Study
O’Neill, Clark, and Jones (2011) evaluated the effects of the Michigan Model for Health® (MMH) using a pretest/posttest control group design. Fifty-eight schools (42 from Michigan and 16 from Indiana) were randomly assigned to a group that implemented the MMH (the intervention group) or a group that did not implement the MMH (the control group). The schools were characterized as urban (28 percent), rural (31 percent), and suburban (41 percent). Although schools were the unit of randomization, data about program effects was collected and analyzed at the student level. The final evaluation sample consisted of 2,512 students (1,345 in the intervention schools and 1,167 in the control schools). The sample consisted of 54 percent boys with an average age of 9.56 years. The racial/ethnic composition was 54 percent white, 38 percent African American, and 8 percent other or mixed ethnicity. There were no significant differences between the intervention and control groups on gender, race/ethnicity, and all outcome variables.
The MMH intervention was implemented in classrooms over a 12-week period in grade 4 during the 2006–07 school year, and a 14-week period in grade 5 during the 2007–08 school year. Pretests were administered one week prior to the implementation of the intervention, and posttests were administered 5 to 6 weeks after the conclusion of the intervention.
A self-report questionnaire was developed to assess knowledge, skills, intentions, and behaviors related to the MMH. Health-promoting skills were measured using selected-response items developed from the State Collaborative on Assessment and Student Standards-Health Education Assessment Project. Aggressive behavior and drug use were measured using items from the Youth Risk Behavior Survey. Drug use intentions and prosocial behavior were measured using items from previous research (Hansen and McNeal 1997; Bosworth and Espelage 1995).
Statistical tests examining the effectiveness of the MMH involved two different analytic techniques, depending on the type of dependent variable. Analysis of continuous variables involved using a mixed model approach. The models took the general form of the measures nested within student and school with the four factors (treatment condition, gender, time, and ethnicity) having main effects and interactions, and with time as a repeated measure. Analysis of the four drug use outcomes (lifetime and recent alcohol and tobacco use) involved a different approach. The scores on these variables were converted to dichotomous measures (presence or absence of use) and used as the dependent variables in a binary logistic regression.
One limitation to the study is attrition. A number of students pretested in the fourth grade were not available for study in the fifth grade (n = 919) and were excluded from the study. Fourth-grade pretest demographic and outcomes scores for this group were compared to those who completed all fourth- and fifth-grade tests (n = 749). Attrition was not significantly associated with treatment and age, but was associated with gender and ethnicity (attrition rates were lower for girls compared with boys, and lower for African Americans and other minorities compared with Caucasians). There were also significant differences on the outcome variables. Students not retained through the fifth-grade intervention reported higher levels of lifetime and recent use of alcohol and tobacco, and exhibited lower social-emotional skills, interpersonal communication skills, and drug refusals skills than students who were retained. No subgroup analyses were conducted.
