[Skip to Navigation]
Sign In
March 1999

Preventing Adolescent Health-Risk Behaviors by Strengthening Protection During Childhood

Author Affiliations

From the Social Development Research Group, University of Washington, Seattle.

Arch Pediatr Adolesc Med. 1999;153(3):226-234. doi:10.1001/archpedi.153.3.226

Objective  To examine the long-term effects of an intervention combining teacher training, parent education, and social competence training for children during the elementary grades on adolescent health-risk behaviors at age 18 years.

Design  Nonrandomized controlled trial with follow-up 6 years after intervention.

Setting  Public elementary schools serving high-crime areas in Seattle, Wash.

Participants  Of the fifth-grade students enrolled in participating schools, 643 (76%) were given written parental consent for the longitudinal study and 598 (93%) were followed up and interviewed at age 18 years.

Interventions  A full intervention provided in grades 1 through 6 of 5 days of in-service training for teachers each intervention year, developmentally appropriate parenting classes offered to parents when children were in grades 1 through 3 and 5 through 6, and developmentally adjusted social competence training for children in grades 1 and 6. A late intervention, provided in grades 5 and 6 only, paralleled the full intervention at these grades.

Main Outcome Measures  Self-reported violent and nonviolent crime, substance use, sexual activity, pregnancy, bonding to school, school achievement, grade repetition and school dropout, suspension and/or expulsion, and school misbehavior; delinquency charges from court records; grade point average; California Achievement Test scores; and disciplinary action reports from school records.

Results  Fewer students receiving full intervention than control students reported violent delinquent acts (48.3% vs 59.7%;P=.04), heavy drinking (15.4% vs 25.6%; P=.04), sexual intercourse (72.1% vs 83.0%; P=.02), having multiple sex partners (49.7% vs 61.5%; P=.04), and pregnancy or causing pregnancy (17.1% vs 26.4%; P=.06) by age 18 years. The full intervention student group reported more commitment (P=.03) and attachment (P=.006) to school, better academic achievement (P=.01), and less school misbehavior (P=.02) than control students. Late intervention in grades 5 and 6 only did not significantly affect health-risk behaviors in adolescence.

Conclusions  A package of interventions with teachers, parents, and children provided throughout the elementary grades can have enduring effects in reducing violent behavior, heavy drinking, and sexual intercourse by age 18 years among multiethnic urban children. Results are consistent with the theoretical model guiding the intervention and support efforts to reduce health-risk behaviors through universal interventions in selected communities or schools serving high-crime neighborhoods.

AS THE COSTS of crime, teen pregnancies, an undereducated underclass in poverty, and substance abuse have increased, policymakers and researchers have looked more seriously at prevention as a potentially cost-effective approach to reduce the prevalence of these behaviors.1 Efforts to improve schools, reduce crime and violence, combat substance abuse, and prevent unwanted pregnancies have progressed on separate tracks.

Advances in prevention in public health2 provide a model for prevention of adolescent health-risk behaviors by focusing on risk and protective factors predictive of these behaviors.3,4 Research on the predictors of school failure, delinquency, drug abuse, teen pregnancy, and violence indicates that many of the same factors predict these different outcomes.5,6 Recent research has shown that bonding to school and family protects against a broad range of health-risk behaviors in adoles cence.6 Yet, prevention studies typically have focused narrowly on a specific outcome, such as preventing substance abuse, and on attitudes and social influences that predict that outcome.7,8 Previous studies on prevention have not sought to address the shared risk and protective factors for diverse health-risk behaviors that are the main threats to adolescent health. This article reports results of a nonrandomized controlled trial 6 years following the provision of intervention during the elementary grades. The interventions were designed to increase bonding to school and academic success and to prevent a broad range of health-risk behaviors in a multiethnic urban sample.

The intervention sought to reduce specific empirically identified risk factors for adolescent health and behavior problems: persistent physically aggressive behavior in the early elementary school grades,9-11 academic failure,12 and poor family management practices including unclear rules, poor monitoring of behavior, and inconsistent or harsh discipline.13,14 Because being raised in poverty increases risk for crime, school failure, and school dropout,15-17 effects of the intervention on children from low-income families were of particular interest.

Research has shown that teachers can improve children's attitudes toward school, behavior at school, and academic achievement through the use of effective methods of instruction and management.18-21 Studies have also identified child-rearing methods that reduce conduct problems and improve school performance of elementary school-aged children.22-25

To affect the risk factors of concern, we thought that intervention with teachers, parents, and children themselves would be necessary, suggesting a multiple component strategy. To ensure that the components of the strategy would complement each other, the intervention was grounded in the social development model,26,27 an integrated theory of human behavior.

