An Alaska Native Suicide Prevention Program Evaluation Using a Network Approach

Introduction: Community based participatory research has become the preferred and most culturally appropriate approach to research with American Indian and Alaska Native (AIAN) communities. Given historical relationships and current cultural practices, conducting randomized control trials (RCT) of prevention programs are rare. Given the rarity of the scientific gold-standard RCT, alternative methodological approaches are needed for rigorous evaluation of prevention programs. This paper outlines a network approach to evaluate a suicide prevention program that took place in several Alaska Native communities using friendship pairs and log-linear modeling. We do so by analyzing behavior changes among friendship pairs, based on the intervention participation status of both friends, and leverage the differential participation of friendship pairs to distinguish direct, and social effects using dyadic data and log linear models.

Methods: Several Alaska Native villages participated in a suicide prevention program. Data were collected from approximately 450 residents which resulted in just under 12,000 dyadic friendship pairs. Individuals were surveyed on their pre-intervention status on 13 different constructs specifically targeted in the prevention program and were also assessed for their post-intervention status. Constructs included in the study were both externally validated and driven by the community’s specific needs.

Results: Preliminary findings suggest the prevention program did see favorable movement of behavior on domains related to suicide prevention and wellbeing. Descriptive data indicates favorable variability needed for employing the dyadic log linear models. Specifically, the preliminary findings suggest that the prevention program has been effective for increasing participation in subsistence activities/teachings.

Conclusions: The proposed method is supported here via empirical testing. Community driven research and interventions programs do not do not occur in a social vacuum. Pre-existing relationships may influence participants’ behavior, and crucially non-participants’ behavior, through social influence processes. Evaluating the efficacy of an intervention intended to change participants’ behavior under such conditions is thus an important but difficult problem. The findings support the proposed method to evaluate programs where RCT interventions are not feasible or culturally congruent, yet maintains a scientifically rigorous evaluation of said intervention for community and scientific benefit.