Familias Unidas

Study 1

The authors stated that "the Familias Unidas program lasted approximately 9 months."

Summary

Pantin et al. (2003) conducted a randomized controlled trial that examined 167 Hispanic families of sixth and seventh-grade students from three South Florida public schools. After random assignment to intervention and no-intervention control conditions, the students and parents completed five assessments, with the last occurring three months after the program ended.

Pantin et al. (2003) found that the intervention group, relative to the control group, showed significant increases in

• Behavior problems
• Parental investment.

Evaluation Methodology

Design:

Recruitment: The sample included 167 youths and their families who were recruited from three middle schools in low-income areas of Miami, Florida. Eligible students were Hispanic 6th and 7th graders with no history of psychiatric hospitalization, and they resided with at least one Hispanic immigrant parent planning to remain in the study area. Of the 330 families who indicated interest in the study, 167 agreed to participate and completed the baseline assessment.

Assignment: The study randomly assigned the youths and their families to either the Familias Unidas intervention (n = 96) or a no-treatment control condition (n = 71) in a 60/40 ratio. The assignment was done within school and grade strata.

Assessments/Attrition: The baseline assessments occurred on a rolling basis in the fall and winter of 6th or 7th grade. Four other assessments followed on a three-month interval (i.e., 3, 6, 9, and 12 months). Given the nine-month program, the last assessment came three months after the program ended. About 95% of participants completed assessments at three or more time points, about 88% completed all five assessments, and 98% completed the final 12-month assessment.

Sample:

The adolescents in the sample were 61% male and averaged 12.4 years of age. The largest percentage of participants were Cuban (39%), followed by Central and South Americans (29% and 17%, respectively), and a small proportion of Puerto Ricans/Dominicans (5%). The remaining 10% identified themselves as "Other" Hispanic. The majority of parents (94%) and half of the adolescents (49%) were born outside of the United States. The median annual household income was between $15,000 and $20,000, and the modal level of parent education was 12th grade (35%).

Measures:

The study examined three outcome composites:

• parental investment (e.g., positive parenting, perceived child support) with an alpha reliability of .90
• adolescent behavior problems (e.g., anger control, aggression, antisocial behavior) with an alpha reliability of .96
• adolescent school bonding/academic achievement with an alpha reliability of .80.

Each composite was derived by summing adolescent and parent-reported measures. Overall, parents provided four measures, and students provided eight measures. With parents helping to deliver the program, the composite measures were not fully independent.

 Analysis:

The analysis used mixed models for repeated measures with time-by-condition terms to test for condition differences in changes across the five assessments. Mixed models typically use FIML estimation to include participants with missing data.

Missing Data Method: Along with FIML estimation, missing responses for each variable at each assessment time were imputed with an expectation maximization algorithm, provided that the adolescent and/or parent provided some valid data at that assessment point.

Intent to Treat: The analyses included all participants regardless of the number of sessions or assessments they completed.

Outcomes

Fidelity:

A quarter of the group sessions were randomly selected for videotape rating by independent adherence raters. Of the 732 facilitator interventions rated, 89.4% were prescribed, whereas 10.6% were proscribed.

Baseline Equivalence:

Table 2 presents tests for three outcome measures. None of the differences reached the .05 significance level, but using a .20 significant level, the authors noted that the intervention group had significantly lower mean levels of adolescent problem behavior compared to the control group. The authors also mention one other test, finding no significant difference in the gender distribution of the two conditions.

Differential Attrition:

Attrition was low at the final assessment, only 2%. About 12% failed to complete all five assessments, but the rates did not differ significantly across conditions (14% in the control group and 10% in the intervention group) and met both the WWC cautious and optimistic standards. Still further, the use of FIML and the imputation of missing data likely moderated any attrition bias.

Posttest:

Tests for the time-by-condition terms showed two significant effects in three tests. The intervention group reported significantly greater improvement than the control group on parental investment and adolescent behavior problems. However, based on Figure 2, the two groups appear to not differ on behavior problems at the last assessment. Moderation tests suggested stronger program benefits for parents with high baseline parental investment.

Mediation tests indicated that the intervention effect on adolescent behavior problems was partly mediated by changes in parental investment.

Long-term:

Not examined.

Study 2

This study evaluated Familias Unidas when combined with another program and changed to address substance use and HIV risk behaviors. This study cannot be considered a replication of the Familias Unidas program.

