IMPLEMENTATION OF EVIDENCE-BASED PREVENTION PROGRAMMING: VARIATIONS IN PRACTICE SETTINGS THAT AFFECT IMPLEMENTATION QUALITY
Chair: Mary Utne-O'Brien
IMPLEMENTATION OF EVIDENCE-BASED PREVENTION PROGRAMMING: VARIATIONS IN PRACTICE SETTINGS THAT AFFECT IMPLEMENTATION QUALITY. Mary Utne-O'Brien1, Jennifer Axelrod1, Elizabeth Devaney1, Ed Dulaney1, Kristy Ogren1, Manolya Tanyu1, Roger Weissberg1, 1CASEL- University of Illinois at Chicago, Chicago, IL United States
A critical question facing the prevention field is how to assure high-quality implementation in real-world settings of prevention programs found to work in efficacy trials (Kaftarian et al., 2004; Greenberg, 2004; Greenberg et al, in press; Dusenbury and Hansen, 2004). CASEL (the Collaborative for Academic, Social, and Emotional Learning) has examined this issue as it plays out in school settings for almost a decade (e.g., Elias et al, 1997; Zins, Elias and Greenberg, 2000a, 2000b; Elias et al, 2003). In addition to publishing on the topic, CASEL has:
Assembled an Implementation Working Group of very senior scientists and government officials with vast experience in prevention programming implementation and impacts.
Held a Working Conference to identify necessary next steps in implementation research and practice, and develop an agenda for the field to advance both implementation science and practice.
Convened educational leaders to identify factors that support and hinder implementation of SEL programming.
Conducted a cross-disciplinary literature review on implementation and sustainability.
Conducted a retrospective study of implementation and sustainability of SEL programs in 15 schools, one decade after initial implementation.
Working group members and educators reported their experiences with impediments to successful implementation. The eleven factors they reported will be discussed and include: Lack of clear or adequate conceptualization behind the program, and poor translation of the concepts into program components; the program and the needs it addresses are not the central focus of the school, and are not linked to the issues for which teachers are held accountable; inadequacy of staff training, staff development, and ongoing technical assistance; and school leadership—whether the principal acts as an instructional leader, publicly supports the program, goes into classrooms, makes clear the work is a school priority.CASEL leaders have worked to test these practice wisdoms in field trials (e.g., Kam et al., 2003), retrospective studies (Elias and Kamarinos, 2003) and in action research. CASEL´s current work in nine schools in inner-city, suburban, and rural Illinois, designed to provide implementation supports to schools committed to schoolwide SEL programming, is presented in this roundtable discussion. Cross-site observations of the ecological factors expected to affect implementation quality, the natural range of variation observed in these factors, efforts to ameliorate them, and impacts, are described. CASEL´s project team in this work and roundtable includes a 30-year principal, school psychologist, classroom teacher, community psychologist, sociologist and public health professional.
CONCURRENT 9, METHODS, Grouped Papers Application and Appropriateness of Mixture Models in Longitudinal Analysis of Substance Use Chair: Hanno Petras
APPLICATION OF MULTILEVEL LATENT CLASS ANALYSIS TO SCHOOL-BASED PREVENTION STUDIES. M. Van Horn1, Eric Brown2, J. David Hawkins2, Michael Arthur2, 1University of South Carolina, Columbia, SC United States; 2University of Washington, Seattle, WA United States
Recently, prevention scientists have begun using mixture models to identify different typologies of individuals and examine whether interventions may be differentially efficacious across types. This study extends the use of mixture models to cross sectional data where students are nested within schools using multilevel latent class analysis (MLCA; Vermunt, 2003). Analyses in this study focus on first identifying latent classes of students who engage in problem behaviors: second, evaluating the validity of these classes by examining differential prediction of class membership; and third, assessing the degree to which class membership varies across schools. Results may be used as the basis for examining differential intervention effectiveness across groups of students.
Data come from a sample of 494 schools from seven states and include 44,334 eighth grade students. Students´ self-reported use of alcohol, cigarettes, and marijuana, as well as self-reported binge drinking and eight serious delinquent acts (ranging from attacking someone to carrying a handgun to school) were included as indicators of the latent classes.
