Amer. J. Orthopsychiat., 56(4), October 1986, 589-601.
Child Abuse and Pediatric Social Illness: An Epidemiological Analysis and Ecological Reformulation
Eli H. Newberger, M.D., Robert L. Hampton, Ph.D., Thomas J. Marx, Ed.D., Kathleen M. White, Ed.D.
Family Development Program. Oepartment of Medicine, Children’s Hospital and the Department of Pediatrics. Harvard Medical School. Boston.
Children under four years of age hospitalized for child abuse, domestic accidents, failure to thrive, and ingestions were matched with controls admitted for comparably acute medical conditions. A structured parental interview yielded significant case-control differences. Discriminant analysis suggested interrelationships among the case groups and cluster analysis identified three cohesive groups in terms of severity of symptoms. This reformulation provides a matrix for organizing data and an alternative to the present manifestational classification system.
The understanding of child abuse is limited by insufficient attention to issues of definition and etiology and by a particular focus on individual psychiatric disturbance.19 The few existing controlled studies suggest that maltreatment has diverse concomitants in child development,7 family stress,18 parental and child conflicts,3 and environmental difficulties.10 Case-control studies, however, have been plagued with difficulties in achieving adequate matching, which lead in turn to confounding by social class and age and to forming causal inferences based on such commonly accepted risk factors as low birth weight, maternal depression, and the intergenerational transmission of violence.12, 17
Advances in knowledge of child abuse suggest three areas in which past knowledge is challenged by newer findings and critical analyses. First. important associations may exist between violence toward children and violence between parents.12, 22 Second. the distinctive visibility. if not vulnerability. of poor, socially marginal. and minority children and families cannot be separated from a substantial bias which favors them for child abuse identification.11-14, 20, 22 Third, the theoretical orientations of studies of child abuse define, select, and. to some extent, distort what we may understand as the nature, causes, and consequences of the problem; narrow and unitary conceptions of cause and effect may confine the available range of preventive and therapeutic interventions.19
More recent research appears to favor an ecological approach to child maltreatment, i.e., the progressive mutual adaptation of the individual and an environment conceptualized as a set of “nested” interactive systems.2 Proponents of this ecological approach argue that we cannot ignore the influence of familial economic and demographic factors on the quality of life of children and their families.1, 5, 8, 9
Ecological analysis focuses on the multiple levels of individual and family characteristics which interact with characteristics of the broader social context.1, 5, 8, 9 This approach to child abuse suggests, for example, that informal support networks can mediate the effects of economic stress, social isolation, and family structure by providing information, relief, and direct support. The existence of reciprocal helping relationships within a community appears to be associated with a lower prevalence of reported child abuse. 1, 5, 8, 9 The framework argues for the collection and analysis of a rich array of data on child, family, and community.
The concept of pediatric social illness articulated previously18 emphasizes the view that family environments and social conditions contribute importantly to the children’s health and may form a common etiologic substrate for certain conditions. To understand more adequately the process by which a child might suffer inflicted trauma, accidental trauma, failure to thrive, or poisoning requires a capacity to organize the complex data that characterize a child’s life setting.
The object of this investigation is to broaden the focus of inquiry with respect both to classification (from child abuse to pediatric social illness) and to the elements of risk (from personal pathologies to the family in its ecologic niche). Previous research on the pediatric social illnesses suggests that this conceptualization may be useful.5, 18
METHOD
Sample
All children under four years of age hospitalized at Children’s Hospital Medical Center, Boston, between July 1975 and April 1977 were eligible for selection as cases if they bore the diagnoses of child abuse, accident, ingestion, or nonorganic failure to thrive (FTT). Children hospitalized for comparably acute medical illnesses (such as pneumonia or meningitis) were eligible for selection as controls. Cases and controls were individually matched on three attributes: age, race, and socioeconomic status (determined by the Hollingshead two-factor social index15). The final sample consisted of 209 cases and 209 controls. Salient characteristics of the sample are presented in TABLE 1. A previous report compared the child developmental and family characteristics of the accident and child abuse victims in black families.5
Table 1
|
|||
SAMPLE CHARACTERISTICS (N=4181)
|
|||
VARIABLE |
PERCENT
|
VARIABLE |
PERCENT
|
Child’s Diagnosis | Age of Child | ||
Abuse Failure to thrive Accident Ingestion |
22.9
19.6 46.4 11.0 |
0 to 6 months 7 to 12 months 13 to 18 months 19 to 24 months 25 months and older |
19.7
17.8 18.7 13.8 30.1 |
Sex of Child | Child’s Usual Health | ||
Male Female |
58.0
42.0 |
Excellent Good Fair Poor |
44.0
39.0 13.0 4.0 |
Race of Child | |||
White Nonwhite |
69.0
31.0 |
||
Mother’s Marital Status | |||
Family Receiving Public Assistance |
Married Single Separated, divorced or widowed |
63.6
19.3 16.9 |
|
Yes No |
45.0
55.0 |
||
1 Each case was matched with a control of similar age, race, and SES. The sample contained a total of 209 cases and 209 controls. |
Measures
The principal instrument was a standardized, precoded maternal interview lasting approximately one hour and conducted by specially trained interviewers at the hospital. (The interviewers were women with human service backgrounds, who were given supervision individually and in groups by two research psychologists. Although inter-interviewer reliability was not measured because of the constraints of the clinical setting, the systematic attention to interview technique and findings yielded an impression of a high order of reliability.) Questions focused on a wide range of variables including family structure, housing, employment, finances, support, mobility, psychological stresses, child rearing, and the parents’ own childhoods. Data on the child included a Vineland Social Maturity Index derived from a maternal report, Carey’s adaptation of the Thomas, Chess and Birch infant scales was employed to assess the mother’s perceptions of her child’s temperament.4, 23 From the medical records, physical characteristics were obtained, i.e., height, weight, hematocrit, head circumference, and duration of hospitalization.