Summary

Prado et al. (2007) evaluated the efficacy of Familias Unidas combined with Parent-Preadolescent Training for HIV Prevention (PATH), a Hispanic-specific, parent-centered intervention, in preventing adolescent substance use and unsafe sexual behavior. Adolescents and their families from three Florida middle schools were randomly assigned to one of three conditions: Familias Unidas Plus PATH, English for Speakers of Other Languages plus PATH, or English for Speakers of Other Languages plus HeartPower! for Hispanics.

Prado et al. (2007) found that the intervention group, relative to the control group, showed significantly lower

• Cigarette use at posttest and long-term
• Illicit drug use at posttest and long-term
• Family functioning at posttest and long-term.

Evaluation Methodology

Design:

Recruitment: Adolescents and their families were recruited for participation from three middle schools in the Miami-Dade County school system. Eligible students were in seventh grade and would advance in the next year to eighth grade. Eligible families of the students had at least one parent born in a Spanish-speaking country in the Americas, were not planning to leave the area, and could attend sessions on weekday evenings. Neither the adolescent nor the primary parent had ever been hospitalized for psychiatric reasons. Of the 649 potential families, 70 refused to participate. Of the remaining 579 families, 266 met the inclusion criteria and joined the study.

Assignment: This study used urn randomization to assign the students and families within schools to three conditions:

• the Familias Unidas plus PATH (n = 91),
• English for Speakers of Other Languages (ESOL) plus PATH (n = 84),
• ESOL plus HeartPower! for Hispanics (HEART) (n = 91).

The authors stated that Familias Unidas plus PATH served as the intervention group and the remaining two groups (ESOL plus PATH and ESOL plus HEART) served as attention control groups. The urn randomization was designed to balance the following adolescent characteristics: gender; years in the United States; having initiated substance use; and having initiated oral, vaginal, or anal sex.

Assessments/Attrition: Assessments occurred at five time points: baseline, 6, 12, 24, and 36 months post-baseline. The 12-month assessment served as a posttest, and the 24-month and 36-month assessments served as long-term follow-ups. About 86% of participants completed the posttest at 12 months, and 79% completed the long-term follow-up at 36 months.

Sample: The adolescents in the study were 48% male with a mean age of 13.4 years, and their primary caregivers were 87% female with a mean age of 40.9 years. Only 18.6% of the families reported a household income greater than $30,000 per year. Forty percent of the adolescents were born in the United States. Immigrant adolescents (n = 159) and their parents were born in Cuba (40%), Nicaragua (25%), Honduras (9%), Colombia (4%), and other Hispanic countries (22%). Parents of U.S.-born adolescents were born primarily in Nicaragua (33%), Cuba (20%), and Honduras (12%).

Measures: Adolescents and parents completed questionnaires, but parents did not rate their child's behavior.

Parent reports of family functioning were assessed with four indicators: parental involvement, positive parenting, family support, and parent-adolescent communication. Three of the four indicators had good reliability but the alpha value for family support was only .49.

Adolescents reported whether they had ever smoked, drunk alcohol, or used an illicit drug in their lifetime and the past 90 days and whether they had ever had sex in their lifetime and in the past 90 days. Adolescents who reported having had sex in the past 90 days were asked whether they had engaged in unprotected sex during that time and whether they had engaged in unprotected sex at last intercourse, had consumed alcohol or drugs before their last sexual intercourse, and had ever contracted a sexually transmitted disease.

Analysis:

The analysis examined growth curves using structural equation models and used FIML estimation. The models estimated individual trajectories of change over time to test for slope differences among the three study conditions. Models for the distal outcomes of adolescent substance use and sexual behavior included all five assessments. Models for the proximal outcomes of family functioning focused on changes between baseline and 12 months post-baseline.

Missing Data Method: FIML estimation allowed for the inclusion of participants with missing data.

Intent to Treat: The study continued to assess participants whether or not they had dropped out of the intervention and included all participants in the analysis.

Outcomes

Fidelity:

Ratings of videotaped group sessions, parent-adolescent discussion circles, and individualized family visits found that the average extensiveness/quality rating for all three intervention modules was "considerably/good." On a scale ranging from 0 to 6, the average adherence ratings were 3.72 for Familias Unidas, 3.64 for PATH, and 3.70 for HEART.

Baseline Equivalence:

Chi-square tests and analyses of variance for 18 baseline measures indicated no significant differences by condition on demographic characteristics, acculturation, family functioning, or alcohol use, cigarette use, illicit drug use, or unprotected sex in the past 90 days.