MLCA was conducted examining 1- to 9-class models. Information criteria and interpretability of the classes were the criteria for class enumeration. Ultimately, a 6-class solution was selected. Students in Class 1 (51%) reported essentially no problem behaviors. Class 2 (20%) consisted of alcohol and cigarette experimenters. Students in Class 3 (11%) reported current alcohol use, moderate binge drinking, and some delinquent acts. Class 4 (10%) consisted of students who engaged in problematic substance use and were much more likely to have attacked someone, sold drugs, and been suspended. Students in Class 5 (5%) reported the same levels of substance use as those in Class 2 (i.e., experimenters), but also reported high levels of delinquent acts. Class 6 (3%) consisted of students with very high levels of substance use and delinquency items.
To determine whether class formation was influenced by individual-level characteristics, we included gender as a predictor of class membership. Males were significantly more likely to be in Classes 5 and 6 than females (ORs = 3.4 and 4.5, respectively), there were no differences in the remaining classes. Intraclass correlations indicating variability in class membership across schools ranged from .01 in Class 2 to .06 in Class 1.
These analyses begin to provide a basis for using MLCA in school- or community-based prevention studies. They demonstrate that it is possible to identify multiple classes of students with cross sectional multilevel data. Further analysis will focus on examining multiple predictors of class membership and replicating the results in an independent sample.
LATENT TRANSITION ANALYSIS OF SUBSTANCE USE IN ADOLESCENCE. Nicholas Perrine1, Lisa Dierker1, Faiza Vesel1, 1Wesleyan University, Middletown, CT United States
Lifetime substance use data from the National Longitudinal Study of Adolescent Health was analyzed using Latent Transition Analysis (N=1772). Survey items assessing lifetime and recent (i.e., use within the past 30 days) alcohol, tobacco, and marijuana use were used to measure a Latent Class Model of substance use. A five class structure provided the best fit for the data at both time 1 and time 2 (G2 = 473.96, df=521, p=.93 and G2 = 481.99, df=521, p=.89, respectively). Latent Status 1=no substance use, Latent Status 2=tobacco only, Latent Status 3=alcohol only, Latent Status 4=alcohol and tobacco, Latent Status 5=alcohol, tobacco, and marijuana use. Change in latent status membership between time 1, when research participants were between the ages of 15 and 17, and time 2, when research participants were between the ages of 16 and 18, was modeled through the use of Latent Transition Analysis. The data augmentation function from WinLTA software was used to calculate parameter estimates and corresponding 95% confidence intervals so as to determine whether transition to the high risk latent status (LS5) was more likely to occur from the tobacco use only (LS2) or the alcohol use only (LS3) latent statuses. Accelerated progression to the high risk latent status (LS5) was equally as likely to occur from the tobacco only (LS2) and alcohol only (LS3) latent statuses. Specifically, results suggested that members of the tobacco only (LS2) latent status had a .09 (.06 - .13) probability compared to a .006 (.00 -.08) probability among the alcohol only (LS3) latent status of progressing to the high risk latent status. Among those transitioning out of the no use latent status at time 1, transitions to either the alcohol (LS3) or tobacco (LS2) only latent statuses were significantly more likely to occur than transitions to either the alcohol and tobacco (LS4) or high risk use (LS5) latent statuses. Specifically, results suggested that no use status (LS1) had a .10 (.08 - .13) probability of transitioning to the tobacco only (LS2) status and a .08 (.06 – 11) probability of transitioning to the alcohol only (LS3) status compared to a .03 (.01 - .04) and .03 (.02 – 05) probability of transitioning to either the alcohol and tobacco use (LS4) or the high risk (LS5) latent statuses, respectively. Results provide evidence to suggest that delaying early onset decreases the likelihood of progression to high risk substance use in late adolescence.
