Child Abuse & Neglect, Vol. 23, No. 4, pp.

305–319, 1999
Copyright © 1999 Elsevier Science Ltd
Pergamon Printed in the USA. All rights reserved
0145-2134/99/$–see front matter

PII S0145-2134(99)00003-4

PREDICTING CHILD MALTREATMENT IN THE FIRST 4

YEARS OF LIFE FROM CHARACTERISTICS ASSESSED
IN THE NEONATAL PERIOD

JONATHAN B. KOTCH, DOROTHY C. BROWNE, VINCENT DUFORT, AND JANE WINSOR

Department of Maternal and Child Health, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

DIANE CATELLIER
Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

ABSTRACT

Objective: To examine the significance of neonatal risk factors from the individual, family, social, and parenting behavior
domains of the ecological model of child maltreatment in predicting maltreatment reports in the first 4 years of life, and to
examine the extent to which the interactions of life event stress and social support modify those risk factors.
Method: Mothers of 708 predominantly at-risk infants were interviewed in their homes soon after their infants’ discharge
from the hospital. State child abuse and neglect central registry data were tracked every 6 months until the infants reached
their fourth birthdays.
Results: The incidence of maltreatment reports was higher in households where the mothers were depressed, complained
of psychosomatic symptoms, had not graduated from high school, consumed alcohol, participated in public income support
programs, cared for more than one dependent child, or were separated from their own mothers at age 14 years (p , .1). In
interaction models including these seven predisposing variables, there were significant interactions (p , .01) between social
support, as measured by the social well-being index after the birth of the index child, and depression, and between social
well-being and stress, as measured by an increase in total life events.
Conclusion: Some predisposing risk factors measured soon after birth continue to be significant predictors of child
maltreatment reports through the fourth year of life. In general, families with low levels of social support had a higher risk
of a maltreatment report. For families with lower levels of maternal depression and/or life event stress, low social support
significantly increased the risk of a maltreatment report by as much as a factor of four. © 1999 Elsevier Science Ltd

Key Words—Child abuse, Child neglect, Life event change, Stress, Psychological, Social support.

INTRODUCTION

ATTEMPTING TO UNDERSTAND the causes and correlates of child maltreatment has been a
major concern of researchers, clinicians, and policy makers (Starr, MacLean, & Keating, 1991).

Supported in part by Grant #MCJ-37051 from the Maternal and Child Health Research Program, Bureau of Maternal and
Child Health, Health Resources and Services Administration, USDHHS to the Department of Maternal and Child Health,
University of North Carolina at Chapel Hill.
Supported in part by Grant #90-CA1467 from the Children’s Bureau, Administration for Children and Families, USDHHS,
to the Injury Prevention Research Center, University of North Carolina at Chapel Hill.
Received for publication March 9, 1998; final revision received October 16, 1998; accepted October 20, 1998.
Reprint requests should be addressed to Dr. Jonathan B. Kotch, Department of Maternal and Child Health, CB# 7400
Rosenau Hall, University of North Carolina, Chapel Hill, NC 27599-7400.
305
306 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

Unfortunately, little is known about the extent to which factors found to predict child maltreatment
at an early age are also predictors of child maltreatment later in life. For example, results from a
prospective study of pregnant, low-income women reported that mothers who abused or neglected
their children during the first year of life had had the following characteristics during their
pregnancies: two or more children, moved frequently, and denied having aggressive feelings
toward their children (Altemeier, Vietze, Sherrod, Sandler, Falsey, & O’Connor, 1979). Egeland
and Brunnquell (1979) also found strong associations between maternal antenatal variables and
postnatal outcomes. However, the predictive capacity of these characteristics, assessed in the
antenatal period, decreased at 2 years of age. One explanation for these findings is that the mothers
may have changed over time, and therefore the answers given to interview questions during the
antenatal period were no longer pertinent beyond their infants’ first year. Alternatively, mothers
who maltreat their infants in the first year of life may differ from those who commit abuse or
neglect during the second year of life. Using the data from Altemeier and colleagues (1979),
Brayden, Altemeier, Tucker, Dietrich, and Vietze (1992) reported a decrease over time in the
relevance of those factors found to be important predictors of child maltreatment by l2 months of
age. They also found that stress and social support during pregnancy had neither main nor
interaction effects on reports of child maltreatment.
The primary aim of this paper is to increase the understanding of the etiology of maltreatment
by examining the extent to which characteristics identified soon after birth are risk factors for child
maltreatment during the first 4 years of life. The authors, in a study designed to test the ecological
model of child maltreatment (Kotch, Browne, Ringwalt, Stewart, Ruina, Holt, Lowman, & Jung,
1995), demonstrated that the following risk factors were predictive of reports of child maltreatment
during the first year of life: receipt of Medicaid, low maternal education, the presence of any other
dependent children in the home, maternal depression, and mother’s separation as a young adoles-
cent from her own mother.
In addition, unlike the findings of Brayden and colleagues (1992), Kotch and colleagues (1995)
found that life events tended to increase or decrease the risk of maltreatment reports, depending
upon the presence of social support. Similar results were documented for the next 2 years of life
(Kotch, Browne, Ringwalt, Dufort, Ruina, Stewart, & Jung, 1997). Therefore, a second aim of the
present study is to determine whether stressful life events increase the risk of a report, and if so,
whether social support mitigates the risk for yet another year of the child’s life. Previous studies
by other investigators of the role of social networks or social supports in the etiology of child
maltreatment (Coohey, 1996; Corse, Schmid, & Trickett, 1990; Hashima & Amato, 1994; Jennings,
Stagg, & Connors, 1991) are limited by small sample sizes, use of clinical populations, and cross
sectional designs. In this study the authors continue their prospective investigation of a population
followed since birth in order to assess the contributions of predisposing risk factors, life event stress
and social support assessed at birth to child maltreatment reports in the first 4 years of life. If the
early predictors of child maltreatment could be shown to continue to contribute to the risk of
maltreatment until the fourth birthday, then it might make sense to continue to provide services to
at-risk families at least that long.