Differential Attrition:

The study did not test for differential attrition but used FIML in the growth curve analysis to adjust for potential bias from attrition. In addition, the differences in attrition rates between the Familias Unidas plus PATH condition and the two other conditions were small enough to meet both the WWC cautious and optimistic standards.

Posttest and Long-Term:

Alcohol Use: Growth curve analyses showed no significant differences in past 90-day alcohol use between Familias Unidas plus PATH and either of the other two conditions.

Cigarette Use: Growth curve analyses indicated significant differences in past 90-day cigarette use between Familias Unidas plus PATH and each of the two control conditions. The mean trajectory of smoking in Familias Unidas plus PATH decreased, while the mean trajectories of smoking increased for the remaining two groups.

Illicit Drug Use: Growth curve analyses indicated significant differences in past 90-day illicit drug use between Familias Unidas plus PATH and one of the two control conditions. The observed mean frequency of illicit drug use decreased in Familias Unidas plus PATH but increased in ESOL plus HEART between 24 and 36 months post-baseline.

Unprotected Sexual Behavior: Due to the small number of participants engaging in sexual behavior in the past 90 days, growth curve analyses were not estimated for past 90-day unprotected sex. Fisher's exact tests conducted at each time point indicated that there were no significant differences by condition for unprotected sexual behavior.

Family Functioning: Growth curve analyses indicated significant differences in family functioning between Familias Unidas plus PATH and both control conditions. The mean trajectory of family functioning in Familias Unidas plus PATH increased, while the mean trajectories of family functioning decreased in the control conditions..

Mediational Analyses: Mediational analyses found that changes in family functioning partially mediated the effect of the intervention condition on smoking and illicit drug use.

Post-Hoc Analyses: Additional analyses decomposed the Family Functioning factor score into its four component indicators (parent involvement, family support, positive parenting, and parent-adolescent communication) to explore the specific impacts of the programs. Of the four components, parent involvement and parent-adolescent communication were most improved by Familias Unidas plus PATH compared to the control conditions.

Study 3

The program included nine two-hour group sessions and ten one-hour family visits. Families also attended four one-hour booster sessions during the follow-up phase, at approximately 10, 16, 22, and 28 months post-baseline. Given the booster sessions, the study does not meet the criterion for sustainability.

Summary

Pantin et al. (2009) randomly assigned 213 eighth-grade students at-risk for problem behaviors in three Florida middle schools to treatment or control groups. Control families received three referrals to agencies in their catchment area that serve youth with behavior problems. Assessments were completed at baseline and at 6, 18, and 30 months post-baseline.

Pantin et al. (2009) found that the intervention group, relative to the control group, showed significantly

• Lower substance use
• Higher condom use among sexually active youth
• Greater improvements in family functioning (e.g., parent-adolescent communication, positive parenting, and parental monitoring of peers).

Evaluation Methodology

Design:

Recruitment: Adolescents and their families were recruited from three large, predominantly Hispanic middle schools in a single urban, low-income school district in Miami-Dade County, Florida. Eligible students had been rated on a parent screening measure as at least one standard deviation above average on conduct disorder, socialized aggression, or attention problems. Adolescents also had to be of Hispanic immigrant origin, be in the 8th grade, have an adult primary caregiver willing to participate in the study, and live within the catchment areas of one of the three middle schools included in the study. Of the 531 potential participants identified, 213 met the eligibility criteria and completed the baseline assessment.

Assignment: The study randomly assigned participants to the treatment group (n = 109) or a control group (n = 104) that received three referrals to agencies in their catchment area that serve youth with behavior problems. The study used urn randomization to balance adolescent gender, years in the United States, having initiated substance use, and having initiated oral, vaginal, or anal sex.

Assessments/Attrition: Participants were assessed at baseline and reassessed at 6, 18, and 30 months post baseline. However, booster sessions following the program continued through 28 months, making the last assessment a posttest rather than a long-term follow-up. At 30 months, the completion rate was 85%.

Sample:

The sample included 64% boys with a mean age of 13.8 years. Only 13.1% of the families reported household incomes of over $30,000 per year. A slight majority (56.1%) of adolescents were born in the U.S. Immigrant adolescents and parents were primarily born in Honduras (26.9%), Cuba (20.4%), and Nicaragua (16.1%).

