Olds D, Henderson, Jr CR, Cole R, Eckenrode J, Kitzman H, Luckey D, Pettitt L, Sidora K, Morris P, Powers J. Long-term Effects of Nurse Home Visitation on Children's Criminal and Antisocial Behavior15-Year Follow-up of a Randomized Controlled Trial. JAMA. 1998;280(14):1238-1244. doi:10.1001/jama.280.14.1238
From the University of Colorado Health Sciences Center, Denver (Drs Olds and Luckey); Cornell University, Ithaca, NY (Mr Henderson, Drs Eckenrode and Powers, and Ms Morris); the University of Rochester, Rochester, NY (Drs Cole and Kitzman and Ms Sidora); and the University of Denver (Dr Pettitt).
Context.— A program of home visitation by nurses has been shown to affect the
rates of maternal welfare dependence, criminality, problems due to use of
substances, and child abuse and neglect. However, the long-term effects of
this program on children's antisocial behavior have not been examined.
Objective.— To examine the long-term effects of a program of prenatal and early
childhood home visitation by nurses on children's antisocial behavior.
Design.— Fifteen-year follow-up of a randomized trial. Interviews were conducted
with the adolescents and their biological mothers or custodial parents.
Setting.— Semirural community in New York.
Participants.— Between April 1978 and September 1980, 500 consecutive pregnant women
with no previous live births were recruited, and 400 were enrolled. A total
of 315 adolescent offspring participated in a follow-up study when they were
15 years old; 280 (89%) were born to white mothers, 195 (62%) to unmarried
mothers, 151 (48%) to mothers younger than 19 years, and 186 (59%) to mothers
from households of low socioeconomic status at the time of registration during
Intervention.— Families in the groups that received home visits had an average of 9
(range, 0-16) home visits during pregnancy and 23 (range, 0-59) home visits
from birth through the child's second birthday. The control groups received
standard prenatal and well-child care in a clinic.
Main Outcome Measures.— Children's self-reports of running away, arrests, convictions, being
sentenced to youth corrections, initiation of sexual intercourse, number of
sex partners, and use of illegal substances; school records of suspensions;
teachers' reports of children's disruptive behavior in school; and parents'
reports of the children's arrests and behavioral problems related to the children's
use of alcohol and other drugs.
Results.— Adolescents born to women who received nurse visits during pregnancy
and postnatally and who were unmarried and from households of low socioeconomic
status (risk factors for antisocial behavior), in contrast with those in the
comparison groups, reported fewer instances (incidence) of running away (0.24
vs 0.60; P=.003), fewer arrests (0.20 vs 0.45; P=.03), fewer convictions and violations of probation (0.09
vs 0.47; P<.001), fewer lifetime sex partners
(0.92 vs 2.48; P=.003), fewer cigarettes smoked per
day (1.50 vs 2.50; P=.10), and fewer days having
consumed alcohol in the last 6 months (1.09 vs 2.49; P=.03).
Parents of nurse-visited children reported that their children had fewer behavioral
problems related to use of alcohol and other drugs (0.15 vs 0.34; P=.08). There were no program effects on other behavioral problems.
Conclusions.— This program of prenatal and early childhood home visitation by nurses
can reduce reported serious antisocial behavior and emergent use of substances
on the part of adolescents born into high-risk families.
