practical terms, its predictive power was so low (a risk ratio of .97-.99, where 1.0 would be the null hypothesis) that it cannot be a considered an incrementally valid predictor. Depression was found to be more uniquely associated with physical abuse rather than neglect once social factors and substance abuse are statistically controlled. Approximately 4% of depressed parents became abusive during the 1-year follow-up period. This may be related to depressive irritability or, alternately, could be the product of a helpless stance in which parenting stressors are allowed to escalate until they reach explosive proportions. Depression carried the highest adjusted risk of any disorder studied for physical abuse, with depressed parents found to be 3.45 times more likely to initiate physical abuse than their nondepressed counterparts when other factors were statistically controlled. Given that depression is also highly prevalent in the general population (approximately 4.4% of the parent population studied met criteria at Wave I), this is cause for significant concern over time. Although depression was related to neglect univariately, once substance abuse was controlled no significant association remained. This suggests that the relationship between depression and neglect may not be direct, as it appears to be for abuse, but may be mediated by substance abuse, which is a common complication of depression. On a more positive note, there are beneficial and often rapidly effective treatments for depression, so the potential for mitigating this risk factor could be very good. The association of neglect with OCD was unanticipated and presents something of a puzzle, especially given that the relationship persists when controlling for substance abuse. The association has not been well-described in the literature, nor has OCD been hypothesized to be an important causal or risk factor. If anything, the association appears to be counterintuitive and, given the small number of cases involved (N = 6) may be artifactual. However, a few other possibilities could be suggested. First, obsessional rituals may, in fact, actively interfere with child rearing responsibilities. Alternately, parents with OCD may have been overly meticulous or overly self-doubting, a response set that could have led to overreporting. Or, this small number of individuals may have been highly disturbed. Because the prevalence of OCD is rather low (.87% of parents studied), the public health significance of this disorder in understanding child maltreatment would appear to be minimal. However, the results do suggest that mental health professionals involved in treating patients with OCD be aware of their potentially increased risk for neglect. No association was found between schizophrenia and either type of maltreatment. This suggests that concern over the risk posed by mentally ill parents, as a group, may be misplaced, assuming that their condition is not comorbid with depression or substance abuse. This supports the findings of Swanson and colleagues (1990) that persons with schizophrenia are not at a
Onset of abuse
201
greatly increased risk for interpersonal violence in general, including abuse of children. Of the 30 parents with schizophrenia followed, none self-reported new abuse and one self-reported new neglect. However, it was not possible to ascertain how many of these parents were actively engaged in caretaking for their children during the assessment period, and if that number was small, any effect would be masked. Finally, the results are consistent with the idea that child physical abuse and child neglect are distinct problems. Very little overlap was noted between the types of maltreatment, with so few parents endorsing both abuse and neglect that the group could not be meaningfully statistically analyzed. Furthermore, the abuse group and the neglect group were different with respect to both social and psychiatric risk factors. Finally, the abuse group did not differ from controls on race, marital status or SES while the neglect group did. This is consistent with previous findings that have noted, for example, a greater role for depression in physical abuse relative to neglect (Pianta, Egeland, & Erickson, 1989; Zuravin, 1988), but is at odds with a number of other studies that have found considerable comorbidity between physical abuse and neglect, primarily among referred rather than community samples (see Belsky, 1993 for a review). A number of limitations should be considered in interpreting these findings. First, the DIS does not sample