Measures:

Adolescents and parents completed questionnaires. Parents rated their child's behavior for the measure of externalizing.

• A family functioning scale included five parent-reported indicators: parental involvement, positive parenting, family support, parent and adolescent communication, and parental monitoring. The subscales generally had acceptable reliabilities.
• Substance use was assessed from student reports on the use of cigarettes, alcohol, and illicit drugs in their lifetime and in the 30 days before the assessment.
• Sexual risk behaviors were measured from student reports on whether they had ever had sex (including vaginal, anal, and oral sex) in their lifetime and in the 90 days before assessment. Adolescents who reported having had sex were asked how often condoms were used.
• Externalizing disorders were measured with parent reports on attention deficit hyperactivity disorder, oppositional defiant disorder, and conduct disorder.

Analysis:

The analysis used growth curve models to compare trajectories of change in outcomes across conditions and included controls for baseline outcomes. The growth curve models allowed for missing data, with the assumption that the data are missing at random. Chi-square tests and ANOVAs were also utilized for some outcomes.

Missing Data Method: The growth curve models used FIML estimation to include participants with missing data.

Intent to Treat: The analysis included all participants.

Outcomes

Fidelity: Fidelity was assessed through videotaped sessions and was assessed as "considerably/good." The mean rating was 4.98 on a scale ranging from 0 to 6.

Baseline Equivalence:

Table 1 lists 17 tests with three significant differences at p < .05. Family functioning, positive parenting, and parental monitoring were higher in the control condition than in the intervention condition.

Differential Attrition:

The study did not test for differential attrition but used FIML in the growth curve analysis to adjust for potential bias from attrition. In addition, the difference in attrition rates between conditions was small enough to meet both the WWC cautious and optimistic standards.

Posttest:

Growth curve analysis found significantly smaller increases in substance use and significantly greater increases in condom use (among sexually active youth) for the intervention group than the control group. Parents in the intervention group reported significantly greater improvements in family functioning than the control condition. Further analyses suggested that family functioning mediated the effects of the intervention condition on substance use but not on unprotected sexual behavior.

Long-term:

Not examined.

Study 4

The program included eight two-hour parent group sessions and four one-hour family visits (with parents and adolescents) administered over three months. The three-month implementation period was shorter than the usual nine months.

Summary

Prado, Cordova et al. (2012) and Prado, Pantin et al. (2012) conducted a randomized controlled trial that examined 242 delinquent Hispanic adolescents (aged 12-17 years) from the Miami area. After randomly assigning participants to either the Familias Unidas intervention or a community practice control group, the study assessed substance use and risky sexual behavior at posttest (six months post-baseline), and nine-month follow-up (one-year post-baseline).

Prado, Cordova et al. (2012) and Prado, Pantin et al. (2012) found that the intervention group, relative to the control group, showed significantly

• Lower illicit drug use
• Fewer alcohol dependence diagnoses
• Safer sexual activity (e.g., condom use, number of partners)
• More improvement in family functioning (e.g., parent-adolescent communication, positive parenting, and parental monitoring of peers).

Evaluation Methodology

Design:

Recruitment: Study referrals came from the Miami-Dade County's Department of Juvenile Services and the Miami-Dade County Public School system. Recruiters screened 446 self-identified Hispanic youth and their primary caregivers. Of these, 310 met the study's eligibility criteria (Hispanic, 12-17 years old, delinquent, intending to stay in South Florida during the study period), and 242 (78%) consented to participate.

Assignment: The study randomly assigned participants to Familias Unidas (n = 120) or a community practice control group (n = 122) that received standard care services such as individual and family therapy available to youth and parents of delinquent youth in Miami-Dade County.

Assessments/Attrition: Assessments were conducted at baseline, posttest (three months after the intervention), and follow-up (nine months after the intervention). Completion rates were 96% at posttest and 95% at follow-up.

Sample Characteristics:

The sample was predominantly male (64%), with a mean age of 14 years. About 65% of the adolescents were born in the U.S. Immigrant adolescents and parents were primarily born in Cuba (25.0%), Honduras (15.5%), Nicaragua (9.5%), Puerto Rico (8.3%), and Dominican Republic (7.1%). Most of the youths (60%) came from families of relatively low SES, as indicated by a family income below $20,000 per year.

Measures:

The primary outcomes relating to risky sexual behavior and substance use came from adolescents, and secondary outcomes relating to family functioning and parenting came from parents. Individuals supervising the data collection process were blind to the condition assignments. The primary measures had been used by other researchers in published work.