METHODS

Study Sample

The subjects for this analysis were drawn from a sample of participants in a prospective
investigation which focused on determining the causes of child maltreatment in the context of the
ecological model. This project, referred to as the “Stress, Social Support, and Abuse and Neglect
in High Risk Infants Study” (SSS), recruited a cohort of 1,111 mother-infant dyads between late
 Child maltreatment in the first 4 years 307

1985 and early 1987 from North and South Carolina hospitals and health departments. Four out of
every five infants recruited had risk factors for developmental and medical problems which made
them eligible for North Carolina’s High Priority Infant Program (HPIP) or its South Carolina
equivalent. HPIP eligibility criteria included low birth weight (, 2000 grams), low maternal age
(, 14 years), congenital abnormalities, birth defects, and other significant medical or social
problems. In addition to the group eligible for HPIP, mothers between 14 and 17 years of age
(inclusive) or whose infants were between 2000 and 2499 grams (inclusive) were recruited into this
study. The remaining 20% of the cohort consisted of infants having none of the above character-
istics. Eight hundred and forty-two mothers were successfully interviewed at home an average of
7 weeks after their infants were discharged from the hospital. Eighty-three percent of the 842
mother-infant pairs met HPIP criteria.
The subjects for this study were limited to North Carolina mother-infant pairs because, when this
study began, South Carolina did not retain unsubstantiated maltreatment reports in its Central
Registry of Child Abuse and Neglect. Additional subjects eliminated from this analysis included
one member of twin sets (determined randomly) and those infants neither reported for maltreatment
nor known to be residing in North Carolina on their fourth birthday. Analyses were conducted to
determine whether the effective sample for this study, consisting of the remaining mother-infant
dyads, were comparable to the total sample of North Carolina mothers and infants.

Data Collection
The questionnaire was administered in the mothers’ homes by trained interviewers after both
mothers and their infants had been discharged from birth hospitals. The questionnaire contained
items hypothesized to be relevant to the ecological model of child maltreatment from the
individual, family, social, and parenting behavior domains (Howze & Kotch, l984). These items
included mother’s exposure to intrafamily violence as a child; mother’s mental health and
psychosomatic symptoms; mother’s report of her parents’ behaviors; mother’s self-esteem; moth-
er’s physical health; mother’s parenting beliefs; infant’s health status; infant’s temperament; social
support/social networks; life event stress; and everyday stress. In addition to these items, the
questionnaire contained items pertaining to marital status, household structure, receipt of cash and
in-kind public support, and respondent’s views about the safety and friendliness of, and her
satisfaction with, her neighborhood.
In the present study child maltreatment was assessed by regularly reviewing North Carolina’s
Central Registry of Child Abuse and Neglect. Access to the Central Registry was obtained with the
permission of the Director of the Division of Social Services of the State of North Carolina. The
study was approved by the Institutional Review Board for the Protection of Human Research
Subjects of the School of Public Health, University of North Carolina at Chapel Hill.

Instruments
A complete list of variables appears in Appendices A and B. Descriptions of the instruments used
to measure maternal depression, stress, and social support, the principal independent variables of
interest in this analysis, follow. For a more complete description of other measures used, see Kotch
and colleagues (1995).

Depression
Center for Epidemiologic Studies—Depression (CES-D). The CES-D (Radloff, 1977) is a self-
report measure of depressive symptomatology (or “state” depression). This 20-item questionnaire
asks respondents to rate the frequency and duration of certain feeling states during the week
previous to the presentation of the question. Devins and Orme (1985) in an extensive review
308 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

conclude that the CES-D has adequate psychometric properties. A cut-off point of 16 is correlated
well with clinically diagnosed depression (Weissman, Sholomskas, Pottenger, Prusoff, & Locke,
1977) and with symptoms of depression in unselected community groups (Myers & Weissman,
1980).

Stress
Life Experiences Survey (LES). The LES is a 57-item self-report measure of life cycle events that
have occurred in the past year. For each of 38 events designed for use with the general population,
respondents indicate whether or not they experienced the event and the extent to which they viewed
the event as positive or negative. Test-retest correlations for the instrument are moderate (Sarason,
Johnson, & Siegel, 1978). Some items required adaptation to the experiences of a low-income
population (Sarason [1978] standardized the instrument on a population of college students); for
example, foreclosure of a mortgage was replaced with eviction due to failure to pay rent, and the
seven-point scale was reduced to a five-point scale to simplify the instrument for this less
well-educated sample. Six possible scores (positive and negative events, positive and negative
change, total events, and total change) were derived from this one instrument.