Data Analysis
First, we examined each of the independent variables separately in relation to each group of pediatric social illness cases and their matched controls in order to define the associations and their directions. We accomplished this step with a series of cross-tabulations and used t-tests and chi-square tests respectively to assess statistical significance for continuous and categorical variables (see TABLE 2). Next, we used discriminant analysis to determine from all the variables which set of variables-together and relative to others-provided the best means of discriminating cases from controls, Finally, we employed hierarchical cluster analysis to search for homogeneous groups, irrespective of diagnosis, in a random half-sample of the entire population, and we replicated this analysis on the other half-sample.
Table 2
|
|||
ATTRIBUTES OF CASES AND MATCHED CONTROLS
|
|||
VARIABLE | CASES(N=48) | CONTROLS(N=48) | TOTALS |
CHILD ABUSE | |||
Mother has Telephone | |||
Yes | 59.4% | 40.6% | 35.6% |
No | 37.9 | 62.1 | 64.4 |
X2=3.8; df=1; p≤.05 | |||
Child Easy or Difficult | |||
Very difficult | 64.0 | 36.0 | 28.1 |
Not very difficult | 39.1 | 60.9 | 71.9 |
X2=4.5; df=1; p≤.03 | |||
Times/Week Spanks Child | |||
Zero | 53.7 | 46.3 | 61.4 |
One to three | 47.1 | 52.9 | 19.3 |
Four or more | 17.7 | 82.4 | 19.3 |
X2=6.8; df=2; p≤.03 | |||
Mother’s Description of Childhood | |||
Unhappy | 85.7 | 14.3 | 15.7 |
Slightly unhappy | 44.0 | 56.0 | 28.1 |
Slightly happy | 45.0 | 55.0 | 22.5 |
Very happy | 30.0 | 70.0 | 33.7 |
X2=12.0; df=3; p≤.007 | |||
Maternal Grandmother’s Discipline Felt | |||
Severe | 75.0 | 25.0 | 29.3 |
Appropriate | 33.0 | 66.7 | 58.5 |
Lenient | 40.0 | 60.0 | 12.2 |
X2=11.4; df=2; p≤.003 | |||
Mother Rural to Urban Shift | |||
Yes | 52.8 | 47.2 | 80.0 |
No | 16.7 | 83.3 | 20.0 |
X2=7.6; df=1; p≤.006 | |||
Father Rural to Urban Shift | |||
Yes | 30.2 | 69.8 | 47.8 |
No | 59.6 | 40.4 | 52.2 |
X2=7.8; df=1; p≤.005 | |||
Child Below Age Self-help, Dressing | |||
Yes | 80.0 | 20.0 | 11.6 |
No | 39.5 | 60.5 | 88.4 |
X2=5.9; df=1; p≤.02 | |||
Child Below Age of Communication | |||
Yes | 63.6 | 36.4 | 25.6 |
No | 37.5 | 62.5 | 74.4 |
X2=4.5; df=1; p≤.03 | |||
Times/Week Mother Sees Relatives | |||
Once | 65.5 | 36.0 | 27.8 |
Twice | 52.0 | 48.0 | 27.6 |
Three times or more | 25.0 | 75.0 | 40.2 |
X2=11.2; df=2; p≤.03 | |||
Mother Sees Relatives Enough | |||
Yes | 40.3 | 59.7 | 87.5 |
No | 81.8 | 18.2 | 12.5 |
X2=6.7; df=1; p≤.01 | |||
FAILURE TO THRIVE | (N=41) | (N=41) | |
Mother Owns Car | |||
Yes | 41.7 | 58.3 | 55.2% |
No | 66.7 | 33.3 | 44.8 |
X2=5.4; df=1; p≤.02 | |||
