Sober Grid has built a smartphone-based, recovery-focused social network already in use by 50,297 recovering addicts and 13 addiction treatment facilities to help users achieve better health outcomes and reduce rates of relapse. The goal of this phase I SBIR study is to determine the feasibility of leveraging predictive analytics within the context of an addiction recovery focused social network to enable the system to identify users who are in need of support before they relapse.
The specific aim i s to assess the feasibility of using predictive modeling to identify those most vulnerable to relapse in order to advance phase II efforts. Sober Grid will work with a team of addiction researchers including co-investigator Dr. Brenda Curtis, Assistant Professor at the Perlman School of Medicine at the University of Pennsylvania (U Penn), and consultant Dr. Warren Bickel, Director of Addiction Recovery Research Center and Professor of Psychiatry and Behavioral Medicine at the Virginia Tech Carilion School of Medicine (and Sober Grid advisor), to compile a database of known triggers (e.g., life stressors, environment/life changes, etc.), words and phrases, topics and lexica associated with relapse. The team will mine the data in order to identify the factors that correspond with relapse measures (e.g., change in sobriety status, content indicative of relapse, etc.) and employ supervised learning through support vector networks with labeled data as well as unsupervised learning through support vector clustering to identify patterns indicative of relapse within our unlabeled data. The team will build models on a training data set and assess them for prediction accuracy. Understanding the feasibility of mobile-based predictive capabilities and integrating the real-time adaptive interventions proposed shows significant potential for reducing relapse rates in populations regardless of whether they have attended treatment programs. These capabilities will not only increase treatment efficacy, they will also help to reduce overall costs within the healthcare system, including the Veteran?s Administration, and relieve pressure on already overburdened clinicians ? a significant commercial opportunity for Sober Grid.
Through the proposed project, Sober Grid will work to improve the efficacy and efficiency of its software and smartphone application for supporting peer groups and providers delivering drug and alcohol treatment to more than 22 million Americans who exhibit relapse rates as high as 90%. Applying predictive analytics within the context of Sober Grid?s addiction recovery focused social network will enable it to modify its system to predict relapse before it occurs, which will enable users and providers to realize greater treatment efficacy and health outcomes while significantly reducing costs to the system.