Everyday Stressors Index (ESI). The ESI was developed to target common problems faced on a
daily basis by low-income mothers with young children. This 22-item index covers financial
concerns, role overload, parenting worries, employment problems, and interpersonal conflict. The
ESI has been significantly associated with both depressive and psychosomatic symptoms of
unmarried mothers and has good reliability (Hall, Williams, & Greenberg, 1985).

Social Support
Autonomy and Relatedness Inventory (ARI). The ARI assesses the quality of the mother’s rela-
tionship with the individual she has identified as her primary intimate. The instrument has eight
subscales (acceptance, autonomy, control, hostility, listening, detachment/rejection, relatedness,
and support) with internal consistency reliabilities in the .70 to .80 range (Schaefer & Edgerton,
1982). The cumulative score, used in the research reported here, was found to have a Cronbach’s
alpha of .75 by Hall (1983). The validity of the ARI is supported by moderate to strong [VSL1]
correlations with the Spanier Dyadic Adjustment Scale (Schaefer & Edgerton, 1982).

Social Well-being Index (SWI). The SWI is a 10-item measure assessing a person’s social
contacts/activities and her social resources. Through extensive psychometric analysis with a large
sample (N5 4603), Donald and Ware (1982) derived social contacts and group participation
subscales and five single item measures of social well-being. The two subscales had adequate
reliability and homogeneity coefficients. Construct validity studies showed that the social well-
being measures are distinct from other dimensions of health (e.g., mental health) and positively
related to each other, suggesting an underlying dimension (Donald & Ware, 1982). Only the social
contact scale score was used in the present study.

Social Network Index (SNI). The structural aspects of subjects’ social networks were measured by
items concerning marital status, numbers of close friends and relatives, frequency of contacts, and
church and group memberships. The scoring method developed by Berkman classifies respondents
into four groups ranging from few to many social ties. Using the SNI, researchers have predicted
mortality in a random sample of 6298 adults over a 9-year period (Berkman & Syme, 1979) and
have found a strong correlation with both depressive and psychosomatic symptoms among a group
of mothers with kindergarten-aged children (Hall, 1983).
 Child maltreatment in the first 4 years 309

Analysis

To explore the capacity of the individual, family, social and behavioral risk factors to predict
reports of child maltreatment during the first 4 years of life, several analytical steps were taken to
arrive at a final regression model. Each of the executed steps built upon the previous one. All
variables in the maternal, infant, family, and parenting domains were divided into “high” and “low”
categories on the basis of median splits. Stress and social support indicator variables were created
by splitting the population based on the most extreme tertile.
First, individual stepwise logistic regressions were performed within each domain (excluding the
stress and social support domains) to identify the most influential factors among the 54 predictors.
The logistic regression analysis was addressed through PROC LOGISTIC (SAS Institute, 1990).
Second, all the variables which were found to be related to maltreatment in the first step using an
inclusion criterion of 10% (p , .1) were combined with the stress and social support variables in
one logistic regression model. This model was reduced in a stepwise fashion.
Next, we investigated the interaction between stress and social support. Rather than allowing the
statistical procedures to automatically select the “strongest” among the possible stress and social
support variables and their interactions, we created 21 parallel logistic regression models, each
testing one of the seven stress variables, one of the three social support variables, their interaction,
and the significant predictors found in the second step. Within each set of logistic models involving
each of the three social support variables, only the most significant (p , .05) interaction (if any)
was considered for the next step. Fourth, the three chosen social support models, each including the
most significant stress and social support interaction and other predictors, were further reduced to
include only significant (p , .05) main effects and interactions. Finally, the “best” model among
the three reduced models was chosen by considering the simplicity of the model as well as its
overall explanatory capability based on a generalization of r2 used in classical regression analysis
(Negelkerke, 1991).
Given that the study sample included cases who could have been reported for maltreatment at
any time in their first 4 years and controls who were known still to be resident in North Carolina
on their fourth birthdays, there was a strong possibility of systematic bias due to the unknown
report status of those lost to follow-up. To test for such bias we repeated the analysis outlined above
using time to event as the response instead of the dichotomous “ever reported for maltreatment.”
Cox regression was used to examine the relationship between the same set of predictors and time
to maltreatment report using a proportional hazards model.

RESULTS

There were 708 mothers meeting the criteria for inclusion in the sample and whose infants were
born in North Carolina. Of these mothers, 442 (62%) were Black; their average age at the birth of
the index child was 22.8 years; 65% of the mothers had been unmarried at the time of birth; and
approximately 81% received Medicaid or other income support (see Table 1). No significant
differences were found in the distributions of age, sex, race, number of dependent children in the
home, marital status, receipt of income support, and maltreatment reports between the total sample
of 788 North Carolina mothers interviewed soon after their infants’ discharge from hospital and the
sample of 708 eligibles included in this analysis.
One hundred and seventy-two of the index children were reported for child maltreatment during
their first 4 years of life. Eighty three of these children were reported during their first year, 16 in
their second, 39 in the third year, and 34 in year 4 (see Figure 1). The sharp decline in reports in
the second year is attributable in part to the state’s failure to archive unsubstantiated reports during
this period. Many of the children were reported more than once. Of the initial report of each child,
310 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

Table1. Descriptive Statistics for the Study Population at Baseline

(N 5 708)
Mothers

N %
Mean Age (Years) 22.8
Education (Mean Grade Completed) 10.8
Nonwhite 442 62%
Not Married 463 65%
Incomea
,$5,000 241 34%
$5001–10,000 157 22%
$10,001–15,000 102 14%
$15,001–20,000 62 9%
$20,001–30,000 43 6%
$.30,000 38 5%
Missing 365 9%
Public Income Supportb
WIC 504 71%
Medicaid 333 47%
AFDC 234 33%
Food stamps 225 32%
None 137 19%
Missing 11 2%

Infants
Mean Birthweight (Grams) 2729
HPIP Eligible 605 86%
Sex 5 Male 344 49%
a
May not add to 100% due to rounding.
b
May be duplicated.