the entire range of child physical abuse and neglect behaviors, does not distinguish single events from a pattern of maltreatment, and its psychometric adequacy for measuring maltreatment is unknown. The physical abuse determination is made on the basis of a single question that asks about relatively serious occurrences. Thus, the findings for physical abuse could be consistent with Steele's (1987) contention that substance abuse is a greater factor only in more severe maltreatment cases. Also, the study could be effected by self-report biases. It appears unlikely that a generic "yes" response set would have influenced the data given that all newly abusive or neglectful parents denied abuse or neglect at Wave I, and that there was no association between maltreatment and other disorders such as phobias or schizophrenia. On the other hand, social desirability bias or a concern about self-incrimination may have been involved. Combined with the fact that some number of subjects were probably not actively parenting, social desirability bias would decrease obtained incidence rates. If social desirability was more a factor for the child maltreatment items than for the other psychiatric symptom items (as seems likely given that maltreatment is subject to greater social disapproval and sanction than are most psychiatric symptoms), this would cause some maltreating parents to be misclassified as control parents, and the effect on the study would be a conservative one. If, however, social desirability effects were to decrease from Wave I to Wave II, (e.g., as subjects became more trusting of the study) then the effect would be to increase measured incidence rates. It is unclear whether, or in which direction, any change in desirability effects across time would influence risk ratios. If subjects were included who had been falsely identified as nonmaltreating at Wave I (due to response bias), and then identified as maltreating at Wave 1I, this would only impact risk ratios if the risk factor was unrelated to abuse or neglect prospectively (i.e., had developed concomitantly with or subsequent to the maltreatment). Finally, given that no single disorder studied was present in more than 21% of any maltreatment population, these psychiatric risk factors cannot comprise a necessary or sufficient model which could be expected to apply to all or even most new cases of maltreatment. The results do suggest, however, that we cannot expect to understand child maltreatment solely on the basis of social ecologies. Both substance abuse and depressive disorders can predate maltreatment and pose a risk for its development. Both can be chronic relapsing problems with strong constitutional predispositions. Both are relatively common among adults of parenting age. Furthermore, the development of both disorders has been linked to childhood histories of maltreatment (Malinosky-Rummell & Hansen, 1993). Consequently these conditions may function as mediating variables or proxies in the link between early abuse and subsequent parenting behavior. The nature of these relationships, at least in representative community samples, is a question that must be addressed by additional research. In any case, the current findings have clear implications
202
M. Chaffin, K. Kelleher, and J. Hollenberg
for our efforts to understand and prevent physical abuse and neglect. A t a m i n i m u m , the results support the o n g o i n g focus on substance abusing parents as an at-risk population for both forms o f maltreatment, and depressed parents as at-risk for physically abusing. A l t h o u g h the multideterm i n e d nature o f maltreatment is a given, and the e c o l o g y o f maltreatment is broad, it does appear that mental health factors cannot be ignored in either understanding etiology or designing prevention efforts.
REFERENCES
Azar, S. T. (1991). Models of child abuse: A metatheoretical analysis. Criminal Justice and Behavior, 15, 30-46. Belsky, J. (1993). Etiology of child maltreatment: A developmental-ecological approach. Psychological Bulletin, 114, 413-434. Black, R., & Mayer, J. (1980). Parents with special problems: Alcoholism and opiate addiction. Child Abuse & Neglect, 4, 45-54. Crittenden, P. M. (1993). An information processing perspective on the behavior of neglectful parents. Criminal Justice and Behavior, 20, 27-48. Culp, R. E., Culp, A. M., Soulis, J., & Letts, D. (1989). Self-esteem and depression in abusive, neglecting, and nonmaltreating mothers. Infant Mental Health Journal, 10, 243-251. Dinwiddie, S. H., & Bucholz, K. K., (1993). Psychiatric