Consistent with recent research,6 the social development model hypothesizes that strong bonds to school serve as a protective factor against behaviors that violate socially accepted standards. Attachment (ie, a positive emotional link), and commitment (ie, a personal investment in the group) are the component elements of such social bonds. The theory hypothesizes that when social groups produce strong bonds of attachment and commitment in members, and promote clear standards for behavior, these groups increase behavior consistent with those standards and prevent behavior that violates them.

The social development model hypothesizes that the interplay of specific factors during development influences the degree to which children develop strong social bonds to school. The factors that affect children's bonding are the degree of opportunity for active involvement available in the family and classroom, the skills possessed and applied by children during participation in these social groups, and the reinforcements provided to children in response to their behavior in these groups.

We hypothesized that training teachers to teach and manage their classrooms in ways that promote bonding to school, training parents to manage their families in ways that promote bonding to family and to school, and providing children with training in skills for social interaction would positively affect children's attitudes toward school, their behavior at school, and their academic achievement. We thought that these changes would, in turn, set children on a different developmental trajectory observable in more positive academic outcomes and fewer health-risk behaviors later in adolescence.

We have found and reported short-term positive effects of this intervention package on behavior, school bonding, and achievement of children receiving the intervention during grades 1 through 4.28,29 We have also found positive effects on children from low-income families at the end of grade 6.30

Previous delinquency, substance abuse, and teen pregnancy prevention programs have been provided in the late elementary or middle school grades, just prior to the ages when delinquent behavior, substance use, and sexual activity increase in prevalence.6,7,31,32 Yet the social development model that guides the present intervention suggests that early and sustained intervention through the elementary grades should put children on a different developmental trajectory leading to positive outcomes over the long term. An important question is whether delivering the intervention package over the full course of elementary school ("full intervention") had greater effects than delivering the intervention just prior to adolescence ("late intervention"). This study examines this question by comparing outcomes for 3 separate groups: a "full intervention group," exposed to the interventions throughout the elementary grades; a "late intervention group," exposed to the interventions only in grades 5 and 6; and a control group, 6 years after the interventions ended, when children progressing normally were completing high school.

Participants and methods

Sample and design

This study of intervention effects is part of a larger ongoing longitudinal study of all consenting fifth-grade students in 18 public schools serving high-crime areas of Seattle, Wash. To assess the effects of full intervention and late intervention, a nonrandomized controlled trial with 3 conditions was created. The full intervention group received the intervention package from grade 1 through grade 6. The late intervention group received the intervention package in grades 5 and 6 only, and the control group received no special intervention. This design was created in 1985 by nesting an intervention initiated in 1981 at first-grade entry, within the longitudinal panel study. For the present study, schools were assigned nonrandomly to conditions in the fall of 1985, and from that point, all fifth-grade students in each school participated in the same interventions. New schools added for the panel study were matched to the intervention schools with respect to grades served and inclusion of students drawn from high-crime neighborhoods of Seattle. Schools added for the panel study were assigned to conditions to achieve balanced numbers across conditions. It should be noted that during this study the Seattle School District used mandatory busing to achieve racial equality in schools. As a result, all schools in this study served heterogeneous population of students drawn from at least 2 different neighborhoods of the city. This practice reduced the risk that outcomes observed in this study reflected contextual or neighborhood differences in the populations attending different schools.

In the present analyses, the full intervention group consists of all students who were randomly assigned to intervention classrooms in grades 1 through 4 in 8 elementary schools participating in the earlier experimental study, and who remained in schools assigned to the intervention condition in grades 5 or 6 in the present study. The late intervention group consists of students in intervention schools who were in intervention classrooms in grades 5 and 6 only, some of whom were controls in the earlier intervention study. The control condition consists of students in schools assigned to receive no intervention in grades 5 and 6 and who were not in intervention classrooms in grades 1 through 4.

When the longitudinal study was funded by the National Institute on Drug Abuse in 1985, school-based substance abuse prevention research trials had been limited to studies of social influence resistance curricula provided to students in grades 5 through 10.3 In this context, reviewers desired that the study include a condition assessing effects of only 2 years of intervention in the late elementary grades as well as effects of the full intervention, since they questioned whether intervention in the early elementary grades was necessary. In addition, reviewers concerned with the possibility of contamination of control students by experimental students in our earlier randomized within-building study design preferred a nonrandomized design in which whole schools, rather than classrooms, were assigned to intervention or nonintervention conditions. As a result some students who had served as controls in the earlier study became part of the late intervention group in the longitudinal study.