68 or 40% were substantiated. The total unduplicated count of children whose first or subsequent
reports were substantiated was 90 or 52% of the children ever reported.
There were a total of 290 reports involving these 172 children, of which 128 (44%) were
substantiated, 107 in the first year, 35 in the second, 60 in the third year, and 88 in the fourth.
Neglect alone constituted 241 (83%) of the total reports, followed by 27 (9%) for abuse alone, and
22 (8%) for abuse and neglect. In North Carolina the Central Registry does not further break down
abuse reports unless they are substantiated, so the exact number of reports of physical, sexual, and
emotional abuse is unknown. Only six of the 128 substantiated reports were for abuse alone, and
three were abuse and neglect.
Bivariate comparisons between maltreatment report status and baseline variables revealed
differences at the p , .1 level for 15 predisposing factors (see Table 2). Seven of these factors were
identified at the p , .1 level as predictors of a child maltreatment report in the main effect model
(4 maternal, 1 family, 1 social, and 1 parenting) (see Table 3). In the final regression models with
interaction terms, the variable for number of dependent children in the home was the only variable
of the seven significant main effect predictors that was not also significant in the interaction model.
The interactions between the Social Well-being Index (Donald & Ware, 1982) and depression
(Radloff, 1977), and between social well-being and total life event change (Sarason et al., 1978)
had significant associations (p , .01) with the reporting of child abuse and neglect (see Table 4).
In order to interpret the joint effect of the three variables involved in the interactions, we
examined the effects of maternal depression and life event change (comparing the high versus low
categories) on the risk of a maltreatment report within each of the two levels of social well-being.
As shown in Figure 2 for subjects at high levels of social well-being, a greater depression score was
 Child maltreatment in the first 4 years 311

Figure 1. Cumulative number of children reported for maltreatment, N 5 708.

significantly (p 5 .005) associated with a higher probability of being reported for child abuse and
neglect, with the corresponding odds ratio (OR) being 2.04 (95% CI: 1.2–3.4). However, at low

Table 2. Means and Standard Errors (SE) or Percentage Distributions for Variables Associated (p < 0.10) with
Maltreatment Report
No Report Report

Mean SE Mean SE

Maternal Characteristics
Completed education (years) 11.0 .10 10.1 .17
Psychosomatic symptoms (HOS scores) 27.7 .21 28.7 .37
Depression score 14.1 .44 18.8 .88
Alcohol use 0.88 .06 1.37 .14
Sexual abuse prior to age 18 28% 41%
Infant Characteristics
Health status score (1 5 excellent, 4 5 poor) 1.5 .03 1.6 .05
Activity limitation 11% 15%
Family Characteristics
Married 38% 26%
Two or more dependent children in the home 0.37 .02 0.49 .04
Social Characteristics
Poverty index 1.9 .07 2.9 .10
Warmth of neighborhood 1.4 .05 1.1 .08
Parenting Behavior
Mother present when respondent was age 14 90% 79%
Respondent accepted/rejected by mother at 14 3.9 .23 2.0 .43
Mother empathized with child’s needs 28 .25 26 .43
Witnessed severe violence in family as child 5.2 .31 6.2 .59
312 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

Table 3. Odds Ratios and Confidence Intervals for Variables Predicting Child Maltreatment Reports (N 5 694)
95% CI

Variable Odds Ratio Lower Upper

Social Well-being Index 0.72 0.48 1.08

Total Life Event Change Score 1.20 0.77 1.87
Health Opinion Survey* 1.60 1.05 2.44
Maternal Depression Score 1.41 0.93 2.14
Maternal Education $ 12 Years* 0.63 0.43 0.94
Any Alcohol Use** 1.77 1.18 2.67
Two or More Dependent Children in the Home* 1.52 1.03 2.26
Receiving Income Support*** 1.48 1.30 1.69
Presence of Mother at Respondent Age 14** 0.47 0.28 0.79

* p # 0.05, **p # 0.01, ***p # 0.001.

levels of social well-being, depression had no significant effect on reported maltreatment (see
Figure 3). Similarly, for the high social well-being group, high life events nearly doubled the
probability of a maltreatment report (p 5 .015, OR 5 1.9, 95% CI: 1.1–3.3), whereas at low levels
of social well-being, increasing stress had no significant effect (see Figure 3).
Finally, we also examined the effect of social well-being (high versus low) within each of four
depression by life event change categories. Low social well-being had its greatest effect on
maltreatment reports in the first 4 years of life for those reporting both low life event stress and low
depression at the postnatal interview (p , .001, OR 5 3.8, 95% CI: 2.0 –7.2). For the other three
combinations (high depression/low stress, low depression/high stress, and high depression/high
stress), social well-being had no significant effect.
As described in Methods, the above analysis was repeated with time to event as the outcome in
Cox regressions using proportional hazards models. The model parameter estimates and signifi-
cance levels differed very little from those of the final logistic models reported here.

