Scientific EvaluationFollowing the content revisions to the original SFP and the subsequent feasibility studies (Kumpfer, Molgaard, and Spoth, 1996), a large-scale prevention trial of the ISFP, including long-term followup evaluations, was conducted in public schools in the rural Midwest. In addition to a posttest, followup data collections were completed approximately 1½, 2½, and 4 years after pretesting. Selected schools were located in rural communities with populations of less than 8,500; these communities had a relatively high percentage of low-income families participating in the school lunch program.
The experimental design entailed random assignment of 33 schools to 1 of 3 conditions: (1) the ISFP; (2) Preparing for the Drug Free Years (PDFY) (Catalano and Hawkins, 1996), a 5-session youth and family program; or (3) a minimal-contact control condition, with a total of 22 schools in the ISFP and control conditions. Families in the control condition received a set of four parenting guidelines written by Cooperative Extension Service personnel; the guidelines provided information on developmental changes of preteens and teens in physical, emotional, cognitive, and relational domains. (See Spoth, Redmond, and Shin, 1998, for further information on evaluation design and methods.) The results that follow include those from ISFP families and control families.3
Measures and Data Collection
A total of 161 families participated in 21 ISFP groups at 11 different schools. Groups ranged in size from 3 to 15 families; the average group consisted of 8 families, with an average of 12 adults and 8 youth. Both single-parent and two-parent families participated. Participation rates were high among pretested families. Among more than half of the two-parent families, both parents attended at least some of the sessions. Ninety-four percent of attending pretested families were represented by a family member in five or more sessions.4 Unannounced observations of the program implementation procedures of each team of group leaders confirmed that the teams covered all of the key program concepts.
Following confirmation that randomization resulted in equivalent groups at pretest, differential attrition was assessed and found to be nonsignificant. Subsequent posttest analyses of parenting outcomes included indicators for three constructs: intervention-targeted parenting behaviors (ITPB's), general child management, and parent-child affective quality. A latent variable measurement model incorporating multiple self-report and observational indicators of the parenting outcomes was developed. This measurement model was then employed in a test of a theory-based structural model examining a sequence of direct and indirect intervention effects. According to this model, ITPB's were expected to be directly affected by the intervention, whereas parent-child affective quality and general child management were expected to be primarily indirectly influenced by the intervention through effects on ITPB's. Results supported the hypothesized model. When controlling for pretest levels and measurement method effects, all hypothesized effect paths at posttesting were significant at the 0.01 level. The effect size of the intervention on the ITPB's was moderate, and the indirect effects of the intervention on the global parenting constructs were statistically significant (Russell et al., 1998; Spoth, Redmond, and Shin, 1998). Similar results were obtained when the model was evaluated with 1½-year followup data (Redmond et al., 1999).
The two general parenting constructs employed in the structural equation model (general child management and parent-child affective quality) were also assessed individually through multilevel analysis of covariance (ANCOVA), employing 1½- and 2½-year postbaseline followup assessment data. Reports from the target child, mother, and father, along with observer ratings, were standardized and combined to construct the measures for the 1½-year followup analyses; all except observer ratings were also available for the 2½-year followup analyses. Results indicated significant parenting outcome differences between the intervention and control groups at both the 1½-year followup (on ITPB's and parent-child affective quality) and the 2½-year followup (parent-child affective quality) (Spoth, Redmond, and Project Family Research Group, 1997).
Because of the young age of children in the study, significant intervention-control differences in problem behavior outcomes were expected to be initially detectable at the 1½-year postbaseline followup and in subsequent assessments. Consistent with this expectation, multilevel ANCOVA's showed significant intervention-control differences in substance use, conduct problems (e.g., physical aggression, minor theft, property damage), school-related problem behaviors (e.g., truancy, cheating), peer resistance, and affiliation with antisocial peers at 1½ and 2½ years following pretesting (Spoth, Redmond, and Project Family Research Group, 1997, 1998). Detailed analyses of individual substance use behaviors showed noteworthy differences between the intervention and control groups. For example, at the 1½-year followup, there was a 60-percent relative reduction in the first-time use of alcohol without parental permission in the intervention group (Spoth, Redmond, and Lepper, 1999). Continued divergence of the intervention group and the control group on this outcome variable has been observed at the 2½-year postbaseline followup and beyond (see figure 1).
In addition, the probability of transitioning from nonuse of tobacco, alcohol, or other drugs was examined. This examination focused on a five-status, stage-sequential model of substance use initiation and progression, adapted from a model previously tested by Graham and colleagues (1991). Findings suggested that ISFP group children who had not initiated substance use at the 1½-year followup assessment were significantly less likely to initiate use by the 2½-year followup assessment than were control group children (Spoth et al., 1999). Finally, recent analyses of substance use initiation and levels of use (alcohol, tobacco, and marijuana) and other problem behaviors (hostile and aggressive behaviors) at 4 years postbaseline have also shown positive results. Specifically, as compared with youth in the control group, those in the ISFP group showed significantly delayed initiation of alcohol, tobacco, and marijuana use (Spoth, Redmond, and Shin, 2000a); lower frequency of alcohol and tobacco use (Spoth, Redmond, and Shin, 2000a); and lower levels of overt and covert aggressive behaviors and hostility in interactions with parents (Spoth, Redmond, and Shin, 2000b). (See figure 1 above for a comparison of intervention and control group alcohol use initiation rates across data collection points.)
3 For results from PDFY, see Haggerty et al. (1999).
4 Other findings concerning participation in ISFP can be found in Spoth, Goldberg, and Redmond (1999); Spoth et al. (1996, 1997); and Spoth (1999).