The present study included all fifth-grade children assigned to the 3 conditions in 1985 whose parents provided annual written consent to their involvement in the longitudinal follow-up study (N=643). Seventy-six percent of the total eligible fifth-grade population in these schools consented. Children also provided consent annually. Participants were distributed across the 3 conditions as shown in Table 1. The sample included roughly equal proportions of boys and girls. More than 56% of the sample were from poor families as evidenced by their participation in the National School Lunch/School Breakfast Program. About 44% of the sample were white, 26% were African Americans, 22% were Asian Americans, 5% were Native Americans, and 3% were classified as other racial/ethnic groups.

Table 1. 
Comparisons of Baseline and Analysis Samples With Tests of External and Internal Validity
Comparisons of Baseline and Analysis Samples With Tests of External and Internal Validity

In the spring of 1993 when the participants were aged 18 years, 598 participants (93.0% of the original consenting sample assigned to the 3 conditions) were successfully interviewed. All self-reported follow-up data came from these interviews. Standardized achievement test scores, grade point averages, and school disciplinary records were collected in the spring of 1992 when participants were 17 years old. Delinquency charges were collected in 1993 from juvenile court records, through age 17 years.

Because analyses reported below revealed significant differences between the full intervention and control groups at age 18 years, analyses were conducted to ascertain the equivalence of these 2 groups on factors that might influence outcomes. These are reported in Table 2. The groups did not differ on residential stability as measured by mean number of years living in Seattle by age 12 years and by the mean number of residences in which participants lived from age 5 to 14 years; socioeconomic status, as measured by years of parental education or proportion eligible for the school lunch program; proportion from single-parent families; proportion of boys; or proportion of whites or nonwhites. Table 2 also shows that roughly equivalent proportions of students in both the full intervention and control groups were living in disorganized neighborhoods at age 16 years as indicated by students' self-reports of rundown housing, crime, poor people, drug-selling, gangs, and disorderly and undesirable neighbors in their neighborhoods. To ascertain whether later school assignments might have affected observed outcomes differentially for the full intervention and control groups, contingency tables comparing high schools attended in grades 9, 10, and 11 were computed. No significant differences in high schools attended were found between the 2 groups, again suggesting that Seattle's mandatory busing program overcame the separation of students from different neighborhoods and elementary schools sometimes found in large urban school districts. This decreases the plausibility that observed outcomes reflected later differences in schools attended by students in the 2 conditions. These results suggest that a number of factors that might have affected outcomes were equivalent across intervention conditions, and do not pose major threats to the internal validity of the present study.

Table 2. 
Characteristics of Control and Full Intervention Groups
Characteristics of Control and Full Intervention Groups


The intervention package included the following components, elaborated in Table 3.

Table 3. 
Seattle Social Development Project Interventions
Seattle Social Development Project Interventions

Classroom Instruction and Management

Each year, as the panel moved through the elementary grades, teachers in intervention classrooms received 5 days of in-service training in a package of instructional methods33 with 3 major components: proactive classroom management,34 interactive teaching,35 and cooperative learning.36

Teachers of control students did not receive training in instructional or classroom management skills from the project. However, both intervention and control teachers were observed by evaluators, uninformed as to intervention condition of the classrooms they observed. Observations lasted for 50 minutes on 2 different days in the fall and spring each year using the interactive teaching map to document the use of the targeted teaching strategies in all conditions.37 Greater use of the experimental instructional and management methods was observed in intervention classrooms, as discussed elsewhere.38 Teachers' use of the experimental instructional and management methods has been found to predict short-term variation across classrooms in students' levels of social development constructs related to school bonding.38

Child Skill Development

First-grade teachers of the full intervention group also received instruction in the use of a cognitive and social skills training curriculum, interpersonal cognitive problem solving,39,40 which teaches skills to children to think through and use alternative solutions to problems with peers. This curriculum developed children's skills for involvement in cooperative learning groups and other social activities without resorting to aggressive or other problem behaviors.

In addition, when students in both intervention conditions were in grade 6, they received 4 hours of training from project staff in skills to recognize and resist social influences to engage in problem behaviors, and to generate and suggest positive alternatives to stay out of trouble while keeping friends.41

Parent Intervention

Parent training classes appropriate to the developmental level of the children were offered on a voluntary basis to parents or adult caregivers of children. Parents of children in the full intervention condition were offered training in child behavior management skills when their children were in the first and second grades through a 7-session curriculum called "Catch 'Em Being Good,"42 which is grounded in the work of Patterson.43 In the spring of the second-grade school year and again in the third grade, parents of children receiving the full intervention also were offered a 4-session curriculum, called "How to Help Your Child Succeed in School,"38 to strengthen their skills for supporting their children's academic development. When their children were in grades 5 and 6, parents of children in both the full and the late intervention conditions were offered a 5-session curriculum, "Preparing for the Drug (Free) Years,"44 to strengthen their skills to reduce their children's risks for drug use. Professional multiethnic project staff provided parenting workshops in collaboration with participating schools and parent councils.