Primary outcomes: Risky sexual behavior was measured with items relating to sex during the past 90 days, inconsistent condom use, sexual partners, sex under the influence of drugs and alcohol, alcohol use, illicit drug use, and alcohol and marijuana dependence.

Secondary Outcomes (mediators): Family functioning was assessed using an index composed of three subscales: Parenting Practices Scale (alpha = .72), Parent-Adolescent Communication Scale (alpha = .77), and Parent Relationship with Peer Group Scale (alpha = .84). Other measures included Parental Stress (alpha = .93) and Social Support for Parents (alpha = .91).

Analysis:

The posttest analysis (Prado, Pantin et al., 2012) employed generalized linear models with controls for baseline outcomes. Only participants who reported having had sex in the past 90 days were included in the analyses of sexual risk behaviors in the past 90 days.

The follow-up analysis (Prado, Cordova et al., 2012) used growth curve models to test for differences across the conditions in outcome trajectories over all three time points (pretest, posttest, follow-up). The estimates used full information maximum likelihood estimation to include missing data (under the assumption that the data are missing at random). Only sexually active youth were included in the analysis of having sex under the influence of alcohol or drugs.

Missing Data Method: Parado, Pantin et al. (2012) used complete-case analysis, and Pantin, Cordova et al. (2012) used FIML estimation.

Intent to Treat: The analyses included either all participants with complete data or all participants.

Outcomes

Implementation Fidelity:

Based on ratings of sample videotapes of family sessions, facilitators covered 91% of the behaviors detailed in the program manual. Of the eight sessions offered, the mean number of sessions attended was 6.88. Among those who attended at least one session (87%), the mean number of sessions was 7.9.

Baseline Equivalence:

Prado, Pantin et al. (2012) presented condition means at baseline but did not report tests for the significance of differences. Table 1 (p. 130) reveals substantial differences in sociodemographic and behavioral characteristics. For example, among foreign-born participants, 26.7% of the intervention group had been living in the U.S. for more than 10 years, while only 12.8% of the control group had been living in the U.S. for more than 10 years. For the outcome measures, a lower number in the intervention group (18.8%) reported unprotected anal sex at last sexual intercourse compared to the control group (42.9%).

Prado, Cordova et al. (2012) found one significant difference in ten tests for sociodemographic and outcome measures (Table 1). A higher proportion of youth in Familias Unidas had an alcohol dependence diagnosis, which was added as a control variable in the analysis. However, the analysis of baseline equivalence did not examine the measures in Prado, Pantin et al. (2012) that showed the large deviations.

Differential Attrition:

At posttest, attrition was low, only 4% (Prado, Pantin et al., 2012). The attrition rate did not differ significantly across the two conditions, and the condition difference was small enough to meet the WWC cautious and optimistic standards.

At the nine-month follow-up, attrition was low, only 5% (Prado, Cordova et al., 2012). The condition difference in attrition rates was not significant and was small enough to meet the WWC cautious and optimistic standards. Also, comparing completers and dropouts with main effects attrition analysis showed no significant differences in demographic variables or baseline outcomes.

Posttest:

At posttest, Prado, Pantin et al. (2012) reported significant effects favoring the intervention group on five of seven measures of HIV risk behavior, including inconsistent condom use during vaginal intercourse and the number of sexual partners. In addition, the results showed significant effects on the family functioning scale and positive parenting.

At posttest and nine-month follow-up, Prado, Cordova et al. (2012) found that the growth curves for the conditions differed significantly for four outcomes. The intervention group reported significantly better trends for substance use (alcohol or drug use), drug use, alcohol dependence diagnosis, and having had sexual intercourse under the influence of alcohol or drugs in the past 90 days.

Tests for moderation showed significant results for only 2 of 12 interactions tested. For the alcohol dependence outcome, the program worked for youth with low social support from parents but not for youth with high social support. For the illicit drug use outcome, the program worked for youth with high parental stress but not youth with low parental stress.

Long-Term:

Not examined.

Study 5

The structure of the Familias Unidas program was shortened from 12 weeks to six weeks in this study, though it appears the content was unchanged. Specifically, three of the eight parent group sessions were eliminated; the number of parent-homework assignments was reduced from six to three; and one of four family visits was delivered.