Child’s Usual Health | |||
Poor to Fair | 79.0 | 21.5 | 22.6 |
Good | 65.8 | 34.2 | 45.2 |
Excellent | 22.2 | 77.8 | 32.1 |
X2=17.9; df=2; p≤.0001 | |||
VARIABLE | CASES(N=48) | CONTROLS(N=48) | TOTAL |
Times/Month Mother Has Headaches | |||
None | 41.5 | 58.5 | 47.7 |
One to seven | 70.6 | 29.4 | 39.5 |
More than seven | 45.5 | 54.6 | 12.8 |
X2=6.7; df=2; p≤.04 | |||
Mother’s Emotional Support | |||
Enough | 59.4 | 40.6 | 82.1 |
Sometimes enough | 27.3 | 72.7 | 13.1 |
Not enough | 0.0 | 100.0 | 3.6 |
None | 100.0 | 0.0 | 1.2 |
X2=8.3; df=3; p≤.04 | |||
Child Below Age Locomotion | |||
Yes | 92.3 | 7.7 | 15.3 |
No | 48.6 | 51.4 | 84.7 |
X2=8.5; df=1; p≤.03 | |||
Number of Categories Below Age Group | |||
Zero | 50.0 | 50.0 | 44.7 |
One | 29.2 | 70.8 | 28.2 |
Two | 92.9 | 7.1 | 16.5 |
Three or more | 100.0 | 0.0 | 10.6 |
X2=20.1; df=4; p≤.03 | |||
Vineland Social Quotient | |||
Low | 80.0 | 20.0 | 29.8 |
High | 45.8 | 54.2 | 70.2 |
X2=6.9; df=2; p≤.03 | |||
ACCIDENT | (N=97) | (N=97) | |
Times/Week Mother Sees Relatives | |||
No relatives | 30.0% | 70.0% | 15.9% |
Rarely | 48.3 | 51.7 | 30.7 |
Sometimes | 59.7 | 40.4 | 30.2 |
Often | 68.8 | 31.3 | 8.5 |
Very often | 42.9 | 57.1 | 14.8 |
X2=9.8; df=4; p≤.04 | |||
Child’s Usual Health | |||
Poor to fair | 20.8 | 79.2 | 12.6 |
Good | 45.3 | 54.7 | 39.3 |
Excellent | 60.9 | 39.1 | 48.2 |
X2=14.6; df=3; p≤.002 | |||
Religion Specified | |||
Yes | 45.2 | 54.8 | 81.4 |
No | 69.4 | 30.6 | 18.6 |
X2=6.9; df=1; p≤.01 | |||
INGESTION | (N=23) | (N=23) | |
Below Age Communication | |||
Behind | 100.0% | 0.0% | 11.1% |
Not behind | 42.5 | 57.5 | 88.9 |
X2=5.9; df=1; p≤.02 | |||
Below Age Any Category | |||
Behind | 70.6 | 29.4 | 37.0 |
Not behind | 37.9 | 62.1 | 63.0 |
X2=4.6; df=1; p≤.03 | |||
Extended Family | |||
No | 56.4 | 43.6 | 84.8 |
Yes | 14.3 | 85.7 | 15.2 |
X2=5.9; df=1; p≤.02 | |||
NOTE: Full contingency tables omitted because of space limitations. Percentages may not add to 100 because of rounding. |
Discriminant Function Analyses
The most important results of the four multiple discriminant function (MDF) equations we constructed can be seen in TABLE 3, which summarizes the variables that emerged as the most powerful discriminators between each case category and its controls and lists the variables considered for entry by the MDF program. Characteristics of the children served in all instances to differentiate between cases and controls. Child characteristics, aspects of the parent-child relationship, past stresses, and life-context variables emerged as significant variables in some but not all of the discriminant function equations.
