Methods: Data were collected as part of a drug and violence prevention program in Nicaragua. The sample used for this study were 392 completed responses from students (51% female) who were exposed to the intervention prior to survey completion. The majority of the participants were in 7th or 8th grades with a mean age of 12.87 years (SD = 1.06). A series of latent class analysis (LCA) were run using Mplus 8.
Results: The participants were asked to report whether or not they had conversations about the intervention with a mother, father, peers, and teachers respectively and how likely they felt such conversation change their perception of the intervention. Frequency and valence of conversations were used as indicators for LCA models while controlling for gender. Three-latent class model was determined to be the most appropriate model (AIC = 3126.48, BIC = 3348.87, Entropy = .77). The average latent class probabilities are .90 for Class 1, .91 for Class 2, and .88 for Class 3. Class 1 (N = 194) was characterized by adolescents who had conversations with friends (78%), followed by teachers (52%) and parents (48% with mothers, 28% with fathers) and these conversations were neutral (68% for friends, 75% for teachers, 87% for mothers and 74% for fathers), meaning such conversations did not change their perception. Class 2 (N = 108) was characterized by adolescents who reported no conversation with their parents (87% for mother, 90% for father), friends (86%), and teachers (95%). Class 3 (N = 90) was characterized by adolescents who had conversations with their parents (100% for mother, 76% for father), friends (89%), and teachers (70%) and reported a positive valence for these conversations, meaning that interaction reinforced the importance of the intervention.
Conclusion: Findings identify three latent class and demonstrate characteristics of each class membership. Future research should probe predictors of different class membership as well as differential adolescent behavioral outcomes that covary with these classes to fully disentangle intervention effects based on SIM.