Summary

Estrada et al. (2015) randomly assigned 160 ninth-grade Hispanic students to either a shortened six-week Familias Unidas intervention condition or a community-practice control group. Assessments were completed at baseline and at six, 12, and 24 months post-intervention.

Estrada et al. (2015) found that the intervention group, relative to the control group, showed significantly

• Lower sexual initiation
• Improved adolescent-reported positive parenting scores at six-month follow-up.

Evaluation Methodology

Design:

Recruitment: The study recruited 160 Latino ninth-grade students and their primary caregivers in 2011 from four Florida public high schools.

Assignment: The families were randomly assigned to either the Familias Unidas intervention condition (n = 72) or the community practice condition (n = 88). The community practice condition functioned as a control group and consisted of the school-based HIV risk-reduction course offered to all students and delivered by health science teachers in a classroom format.

Attrition: Participants were assessed at six, 12, and 24 months from baseline (or at six-month posttest, six-month follow-up, and 18-month follow-up). The completion rate at the 18-month follow-up was 92.5%.

Sample:

The sample had an average age of 15.3 and was 51% male. Over half (54.5%) of the participants were born in the U.S., and a majority had a family income under $30,000 (86%). Foreign-born participants and their parents were predominantly born in Cuba (37%), Honduras (12.3%), and Nicaragua (9.6%). Most caregivers' preferred language was Spanish (89.4%), while most adolescents preferred English (73.1%).

Measures:

All measures came from adolescent self-reports.

Adolescent substance abuse was assessed using an adapted instrument that indicated any substance use in the previous 90 days, as well as lifetime numbers for cigarettes, alcohol, and other drugs. Adolescent sexual risk behavior was assessed using another adapted instrument that indicated whether adolescents had engaged in sexual activity in their lifetime, and the frequency of condom use during sexual activity during the 90 days prior to assessment. Parental involvement (alpha = .86) and positive parenting (alpha = .80) were measured with the Parenting Practices Scale, and parent-adolescent communication (alpha = .91) was measured with the Parent-Adolescent Communication Scale.

Analysis:

The analysis used growth curve modeling to estimate individual trajectories of change from baseline and to test for condition differences over time. Regression analyses were used to examine intervention effects on the family functioning variables.

Missing Data Method: Growth curve models typically use FIML estimation to include participants with missing data, but the authors say nothing about missing data and likely used complete cases in the analysis.

Intent-to-Treat: The authors did not discuss whether they dropped missing data or used maximum likelihood estimation to account for the missing data but noted that "for each of the outcomes included in the analyses, data from all four assessment time points were used." This likely means that the models included data from all four time points but could mean that they only analyzed cases with data at all four time points.

Outcomes

Implementation Fidelity:

The authors made no mention of implementation fidelity.

Baseline Equivalence:

Tests in Table 1 for 17 baseline demographic and outcome measures showed no significant differences by condition.

Differential Attrition:

The condition difference in attrition rates was not significant and met the WWC cautious and optimistic standards. The study did not test for differential attrition by baseline measures or by condition-times-baseline measures but may have used FIML to adjust for missing data bias.

Posttest and Long-term:

Primary outcomes:

From baseline to 18 months postintervention, the program showed no improvement on curbing unsafe sex practices, but youth in the intervention group showed a significantly lower sexual initiation rate (34%) compared to the control group (55%) at the last assessment. There were no significant differences in substance use or substance use initiation rates between the intervention and control groups. At the 6-month follow-up youth in the intervention group scored significantly higher on the positive parenting scale, but it is not discussed whether this continued to 24 months.

Moderating effects:

Gender: The program was more effective at preventing substance use initiation among girls in the intervention group (28.6%) compared to the control group (65.2%) than it was for boys. The program was also significantly associated with decreased alcohol use initiation among girls in the intervention group (30.4%) compared to the control group (64%).

Age: The program was significantly associated with reduced unsafe sex among adolescents aged 15 years or less, but not among older adolescents.

Parenting: The program was effective at increasing positive parenting scores for adolescents who reported low to moderate levels of positive parenting at baseline but not among adolescents who reported high positive parenting at baseline.

Study 6

Like Studies 4 and 5, Estrada et al. (2017) and Vidot et al. (2016) used the three-month program (instead of the original nine).

Summary

Estrada et al. (2017) randomly assigned 746 eighth-grade Hispanic students to either a three-month Familias Unidas intervention (n = 376) or a prevention-as-usual control group (n = 370). The authors assessed past 90-day substance use and risky sexual behavior at baseline and three, fifteen, and twenty-seven months after the program end. A supplemental study (Vidot et al., 2016) explored the role of parent-adolescent communication on suicide thoughts and attempts.