diagnoses and self-reported child abusers. Child Abuse & Neglect, 17, 465-476. Eaton, W. W., & Kessler, L. G. (1985). Field methods in psychiatry: The NIMH Edidemioligic Catchment Area Program. Orlando, FL: Academic Press. Failer, K. C., & Ziefert, M. (1981). Causes of child abuse and neglect. In K. C. Failer (Ed.), Social work with abused and neglected children (pp. 32-52). New York: Free Press. Gelles, R. J. (1980). A profile of violence toward children in the United States. In G. Gerbner, C. Ross, & E. Zigter (Eds.), Child abuse: An agenda for action (pp. 82-105). New York: Oxford. Gelles, R. J. (1993). Alcohol and other drugs are associated with violence--they are not its cause. In R. J. Gelles, & D. L. Loseke (Eds.), Current controversies on family violence (pp. 182-196). Newbury Park, CA: Sage. Gelles, R. J., & Cornell, C. P. (1990). Intimate violence in families (2nd ed.) (pp. 11-20). Newbury Park, CA: Sage. Kelleher, K., Chaffin, M., Hollenberg, J., & Fischer, E. (1994). Alcohol and drug disorders among physically abusive and neglectful parents in a community-based sample. American Journal of Public Health, 84, 1586-1590. Malinosky-Rummell, R., & Hansen, D. J. (1993). Long-term consequences of childhood physical abuse. Psychological Bulletin, 114, 68-79. Margolin, L. (1992). Beyond maternal blame: Physical child abuse as a phenomenon of gender. Journal of Family Issues, 13, 410-423. Mash, E. J., & Wolfe, D. A. (1991). Methodological issues in research on physical child abuse. Criminal Justice and Behavior, 18, 8-29. Milner, J. S., & Chilamkurti, C. (1991). Physical abuse perpetrator characteristics: A review of the literature. Journal of Interpersonal Violence, 6, 345-366. Murphy, J. M., Jellinek, M., Quinn, D., Smith, G., Poitrast, F. G., & Goshko, M. (1991). Substance abuse and serious child mistreatment: Prevalence, risk, and outcome in a court sample. Child Abuse & Neglect, 15, 197-211. Newberger, E. H., Reed, R. B., Daniel, J. H., Hyde, J. N., & Kotelchuck, M. (1977). Pediatric social illness: Toward an etiologic classification. Pediatrics, 60, 178-185. N1S-II Study findings: Study of national incidence and prevalence of child abuse and neglect: 1988. Washington, DC: U.S. Department of Health and Human Services--National Center on Child Abuse and Neglect. O'Leary, K. D. (1993). Through a psychological lens: Personality traits, personality disorders, and levels of violence. In R. J. Gelles, & D. R. Loeske (Eds.), Current controversies on family violence (pp. 7-30). Newbury Park, CA: Sage. Pianta, R., Egeland, B., & Erickson, M. (1989). The antecedents of maltreatment: Results of the mother-child interaction research project. In D. Cicchetti & V. Carlson (Eds.), Child maltreatment: Theory and research on the causes and consequences of child abuse and neglect (pp. 203-253). Cambridge, MA: Cambridge University Press. Robins, L., Helzer, J., Croughan, J., & Ratcliff, K. (1981). National Institute of Mental Health diagnostic interview schedule: Its history, characteristics, and validity. Archives of General Psychiatry, 38, 381-389. Steele, B. (1987). Psychodynamic factors in child abuse. In R. Helfer & C. Kempe (Eds.), The battered child (4th ed.) (pp. 81-114). Chicago, IL: University of Chicago Press. Straus, M., Gelles, R., & Steinmeetz, S. (1980). Behind closed doors: Violence in the American family. Garden City, NJ: Anchor Books. Straus, M. A. (1979). Measuring intrafamily conflict and violence: The conflict tactics (CT) scales. Journal of Marriage and Family Therapy, 41, 75-88. Swanson, J. W., Holzer, C. E., Ganju, V. K., & Jono, R. T. (1990). Violence and psychiatric disorder in the community: Evidence from the Epidemiologic Catchment Area surveys. Hospital and Community Psychiatry, 41, 761-770. Widom, C. S. (1993). Child abuse and alcohol use and abuse. In Alcohol and interpersonal violence: Fostering
Onset of abuse
203
multidisciplinary perspectives (pp. 291-314). National Institute on Alcohol Abuse and Alcoholism, Research Monograph 24. Washington, DC: U.S. Department of Health and Human Services. Wolfe, D. A. 0985). Child abusive parents: An empirical review and analysis. Psychological Bulletin, 97, 462-482. Wolfner, G. D., & Gelles, R. J. 0993). A profile of violence toward children: A national study. Child Abuse & Neglect, 17, 197-212. Zuravin, S. (1988, February). Child abuse, child neglect, and maternal depression: Is there a connection? Research Symposium on Child Neglect, U.S. Department of Health and Human Services, National Center on Child Abuse and Neglect, Washington, DC.