Parents of 43% of children in the full intervention condition attended parenting classes. Forty-six percent of parents who attended parenting classes were from low-income families.


Data were collected from group-administered questionnaires completed by children in the fall of 1985 just after entering fifth grade, and from Seattle School District records. Outcome data were collected through individually administered interviews conducted in the spring of 1993. Self-reported outcomes were supplemented with data on delinquency charges in the King County Juvenile Court, and California Achievement Test results, grade point averages, and school disciplinary action reports through age 17 years from the Seattle School District. Measures used to assess constructs addressed by the intervention are listed in Table 4. (The scales, items, and reliabilities are available from one of us [J.D.H.].)

Table 4. 
Comparisons of Control, Late, and Full Intervention Groups Across Targeted School Outcomes*
Comparisons of Control, Late, and Full Intervention Groups Across Targeted School Outcomes*


Attrition Analysis

Attrition analyses shown in Table 1 indicate that effects reported below are not likely to be an artifact of differential attrition from conditions. Forty-five participants from the original sample were not interviewed at age 18 years. As shown in Table 1, they did not differ significantly from those remaining in the study with respect to gender, ethnicity, or poverty. Also, as shown in Table 1, attrition was unrelated to intervention condition. The participants missing at follow-up (14 from the control, 24 from the late, and 7 from the full intervention groups) were randomly distributed among the 3 intervention conditions. Moreover, after accounting for the attrition of the 45 participants, the overall distribution of the sample with respect to gender, socioeconomic status, or race was unrelated to intervention condition. Separate χ2 analysis for gender, socioeconomic status, and race indicated no significant relationship between attrition and intervention condition for these variables. In an additional test of attrition effects, the interaction of intervention condition with attrition was assessed with respect to each outcome variable reported inTable 3 by examining attrition-by-condition interaction effects on corresponding measures at fifth-grade entry. An interaction of attrition and condition would indicate possible bias introduced by differential attrition by condition. Effects on fifth-grade entry measures of school bonding, self-reported grades, California Achievement Test scores, school misbehavior, fighting, delinquency, arrests, and substance use were estimated. No significant interaction of intervention condition (control vs full) with the attrition variable was found for effects on related fifth-grade measures, reinforcing the internal validity of the study with respect to all comparisons of the full intervention and control groups. Observed outcomes are unlikely to have been an artifact of differential attrition across conditions. It should be noted that regardless of condition, participants missing California Achievement Test scores at age 17 years had significantly lower California Achievement Test scores at age 9 years (F1,359 = 6.66,P<.05), participants missing school records of grade point average at age 17 had significantly lower self-reported grades in fifth grade (F1,345 = 4.23, P<.05), and those missing self-reports of school misbehavior at age 18 reported significantly more school misbehavior at fifth-grade entry than did their counterparts who were retained in the study (F1,370 = 4.22, P<.05), suggesting care in generalizing the observed results on these outcomes to all low-achieving or poorly behaved students.

Intervention Analysis

Continuous measures were analyzed by comparing means across the 3 conditions using analysis of variance procedures.

Prevalence measures were compared using the χ2 testwith the continuity correction to contrast each intervention condition with the control condition. As shown later, significant effects on health-risk behaviors were found only for the full intervention group.

Because poverty predicts risk for school adjustment problems, low achievement, crime, and other problem behaviors, the effects of the full intervention on children from poor families were investigated using logistic and linear regression methods as appropriate, with terms for intervention and free lunch eligibility as main effects and an interaction term for intervention by participation in the free lunch program. Similar analyses were conducted to assess possible differential effects of the full intervention on boys and girls using an interaction term for treatment by gender.

Although this study makes specific directional hypotheses regarding intervention effects, a conservative approach is used, and 2-tailedP values are reported.


School behavior, school bonding, and achievement

Table 4 compares the prevalences and scale means for school outcome measures between the intervention groups and the control group. Table 4 shows that students in the full intervention condition reported significantly stronger commitment (P=.03) and attachment (P=.006) to school at age 18 years, suggesting that the full intervention had lasting effect on school bonding, an important mediating social development construct and a protective factor identified in descriptive research.6 This effect was paralleled by significant improvement in self-reported achievement (P=.01), a near significant improvement in school-reported grade point average (2.18 for controls vs 2.42 for full intervention participants; P=.09), and proportion who had repeated a grade (22.8% of controls vs 14.1% of the full intervention participants;P=.05) and a nonsignificant reduction in the proportion who had dropped out of school (26.2% vs 18.9%; P=.14). No effect on California Achievement Test scores at age 17 years was observed.