Estrada et al. (2017) and Vidot et al. (2016) found that the intervention group, relative to the control group, showed significantly

• Less frequent drug use in the last 90 days
• Increased condom use in the last 90 days
• Parental monitoring of peers and overall family functioning.

Evaluation Methodology

Design:

Recruitment: Estrada et al. (2017) examined students from 18 middle schools with Hispanic majorities in Miami-Dade County, Florida. Recruitment began in September 2010 while the intervention and follow-up assessments proceeded through June 2014. Inclusion criteria required students to claim Hispanic origin, be enrolled in eighth grade, have a participating primary caregiver, live in the catchment area of a participating middle school, and plan to live in South Florida during the study. Of the 989 assessed, 746 (75%) were eligible and consented.

Assignment: The authors used stratified randomization within school to assign the students to the Familias Unidas intervention group (n = 376) or the "prevention as usual" control group (n = 370).

Attrition: Participants were assessed at six, 18, and 30 months from baseline. Given the three-month program, the posttest came three months after the intervention ended, and follow-ups occurred 15 and 27 months after the intervention ended. Familias Unidas had a 13.3% attrition rate, while the control group had a 12.2% attrition rate. Vidot et al. (2016) reported overall retention rates of 93.4%, 90.5%, and 87.3% for each of the assessments.

Sample:

Most students self-reported as slightly less than 14 years old (about 13.8 years). The sample had a majority proportion of males (52.14%) and students born in the U.S. (54.83%). Most parents (67.56%) reported earning less than $30,000 annually and being born in a Spanish-speaking country in the Americas. Mothers were more likely to participate (83%), and the average parental age was 41.

Measures:

Students self-reported substance use and sexual behaviors. Youth were asked about alcohol use in the past 90 days and in their lifetime and about the use of illicit drugs (e.g., marijuana, LSD, cocaine) in the past 90 days and in their lifetime. They also reported sex in the past 90 days and in their lifetime, as well as frequency in the past 90 days of sex without a condom. The study pre-registration (NCT01038206) listed the following primary outcome measures:

·       Substance use will be assessed using items extracted from the Monitoring the Future Study.

·       Adolescent unsafe sexual behavior will be measured using Jemmott, Jemmott, and Fong's (1998) 37-item Sexual Behavior instrument.

Parents reported on family functioning using three indicators: monitoring and knowing their youths' friends, using positive parenting practices (e.g., rewards, acknowledgment) to respond to positive behaviors, and family communication.

Additionally, Vidot et al. (2016) assessed student self-reported suicide ideation ("thought seriously about killing yourself") and suicide attempts ("tried to kill yourself in the last year"). They also used a 20-item parent-reported measure of communication with their child (alpha = .82).

Analysis:

Outcomes were analyzed with linear growth curve models. The authors used time-varying zero-inflated Poisson growth models for continuous drug use outcomes, while the probit-LINK function (Mplus) was used to test binary outcomes for past 90-day alcohol use outcome. Vidot et al. (2016) also used growth curve models. Both studies adjusted for the clustering of students within schools.

Missing Data Method: The analysis used FIML estimation to adjust for missing data.

Intent-to-Treat: All subjects were included in most analyses. Estrada et al. (2017) dropped seven student participants (<1%) from the analysis of drug use in the past 90 days for outlying or invalid response patterns. Also, only those reporting past 90-day sexual activity were included in analyses for sex without a condom.

Outcomes

Implementation Fidelity:

Fidelity ratings were slightly higher for family visits (from 2.60 to 5.20) than parent groups (from 2.04 to 4.30) and averaged 3.61, but both scores appear low. Also, parents attended 6.4 of 12 sessions on average, and 12.9% of participants attended no sessions.

Baseline Equivalence:

Tests for baseline equivalence for 14 measures in Estrada et al. (2017) found two significant differences. The control student group was significantly more likely to report sex without condom use in the last 90 days than the intervention group. Control parents also reported higher family communication than parents in the Familias Unidas group. Vidot et al. (2016) noted no differences in demographic or suicide characteristics.

