Strong African American Families – Teen

Study 1

Summary

Brody et al. (2012), Kogan, Brody et al. (2012), Kogan, Yu et al. (2012), and Ingels et al. (2013) conducted a randomized controlled trial to examine 502 African American families with a youth aged 15-16 and residing in six rural counties in Georgia. The families were randomized to intervention and control groups. Assessment of self-reported parent management skills occurred at pretest and posttest (two months after the intervention ended), and assessment of self-reported youth outcomes occurred at pretest and an average of 22 months after baseline.

Brody et al. (2012), Kogan, Brody et al. (2012), Kogan, Yu et al. (2012), and Ingels et al. (2013) found that, relative to the control group, the program significantly improved self-reported youth:

• Substance use
• Substance use problems
• Conduct problems
• Alcohol use
• Depression
• Frequency of unprotected sex.

Risk and protective factors:

• Parent-reported family management skills
• Youth-reported condom efficacy.

Evaluation Methodology

Design:

Recruitment: Participating families residing in six rural counties in Georgia were recruited from lists of 10th graders provided by public high schools. Eligibility requirements included having a youth aged 15-16 and self-identification as African American. Of the 692 families screened, 632 (91%) were eligible to participate and 502 (79%) of those eligible agreed to take part in the study. The statement (Kogan, Brody et al., 2012, p. 211) that "refusal rates were similar across conditions" implies consent came after assignment to conditions, but the participant flow diagram shows assignment after agreement to participate. Data collection began in 2007 and ended in February 2010.

Assignment: The study randomly assigned the 502 families to the intervention (N = 252) and control (N = 250) groups. The control group consisted of an alternative treatment call Fuel for Families, which was identical to SAAF-T in duration, format, and use of videotaped presentations and interactive components. The control treatment, however, focused on health promotion topics that, unlike the intervention, did not include content about substance use or sexual risk reduction and did not target family relationships. The design thus controlled for non-specific attention factors that can influence outcomes.

Attrition: Families provided data at pretest, 5-month posttest administered 2 months after the intervention ended, and 22-month (on average) follow-up (about 1.5 years after the program ended). Of the 502 randomized families, 482 (96%) completed the posttest (Kogan, Brody et al., 2012), and 478 (95%) completed the long-term follow-up (Brody et al., 2012). Ingels et al. (2013) included 473 participants (94%) "who completed baseline and follow-up assessments."

Sample: The mean age was 16.0 years for the youth and 43.1 for the primary caregivers. In 55.8% of the families, the target youth was a girl. Single mothers headed 55.5% of the families. A majority of the primary caregivers, 74.6%, had completed high school or earned a GED; 25.4% did not complete high school. With mean household monthly gross income of $1,482, approximately 64% of study families lived below the poverty threshold. They had an average of 2.5 children. As most caregivers worked, they can be described as working poor.

Measures: Self-report questionnaires were administered at posttest to a primary caregiver and the target youth via audio computer-assisted self-interviewing technology on laptop computers. Researchers assisting in data collection were blind to condition.

The posttest study (Kogan, Brody et al., 2012) measured parent-reported protective family management skills as a single latent construct with four component scales:

• Parental communication of expectations on substance use and sexual behavior (e.g., "I have clear and specific rules about my teen's association with peers who use alcohol"). Cronbach's alpha exceeded .92 at pretest and posttest.
• Discussion Quality Scale with two items on how often parents discussed sex and substance use with their teens. The items comprising the subscale were intercorrelated at .36 (p<.001) at pretest and .46 (p<.001) at posttest.
• Parental Academic Involvement on behaviors such as "Talk about the importance of finishing high school," "Discuss school activities with your teen," and "Tell your teen that education is the key to being successful." Cronbach's alpha exceeded .89 at pretest and posttest.
• Effective Problem Solving Scale of behaviors such as, "Listen to his/ her ideas about how to solve the problem," "Insist that your teen agree to your solution to the problem," and "Show a real interest in helping him/her solve the problem." Cronbach's alpha was .79 at both pretest and posttest.

The follow-up study of youth (Brody et al., 2012) measured four self-reported outcomes. The measures had been used previously, but the study reported little information on validity or reliability.

• Conduct problems were measured with 14 items on the frequency during the past 6 months that the youth had fought, stolen, been truant from school, or been suspended from school.
• Substance use was measured as the frequency over the past 3 months that the youth had consumed alcohol, consumed three or more drinks at one time, smoked marijuana, and smoked cigarettes. The items were summed to a single composite index.
• Substance use problems were measured with a composite index on the number of times in the past 12 months that youth had used substances in hazardous situations; failed to fulfill role obligations because of substance use; or experienced legal, social, or interpersonal problems because of substance use.
• Depressive symptoms were assessed with the 20-item Center for Epidemiologic Studies Depression Scale, a self-rated measure of symptoms occurring during the previous week.