DISCUSSION

This study examined the extent to which characteristics assessed in infancy are predictive of
child maltreatment during the first 4 years of life. A previous study conducted by these researchers

Table 4. Final Interaction Model with Total Life Event Change, Social Well-being Index, and Significant Predis-
posing Variables (N 5 694)
95% CI

Variable Odds Ratio Lower Upper

Social Well-being Index*** 0.26 0.14 0.50

Total Life Event Change Score 0.50 0.22 1.12
Health Opinion Survey* 1.62 1.05 2.49
Maternal Depression Score 0.68 0.33 1.39
Education $ 12 Years** 0.58 0.38 0.87
Any Alcohol Use** 1.83 1.20 2.77
Receiving Income Support*** 1.45 1.27 1.65
Presence of Mother at Respondent Age 14** 0.45 0.27 0.76
Social Well-being 3 Depression** See Figure 1
Social Well-being 3 Total Change** See Figure 2

* p # 0.05, **p # 0.01, ***p # 0.001.

 Child maltreatment in the first 4 years 313

Figure 2. Probability of maltreatment report according to levels of Social Well-being Index and depression.

(Kotch et al., l995) revealed that low maternal education, presence of other dependent children in
the home, receipt of Medicaid, maternal depression, and whether the maternal subject at age 14
years was separated from her own mother were the best predictors of a maltreatment report in the
first year of life. Further examination revealed that the interaction between stressful life events and
social well-being was also a significant predictor. Similar findings were reported by Kotch and
colleagues (1997) for the second and third years of life.
The present study also found that maternal education, maternal psychosocial problems such as
depression, psychosomatic symptoms, and alcohol use, participation in public support programs
such as Aid to Families with Dependent Children (AFDC), Medicaid, Food Stamps, or the Special
Supplemental Food Program for Women, Infants and Children (WIC), other dependent children in
the home, and whether the maternal subject at age 14 lived with her own mother were predictive
of child maltreatment reports up to the child’s fourth birthday. The roles of mother’s psychosomatic
symptom status and her use of alcoholic beverages, both as reported at the birth of her infant, are
new findings. Comparable data for fathers are lacking due to the preponderance of single parent,
female-headed households in the study population.
Another new finding is the interaction between social support, as measured by the Social
Well-being Index, and maternal depression. Both the interactions between social support on the one
hand and either depression or stress on the other offer further explanation of the roles of all three
in the etiology of maltreatment. In general it may be said that families with low social support are
at higher risk of abuse and neglect regardless of the level of stressful life events or depression. But
it is among the families with either lower levels of stress or depression that the impact of social
314 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

Figure 3. Probability of maltreatment report according to levers of Social Well-being Index and total change.

support is most apparent, reducing the risk of a maltreatment report by half in each case. For those
families who would otherwise be at the lowest risk by virtue of their low levels of stress and
depression, low social support increases the risk of a maltreatment report in the first 4 years of life
by nearly a factor of four.
It is remarkable that the same risk factors measured soon after birth continue to predict
maltreatment reports, and that new risk factors emerge as significant predictors 4 years later. In part
this may be a consequence of the increasing numbers of outcomes over time, increasing the
statistical power of our models. Given the fact that the single largest number of annual reports
occurred in the children’s first year of life it may be argued that the persistence of many of the same
risk factors is no surprise. On the other hand, by the end of the fourth year of life those first-year
reports represent fewer than half of all reports. In addition, the new finding that mother’s drinking
behavior and her psychosomatic symptoms predict maltreatment lends weight to the validity of the
overall models and to the contribution of reports subsequent to the first year.
Among several competing measures of social support, the significance of the Social Well-being
Index is that it measures the subject’s own assessment of the social resources available to her. This
is consistent with Thompson’s emphasis on the function of social support, as opposed to measures
of social networks (Thompson, 1995). Among the functions listed by Thompson are emotional
sustenance; counseling, advice or guidance; access to information, services, and material resources
and assistance; and skills acquisition (Thompson, 1994). In the case of families at risk of abuse or
neglect, the promising work of Olds and colleagues (1997) suggests that home visitors may
successfully provide some or all of these functions.
 Child maltreatment in the first 4 years 315