JUVENILE CRIME is a significant problem in the United States. In 1996,
law enforcement agencies made 2.9 million arrests of juveniles (children <18
years). Moreover, 19% of all arrests and 19% of all violent crime arrests
were accounted for by juveniles. Although the number of juvenile Violent Crime
Index arrests (ie, for murder, forcible rape, robbery, and aggravated assault)
declined in both 1995 and 1996, the rate in 1996 was still 60% higher than
the 1987 level.1
Antisocial behavior can be classified according to its time of onset:
prior to puberty (childhood onset) vs after puberty (adolescent onset).2,3 Childhood onset is characterized by
more serious behavioral disruption, such as violent behavior toward classmates
and cruelty toward animals beginning as early as age 3 years, but occurs less
frequently. The adolescent-onset variety, although sometimes expressed as
aggression toward peers, is generally less serious (eg, shoplifting, lying
to teachers and parents) and occurs so frequently that some consider it normative.3 Childhood-onset antisocial behavior is associated
with neuropsychological deficits (eg, impaired language and intellectual functioning,
attention-deficit/hyperactivity disorder) and harsh, rejecting parenting early
in the child's life.4,5 The adolescent-onset
type has been hypothesized to be a reaction to the limited number of responsible
roles for adolescents in Western societies.3
In earlier articles, we have shown that a program of prenatal and infancy
home visitation by nurses improved women's prenatal health-related behavior6 and reduced the rates of child abuse and neglect,7,8 maternal welfare dependence, closely
spaced successive pregnancies, maternal criminal behavior and behavioral problems
due to use of alcohol and other drugs,8 and
children's intellectual impairment associated with prenatal exposure to tobacco.9,10 These aspects of maternal and child
functioning represent significant risks for early-onset antisocial behavior.11
This article examines the extent to which this program produced a reduction
in children's criminal and antisocial behavior. We expected that the program
would reduce antisocial behaviors indicative of the early-onset type but did
not expect it to have as dramatic an effect on adolescent-onset antisocial
behavior.11 We expected that program effects
would be concentrated on children born to women who were unmarried and from
low-income families at registration during pregnancy. One of the treatment
conditions used in this study consisted of prenatal home visitation with no
postpartum follow-up. We expected that the group receiving only prenatal home
visitation would function better than the comparison group but not as well
as the group that received prenatal and postnatal home visitation.
The details of this study's design can be found in an earlier article.8 A summary of the design is given herein.
Pregnant women were recruited from a free antepartum clinic sponsored
by the Chemung County, New York, health department and the offices of private
obstetricians in Elmira, NY. We actively recruited women with no previous
live births who were less than 25 weeks pregnant and who were young (aged
<19 years at registration), unmarried, or of low socioeconomic status (SES).
Women without these sociodemographic risk characteristics were permitted to
enroll if they had no previous live births. From April 1978 through September
1980, 500 women were invited to participate and 400 enrolled. Eighty-five
percent were young, unmarried, or from low-SES households (August Hollingshead,
PhD, unpublished manuscript, 1976). After completing informed consent and
baseline interviews, women were stratified by sociodemographic characteristics
and randomized to 1 of 4 treatment conditions. Persons involved in data gathering
were blinded to the women's treatment conditions.
Families in treatment group 1 (n=94) were provided sensory and developmental
screening for the child at 12 and 24 months of age. Based on these screenings,
the children were referred for further clinical evaluation and treatment when
needed. Families in treatment group 2 (n=90) were provided the screening services
offered those in treatment group 1 in addition to free transportation (using
a taxicab voucher system) for prenatal and well-child care through the child's
second birthday. There were no differences between treatment groups 1 and
2 in their use of prenatal and well-child care (both groups had high rates
of completed appointments). Therefore, these 2 groups were combined to form
a single comparison group as in earlier articles. Families in treatment group
3 (n=100) were provided the screening and transportation services offered
to treatment group 2 and in addition were provided a nurse who visited them
at home during pregnancy. Families in treatment group 4 (n=116) were provided
the same services as those in treatment group 3 except that the nurse continued
to visit through the child's second birthday.
In the home visits, the nurses promoted 3 aspects of maternal functioning:
(1) positive health-related behaviors during pregnancy and the early years
of the child's life, (2) competent care of their children, and (3) maternal
personal development (family planning, educational achievement, and participation
in the workforce). In the service of these 3 goals, the nurses linked families
with needed health care and human services and attempted to involve other
family members and friends in the pregnancy, birth, and early care of the
child. The nurses completed an average of 9 visits during pregnancy (range,
0-16) and 23 visits from birth to the child's second birthday (range, 0-59).