In the study of sexual risk behavior, Kogan, Yu et al. (2012) examined unprotected intercourse and condom efficacy. Youth reported the number of times they had sexual intercourse during the past 3 months and the number of times condoms were used. Subtracting protected episodes from total episodes yielded a count of unprotected episodes. Condom efficacy was assessed using a 6-item scale (e.g., "How much of a problem would it be for you to unroll a condom down correctly on the first try?"), with Cronbach's alpha ranging from .87 to .90.

Ingels et al. (2013) measured two outcomes over the previous three months: the number of times having consumed at least one drink of alcohol and the number of times having consumed three or more drinks at one time.

Analysis: In the analysis of family management skills, Kogan, Brody et al. (2012) used structural equation models with full information likelihood estimation to deal with missing data and control for the pretest outcome. Effect sizes were reported as standardized betas.

In the analysis of long-term youth outcomes, Brody et al. (2012) used zero-inflated Poisson regression models for count measures; the models present coefficients for both a binary yes/no indicator and the frequency of occurrence. Ordinary least squares was used for the logarithm of depressive symptoms. All models controlled for baseline outcomes.

In the analysis of sexual risk behavior, Kogan, Yu et al. (2012) used complier average causal effect models to "provide unbiased estimates of causal effects for a full dose of an intervention while accounting for self-selection factors that would bias 'as-treated' analyses." The models used full information likelihood estimation for all available data to identify a group of compliers in the intervention condition and an equivalent group of compliers in the attention-control condition. In each model, intervention dose was included as a covariate to increase the precision of the model estimates.

The complier average causal effect models used zero-inflated Poisson estimation for any unprotected sex (yes/no) and for the count of unprotected sex. Ordinary least squares regression was used for the logarithm of condom efficacy. All models included the baseline outcome as a covariate.

Ingels et al. (2013) used interval regression for the two ordinal outcome measures of episodes of alcohol use and binge drinking. The regressions included pretest measures of alcohol use or binge drinking as covariates. The authors checked for clustering within sites, which differed by the county and day of the week of the sessions. They found "no significant intra-class correlation among the participants" and did not adjust for clustering.

Intent-to-Treat: The analysis of parents in Kogan, Brody et al. (2012) used all randomized subjects. Brody et al. (2012) dropped only the 5% without follow-up data, with one exception. Ingels et al. (2013) used all participants with complete data at baseline and follow-up. The analysis of substance use problems excluded subjects who had not used substances at baseline. In the evaluation of the optional unit on sexual health, Kogan, Yu et al. (2012) used the complier average causal effect analysis of attenders to adjust for not having an intent-to-treat sample.

Outcomes

Implementation Fidelity: All sessions were videotaped and used to provide constructive feedback to group leaders. Coverage of the curriculum components exceeded 80% for both intervention and control sessions. Mean attendance was approximately four of five sessions; only 32 families (6.3%) declined to attend any intervention sessions.

Of the 252 intervention families, 175 (69.4%) participated in the optional condom skills unit.

Baseline Equivalence: Kogan, Brody et al. (2012, p. 212) found no significant differences on three sociodemographic and four outcome measures (Table 2). Brody et al. (2012) reported finding no condition differences on baseline variables, including socioeconomic risk, gender, and five outcomes. Kogan, Yu et al. (2012) examined baseline equivalence for the complier intervention and control samples. These groups did not differ significantly on eight background and outcome measures.

Differential Attrition: Attrition was low (4-5%), and no differences emerged on study variables or demographic characteristics based on attrition status at posttest (Kogan, Brody et al., 2012) or follow-up (Brody et al., 2012; Kogan, Yu et al., 2012). In Ingels et al. (2013), the 6% without both baseline and follow-up data did not differ on baseline alcohol use or demographic variables from those with complete data.

Posttest: In Kogan, Brody et al. (2012), the program significantly improved parent-reported family management skills at posttest relative to the attention-intervention control group. The program effect was small (beta = .10).

Long-Term: In Brody et al. (2012), the results for long-term youth outcomes showed significant effects on the count portion of the zero-inflated Poisson regression analyses for self-reported conduct problems, substance use, and substance use problems. Participation in the program was associated with a 36% decrease in the frequency of conduct problems, a 32% decrease in substance use, and a 47% decrease in substance use problems (among those having used substances at baseline). However, the program did not affect the binary outcomes of having a conduct problem, using substances, or having substance use problems. In a regression analysis, intervention subjects reported significantly fewer depressive symptoms (4.5%) than the control group.

A validity check of the results examined program effects on a measure of healthful behaviors not related to program content. That the control group did significantly better on this outcome suggests that the attention-control group served as an effective comparison.

The sexual behavior results for the complier analysis showed a significant reduction in the frequency of unprotected intercourse but not in the binary (yes, no) outcome. They also showed a significant positive effect on condom efficacy. No gender interactions emerged, suggesting that the intervention worked equally well for male and female youth.

Ingels et al. (2013) found that intervention adolescents reported significantly fewer episodes of alcohol use and binge drinking than control adolescents.