R~umr----Les 6tudes des facteurs de risque psychiatriques et sociaux sont limitres par l'utilisation de mrthodologies rrtrospectives et la drpendance d'rchantillons identifirs ou rapportrs officiellement. Sept mille cent et trois parentsont 6t6 identifirs sur la base des donnres provenant des premieres et deuxi~mes vagues de l'rtude "Epidemiologic Catchement Area" de l'Institut National pour la Sant6 Mentale. IIs sont issus d'un 6chantillon reprrsentatil de la communautr, qui n'avaient pas spontanrment signalrs d'abus physiques ou de nrgligence de leurs enfants au cours de la premiere vague. Ils ont 6t6 suivis pour drterminer les facteurs de risque associrs h l'apparition d ' u n autosignalement d'abus physique ou de nrgligence, identifi6 darts la seconde vague. Les facteurs sociaux considrrrs incluaient l'Age, le statu socio-rconomique, le soutien social, 1"rducation, la taille du mrnage et le sexe. Plusieurs affections psychiatriques parmi lesquelles les abus de drogue et la drpression ont aussi 6t6 examinres. Des modules de risque ont 6t6 drvelopprs en utilisant une rrgression Iogistique hirrarchisre. Les abus sexuels et la nrgligence se sont rrvrlrs avoir des groupes distincts de facteurs de risque avec une superposition minime entre les groupes. Les variables sociales et drmographiques se sont rrvrlres peu prrdictives de maltraitance par opposition aux abus de drogue, qui 6taient fortement associrs avec l'rmergence d'abus comme la nrgligence (risque relatif = 2.90 et 3.24). La drpression est appartie comme un facteur de risque important de I'abus physique (risque relatif 3.45). Les implications de ces donnres sont discutres sous l'angle des modules de causes majeures de la maltraitance. Resumen---Los estudios psiqui~itricos y de factores de riesgo social del maltrato a los nifios, hart estado limitados por metodologias retrospectivas y por apoyarse en muestras reportadas o identificadas oficialmente. Utilizando los datos del " W a v e s I" y el " W a v e s lI" del area de investigacirn epidemiol6gica del Instituto Nacional de Salud Mental, de 7,103 padres y madres de una muestra comunitaria probabillstica que no se auto-report6 como habiendo abusado fisicamente ni ser negligentes de sus hijos e hijas en el " W a v e I"; se les hizo el sequimiento para determinar los factores de riesgo asociados con la presentaci6n de abuso fisico o negligencia autoreportado que fue identificado en el " W a v e lI." Los factores sociales considerados incluyeron edad, status socioeconrmico, manuntencirn social, educacirn, tamafio de la vivienda y grnero. Ademfis, se examinaron varios desrrdenes psiquilitricos, incluyendo des6rdenes de abuso de sustancias y depresi6n. Se desarrollaron los modelos de riesgo utilizando regresirn loigistica en jerarquia. Se encontr6 que el abuso fisico y la negligencia tienen conjuntos distintos de factores de riesgo, con obreposici6n minima entre los grupos. Las variables sociales y demogr~ificas se encontr6 que eran predictores iimitados de maltrato, mientras que los desrrdenes de abuso de sustancias estaban fuertemente asociados con la presentacirn tnato del abuso como de la negligencia (riesgos relativos = 2.90 y 3.24 respectivamente). Se encontr6 que la depresi6n era un factor de riesgo fuerte para el abuso ffisico (riesgo relativo = 3.45). Se discuten las implicaciones de los resultados en trrminos de los modelos causales principales de tratamiento.