Details of the program can be found elsewhere.12,13
The current phase of the study consists of a longitudinal follow-up
of the 400 families who were randomized to treatment and control conditions
and in which the mother and child were still alive and the family had not
refused participation at earlier phases. The flow of patients from recruitment
through the 15-year follow-up is presented in Table 1. Interviews were conducted with the adolescents, their biological
mothers, and their custodial parents if the biological mother no longer had
custody. Assessments using parent reports used interview data from the parent
who was judged to have had the greatest amount of recent experience with the
Assessments conducted at earlier phases are specified in previous articles.7,8 At the 15-year follow-up assessment,
adolescents completed interviews that measured whether they had been adjudicated
a person in need of supervision (PINS) resulting from incorrigible behavior
such as recurrent truancy or destroying parents' property; their frequency
of running away from home; and the number of times they had been stopped by
the police, arrested, convicted of a crime or of probation violations, and
sent to youth correctional facilities.14 They
also reported on their disruptive behavior in school; number of school suspensions;
delinquent and aggressive behavior outside school; experience of sexual intercourse;
rates of pregnancy; lifetime number of sexual partners; and frequency of using
cigarettes, alcohol, and illegal drugs during the 6-month period prior to
the 15-year interview.15
Variables were created to summarize the number of occurrences of being
stopped by the police, arrested, convicted (adjudicated) of the original crime
or of probation violations, and sent to a youth correctional facility. Although
we asked the children to report their number of school suspensions and disruptive
behaviors in school, we used archived school data and teacher reports to measure
these outcomes because they are less subject to reporting bias than are self-report
A variable was constructed to characterize the total number of cigarettes
currently smoked per day. Separate variables were constructed to count the
number of days the children had consumed alcohol or used illegal drugs during
the 6-month period prior to the interview. The adolescents were asked questions
regarding the effect of alcohol on 5 domains of their lives (trouble with
parents, trouble at school, problems with friends, problems with someone they
were dating, trouble with police).16 These
data were summarized in an alcohol-use behavioral problem scale (range, 0-5).
Corresponding questions regarding use of illegal drugs were omitted because
of clerical error.
The self-reports of antisocial and delinquent acts were factor analyzed
and found to produce 2 factors, major delinquent acts and minor antisocial
acts, with Cronbach α coefficients of .82 and .68, respectively. The
adolescents also completed the Achenbach Youth Self-Report of Problem Behaviors,
which produces 2 broadband scales: internalizing (anxiety/depression, social
withdrawal, and somatic complaints) and externalizing (delinquency and aggression)
Parents were asked questions about their children's behavioral problems
(the Achenbach scale); school suspensions; arrests; and use of cigarettes,
alcohol, and illegal drugs, including the effect of alcohol and other drugs
on their children's lives. Variables were constructed to coincide with those
based on the child's self-report of behavior. Parents' reports of their children's
behavioral problems caused by substance use included children's use of illegal
drugs (range, 0-10).
The number of short-term and long-term suspensions were counted from
an abstraction of the children's school records for grades 7 through 9. In
New York State, long-term suspensions require a hearing and usually are for
serious infractions such as assaulting a student or teacher. Records were
analyzed when there were complete school data for 2 of the 3 years. The students'
current teachers in English and mathematics completed an "acting out" scale
that rated children's disruptive behavior in the classroom (eg, disruptive
in class, defiant, obstinate, stubborn).18
Finally, the records of 116 children who lived in Chemung County for
their entire lives were reviewed by the Chemung County Probation Department
and the Chemung County Family Court. Identifying information on the adolescents
(name, birth date, sex, Social Security number) was provided to these departments
for purposes of matching their records with the participants in this study.
The department staff summarized the counts of arrest and PINS records within
treatment and risk-status groups. Individual identifiers were not returned
in the abstraction of these data, although the children's treatment group,
sex, and risk status (ie, whether they were born to an unmarried mother from
low SES) were returned.
The study was conducted with an intention-to-treat approach. A core
statistical model was derived that was consistent with the one used in the
earlier phases of this research. It consisted of a 3 × 2 × 2 ×
2 factorial structure and 6 covariates. The classification factors were treatment
groups (1 and 2 vs 3 vs 4), maternal marital status (married vs unmarried
at registration), social class (Hollingshead I and II vs III and IV at registration),
and sex of child. All interactions among the first 3 factors were included.