ASSESSING EFFECTIVENESS OF PREVENTION PROGRAMS USING GENERAL GROWTH MIXTURE MODELS: INFERRING CAUSE FROM TRAJECTORY CLASSES OF PROGRAM OUTCOMES. Sanjeev Sridharan1, Andrea Piesse1, 1Westat, Rockville, MD United States
In evaluation of prevention programs, prevention scientists are increasingly interested in developmental trajectories (growth patterns) of social and behavioral outcomes. In recent years, there has also been a focus on “latent classes” of trajectories (Muthen and Muthen, 2001; Nagin, in press) using General Growth Mixture Models. While such models have been used to describe longitudinal patterns of change, these approaches have also begun to be applied to study the causal effectiveness of prevention programs. The underlying causal argument is that programs can modify “natural” trajectories of program outcomes. We discuss the conditions that need to be satisfied before causal attribution could be made using such methods. While our focus is on randomized prevention trials we pay attention to issues of compliance to treatment and differential attrition. The strengths and limitations of General Growth Mixture Models in estimating program effectiveness is highlighted by studying the effectiveness of a preventions program focused on drinking behavior in college fraternities. As point of comparison, results from the General Growth Mixture Models are compared with : (1) Approaches using a propensity scoring framework (Rosenbaum, 2004); (ii) Targeted cluster or tree structured regression method (Friedman, 2002); (c) Standard multilevel models (Raudenbush and Bryk, 2002).
CONCURRENT 10, PROMOTING WELL-BEING, EMERGING OPPORTUNITES IN PREVENTION RESEARCH, Poster Forum
PARENTING PRACTICES AND ADOLESCENT DEVELOPMENT AND SEXUAL BEHAVIOR AND RELIGOSITY AS A MODERATOR ON COMMUNITY VIOLENCE AND SUBSTANCE ABUSE
Chair: Felipe Castro
THE TIME-VARYING EFFECTS OF PEER AND FAMILY SUPPORT ON ADOLESCENT DAILY MOOD STATES ACROSS ENTRY TO HIGH SCHOOL AND BEYOND. Sally Weinstein1, Donald Hedeker1, Robin Mermelstein1, Benjamin Hankin1, Brian Flay1, 1University of Illinois at Chicago, Chicago, IL United States
Much research suggests that social support promotes well-being in adolescence, but less is known about the time-varying effects of peer and family support on emotional functioning across adolescent development. Additionally, few longitudinal studies have examined mediators of the relationship between support and well-being using real-time measures of mood. The present study examined how peer and family support predict and change with daily mood over time, comparing a cohort of adolescents transitioning to high school (8th -9th grade, N = 268, 52% female) with a cohort transitioning from 10th-11th grade (N = 240, 58% female). Self-esteem was analyzed as a mediator of the support-affect relationship. It was hypothesized that (1) peer and family support would significantly predict adolescent positive and negative mood patterns over time, but (2) these effects would be moderated by age cohort, and (3) self-esteem would mediate the relationship between support and affect among all adolescents.
The present study examined real-time measures of daily positive and negative affect using Ecological Momentary Assessments via palmtop computers at baseline, 6-, and 12-months. Participants responded to random prompts 5 to 7 times/day for 7 consecutive days; when prompted, participants rated 12 mood adjectives on a Likert-type scale. Across participants, 49,000 observations of mood were analyzed. In addition, self-report questionnaires were administered at each wave to assess demographic information and perceptions of peer and family support. Data come from a longitudinal study on adolescent smoking; participants were selected based on smoking intentions and/or experience. 562 participants completed the baseline assessment wave, and 507 (90.2%) participated in the final wave at 12 months.
Mixed-effects regression models were used to examine peer and family support as time-varying covariates of longitudinal mood patterns. Results revealed that peer and family support contributed to well-being, but in complex ways. Significant peer support by age cohort by time interactions were found for positive and negative moods, with the 8th grade cohort showing greater increases in the support-affect relationship over time than the 10th grade cohort. Family support did not interact with cohort or time. Findings suggest that entry to high school marked a shift in the importance of peers to mood, but family support´s effects remained constant across development. Self-esteem, in part, explained these effects on well-being: multiple regression analyses indicated that self-esteem completely or partially mediated all support-affect relationships. Implications of findings for the content and timing of health promotion and prevention efforts will be discussed.