Differential Attrition:

Estrada et al. (2017) reported that attrition rates did not differ significantly across conditions. The difference in condition attrition rates was small enough to meet the WWC cautious and optimistic standards. Although Estrada et al. (2017) did not test for attrition differences by any of the baseline measures or their interaction with condition, Vidot et al. (2016) reported that, based on the Little test, the data were missing completely at random. Both reports used FIML estimation, which likely moderates the threat from differential attrition.

Posttest and Long-Term:

In Estrada et al. (2017), intervention students reported improved drug use frequency in the last 90 days compared with control students (d = .27). Among those ever having had sex, control participants reported significantly greater increases in past 90-day sexual activity without a condom than intervention participants (d = .98). The program did not significantly affect alcohol use or ever using drugs.

Also in Estrada et al. (2017), parents of youths assigned to the Familias Unidas group reported greater oversight of their child's friends (d = .14) and greater family functioning (d = .47) compared with those in the control group at the six-month follow-up. In addition, parental peer monitoring significantly mediated the intervention effect on drug use frequency.

Vidot et al. (2016) found no program effects overall on suicide ideation or suicide attempts. They did find a significant positive effect on parent communication overall and a significant negative effect on suicide attempts but only for students with low parent-adolescent communication scores at baseline.

Study 7

Like Studies 4, 5, and 6, Molleda et al. (2016) used the three-month program (instead of the original nine). The program was adapted to fit the language and culture of Ecuador but without changing any of the core components.

Summary

Molleda et al. (2016) randomly assigned 239 families from two schools in Ecuador to the 3-month Familias Unidas intervention (n = 129) or Community Practice control group (n = 110). Data were collected at baseline and at the conclusion of the three-month program. The authors assessed parent reports of adolescent conduct problems and family functioning (parent-adolescent communication and parental monitoring of peers).

Molleda et al. (2016) found that the intervention group, relative to the control group, showed significant

• Reduction in conduct problems at three months
• Improvement in parent-adolescent communication.

Evaluation Methodology

Design:

Recruitment: The study examined students from two public schools in the largest city in Ecuador, both of which were located in neighborhoods associated with low income, high crime rates, and high risk of substance use. Recruitment was open to all students ages 12-14 who consented to participate, except those with prior psychiatric treatment. Of 248 screened, 239 were eligible.

Assignment: The 239 students were randomly assigned within schools to the Familias Unidas intervention group (n = 129) or the Community Practice control group (n = 110). The authors noted that the randomization was done "without stratifying by school" but do not explain the uneven number of subjects in each group.

Attrition: Participants were assessed at baseline and after the 3-month intervention. About 88.7% of the randomized subjects completed the posttest.

Sample:

Most students self-reported as slightly less than 13 years old (about 12.87 years). Male participation (48.3%) was slightly less common than that of females. About one-third of parents were married (33.9%). Mothers were more likely to participate (93.7%), and average parental age was 37.21.

Measures:

The outcome measures came from parents who were involved in the program and learned strategies for dealing with their children; none came from the adolescents. Parent responses assessed family functioning on three indicators: parent-adolescent communication, monitoring or knowing their child's friends, and conduct problem (open disobedience, defiance, oppositional disorders, physical aggression, and difficulty controlling anger).

Analysis:

Outcomes were analyzed using a cross-lagged design to explore the direct and indirect effects of the intervention on the family functioning measures. The model controlled for baseline outcomes, and estimates from MPlus used robust standard errors.

Missing Data Method: Maximum likelihood estimation adjusted for missing data.

Intent-to-Treat: All subjects were included in the analyses.

Outcomes

Implementation Fidelity:

Trained staff members from the Catholic University of Santiago of Guayaquil evaluated all videotaped group sessions and 72% of the family sessions using a 7-point scale. Family sessions were rated slightly higher (4.58) than group sessions (4.5), though both session types suggest only modest fidelity to the program. Participants averaged attending 9.57 sessions of the 12 offered.

Baseline Equivalence:

Table 1 shows no significant differences for six demographic and outcome measures, but the table did not include the peer monitoring measure.

Differential Attrition:

The study did not test for attrition differences by any of the baseline measures or their interaction with condition but used maximum likelihood estimation to adjust for missing data. Also, the difference in condition attrition rates was small enough to meet the WWC cautious and optimistic standards.

Posttest and Long-Term:

Intervention group assignment resulted in a significant and direct effect on conduct problems at posttest (d = -.262) and a significant direct effect on parent-adolescent communication. In addition, the mediation analysis found a significant indirect treatment effect predicting behavior problems at three months based on parent-adolescent communication at three months.