Interactions with sex of the child also were examined. Although sex
was a significant predictor of the antisocial behavior outcomes, it did not
interact in a fully interpretable way with other terms in the model for some
outcomes, so it was included without interactions. Where program effects were
moderated by the child's sex in a coherent way, we have noted this in the
presentation of the findings, in which case the model includes SES as a covariate
rather than a classification factor and includes all interactions among treatment,
marital status, and sex. This model was preferable to a 4-factor classification
structure with all interactions because of the low incidence of some outcomes
for certain subclasses, compromising the stability of the Poisson log-linear
models used in the analysis. In addition, for 2 variables, the core log-linear
model produced unstable variance estimates for the tests of treatment main
effects. In these cases, SES and marital status were included without interactions
for that test.
Race of the mother was among a number of additional classification factors
examined in deriving the core model but was not included because it was not
a significant predictor of outcomes once other terms were included.
The 6 covariates included in the final models were maternal age, maternal
education, locus of control,18 support from
husband or boyfriend, maternal employment status, and paternal public assistance
status. All covariates were measured at registration and tested for homogeneity
of regressions for the hypothesized contrasts.19
Dependent variables with normal distributions were analyzed in the general
linear model and low-frequency count data (eg, number of arrests) in the log-linear
model (assuming a Poisson distribution). In the log-linear model, the analysis
was carried out and estimates were obtained in terms of the logs of the incidence.
We use the term incidence to refer to the actual
count or mean of counts during specific periods of measurement. A careful
examination of the distributions of each of the dependent variables was carried
out. Low-incidence count variables with values higher than 20 were analyzed
in a log-linear model, correcting for overdispersion.
For outcomes reported by more than 1 respondent (eg, child and parent
or teachers), we carried out repeated-measures analyses, adding to the basic
model fixed factors for respondent and a random factor for individuals. The
focus of interpretation was on the average of the 2 sources of information.
All treatment contrasts focused on the comparison of the combination
of treatment groups 1 and 2 (the comparison group) with treatment group 4
(the pregnancy and infancy nurse-visited group) because we hypothesized that
the greatest treatment effect would be exerted by the combination of prenatal
and postnatal home visitation, as found in earlier evaluations.7,8
The results for the group that received prenatal home visitation only (treatment
group 3) are included to report whether that group had intermediate levels
of functioning. To address our primary hypotheses, treatment effects also
are shown for adolescents whose mothers were unmarried and from low-SES households
at registration during pregnancy. All estimates of treatment main effects
and effects for the unmarried, low-SES group are derived from a common statistical
As indicated in Table 2,
for those families for which 15-year assessments were completed, the treatment
groups were essentially equivalent on background characteristics for both
the sample as a whole and for women who were unmarried and from low-SES households.
Small differences on some background variables (such as paternal receipt of
public assistance) led us to include them as covariates.
Encounters With the Criminal Justice System.Table 3 shows that adolescents
born to nurse-visited women (treatment group 4) reported more frequent stops
by police (P<.001) but fewer arrests and convictions
and violations of probation (P=.005 and .001, respectively);
the arrest and convictions and probation violation effects were concentrated
among children born to women who were unmarried and from low-SES families
(P=.03 and <.001, respectively). For the subsample
of children who lived in Chemung County for their entire lives, nurse-visited
children (treatment group 4) had fewer official PINS records (P=.007). Nurse-visited children whose mothers were unmarried and from
low-SES families were reported by their parents to have been arrested less
frequently than were their counterparts in the comparison group (P=.05). In addition, among adolescents born to unmarried women from
low-SES households, those in treatment group 4 reported fewer instances of
running away (P=.003). As indicated in Table 3, with the exception of parent report of child arrests, most
of these effects were present for children whose mothers were visited only
during pregnancy (treatment group 3).
