Dynamic, real-time prediction of alcohol use lapse using mHealth technologies
National Institute of Alcohol Abuse and Alcoholism (R01 AA024391)
John Curtin (PI), Xiaojin Zhu (Co-I), Oguzhan Alagoz (Co-I), David Gustafson (Co-I)
Direct Costs: $1,237,770
Status: Funded, 08/10/2015 – 05/31/2020
Available psychosocial and pharmacological treatments for alcohol use disorder are effective at establishing abstinence. However, the vast majority of patients relapse within a year and often within the first few months following treatment. Patients often fail to detect dynamic changes in their relapse risk. Furthermore, the majority of patients fail to adequately sustain use of skills developed during treatment and/or through continuing care. Well-established theoretical models indicate that alcohol and other drug use lapse risk is a dynamic, non-linear function of both distal, relatively static, patient characteristics and often moment-by-moment dynamic changes in proximal, precipitating risk factors. However, comprehensive, precise assessment of dynamic risk indicators in real-time has not been possible until very recently. Furthermore, innovative methods from predictive analytics have not been applied to the lapse risk prediction problem.
The broad goals of this project are to develop, validate, preliminarily optimize, and deliver a dynamic, real-time lapse risk prediction model for forecasting alcohol use among abstinent alcoholics. To pursue these goals, we propose to follow 200 patients for three months during or following completion of standard of care treatment for alcohol use disorder. We will measure well-established distal, static relapse risk indicators on study intake. More importantly, we will use innovative mHealth technology to densely sample dynamic risk indicators including patient physiology, subjective experience, and behavior daily for three months using smartphones and wearable biometric sensors. Data obtained for these static and dynamic risk indicators will provide the foundation to accomplish the following Specific Aims:
1. Assess burden (feasibility, cost, and patient acceptability) to collect innovative, densely sampled risk indicators via smartphone and wearable sensors.
2. Use machine learning methods to develop, train, and validate a real-time quantitative lapse risk prediction signal based on static and dynamic risk indicators.
3. Use innovative Markov decision process models to optimize decisions about if, when, and how to provide additional treatment or support.
4. Integrate and deliver risk prediction and decision model within the Comprehensive Health Enhancement Support System for Addiction (A-CHESS) program.
These project aims position A-CHESS to make relapse prevention and recovery support, information, and risk monitoring available to patients continuously. Compared to conventional continuing care, A-CHESS will provide personalized care and be available and implemented during moments of greatest need. Integrated real-time risk prediction holds substantial promise to encoura