This study has a number of limitations. First, the population was overwhelmingly poor and at
higher risk of abuse and neglect due to social and medical problems. Overall, about 2% to 3% of
the newborns in North Carolina are reported in their first year of life. Kotch and colleagues (1995)
found that approximately 10% of this sample was reported for child maltreatment during the first
year of life. Approximately 24% of the 708 infants studied here were reported during their first 4
years of life. That this number of children was reported confirms the expectation that they are at
increased risk. Therefore, these results cannot be generalized to more representative populations of
children. Nevertheless, the risk factors documented here can be useful in the second of a two step
screening procedure such as that currently recommended for home visiting programs such as
Hawaii’s Healthy Start initiative (Fuddy, 1992). Selecting families for home visiting services using
indicators that have been demonstrated empirically to predict maltreatment in the first 4 years of
life would increase the chances of significantly reducing the risk of maltreatment.
Second, this study utilized all reports of child maltreatment rather than substantiated reports as
the dependent variable. Some researchers argue that there is a qualitative distinction between
substantiated and unsubstantiated reports. However, it has also been suggested that cases of child
maltreatment may represent a continuum which includes a number of unsubstantiated cases that, in
truth, should have been substantiated. In this population 21% of subjects whose first report was not
substantiated had a subsequent substantiated report in the first 4 years of life. By their eighth
birthday, that percentage had risen to 44%. In a previous report (Kotch et al., 1997), and in a series
of analyses paralleling those presented here but using substantiated reports as the outcome, our
results were essentially the same, any differences being attributable to fewer cases and less
statistical power.
Finally, a large proportion of this sample was known to the Department of Social Services as
recipients of Food Stamps, WIC, Medicaid, and/or AFDC. Given this fact, many families were
exposed to a system of formal and informal surveillance, increasing the likelihood that they would
be reported. On the other hand, univariate statistics do not show significant differences in reporting
according to whether or not subjects had Medicaid, Food Stamps, WIC, or AFDC.
Despite these limitations the present study suggests some strategies for the prevention of child
maltreatment. The variables that are predictive of child maltreatment are those that suggest that
amelioration of the social circumstances of poor families can prevent maltreatment reports. Such
system changes include the provision of educational opportunities and the elimination of the
adverse consequences of poverty. Additionally, the risk factors of depression, psychosomatic
symptoms, and drinking suggest the need for prevention services that address depression and
problems associated with lack of self-esteem. Finally, social support and life event stress play
significant roles. For women without social support from friends, neighbors, and relatives, com-
munity-based support systems, and social services emphasizing prevention must be developed.

Acknowledgement—The authors wish to thank Carol Cuatrecasas, Ellen Ruina, and Mary Harris for their contributions to
the research.

REFERENCES

Altemeier, W. A., Vietze, P., Sherrod, K. B., Sandler, H. M., Falsey, S., & O’Connor, S. (1979). Prediction of child
maltreatment during pregnancy. Journal of the American Academy of Child Psychiatry, 18, 205–218.
Bates, J., Freeland, C., & Lounsbury, M. L. (1979). Measurement of infant difficultness. Child Development, 50, 794 – 803.
Bavolek, S. (1984). Handbook of the Adult-Adolescent Parenting Inventory. Schaumburg, IL: Family Development
Associates.
Berkman, L. F., & Syme, S. L. (1979). Social networks, host resistance, and mortality: A 9-year follow-up study of Alameda
County residents. American Journal of Epidemiology, 109, 186 –204.
Brayden, R. M., Altemeier, W. A., Tucker, D. D., Dietrich, M. S., & Vietze, P. (1992). Antecedents of child neglect in the
first two years of life. Journal of Pediatrics, 120, 426 – 429.
Coohey, C. (1996). Child maltreatment: Testing the social isolation hypothesis. Child Abuse & Neglect, 20, 241–254.
316 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier

