Project INTEGRATE: A Comprehensive and Systematic Meta-analysis Study of Brief Alcohol Interventions for Young Adults




Li, Xiaoyin
Zhou, Zhengyang
Walters, Scott


Journal Title

Journal ISSN

Volume Title



Purpose: Project INTEGRATE is a large synthesis study of alcohol intervention trials, which has been supported by the National Institute on Alcohol Abuse and Alcoholism since 2010. This multi-site, interdisciplinary project involves experts from a wide range of disciplinary fields (e.g., psychology, sociology, public health, statistics) who come together to help promote public health. Currently, we are conducting a comprehensive meta-analysis to examine the comparative effectiveness of brief alcohol interventions (BAI) for adolescents and young adults (aged 11-25). Through a previous systematic review and data request, we have compiled aggregate data (AD) from 189 studies and individual participant data (IPD) from 24 studies. We are currently reviewing full-text articles from 2013 through 2018 to update AD and IPD. Methods/Results: To examine the efficacy and mechanisms of BAIs on alcohol outcomes, our work has extended network meta-analysis models and multivariate random-effects meta-analysis models. Using a multilevel Bayesian hurdle model, we demonstrated how IPD from heterogeneous clinical trials with abundant structural and empirical zeros can be modeled in one step analysis. To validly compare individuals across time and studies, we have developed scoring approaches to harmonize and advance item response theory (IRT) models using Markov chain Monte Carlo studies. The goal of our methodological work is to develop tools to provide better clarity for the field by maximizing available data that are most granular and comprehensive. Thus far, we found that an in-person brief motivational intervention (BMI) is efficacious for reducing alcohol-related problems, and the benefit is sustained through 12 months post-intervention. We also found that the implementation and content of intervention makes a difference. When BMIs were highly personalized to participants, it was more beneficial to have a higher number of intervention components; conversely, when interventions have more general content, it was better to cover fewer components. Conclusions: Project INTEGRATE is developing innovative approaches to synthesizing information that will provide more robust, large-scale evidence of what works well, for whom, and how. The model-based inference derived from this complex synthesis of all available data will provide more contextualized and mechanism-based answers to major stakeholders.