The effect of the treatment group 4 program on the children's reports
of running away was concentrated in girls, whereas the effect on parents'
reports of the children's arrests and children's reports of convictions and
probation violations was greater for boys (data not shown). The effect of
the program on arrests was not limited to any specific type of crime, although
property crimes were more frequent and, therefore, accounted for a larger
portion of the program effect on arrests overall.
School Suspensions, Behavior Problems, and Use of Substances.Table 4 shows that children
born to nurse-visited (treatment group 4) women who were unmarried and from
low-SES households reported having fewer sexual partners (P=.003), smoking fewer cigarettes per day (P=.10),
and consuming alcohol fewer days during the 6-month period prior to the 15-year
interview (P=.03). Parents of children born to nurse-visited,
unmarried women from low-SES families reported that their children had fewer
behavioral problems related to their use of alcohol and other drugs (P=.08). For these outcomes, there was some indication that
the group visited by nurses only during pregnancy (treatment group 3) did
not do as well. Although adolescents in the unmarried, low-SES group reported
smoking fewer cigarettes, they also reported higher levels of illegal drug
use and their parents reported more behavioral problems due to the use of
alcohol and other drugs than did their counterparts in the comparison group.
There were no treatment differences in teachers' reports of the adolescents'
acting out in school; short-term or long-term suspensions; the adolescents'
initiation of sexual intercourse; or the parents' or children's reports of
major delinquent acts, minor antisocial acts, or other behavioral problems.
Adolescents born to nurse-visited (treatment group 4) women who were
unmarried and from low-SES families had fewer episodes of running away from
home, arrests, and convictions and violations of probation than did their
counterparts in the comparison group. They also had fewer sexual partners
and engaged in cigarette smoking and alcohol consumption less frequently.
Their parents reported that they had fewer behavioral problems related to
their use of drugs and alcohol. There were no program effects on less serious
forms of antisocial behavior, initiation of sexual intercourse, or use of
illegal drugs. Children in treatment group 4, irrespective of risk, reported
being stopped by police more frequently, but they reported fewer arrests and
convictions and violations of probation, and the official PINS records corroborated
this pattern. The higher rates of being stopped by police is an anomalous
finding that has no coherence with any other effects and is likely to be either
a sampling or reporting artifact.
The concentration of beneficial effects among children born to unmarried
women of low SES is consistent with the results of other preventive interventions
that have shown greater effects for children of families at greater social
risk.20 This suggests that these kinds of services
ought to be focused on families in greater need by virtue of the mothers'
being unmarried and poor.
In general, these findings are consistent with program effects on early-onset
antisocial behavior rather than on the more common and less serious antisocial
behavior that emerges with puberty.3 The mere
presence of arrests, convictions, and probation violations by the time the
children were 15 years old suggests that these children started offending
early and that they may be on life-course trajectories that portend recurrent
and more serious offenses in the future. Given that early-onset antisocial
behavior is associated with (1) subtle neurological impairment, (2) harsh,
punitive, and neglectful parenting, and (3) family contexts characterized
by substance abuse and criminal behavior,2- 5
it is important to note that this program has affected these aspects of maternal,
child, and family functioning at earlier phases in the child's development.6- 11
Moreover, genetic vulnerability to impulsivity and aggression is expressed
much more frequently when vulnerable rhesus monkeys experience aberrant rearing21 (also Allyson J. Bennett, PhD, K. Peter Lesch, Armin
Heils, et al, unpublished data, 1998), adding to the plausibility of the findings
The prenatal phase of the program reduced fetal exposure to tobacco,
improved the qualities of women's prenatal diets, reduced rates of pyelonephritis,
improved levels of informal social support, and reduced intellectual impairment
and irritable behavioral styles associated with fetal exposure to tobacco.6,10,11,22 Prenatal
exposure to tobacco is a risk factor for early behavioral dysregulation, problems
with attention, and later crime and delinquency.22
Moreover, the combination of birth complications (and, by implication, neurological
impairment) and rejecting parenting substantially increases the likelihood
of violent offenses by the time children are 18 years old.5
We did not expect prenatal home visitation (treatment group 3) by itself
to be as effective as it was in preventing criminal behavior among children
born to low-SES, unmarried women. This occurred even though these children's
mothers showed almost none of the postnatal benefits observed for those visited
during pregnancy and infancy (such as reduced welfare dependence, substance
abuse, criminal behavior, and child abuse and neglect).8
The mechanisms through which these beneficial effects occurred will be examined
in future reports, with a focus on the alteration of maternal prenatal health
and the children's corresponding neuropsychological functioning,22,23
as well as prenatal stress, given that stress during pregnancy affects the
social and neuromotor development of nonhuman primates.24,25
The impact of the full program (prenatal and infancy home visitation)
on children's use of alcohol and number of sexual partners is important because
recent evidence indicates that alcohol use prior to age 15 years multiplies
the risk of alcoholism in adulthood26 and multiple
partners increase the risk for sexually transmitted diseases, including human
immunodeficiency virus infection.27,28
The effect of the program on alcohol use is consistent with greater alcohol
consumption observed among adult rhesus monkeys who experienced aberrant rearing.29 These findings must be tempered, however, with an
acknowledgment of their limitations.