Corse, S. J., Schmid, K., & Trickett, P. K. (1990). Social network characteristics of mothers in abusing and nonabusing
families and their relationships to parenting beliefs. Journal of Community Psychology, 18, 44 –59.
Devins, G. M., & Orme, C. M. (1985). Center for epidemiological studies depression scale. In D. J. Keyser & R. C.
Sweetland (Eds.), Test critiques, II (pp. 144 –160). Kansas City, MO: Test Corporation of America.
Donald, C., & Ware, J. (1982). The quantification of social contacts. Santa Monica, CA: The Rand Corporation.
Egeland, B., & Brunnquell, B. A. (1979). An at-risk approach to the study of child abuse. Journal of the American Academy
of Child Psychiatry, 18, 219 –235.
Evans, G. A. (1988). Childhood sexual trauma and psychological distress in a postpartum population [dissertation]. Chapel
Hill, NC: University of North Carolina at Chapel Hill.
Fuddy, L. (1992). Hawaii’s Healthy Start’s success shared at the Ninth International Congress on Child Abuse and Neglect.
Unpublished paper.
Hall, L. (1983). Social supports, everyday stressors, and maternal mental health [dissertation]. Chapel Hill, NC: University
of North Carolina at Chapel Hill.
Hall, L., Williams, C. A., & Greenberg, R. S. (1985). Supports, stresses, and depressive symptoms in mothers of young
children. American Journal of Public Health, 75, 518 –521.
Hashima, P., & Amato, P. R. (1994). Poverty, social support, and parental behavior. Child Development, 65, 394 – 403.
Howze, D. C., & Kotch, J. B. (1984). Disentangling life events, stress and social support: Implications for the primary
prevention of child abuse and neglect. Child Abuse & Neglect, 8, 401– 409.
Jennings, K. D., Stagg, V., & Connors, R. E. (1991). Social networks and mothers’ interactions with their preschool
children. Child Development, 62, 966 –978.
Kotch, J. B., Browne, D. C., Ringwalt, C. L., Stewart, P. W., Ruina, E., Holt, K., Lowman, B., & Jung, J.-W. (1995). Risk
of child abuse or neglect in a cohort of low-income children. Child Abuse & Neglect, 19, 1115–1130.
Kotch, J. B., Browne, D. C., Ringwalt, C. L., Dufort, V., Ruina, E., Stewart, P. W., & Jung, J.-W. (1997). Stress, social
support, and substantiated maltreatment in the second and third years of life. Child Abuse & Neglect, 21, 1025–1037.
MacMillan, A. M. (1957). The health opinion survey: Techniques for measuring communities. Psychological Reports, 3,
325–339.
Myers, J. K., & Weissman, M. M. (1980). Use of self-report symptom scale to detect depression in a community sample.
American Journal of Psychiatry, 137, 1081–1083.
National Center for Health Statistics, Centers for Disease Control, Public Health Service, US Department of Health and
Human Services. (1990). Questionnaires from the National Health Interview Survey, 1980 –1984, PHS 90 –1302.
Hyattsville, MD: Author.
Negelkerke, N. J. D. (1991). A note on a general definition of the coefficient of determination. Biometrika, 78, 691– 692.
Newberger, E. H., Marx, T. J., White, K. M., & Hampton, R. L. (1983). Child abuse and pediatric social illness: An
ecological reformulation. Research paper prepared with support from the administration for Children, Youth and Families
(OCD-CB-141) and the National Institute of Mental Health (T01 MHI551176 – 03CD). Washington, DC: US Department
of Health and Human Services.
Olds, D. L., Eckenrode, J., Henderson, C. R., Jr., Kitzman, H., Powers, J., Cole, R., Sidora, K., Morris, P., Pettitt, L. M.,
& Luckey, D. (1997). Long-term effects of home visitation on maternal life course and child abuse and neglect. Journal
of the American Medical Association, 278, 637– 643.
Radloff, L. F. (1977). Sex differences in depression: The effects of occupation and marital status. Sex Roles, 1, 249 –265.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.
Sarason, J., Johnson, J., & Siegel, J. (1978). Assessing the impact of life changes: Development of the life experiences
survey. Journal of Consulting and Clinical Psychology, 46, 932–946.
SAS Institute, Inc. (1990). SAS/STAT user’s guide (Vol. 2, Ver. 6). Cary, NC: SAS Institute Inc.
Schaefer, E. S., & Bayley, N. (1967). Validity and consistency of mother-infant observations, adolescent maternal
interviews, and adult retrospective reports of maternal behavior. Proceedings of the 75th Annual Convention of the
American Psychological Association, 2 (pp. 147–148). Washington, DC: American Psychological Association.
Schaefer, E. S., & Edgerton, M. (1982). The autonomy and relatedness inventory (ARI). Chapel Hill, NC: Department of
Maternal and Child Health, University of North Carolina at Chapel Hill.
Starr, R. H., Jr., MacLean, D. J., & Keating, D. (1991). Life span developmental outcomes of child effects of child abuse
and neglect maltreatment. In R. H. Starr, Jr., & D. A. Wolfe (Eds.), The effects of child abuse and neglect (pp. 1–32).
New York: The Guilford Press.
Straus, M. (1979). Measuring intrafamily conflict and violence: The Conflict Tactics Scale. Journal of Marriage and the
Family, 41, 75– 88.
Thompson, R. A. (1994). Social support. In G. B. Melton, & F. D. Barry (Eds.), Protecting children from abuse and neglect
(pp. 40 –130). New York: The Guilford Press.
Thompson, R. A. (1995). Preventing child maltreatment through social support. Thousand Oaks, CA: Sage.
Weissman, M. M., Sholomskas, D., Pottenger, M., Prusoff, B. A., & Locke, B. A. (1977). Assessing depressive symptoms
in five psychiatric populations: A validation study. American Journal of Epidemiology, 106, 203–214.

RÉSUMÉ

Objectif: Examiner l’importance des facteurs de risque néonatals au niveau individuel, social, familial et du comportement
parental, dans une perspective écologique de la maltraitance. Ceci pour aider à prédire s’il y aura maltraitance au cours des
 Child maltreatment in the first 4 years 317

quatre premières années de la vie, et pour examiner dans quelle mesure les interactions qui se déroulent dans le contexte
des événements stressants de la vie et les appuis sociaux peuvent modifier ces facteurs de risque.
Méthode: On a interviewé 708 mères dont les enfants étaient considérés à risque élevé. Les entrevues ont eu lieu au foyer,
peu après la rentrée de l’hôpital. On a suivi les registres de mauvais traitements et de négligence à tous les six mois jusqu’à
ce que les enfants aient atteint l’âge de quatre ans.
Résultats: L’incidence des signalements de mauvais traitements étaient plus élevée dans les foyers où la mère était
déprimée, se plaignait de symptômes psychosomatiques, n’avait pas complété l’école secondaire, prenait de l’alcool, faisait
partie d’un groupe d’appui financé par l’état, avait plus d’un enfant à sa charge ou avait été séparée de sa propre mère à
l’âge de 14 ans. Une analyse de ces sept variables indiquent des liens importants entre, d’une part, l’appui social (tel que
mesuré selon l’indice de bien-être après la naissance) et la dépression; puis entre le bien être et le stress, ceci à la suite d’une
hausse des événements quotidiens stressants.
Conclusions: Certains facteurs de risque notés peu après la naissance demeurent des facteurs importants capables de prédire
les mauvais traitements au cours des quatre années qui suivront. En général, les familles qui jouissent de peu d’appuis
sociaux sont les plus souvent signalées aux autorités pour cause de maltraitance. Pour les familles qui ont des niveaux de
dépression maternelle inférieure et/ou qui connaissent peu de stress dans leur vie, l’absence d’appuis sociaux augmente
jusqu’à quatre fois le risque d’un signalement.

RESUMEN

Spanish abstract not available at time of publication.