Table 3
|
|
DISCRIMINANT FUNCTION ANALYSIS: VARIABLES DISCRIMINATING BETWEEN EACH CASE CATEGORY AND ITS CONTROLS
|
|
ACCIDENT | (N=194) |
Few medical visits in last year*** | |
Recent moves*** | |
Healthy | |
Religion specified less often** | |
Child spanked more often than on hands and bottom* | |
Mother’s family residence in urban area** | |
Mother estimates young age for toilet training* | |
Advocacy needs higher* | |
Relatives seen less often | |
Less access to shopping and recreation | |
Fewer social agencies | |
Mother less likely to be born in U.S. | |
Wilkes Lambda: | .739 |
Canonical R: | .510 |
FAILURE TO THRIVE | (N=82) |
Unhealthy*** | |
Reactive to visual and auditory changes** | |
Mother sees relatives less often per week* | |
Mother gets away by self less often* | |
Mother watches TV more* | |
Mother has fewer adult relatives in Boston | |
Advocacy needs higher | |
Mother less likely to have own car | |
When mother and father disagree, more likely to hit and throw objects | |
Wilkes Lambda: | .423 |
Canonical R: | .832 |
INGESTION | (N=46) |
Low Vineland Social Maturity score** | |
Healthy* | |
Extended family unavailable | |
Mother gets away by self less often | |
As a child, mother was more often spanked with objects | |
Advocacy needs higher | |
Wilkes Lambda: | .461 |
Canonical R: | .733 |
ABUSE | (N=96) |
Mother sees relatives less per week*** | |
Mother’s childhood unhappy*** | |
Mother reports spanking child less** | |
Mother feels no one interested in problems* | |
Child easy or difficult* | |
Fewer months since mother had last job | |
Religion specified less often | |
Recent death in family* | |
More social agencies* | |
Mother’s family residence in urban area* | |
Mother doesn’t see relatives enough** | |
Mother less likely to have own car | |
Mother and father disagree on discipline** | |
Little help with child care* | |
Low Vineland Social Maturity score** | |
Mother spanked more often than on hands and bottom** | |
Mother has fewer relatives in Boston* | |
Mother and father disagree on child-rearing | |
Wilkes Lambda: | .141 |
Canonical R: | .926 |
*p<.05 **p<.01 ***p<.001 NOTE: Variables listed in order of their inclusion. VARIABLES ENTERED INTO THE MULTIPLE DISCRIMINANT FUNCTION PROGRAM: Times week mother sees relatives (Abuse, Accidents); Mother describes childhood as happy (Abuse); Times week mother spanks child (Abuse, FTT); Mother feels no one interested in her problems (Abuse, Accidents); Child easy or difficult (Abuse, Accidents); Months since mother’s last job (Abuse, Accidents); Religion specified (Abuse, Accidents); Parent recent death stressful (Abuse, FTT, Accidents); Times month mother has headaches (Abuse); Mother’s family intact (Abuse); Number of social agencies (Abuse, Accidents); Mother happy with pregnancy (Abuse); Mother’s family residence urban area (Abuse, Accidents); Mother born in U.S. (Abuse, Accidents); Mother and father disagree violently (Abuse, FTT, Accidents); Mother sees relatives enough (Abuse, Ingestions); Parents consider neighborhood safe (Abuse); Mother owns car (Abuse); Mother’s number of friends in neighborhood (Abuse); Child reacts to new foods (Abuse); Mother and father .disagree on discipline (Abuse, FTT, Accidents); Marital status (Abuse); Mother’s number of relatives in city (Abuse, Accidents); Age of mother (Abuse, Accidents); Child premature (Abuse); Mother’s parents agree on child rearing (Abuse); Mother physically punished in childhood (Abuse, Ingestions); Times child at MD last year (Accidents); Child’s usual health (FTT, Accidents, Ingestions); Child’s reactivity to auditory and visual stimula (Abuse, FTT); Child’s tempo of play (FTT); Total number of children in family (FTT); Male present (FTT); Social isolation (FTT, Ingestion); Mother’s education (FTT, Accidents); Family stress (FTT); Maternal use of drugs and alcohol (FTT); Child’s sleep pattern (FTT); Difficult child to feed (FTT); Unhappy maternal childhood (Abuse, FTT); Times per week child is spanked (FTT, Ingestions); Number of moves within last year (FTT, Accidents); Planned vs. unplanned pregnancy (FTT); Mother views current relationship as supportive (FTT); Parents agree on child rearing practices (FTT); Mother’s usual health (FTT); Mother sees herself as happy (FTT); Mother has telephone (Abuse, Ingestions); Total family advocacy needs (Ingestions, Accidents); Mother just can’t get going (Accidents); Mother’s ease to shopping and recreation (Accidents); Child rearing problems (Accidents); Private pediatrician (Accidents); Mother spanked into adolescence (Abuse, Accidents); Extended family available (Abuse, FTT, Ingestions, Accidents); Child’s locomotion skills (FTT). |
Although we found evidence that all categories of pediatric social illness are characterized by some isolation of case families from their kin and communities, this isolation was most pervasive in the families that bear the diagnosis of abuse. Of the 18 variables that entered the abuse discriminant function equation at a significance level of 0.05 or more, seven reflected directly the mother’s current social isolation. An eighth variable, recent death of a family member, may contribute to isolation as well as to general stress.