The first limitation is that most of the positive results were concentrated
among children born to women who were unmarried and from low-SES households.
Although we hypothesized originally that the effects would be greater for
women who experienced higher levels of stress and who had fewer personal resources,
we did not fully operationalize the stress and resource variables prior to
the beginning of the trial. We chose to use characteristics used for sample
recruitment as indicators of long-term stress (eg, coming from a low-SES household)
and having few personal resources (eg, being unmarried), 2 factors associated
with a host of adverse outcomes. However, positive early results from a large
urban replication of this study focusing almost exclusively on unmarried,
low-income women support our interpretation that the effects observed in the
current study are due to the program.30
The second limitation is that the arrest and conviction data were based
primarily on the children's and parents' reports, which may be subject to
treatment-related reporting bias. To validate the children's and parents'
reports of undesirable behavior, we compared the rates of school suspensions
derived from the school records with the parents' and children's reports of
suspensions and found no treatment differences in accuracy. We also regressed
the English and mathematics teachers' averaged reports of the adolescents'
acting out in school on the adolescents' self-reports of their acting out
in school separately for the nurse-visited and comparison group children and
found no treatment differences in the slopes of these regressions.
Importantly, the pattern of mean differences for treatment groups 1
and 2 vs treatment group 4 for PINS records on the subsample of children who
lived in Chemung County for their entire lives corroborated the pattern of
the children's reported arrests. The PINS finding increases our confidence
that the treatment differences in the adolescents' reported involvement with
the criminal justice system are not the result of the nurse-visited children
and their parents simply underreporting their actual levels of involvement.
The absence of program effect with the official arrest data may be explained
by a significant, 9-fold higher rate of official arrest records prior to randomization
(0.44 vs 0.05) found for treatment group 4 mothers who were unmarried and
of low SES and whose children remained in Chemung County compared with their
treatment group 1 and 2 counterparts.
Finally, we note that the adolescents' self-reports of delinquent and
antisocial behavior are not completely consistent with the data on reports
of arrests and convictions. A survey that used follow-up questions to the
assault questions asked in the current study showed that the answers to the
questions we used produced responses that frequently were trivial (eg, 33%
of the serious violent responses and 64% of the self-reported minor assaults
were too insignificant to lead to arrests).31
This suggests that the particular questions used in this study regarding delinquent
behavior did not adequately assess the severity of delinquency. Thus, the
treatment differences found in reports of arrests and convictions are likely
to be indications of underlying treatment differences in the severity of antisocial
behavior that were not assessed adequately by the set of questions asked about
particular antisocial behaviors.
This program prevented only the more serious forms of antisocial behavior
leading to arrests and convictions. Other types of prevention programs may
be necessary to reduce more normative types of disruptive behavior among young
adolescents.32 In light of the impact of this
program on maternal and youth crime and corresponding government expenditures,8,33 the US Department of Justice is now
supporting an effort to make this program available to a larger number of