PARENTING PRACTICES AND ADOLESCENT SEXUAL BEHAVIOR. Melina Bersamin1, Deborah Fisher2, Doug Hill2, Joel Grube1, 1Prevention Research Center, Berkeley, CA United States; 2Pacific Institute for Research and Evaluation, Calverton, MD United States
Parents may strongly influence their adolescent´s sexual behavior through their ability to monitor, supervise, communicate and model healthy sexual behavior (Kotchick et al. 2001). For example, parental attitudes towards sexuality and mother-daughter sexual risk communications are strong predictors of adolescent sexual behavior (Hutchinson, 2002; Jaccard et al. 1996). Another potential influence on youth sexuality is exposure to televised sexual content, which parents may affect either directly through rules limiting viewing or indirectly through co-viewing and communication about content. Few studies have explored what specific facets of parenting are associated with the initiation of adolescent sexual behavior when a number of parenting variables are considered together. The current study builds upon previous research by examining whether and to what degree general factors related to parenting (quality of overall parent-child communication, monitoring/supervision of behavior) and parental behaviors specific to sex and media (communication about sex, attitudes, limitation of television viewing [LTV], discussion of television content) are associated with both initiation and age of initiation of oral sex and vaginal intercourse in early and middle adolescence. Based on previous research, we expected parental variables related to sex—parental communication and parents´ attitudes—to be the strongest predictors. Additionally, we expected that efforts to limit the amount and types of television programming watched will be more strongly associated with sexual outcomes than efforts to discuss television content. The hypotheses were tested using both multiple and logistic regression with data from a longitudinal survey of 1011 adolescents. The data were collected on two occasions one year apart using confidential computer assisted self-administered interviews. All explanatory variables are from the first wave whereas dependent variables are based on the second wave. The results indicate that after controlling for gender, race, age and sexual behavior at wave 1, LTV was negatively associated with adolescent initiation of oral sex (OR = .69, p<.01) and vaginal intercourse (OR= .62, p < .01). Parental communication about sex was negatively associated with vaginal intercourse (OR = .87, p < .05). Interestingly, none of the variables that were significantly associated with age of initiation of oral sex or vaginal intercourse in bivariate analyses were uniquely associated with age of initiation in multivariate analyses. These findings suggest that in addition to clear communication about sexual issues and monitoring of children´s general behavior, parents´ efforts to reduce youth sexual behavior should also focus on media influences and include limiting access.
RELIGIOSITY AS A MODERATOR ON THE RELATIONSHIP BETWEEN EXPOSURE TO COMMUNITY VIOLENCE AND SUBSTANCE ABUSE. Patrick Fowler1, Sawssan Ahmed1, Debra Jozefowicz-Simbeni1, 1Wayne State University, detroit, MI United States
This study tested the moderating role of public religiosity and private religiosity on the relationship between exposure to community violence and substance abuse/ dependence symptoms.
Participants were 273 at-risk adolescents, aged 18 to 22, from a large Midwestern metropolitan area. Analyses were based on 5-year follow up data from a larger, longitudinal study of homeless and housed adolescents. Initially, a probability sampling design was used to obtain a representative sample of homeless 13 to 17 year olds using shelters, in-patient and outpatient substance abuse treatment centers, psychiatric facilities, and street settings. A housed sample was obtained through peer nominations provided by the homeless adolescents. The two samples were matched on gender, age, ethnicity, and neighborhood socioeconomic characteristics. The five-year follow up represented the fifth wave of data collection.
Hierarchal regression analyses were used to test moderation, while controlling for differences in age, ethnicity, gender, prior homelessness, and concurrent life stressors. The Things I Have Seen or Heard scale measured exposure to community violence, including witnessing or being victimized by violence within participants´ neighborhoods. The General Religiosity Index assessed two aspects of religiosity. First, public religiosity referred to the amount of involvement in religious activities. Second, private religiosity indicated spirituality, including prayer, faith in God, etc. Substance abuse/ dependence symptoms were assessed using the 2nd Edition of the Diagnostic Interview Schedule for Children (DISC).
Results suggested that males, European Americans, and adolescents experiencing more concurrent stressful life events exhibited greater substance abuse/ dependence symptoms. Exposure to community violence predicted more substance abuse symptoms, while higher public religiosity predicted fewer substance abuse symptoms. Furthermore, public religiosity moderated the relationship between exposure to community violence and substance abuse symptoms. Adolescents in high violent neighborhoods endorsed significantly more substance abuse symptoms if they exhibited low public religiosity.
These findings suggested that a lack of public religiosity puts adolescents at greater risk of substance abuse associated with exposure to community violence. Interventions aimed at youth in violent neighborhoods should recognize and address this risk factor. In addition, the protective qualities of religiosity should be further examined.