Students receiving the full intervention also reported significantly less involvement in school misbehavior than did control students (P=.02). This was consistent with the trend for school district disciplinary action reports. Fifty-eight percent of control participants compared with 45.8% of full intervention participants had reports of misbehavior in Seattle School District records (P=.07). Levels of suspension or expulsion did not differ significantly across the groups, although the pattern of results parallels other outcomes; ie, 48.3% of controls compared with 39.6% of full intervention students had been suspended or expelled in their lifetimes (P=.13).

Table 5 compares the lifetime prevalence of health-risk behavior outcomes for the full and late intervention groups with that of the control group. Significantly more participants in the control group than in the full intervention group had committed violent delinquent acts by age 18 years (59.7% vs 48.3%; P=.04). Patterns of prevalence for nonviolent delinquency, self-reported arrests, and official juvenile court delinquency records paralleled the results for violence, although the differences did not achieve statistical significance.

Table 5. 
Comparisons of Control, Late, and Full Intervention Groups Across Targeted Delinquency, Substance Use, and Sexual Behavior Outcomes
Comparisons of Control, Late, and Full Intervention Groups Across Targeted Delinquency, Substance Use, and Sexual Behavior Outcomes

No significant effects were found for any of the lifetime measures of drug use, nor did measures of heavy cigarette and marijuana use differ significantly across the groups at age 18 years (data not shown in Table 5). However, significant differences were found between the control and full intervention groups for heavy alcohol use in the past year. Twenty-five percent of controls compared with 15.4% of full intervention participants reported having drunk alcohol 10 or more times in the past year (P=.04) (data not shown in Table 5). This was paralleled by a nonsignificant trend toward less drinking and driving in the full intervention group. More than 17% of controls compared with 10.9% of full intervention students reported drinking and driving in the past year (P=.14) (data not shown in Table 5).

Significantly more controls than students receiving full intervention had engaged in sexual intercourse (83.0% vs 72.1%; P=.02) and had multiple sex partners by age 18 years (61.5% vs 49.7%;P=.04). More control students had been pregnant or gotten someone pregnant (26.4% vs 17.1%; P=.06), although this finding only approached statistical significance. More control participants than full intervention participants had fathered or had had a baby (14.7% vs 9.5%; P=.20), although this difference was not statistically significant.

In sum, participation in the full intervention during the elementary grades was predictive of enduring significant positive effects through age 18 years on students' bonds to school, achievement, and school behavior. The full intervention group also experienced lower rates of lifetime violence, frequent drinking, and sexual behavior through age 18 years.

Note that "dose effects" are observable for many of the outcomes in Table 4 and Table 5, with the full intervention group showing the most positive outcomes, followed by the late intervention group, followed by the control group.

Analyses of the interactions between poverty and intervention condition revealed that the full intervention was more effective for children from poor families with respect to several outcomes. As shown in Figure 1, significant interaction effects were found for attachment to school (P=.03) and repeating a grade (P=.04). Post hoc tests confirmed that, by age 18 years, children from poor families who had participated in the full intervention were significantly more attached to school than their control counterparts from poor families (P=.001). Further, only 13% of the full intervention children from poor families had repeated a grade compared with 30% of their control counterparts (P=.007). Similarly, trends in interactions indicating stronger intervention effects among children from poor families were found for better school achievement (P=.06; post hoc comparison P=.002), less school misbehavior (P=.05; post hoc comparison P=.003), and less drinking and driving in the past year (P=.06; post hoc comparison P=.03) (data not shown in Figure 1).

Significant interactions of intervention effects with poverty.

Significant interactions of intervention effects with poverty.

In contrast, significant poverty-by-intervention interactions for pregnancy (P=.02) and parenthood (P=.005) revealed that the intervention had significantly greater effects for working and middle class youths in reducing the lifetime prevalence of pregnancy (P=.008 for post hoc comparison) and of having or fathering a baby (P=.02 for post hoc comparison).

The preponderance of observed intervention effects did not differ for boys and girls. However, the full intervention had significantly greater effects in preventing boys from repeating a grade (χ2 = 4.00, P=.05), and engaging in sexual activity (χ2 =4.44, P=.04) (results not shown in Figure 1). We investigated the possibility of a gender-by-poverty interaction in predicting grade repetition, but found none (χ2 =0.01, P=.92).


The results show that intervention through the elementary grades can have enduring positive effects on the academic development and health-risk behaviors of urban children. A package of interventions that trained parents and teachers to promote children's academic competencies and bonding to school, and that developed children's social competencies and skills to resist health-compromising influences produced greater commitment and attachment to school, less school misbehavior, and better academic achievement 6 years after intervention. It seemed particularly effective in improving attachment to school, achievement, and school behavior of children from poor families. Further, the full intervention reduced the lifetime prevalence of violent criminal behavior, heavy drinking, sexual intercourse, and pregnancy.