APPENDIX A

Predisposing Variables Included in the Analysis, by Domain

Variable Measure

Mother

Age Maternal Report

Marital Status Maternal Report
Race Maternal Report
Completed Education Maternal Report
Psychosomatic Symptoms Health Opinion Survey (HOS; Macmillan, 1957)
Self-esteem Self Esteem Scale (Rosenberg, 1965)
Depression Center for Epidemiologic Studies-Depression (Radloff, 1977)
Sexual Abuse Prior to Age 18 Years Sexual experiences while growing up (Evans, 1988)
Self-report of General Health Status National Health Interview Survey (National Center for Health
Statistics, Centers for Disease Control, Public Health Service,
US Department of Health and Human Services, 1990)
Self-report of Activity Limitations National Health Interview Survey (National Center for Health
Statistics, Centers for Disease Control, Public Health Service,
US Department of Health and Human Services, 1990)
Type and Amount of Alcohol Use Maternal Report
Self-reported Disability National Health Interview Survey (National Center for Health
Statistics, Centers for Disease Control, Public Health Service,
US Department of Health and Human Services, 1990)
Pregnancy with Index Child Planned Maternal Report
Happy with Pregnancy with Index Child Maternal Report
Infant

Birthweight Medical records

Fussy Temperament Infant Characteristics Questionnaire (Bates, Freeland, &
Lounsbury, 1979)
Unpredictable Temperament Infant Characteristics Questionnaire (Bates et al., 1979)
(Continued)
318 J. B. Kotch, D. C. Browne, V. Dufort, J. Winsor, and D. Catellier
APPENDIX A (continued)
Variable
Unresponsive Temperament
Mother’s Report of Infant’s General Health
Status
Any Activity Limitation Due to Illness or Injury
Days in Which Normal Activities Were Limited
Due to Illness or Injury
Neonatal Intensive Care Days
Any Assisted Ventilation
Any Medical Diagnoses on Discharge Summary
Family
Number of Other Child Dependents in the Home
Number of Related Adults in the Home Other
Than Index Child’s Mother or Father
Social
On an Income Support Program (AFDC,
Medicaid, WIC, Food Stamps)
Satisfaction with, and Safety and Friendliness of,
Neighborhood
Parenting Behavior
Assuming Appropriate Parent–Child Roles
Rejection of Corporal Punishment
Mother Empathizes with Child’s Needs
Appropriate Expectations for Child Development
Mother-Figure Present When Respondent was 14
Years Old
Accepted/Rejected by Mother-Figure When
Respondent was 14 Years Old
Father-Figure Present When Respondent was 14
Years Old
Accepted/Rejected by Father-Figure When
Respondent was 14
Victim of Severe Physical Violence as a Child
Witnessed Severe Physical Violence in Family
as a Child
Victim of Minor Physical Violence as a Child
Witnessed Minor Physical Violence in Family as
a Child
Victim of Verbal Aggression as a Child
Witnessed Verbal Aggression in Family as a
Child
Adults in Mother’s Family Used Reasoning to
Solve Problems with Each Other
Adults in Mother’s Family Used Reasoning to
Solve Problems with Respondent
Measure
Infant Characteristics Questionnaire (Bates et al., 1979)
National Health Interview Survey (National Center for
Health Statistics, Centers for Disease Control, Public
Health Service, US Department of Health and
Human Services, 1990)
National Health Interview Survey (National Center for
Health Statistics, Centers for Disease Control, Public
Health Service, US Department of Health and
Human Services, 1990)
National Health Interview Survey (National Center for
Health Statistics, Centers for Disease Control, Public
Health Service, US Department of Health and
Human Services, 1990)
Medical records
Medical records
Medical records
Maternal report
Maternal report
Maternal report
Based on the Maternal Inventory (Newberger, Marx,
White & Hampton, 1983)
Adult-Adolescent Parenting Inventory (Bavolek, 1984)
Adult-Adolescent Parenting Inventory (Bavolek, 1984)
Adult-Adolescent Parenting Inventory (Bavolek, 1984)
Adult-Adolescent Parenting Inventory (Bavolek, 1984)
Retrospective Report of Parental Behavior (Schaefer &
Bayley, 1967)
Retrospective Report of Parental Behavior (Schaefer &
Bayley, 1967)
Retrospective Report of Parental Behavior (Schaefer &
Bayley, 1967)
Retrospective Report of Parental Behavior (Schaefer &
Bayley, 1967)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
Conflict Tactics Scale (Straus, 1979, modified)
 Child maltreatment in the first 4 years 319

APPENDIX B

Stress and Social Support Variables

Variable Measure

Social Support

Social Support from Significant Other Autonomy and Relatedness Inventory (ARI; Schaefer & Edgerton, 1982)
Social Networks Social Network Index (SNI; Berkman & Syme, 1979)
Social Well-being Social Well-being Index (SWI; Donald & Ware, 1982)
Stress

Total Number of Life Events
Total Change in Life Event Score
Number of Negative Life Events
Total Negative Change Score
Number of Positive Life Events
Total Positive Change Score
Chronic Stress
Life Experiences Survey (Sarason, Johnson, & Siegel, 1978)
Life Experiences Survey (Sarason et al., 1978)
Life Experiences Survey (Sarason et al., 1978)
Life Experiences Survey (Sarason et al., 1978)
Life Experiences Survey (Sarason et al., 1978)
Life Experiences Survey (Sarason et al., 1978)
Everyday Stressors (Hall, Williams, & Greenberg, 1985)