Stresses differed between mothers of abused children and their controls. The former may have seen their relatives less often, felt that no one was interested in their problems, have suffered a recent death in the family, coped alone with child care, and disagreed with their husbands concerning discipline and child rearing. They could count on few relatives to help and had fewer fin in their communities. More than the mothers in the other case categories, they appeared to see themselves as unconnected to others.
Past stresses entered the discriminant function equations at significant levels only for ingestion and abuse cases. Mothers of abuse victims reported unhappy childhoods.
Aspects of the parent-child relationship entered discriminant function equations at significant levels for both accident and abuse victims. Paradoxically, the mothers of abuse victims reported spanking their children less than mothers in other case categories. Although this report may be untruthful, betrayed by the diagnosis, and forced by the hospital context in which the interview was conducted, lack of truthfulness is not the only hypothesis to arise from this finding. Previous study of the interaction between abuse victims and their parents, for example, disclosed interaction patterns that included a high degree of both permissiveness and punitiveness.24 Indeed, the frequency of hitting might have been less, but the child’s provocations might have been met with more severe violence. In the Sears, Maccoby and Levin study of child rearing patterns, this parenting style was associated with the most frequent expression of serious aggression in children’s observed play.21
We also noted significant case-control distinctions between the accident victims’ mothers’ reports of striking the child on vulnerable body parts and the age at which a child could reasonably be expected to be toilet trained. These reports, more commonly associated with clinical and investigative work on abused children, suggest the possibility that a parent’s capacity to protect a child from anger may be connected to the capacity to protect the child from environmental hazards. Preschool children’s own propensities for accidents may be increased by the child’s inability to internalize self-protection, even as the child is not protected by the parent; the possibly exaggerated expectations of the child may indicate a diagnostic overlap with child abuse in more than a few cases.
Mothers of both accident and abuse victims came more frequently from urban areas and they specified their religions less often. In addition, the mothers of abuse victims were in contact substantially more with social agencies than their controls or the mothers of the accident victims.
Table 4
|
||
CORRECT AND INCORRECT CLASSIFICATIONOF CASES FROM DISCRIMINANT FUNCTION ANALYSIS
|
||
PREDICTED GROUP
|
||
ACTUAL GROUP |
CASES
|
CONTROLS
|
Accident | ||
Cases (N=97) |
76%
|
24%
|
Controls (N=97) |
29
|
71
|
Ingestion | ||
Cases (N=23) |
87
|
13
|
Controls (N=23) |
13
|
87
|
Failure to Thrive | ||
Cases (N=41) |
76
|
24
|
Controls (N=41) |
12
|
88
|
Abuse | ||
Cases (N=48) |
92
|
8
|
Controls (N=48) |
10
|
90
|
The correct classification yielded by the multiple discriminant analyses are presented in TABLE 4. Application of the abuse equation to the abuse sample and its controls resulted in correct classification of 92% of the abuse cases and 90% of their controls. Application of the failure-to-thrive equation to the failure-to-thrive cases and their controls resulted in correct classification of 76% of the FTT cases and 88% of their controls. The ingestion equation yielded correct classification of 87% of the ingestion cases and 87% of their controls, and the accident equation yielded correct classification of 76% of the accident cases and 71% of their controls.
As a result of this study, in which we examined how pediatric social illness categories differed and overlapped, we examined whether an alternative to the traditional pediatric classification might provide a more informative grouping of cases and controls.
Hierarchical Cluster Analysis
Discriminant analyses allowed us to differentiate between groups bases on our existing diagnostic categories. In contrast, cluster analysis generated new categorical schemes by partitioning a set of cases into homogenous subgroups. Hierarchical cluster analysis permitted us to explore the extent to which natural clusters may exist within our data set.
We performed a cluster analysis on a random half-sample of cases in our study. The 63 variables upon which the half-sample was clustered made up the set that best discriminated cases from controls within each of the pediatric social-illness categories as well as the set of variables that discriminated families bearing the abuse diagnosis from families bearing the accident diagnosis. Before analysis could take place, each variable was transformed into standard scores (Z-scores).
The initial cluster analysis produced six cohesive clusters, containing 202 of the 209 families. The remaining families could not be placed in any specific cluster. We used a similar procedure on the second half-sample; 199 of 209 families fell into one of six clusters. Even though six clusters were formed for each subsample, the primary test for validity was whether we discovered similar clusters across subsamples.
