A Stage-Based Expert System for Teen Dating Violence Prevention

Male-to-female intimate partner violence accounts for 26% of violence-related injuries in women presenting in hospital emergency departments and 33% of all female homicides. Adolescence provides an excellent "window of opportunity" for the prevention of intimate partner violence. Patterns of relating in intimate relationships are still relatively undifferentiated and open to influence. However, the evidence supporting traditional, school-based programs for the prevention of teen dating violence is mixed. A major problem with existing programs is that they are "one size fits all," making it difficult to meet the diverse needs of students-boys and girls, individuals who are dating and those who are not, individuals who have experienced dating violence as a victim, perpetrator, or both, and those who have not. Perhaps most importantly, these interventions neglect individual differences in readiness to use healthy, non-violent ways of relating to stay violence-free. In Phase I the objective was to use expert system technology to integrate best practices for teen dating violence prevention with the Transtheoretical Model of Behavior Change (TTM), the "stage model," to develop an interactive, multimedia computer-administered change program that delivers individualized intervention sessions and exercises tailored to stage of change and other individual characteristics. In Phase II, the objective was to complete development of the intervention package and assess its efficacy in a randomized clinical trial involving 3,901 teens from 20 Rhode Island high schools randomly assigned to intervention or comparison. Among youth exposed to risk for dating violence, efficacy was assessed by comparing the intervention and comparison groups on dating violence perpetration and victimization at follow-up. Among youth not exposed to risk for dating violence, efficacy was assessed by comparing intervention and comparison on peer violence perpetration and victimization.

Twenty Rhode Island high schools agreed to participate in the study. A Multiattribute Utility Measurement Approach (Graham, Flay, Johnson, Hansen, & Collins, 1984) was used to ensure that schools assigned to the intervention and comparison conditions were approximately equivalent on school size, student ethnicity, socioeconomic status (SES, percent receiving free or reduced-price lunch), attendance, mobility, truancy, dropout, and standardized test performance. The most similar schools were paired, and one school within each pair was randomly assigned to intervention, and the other to comparison. The intervention trial was launched in the Fall of 2009. Both the intervention and comparison groups completed a computerized baseline assessment and two follow-up assessments approximately 6 and 12 months later-in the Spring and Fall of 2010. Students assigned to the intervention condition completed their first Teen Choices intervention session immediately following their baseline assessment. The second and third intervention sessions were administered at approximately 1 and 2 months follow-up. Students assigned to the comparison condition completed an alternative evidence-based online TTM-based intervention, Health In Motion, which targets physical activity, screen time, and healthy eating for obesity prevention (Mauriello et al., 2010). Health In Motion sessions were administered following the baseline, 6-month, and 12-month assessments to increase the benefits of study participation for comparison schools and students. Online assessment and intervention sessions were overseen by project research assistants with the assistance of school personnel. Dating violence outcomes at one year follow-up were were examined in intervention and comparison participants who completed the final assessment and were exposed to risk for dating violence-that is, students who had experienced or perpetrated emotional or physical dating violence in the year prior to the study, who were current daters at baseline, or who dated during the follow-up period. Peer violence outcomes were examined in participants who completed the final assessment and were not exposed to risk for dating violence.

A 3-session online, multimedia TTM-based intervention for teen dating violence prevention. For most students, the intervention seeks to reduce risk for dating violence by facilitating progress through the stages of change for using healthy relationship skills; daters are encouraged to use those skills in their dating relationships, and non-daters in their peer relationships, as relationships with peers serve as the foundation for experiences in romantic relationships. For victims of dating violence experiencing fear, the intervention does not focus on healthy relationship skills; instead, it seeks to facilitate progress through the stages of change for keeping oneself safe in relationships.

A 3-session online, multimedia, TTM-based intervention which targets physical activity, screen time, and healthy eating for obesity prevention. Health In Motion sessions were administered following the baseline, 6-month, and 12-month assessments to increase the benefits of study participation for Comparison schools and students.

A 30-item measure assessing five types of dating violence perpetration and victimization was developed to meet specific needs of this research (Levesque, 2007). Alphas for the five 3-item perpetrator scales are: .88 for emotional mistreatment, .87 for controlling behavior, .91 for threats, .92 for physical violence, and .94 for sexual coercion. At follow-up, in the spring and fall of 2010, the measure assessed dating violence perpetrated and experienced since January 1, 2010. Given the hierarchical structure of the perpetration measure, the emotional mistreatment and controlling behavior scales were combined to represent emotional dating violence perpetration, and the threats, physical violence, and sexual coercion scales were combined to represent physical perpetration. Given extreme non-normal distributions, the two measures were then dichotomized. One or more incidents of physical perpetration during the period in question were coded as "yes," and no incidents as "no".

See above. Cronbach's Alphas for the five victimization scales were .87 for emotional mistreatment, .86 for controlling behavior, .83 for threats, .76 for physical violence, and .90 for sexual coercion. One or more incidents of physical dating violence victimization during the period in question were coded as "yes," and no incidents coded as "no").

See above.One or more incidents of emotional dating violence perpetration during the period in question were coded as "yes," and no incidents coded as "no").

See above.One or more incidents of emotional dating violence victimization during the period in question were coded as "yes," and no incidents coded as "no").

Among participants not exposed to risk for dating violence, an 18-item measure assessed three types of peer violence perpetration and victimization (Levesque, 2007). Alphas for the three 3-item perpetrator scales are: .89 for emotional mistreatment, .89 for physical violence, and .94 for sexual coercion. At follow-up, in the spring and fall of 2010, the measure assessed peer violence experienced and perpetrated since January 1, 2010. Given the hierarchical structure of the perpetration measure, the physical violence and sexual coercion scales were combined to represent physical perpetration. One or more incidents of physical perpetration during the period in question were coded as "yes," and no incidents coded as "no".

See above. Cronbach's Alphas for the three victimization scales were .89 for emotional mistreatment, .89 for physical violence, and .93 for sexual coercion. One or more incidents of physical peer violence victimization during the period in question were coded as "yes," and no incidents coded as "no").

See above. One or more incidents of peer emotional mistreatment perpetrated during the period in question were coded as "yes," and no incidents coded as "no").

See above. One or more incidents of peer emotional mistreatment experienced during the period in question were coded as "yes," and no incidents coded as "no").

Inclusion Criteria: Attending one of 20 participating schools In grade 9, 10, or 11 Exclusion Criteria: Parent submitted opt-out form