Threats to the internal validity of study conclusions seem minimal. The full intervention and control groups did not differ with respect to residential stability, socioeconomic status, gender, proportion from single-parent families, proportion living in disorganized neighborhoods, or race. No evidence of differential attrition by condition was found. Students were not informed of their intervention condition in the study, and it is unlikely that respondents knew 6 years later whether they received an intervention or control condition during the elementary grades, thus decreasing the likelihood that participants' self-reported outcomes reflected awareness of their condition assignment.

The pattern of results also suggests confidence in the findings. Respondents in the intervention condition did not consistently report less involvement in all forms of health-risk behaviors. Although hypothesized, no effects of intervention were observed for the lifetime prevalence of use of cigarettes, alcohol, marijuana, or other illicit drugs at age 18 years. "Halo effects" do not seem to be a plausible explanation for the findings. Further, the pattern of school district grade and disciplinary reports corroborated self-reported data.

The investigators chose schools for this study that serve substantial proportions of children from poor families who live in high-crime neighborhoods. In this respect, this was a selective intervention focused on a population at elevated risk for adolescent health-risk behaviors.4 However, it is important to emphasize that the intervention package was provided universally to all children in intervention classrooms, without regard to risk. It did not identify subsamples of children as "high risk" and did not segregate the "high-risk" children into separate groups for special assistance or intervention. In this way, it avoided possible labeling and stigmatization of children. It also avoided the often observed problem that special "pullout" programs increase interaction among children at greatest risk, which itself can increase risk iatrogenically for crime, drug use, and other health-risk behaviors.45

Starting preventive interventions at elementary school entry and continuing them through grade 6 had greater effects on both educational outcomes and health-risk behaviors than intervening later in the elementary grades. These findings suggest that early and continued intervention in the elementary grades can help put children on a positive developmental course that is maintained through high school. The findings are consistent with our hypothesis that helping parents and teachers to strengthen children's commitment and attachment to school in the elementary grades can produce lasting differences in bonding to school that mediate risk and prevent health-compromising behaviors in adolescence. Studies currently under way are assessing the effects of the intervention on the growth in school bonding over time to more directly test this hypothesis.

The present study does not allow assessment of the relative contribution of the individual elements of the intervention package to the observed outcomes. While it would be desirable to ascertain the specific contributions of teacher training, parent training, and student skills training, it is important to learn that the package of interventions produced positive outcomes in a multiethnic urban sample. Studies currently under way are assessing effects of individual components of this intervention.46

It is noteworthy that the interventions tested here were provided only during the elementary grades without any boosters or follow-up intervention in the following 6 years. One explanation for the durability of these effects in contrast to those observed by others47 is that this intervention focused on increasing school bonding and achievement rather than on developing norms or skills specifically related to avoiding health-risk behaviors. Developing a strong commitment to schooling in the elementary grades may set children on a developmental path toward school completion and success that is naturally reinforced both by teachers responsive to eager students and by the students' own commitment to schooling. In contrast, school curricula that focus on changing students' norms to be less approving of health-compromising behaviors and on developing their skills to resist social influences to violate those norms may require booster sessions through adolescence to maintain healthy norms or standards for behavior. A promising direction for future prevention studies would be to test a combination of the elementary grade interventions used here with middle school curricula that seek to promote norms against health-risk behaviors such as drug use and to enhance life skills to resist health-compromising social influences.3,6,7,31

The present results indicate that a package of parenting education, teacher training, and skills training for children provided during the elementary grades can significantly improve the achievement of urban children and reduce the prevalence of violent crime, heavy drinking, and precocious sexual activity—behaviors with large costs to society. The cost of intervention over 6 years was $2991 per participant. These results suggest that the basic institutions of child socialization, families and schools, can be strengthened through the use of specific theory-grounded methods that empower parents and teachers to more effectively bond children to school during the elementary grades. We are currently following this panel to see the extent to which the differences observed at age 18 years translate into differences in employment, crime, substance abuse, marital stability, and parenting practices in adulthood.

Accepted for publication August 20, 1998.

Funded by the National Institute on Drug Abuse, Rockville, Md; Office of Juvenile Justice and Delinquency Prevention, Washington, DC; and the Robert Wood Johnson Foundation, Princeton, NJ.

An earlier version of this article was presented at a meeting of the Society for Life History Research, London, England, October 5, 1996.