We designed a threefold procedure to determine whether the same clusters appeared across both subsets. First, to be paired, two clusters had to distribute identically, within sampling fluctuation, over the five diagnostic outcomes. Second, a comparison of mean values for two clusters across the 63 variables had to yield few statistical differences. Third, the two clusters had to be relatively similar in size.
Two clusters from the original and reserve half-samples paired handsomely. Their distributions on diagnoses were very close; means did not significantly differ on 59 of the 63 cluster variables, and they were of similar sizes (90 and 92). This group of 182, which turned out to be at the lowest risk for abuse or failure to thrive, we called the ecologic advantage group.
The clusters that formed our group at highest risk for abuse or failure to thrive also paired well, with one exception. They, too, were distributed very similarly over the diagnoses, but with very low counts for most diagnoses, and their means did not differ on 56 out of the 63 cluster variables. Sample sizes, however, were statistically discrepant. The cluster from the original sample included eight families, whereas the cluster from the reserve sample included 18. Despite the discrepancy in subcluster size, this group became our ecologic crisis group.
A third subgroup was formed by combining one group from our original and two from the reserve sample. This cluster was somewhat less homogenous than the others, though both samples were similar with respect to their distributions across categories of pediatric social illness. These 112 cases constituted our ecologic adversity cluster.
FIGURE 1 shows histograms of diagnoses for each of the three risk groups. Note the distribution of abuse and failure to thrive among the three groups. While 35% of families in the ecologic crisis cluster were diagnosed as abusive, only 7% of families in the adversity and advantage clusters were so diagnosed. Thus, abuse among the families in ecologic crisis was five times as prevalent as among the other two groups combined. While 23% of children in the crisis cluster were diagnosed as failure to thrive, 13% of families in the other two groups were so diagnosed. Thus, failure to thrive among families in ecologic crisis was about 1.5 times as prevalent as among the ecologic adversity and advantage groups.
The picture is reversed for accidents and acute medical conditions. Only 8% of children in the crisis group were classified as accidents, whereas 37% of the children in the adversity and advantage groups were identified as accidents, about 4.5 times as many as in the crisis cluster. Acute medical conditions were 2.3 times as prevalent in the adversity and advantage groups as in the crisis group.
Group Profiles Analyses
For each of the 63 variables that defined interfamilial distance and (taken together) formed the three clusters, we executed a one-way analysis of variance (ANOVA) followed by a Duncan Multiple Range Test. The F-test and R-square from the ANOVA, shown on TABLE 5, demonstrated how well each variable differentiated the three groups. The Duncan Test answered a more refined question: for the variable under study and for each of the three possible pairings of two groups, did the two groups differ at or below the 0.05 level of significance?6, 16
VARIABLE |
F-VALUE
|
R2
|
Per capita income |
9.4***
|
.055
|
Welfare |
23.8***
|
.130
|
Mother has own car |
30.7***
|
.162
|
Mother describes childhood as happy |
12.8***
|
.074
|
Maternal grandmother ever hit elsewhere than bottom or hand |
38.8***
|
.196
|
Maternal grandmother’s discipline appropriate |
39.2***
|
.198
|
Maternal grandparents agreed on child rearing |
12.8***
|
.074
|
Mother was hit into adolescence |
31.9***
|
.167
|
Mother’s education |
19.5***
|
.109
|
Mother’s occupation |
33.0***
|
.172
|
Mother’s health problems |
40.5***
|
.203
|
Mother just can’t get going |
11.6***
|
.068
|
Hours mother watches TV |
21.7***
|
.120
|
Mother feels no one interested in problems |
3.7*
|
.022
|
Mother describes living situation as happy |
7.3***
|
.044
|
Recent death stressful |
7.1**
|
.042
|
Marital status |
40.2***
|
.202
|
Male present |
19.3***
|
.108
|
Mother’s years single |
7.4**
|
.044
|
Mother’s number of relatives can count on |
34.9***
|
.180
|
Mother’s ease to bus |
3.9*
|
.023
|
Child’s health |
11.4***
|
.067
|
Vineland social quotient |
22.7***
|
.125
|
Hematocrit |
3.9*
|
.024
|
Below age of self-help in general |
10.3***
|
.060
|
Below age of communication |
4.5*
|
.027
|
Below age of socialization |
12.7***
|
.074
|
* p<.05; ** p<.01; *** p<.001. |
For any set of three tests of a variable based on three pairings, the probability of not erroneously declaring a random difference significant is 0.95. Across the entire set of 189 Duncan tests, however, this probability drops to 0.04. Therefore, it is reasonably certain that at least one of our findings is wrong. But the probability that no more than seven significant results are false is 0.99. On the basis of these error rates, we may say that a few of the differences set forth below are likely to be spurious, but the great majority of them are not.