DOES TELEVISION INFLUENCE ADOLESCENTS´ SEXUAL BELIEFS AND BEHAVIORS?. Samantha Walker1, Joel Grube1, 1Prevention Research Center, Berkeley, CA United States
There is considerable concern about young people's exposure to sexual content on television and the potential effects it may have on their sexual beliefs and behaviors. The evidence regarding the effects of such exposure on adolescents is limited, however. The present paper reports on data from two waves of an ongoing longitudinal survey of adolescents living in the greater San Francisco Bay Area and Los Angeles County in California. The data were collected using computer-assisted self-administered interviews (CASI) conducted in the home at one year intervals. Interviews were completed with 1,105 12-16 year olds at Wave 1 (48% female) for a response rate of 75%. At Wave 2 a total of 1,012 interviews were completed (92%). Cross lagged structural equation model predicting Wave 2 sexual behavior with latent variables was used to investigate the relationship between frequency of exposure to sexuality on television and changes in normative beliefs regarding sex (perceived friend and parent approval, perceived friend behavior), sex expectancies, and frequency of sexual behavior (e.g., oral sex, vaginal intercourse, and anal intercourse. Age, gender, ethnicity, and other important demographic and individual differences were controlled. The findings indicate that frequency of exposure to adult and high sexual content programming on television at Wave 2 was predicted by more positive peer attitudes toward sexual behavior and greater frequency of watch television at Wave 1. Normative beliefs about peer attitudes and sexual behavior, sex expectancies and frequency of exposure to television and frequency of exposure to adult and high sexual content television programs predicted frequency of sexual behavior at Wave 2. Implications for prevention are discussed.
FAMILISM VALUES AS A MODERATOR OF THE RELATION BETWEEN DEVIANT PEER EXPOSURE AND PSYCHOLOGICAL PROBLEMS AMONG MEXICAN-AMERICAN YOUTH. Miguelina Germán1, Nancy A. Gonzales1, Mark Roosa1, Larry Dumka1, 1Arizona State University, Tempe, AZ United States
This study examined the influence of familism values on psychological problems among a sample of low-income, Mexican-American adolescents. Both a mother´s familism values and a youth´s familism values were hypothesized to function protectively, both as a buffering variable working to reduce the risk of deviant peer exposure, and as a promotive factor benefiting all youth, independent of risk status.
Deviant peer exposure has been shown to be a risk factor for Mexican-origin adolescents on a variety of negative outcomes including substance use (Brooks, et al., 1998) and externalizing and internalizing symptoms (Barrera et al., 2002). Traditional Latino family values, often referred to as familism, may be a potential buffer of the effects of deviant peer exposure on mental health outcomes among Mexican-American youth (Brook et al., 1998; Gil et al., 2000). Familism values may work to reduce the negative impact of deviant peer exposure on youth problem behaviors through its influence on intraindividual factors (youth self-control and parental monitoring) and interpersonal factors (parent-child attachment, social support, and family cohesion).
The cross-sectional study consists of 497 youth ages 9-15 (48% female) and their maternal caregivers recruited from public schools in a large, Southwestern city. Mothers and youth completed the Child Behavior Checklist Inventory and the Youth Self Report Inventory, respectively (Achenbach & Edelbrock, 1983). Additionally, each participant completed a 16-item familism scale taken from a larger measure of values associated with acculturation and enculturation for Mexican Americans. Children also completed a deviant peer exposure measure (Barrera et al., 2001).
Hierarchical regression analyses revealed that a mother´s familism values function as a buffer to reduce the negative impact of deviant peer exposure on her child´s externalizing behaviors. Additionally, youth and maternal familism values showed a direct, negative relation to youth externalizing outcomes. The interaction effect suggests that maternal familism values make Mexican-American youth less vulnerable to the negative effects of having friends who engage in delinquent behavior. In this model, the protective buffering factor is internal to the parent but external to the child. This implies that mothers with relatively higher familism values may be engaging in parenting processes which result in a reduced association between deviant peer exposure and the externalizing problem behaviors of their children. This finding has important implications for interventions intended to reduce delinquent and aggressive behavior in Latino youth.