We thank Philip M. Smith, MS, David P. Farrington, PhD, and Alan Leshner, PhD, for comments on drafts of the manuscript, and Michel Janosz, PhD, for his work on early analyses.

Reprints: J. David Hawkins, PhD, Social Development Research Group, University of Washington, 9725 Third Ave NE, Suite 401, Seattle, WA 98115.

Editor's Note: I think the message here is, "Start early, include everyone possible, and [I'd add] don't ever stop."—Catherine D. DeAngelis, MD

Greenwood  PModel  KRydell  CChiesa  J Diverting Children From a Life of Crime: Measuring Costs and Benefits.  Santa Monica, Calif RAND1996;
Farquhar  JFortmann  SFlora  J  et al.  Effects of communitywide education on cardiovascular disease risk factors: the Stanford Five-City Project.  JAMA. 1990;264359- 365Google ScholarCrossref
Hawkins  JCatalano  RMiller  J Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: implications for substance abuse prevention.  Psychol Bull. 1992;11264- 105Google ScholarCrossref
Mrazek  PedHaggerty  Red Reducing Risks for Mental Disorders: Frontiers for Preventive Intervention.  Washington, DC National Academy Press1994;
Dryfoos  J Adolescents at Risk: Prevalence and Prevention.  New York, NY Oxford University Press1990;
Resnick  MDBearman  PSBlum  RW  et al.  Protecting adolescents from harm: findings from the National Longitudinal Study on Adolescent Health.  JAMA. 1997;278823- 832Google ScholarCrossref
Ellickson  PBell  R Drug prevention in junior high: a multi-site longitudinal test.  Science. 1990;2471299- 1305Google ScholarCrossref
Pentz  MDwyer  JMacKinnon  D  et al.  A multicommunity trial for primary prevention of adolescent drug abuse: effects on drug use prevalence.  JAMA. 1989;2613259- 3266Google ScholarCrossref
Farrington  D Childhood aggression and adult violence: early precursors and later-life outcomes. Pepler  DRubin  Keds The Development and Treatment of Childhood Aggression. Hillsdale, NJ Lawrence A Erlbaum Associates1991;5- 29Google Scholar
Robins  L Sturdy childhood predictors of adult anti-social behavior: replications from longitudinal studies.  Psychol Med. 1978;8611- 622Google ScholarCrossref
Kellam  SRebok  GIalongo  NMayer  L The course and malleability of aggressive behavior from early first grade into middle school: results of a developmental epidemiology-based preventive trial.  J Child Psychol Psychiatr Allied Disciplines. 1994;35259- 281Google ScholarCrossref
Maguin  ELoeber  R Academic performance and delinquency. Tonry  Med Crime and Justice A Review of Research Chicago, Ill University of Chicago Press1996;20145- 264Google ScholarCrossref
Yoshikawa  H Prevention as cumulative protection: effects of early family support and education on chronic delinquency and its risks.  Psychol Bull. 1994;11528- 54Google ScholarCrossref
Widom  C The cycle of violence.  Science. 1989;244160- 166Google ScholarCrossref
Goodlad  J A Place Called School: Prospects for the Future.  New York, NY McGraw-Hill Book Co1984;
Blumstein A, Cohen J, Roth J, Visher C, eds., Criminal Careers and "Career Criminals."  Washington, DC National Academy Press1986;
Elliott  DHuizinga  DMenard  S Multiple Problem Youth: Delinquency, Substance Use and Mental Health Problems.  New York Springer-Verlag NY Inc1989;
Bloom  B Human Characteristics and School Learning.  New York, NY McGraw-Hill Book Co1976;
Brophy  JGood  T Teacher behavior and student achievement. Wittrock  MCed Handbook of Research on Training New York, NY Macmillan Publishers Co Inc1996;328- 375Google Scholar
Mortimore  P The positive effects of schooling. Rutter  Med Psychosocial Disturbances in Young People Challenges for Prevention. Cambridge, England Cambridge University Press1995;333- 363Google Scholar
Slavin  RKarweit  NWasik  B Preventing Early School Failure.  Needham Heights, Mass Allyn & Bacon1994;
Bank  LPatterson  GReid  J Delinquency prevention through training parents in family management.  Behav Analyst. 1987;1075- 82Google Scholar
Dumas  J Treating antisocial behavior in children: child and family approaches.  Clin Psychol Rev. 1989;9197- 222Google ScholarCrossref
Tremblay  RMcCord  JBoileau  H  et al.  Can disruptive boys be helped to become competent?  Psychiatry. 1991;54148- 161Google Scholar