HIGH-RISK FAMILY PROFILE
Family Wealth
Whether family wealth is measured by income or by welfare dependency, families in the ecologic crisis cluster were poor. Even though families in the other clusters were by no means affluent, their mean family incomes were more than twice the mean family income of the crisis cluster. Data with respect to welfare dependency reinforced this income picture. Seven of eight families in the crisis cluster received public assistance, but only one out of three in the advantaged families and one out of six families in the adversity cluster received welfare.
Mother’s Childhood
The parents of mothers in families in ecologic crisis often disagreed about child rearing, according to the mother. These mothers remembered being disciplined more severely than mothers in the other groups. They were more often hit with objects, and on parts of the body other than hands or buttocks. In many cases physical discipline continued into adolescence. When asked to summarize their childhoods, mothers in the crisis group gave a neutral assessment, in contrast to other mothers who felt that they were “somewhat happy” as children.
Mother’s Attainments
The more skilled the mother’s occupation, the less risk there appeared to be that her child would be diagnosed as abused or failing to thrive. In order from highest to lowest risk groups, the mothers’ occupations were: unskilled or semi-skilled labor, skilled labor, and clerical work. This pattern was repeated for education. Among families in crisis, mothers had rarely finished high school. More education was associated significantly with a more favored ecologic state.
Mother’s State
Mothers in the crisis cluster reported minor to moderate health problems while other mothers voiced few health complaints. In addition to health problems, many high-risk mothers sometimes had difficulty in getting started in the morning, as suggested in their reports of watching an average of 7Y2 hours of television per day. Mothers in the other adversity and advantage clusters watched 3 Y2 and 3 hours of television per day respectively.
Mothers in the crisis cluster appeared to be more beset by physical and psychological problems. They were also more likely to feel, at least some of the time, that no one cared what happened to them. In addition, only slightly more than half described themselves as happy, while over 80% of mothers in the lower-risk groups felt they were happy.
Mother’s Relationships
In almost all families in crisis, the mothers were single, separated, or divorced. This absence of a partner diminished as advantage increased; the married state differentiated significantly among the groups. In homes of crisis, there was either no man or a man who did not always live there. More often than not, advantaged mothers were living with a man to whom, most frequently, they were married.
The data also suggest that the more kin one has available, the greater the ecologic advantage.
Mothers in the crisis cluster claimed that they spanked no more than mothers in the other two groups. These mothers, however, found their children more difficult to manage than lower-risk mothers.
Child’s State
The evidence of the child’s state was taken from the mother’s impression after admission of the child to the hospital as a patient. Her interview response might have been influenced by her reactions to doctors, nurses, or social workers who had conferred with her about the child.
Mothers in the crisis group appraised their child’s usual health as less favorable than mothers in the other groups. The Vineland Social Maturity Score was also calculated from information given by the mother. The more advantaged children enjoyed higher scores.
The hematocrits (percentage of red blood cells in centrifugal whole blood) of children in the crisis cluster were slightly lower than the hematocrits of children in the other two groups, but they were not so low that these children, as a group, could be called anemic. Children in the three groups reacted similarly to visual and auditory changes, and they were similarly distractable during fussy periods.
Father’s Background
Though most fathers of families in ecologic crisis were raised in rural locations, most fathers in the two other groups were raised in urban locations. In addition, slightly higher rates of broken marriage were found between parents of fathers in families in crisis.
DISCUSSION
Children admitted to a pediatric hospital with diagnoses of abuse, accidents, ingestions, and nonorganic failure to thrive differed from children with comparably acute medical conditions not only on the basis of presenting symptomatology but also on the basis of a variety of ecological characteristics. Regardless of the particular assigned diagnosis, families of children with a social illness appeared on the whole to be experiencing more stresses in their lives than the families of children with nonfatal acute medical conditions.
Cluster analysis confirmed our notion that abusing families were not characterized by a distinct constellation of variables that set them apart totally from families of children with other diagnoses. Indeed, cases of abuse, like cases of nonorganic failure to thrive, appear in all three of the clusters-ecologic advantage, ecologic adversity, and ecologic crisis-that emerged, although in very different proportions. Information in depth on the children’s family circumstances is essential to define the nature and extent of an individual child’s vulnerability and risk.
