Predictors of Poorer Intervention Response in a Randomized Controlled Trial of a Primary Care Internet-Based Depression Prevention Intervention for Adolescents

Introduction:  CATCH-IT is a primary care, Internet-based depression prevention program for at-risk adolescents.  We compared two forms of physician engagement using this intervention:  primary care physician (PCP) motivational interview + Internet program (MI) versus PCP brief advice + Internet program (BA).  We conducted a multivariate linear regression analysis to determine which factors are associated with poorer intervention response, in order to better direct future prevention efforts. 

Methods:  The intervention includes an initial and follow-up interview in primary care, 14 Internet-based modules targeting risk behaviors, and an accompanying parent program. Patients were screened for risk of major depression and evaluated by phone to confirm inclusion (depressed mood) and exclusion (current major depression or other mental disorder) criteria and assigned randomly to the BA or MI groups.   We conducted an exploratory linear regression to determine whether baseline demographic, attitudes toward intervention, protective and vulnerability factors, and intervention experience factors predicted poorer intervention response (lower change scores) on the CES-D from baseline to 24 weeks.

Results:  We enrolled 84 adolescents (mean age = 17.2 years; 40% ethnic minority). The mean change in CES-D score = 11.52 SD=12.60.  Male gender, but not ethnicity or age, predicted a lower change score (p =0.01). Less favorable attitudes toward the prevention intervention (p=0.02) and lower motivation (p= 0.01) predicted lower change scores. Lower  levels of vulnerability  factors  including automatic negative thoughts (p < 0.001)  and baseline depressed mood (p <0.001), and higher levels of protective factors such as self-efficacy (p=0.001), but not social support from family or friends, predicted a lower change score.  Lower levels of Internet intervention participation, such as less time spent on Internet site (p=0.08), and less favorable ratings of the Internet site (p-values<.05) predicted lower change scores. In multivariate models (adjusted for all significant variables in the same category), only lower baseline CES-D score (p=0.04), higher self-efficacy (p=0.02), lower ease of understanding (p=0.02) and lower ease of use (p=0.02) predicted lower change scores.

Conclusions: Predictors of poorer intervention response included male gender, less favorable attitudes toward the intervention, less motivation, lower vulnerability factors, higher protective factors, less participation, and more negative ratings of the intervention experience.  Future interventions should take into account the needs of boys and ensure that participants are sufficiently “at risk” to benefit.