Tremblay  RVitaro  FBertrand  L  et al.  Parent and child training to prevent early onset of delinquency: the Montreal Longitudinal-Experimental Study. McCord  JTremblay  Reds Preventing Antisocial Behavior Interventions From Birth Through Adolescence. New York, NY Guilford Press1992;117- 138Google Scholar
Catalano  RHawkins  J The social development model: a theory of antisocial behavior. Hawkins  JDed Delinquency and Crime Current Theories New York, NY Cambridge University Press1996;149- 197Google Scholar
Hawkins  JWeis  J The social development model: an integrated approach to delinquency prevention.  J Primary Prev. 1985;673- 97Google ScholarCrossref
Hawkins  JVon Cleve  ECatalano  R Reducing early childhood aggression: results of a primary prevention program.  J Am Acad Child Adolesc Psychiatry. 1991;30208- 217Google ScholarCrossref
Hawkins  JCatalano  RMorrison  DO'Donnell  JAbbott  RDay  L The Seattle Social Development Project: effects of the first four years on protective factors and problem behaviors. McCord  JTremblay  Reds The Prevention of Antisocial Behavior in Children. New York, NY Guilford Press1992;139- 161Google Scholar
O'Donnell  JHawkins  JCatalano  RAbbott  RDay  L Preventing school failure, drug use, and delinquency among low-income children: long-term intervention in elementary schools.  Am J Orthopsychiatry. 1995;6587- 100Google ScholarCrossref
Botvin  GBaker  EDusenbury  LBotvin  EDiaz  T Long-term follow-up results of a randomized drug abuse prevention trial in a white middle-class population.  JAMA. 1995;2731106- 1112Google ScholarCrossref
Kirby  D Programs to Reduce Teenage Pregnancy: Findings From Research.  Washington, DC National Campaign to Reduce Pregnancy1997;
Cummings  CBarber  CCuervo  A School Enhancement Research and Demonstration Project: Methods of Instruction: Teacher's Manual (Secondary).  Washington, DC Office of Juvenile Justice & Delinquency Prevention, US Dept of Justice1982;
Cummings  C Managing to Teach.  Edmonds, Wash Teaching Inc1983;
Hawkins  JDoueck  HLishner  D Changing teaching practices in mainstream classrooms to improve bonding and behavior of low achievers.  Am Educ Res J. 1988;2531- 50Google Scholar
Slavin  R Synthesis of research on cooperative learning.  Educ Leadership. 1991;4871- 82Google Scholar
Kerr  DKent  LLam  T Measuring program implementation with a classroom observation instrument: the interactive teaching map.  Eval Rev. 1985;9461- 482Google ScholarCrossref
Abbott  RDO'Donnell  JHawkins  JDHill  KGKosterman  RCatalano  RF Changing teaching practices to promote achievement and bonding to school.  Am J Orthopsychiatry. 1998;68542- 552Google ScholarCrossref
Shure  MSpivack  G Interpersonal problem solving as a mediator of behavioral adjustment in preschool and kindergarten children.  J Appl Dev Psychol. 1980;129- 44Google ScholarCrossref
Shure  MSpivack  G Interpersonal cognitive problem solving. Price  RCowen  ELorion  RRamos-McKay  Jeds 14 Ounces of Prevention A Casebook for Practitioners. Washington, DC American Psychological Association1988;69- 82Google Scholar
Comprehensive Health Education Foundation,Herman  E Here's Looking at You 2000.  Evanston, Ill Altschul Group Corp1986;
Hawkins  JCatalano  RJones  GFine  D Delinquency prevention through parent training: results and issues from work in progress. Wilson  JQLoury  GCeds From Children to Citizens Families, Schools, and Delinquency Prevention New York Springer-Verlag NY Inc1987;186- 204Google Scholar
Patterson  G Coercive Family Process.  Eugene, Ore Castalia1982;
Hawkins  JCatalano  RBrown  E  et al.  Preparing for the Drug (Free) Years: A Family Activity Book.  Seattle, Wash Comprehensive Health Education Foundation1988;
Dishion  TJKavanaugh  KKiesner  J Prevention of early adolescent substance abuse among high-risk youth: a multiple gating approach to parent interventions. Ahsery  RSRobertson  EBKumpher  KLeds Drug Abuse Prevention Through Family Interventions Rockville, Md National Institute on Drug Abuse1998;229- 259 NIDA Research Monograph 177 National Institute of Health publicaiton99- 4135Google Scholar
Kosterman  RHawkins  JSpoth  RHaggerty  KZhu  K Effects of a preventive parent-training intervention on observed family interactions: proximal outcomes from "Preparing for the Drug (Free) Years." J Community Psychol. 1997;25277- 292
Ellickson  PLBell  RMMcGuigan  K Preventing adolescent drug use: long-term results of a junior high program.  Am J Public Health. 1993;83856- 861Google ScholarCrossref