Our profile of high-risk families in certain vital areas of their lives-family wealth, mother’s childhood, mother’s attainments, mother’s state, mother’s relationships, child’s state, and father’s background-revealed that, in nearly every domain, the crisis cluster appears to suffer. Whether or not abuse or failure to thrive had occurred in these families, they usually had woefully inadequate financial means and major problems in intimate relationships. The mother’s entire life may have been impoverished and punctuated by violence and abandonment.
The three empirically derived clusters identified in this study provide a matrix for organizing data from families whose children suffer pediatric social illnesses. They do not imply judgment of a parent’s adequacy but, rather, they focus on specific aspects of family functioning in the life setting. We believe practice will improve when, with compassion and respect, clinicians can systematically identify and address the strengths and weaknesses of parents, children, and nurturing environments.
REFERENCES
1. BELSKY, J. 1980. Child maltreatment: an ecological integration. Amer. Psycho!. 35:320-335.
2. BRONFENBRENNER, U. 1979. The Ecology of Human Development. Harvard University Press, Cambridge.
3. BURGESS, R. AND CONGER, R. 1978. Family interaction in abusive, neglectful, and normal families. Child Devlpm. 49: 163-173.
4. CAREY, W. 1970. A simplified method for measuring infant temperament. J. Pediat. 188194.
5. DANIEL,J., HAMPTON, R. AND NEWBERGER, E. 1983. Child abuse and childhood accidents in black families: a controlled comparative study. Amer. J. Orthopsychiat. 53:645-653.
6. DUNCAN, D. 1975. T-tests and intervals for comparison suggested by the data. Biometrics 31:339-359.
7. FRIEDRICH, W. AND EINBENDER, A. 1983. The abused child: a psychological review. J. Clin. Child Psychol. 12:244-256.
8. GARBARINO, J. 1977. The human ecology of child maltreatment: a conceptual model for research. J. Marr. Fam. 39:721-736.
9. GARBARINO, J. 1982. Children and Families in the Social Environment. Aldine, New York.
10. GARBARINO, J. AND CROUTER, A. 1978. Defining the community context for parent-child relations: the correlates of child maltreatment. Child Devlpm. 49:604-616.
11. GELLES, R. 1979. Community agencies and child abuse: labeling and gatekeeping. In Family Violence, R. Gelles, ed. Sage Publications, Beverly Hills, Calif.
12. GELLES, R. 1980. Violence in the family: a review of research in the seventies. J. Marr. Fam. 42:875-885.
13. GELLES, R. 1982. Child abuse and family violence: implications for medical professionals. In Child Abuse, E. Newberger, ed. Little, Brown, Boston.
14. HAMPTON, R. AND NEWBERGER, E. 1985. Child abuse incidents and reporting by hospitals: significance of severity’, class and race. Amer. J. Pub. Hlth 75:56-60.
15. HOLLINGSHEAD, A. AND REDLICH, F. 1958. Social Class and Mental Illness. John Wiley, New York.
16. MILLER, R. 1981. Simultaneous Statistical Inference. Springer-Verlag, New York.
17. NEWBERGER, E. AND DANIEL, J. 1976. Knowledge and epidemiology of child abuse: a critical review of concepts: Pediat. Ann. 5: 140-144.
18. NEWBERGER, E. ET AL. 1977. Pediatric social illness: toward an etiologic classification. Pediatrics 60: 178-185.
19. NEWBERGER, E., NEWBERGER, C. AND HAMPTON, R. 1983. Child abuse: the current theory base and future research needs. J. Amer. Acad. Child Psychiat. 22:262-268.
20. O’TOOLE, R., TURBETT, P. AND NALEPKA, c. 1983. Theories, professional knowledge and diagnosis of child abuse. In The Dark Side of Families: Current Family Violence Research, D. Finkelhor, et al, eds. Sage Publications, Beverly Hills, Calif.
21. SEARS, R., MACCOBY, E. AND LEVIN, H. 1957. Patterns of Child Rearing. Row and Peterson, Evanston, Ill.
22. STRAUS, M., GELLES, R. AND STEINMETZ, S. 1980. Behind Closed Doors: Violence in the American Family. Doubleday, New York.
23. THOMAS, A., CHESS, S. AND BIRCH, H. 1978. Temperament and Behavior Disorders in Children. New York University Press, New York.
24. WASSERMAN, G., GREEN, A. AND ALLEN, R. 1983. Going beyond abuse: maladaptive patterns of interaction in abusing mother-infant pairs. J. Amer. Acad. Child Psychiat. 22:245-252.
A revised version of a paper submitted to the Journal in July 1985. Research was supported by grants from the Administration for Children, Youth and Families (OCD-CB-141) and from the National Institute of Mental Health (TO1 MH155117).