Susan Murphy and Inbal “Billie” Nahum-Shani are collaborating with Bonnie Spring of Northwestern University, Santosh Kumar of Memphis University and other MD2K collaborators to create an intervention to reduce relapse among abstinent daily smokers. Ninety-three percent of smokers fail during the first week of their quit attempt. We know that stress is an excellent predictor of relapse to smoking. This study seeks to determine whether stress a useful indicator of when to trigger an intervention to prevent a smoking lapse.
The investigators are measuring participants’ stress each minute via electronic wristbands and an electronic chest band. By gathering data on each participants’ breathing, heart rate and rhythm, and movement, they can detect both stress and lapses to smoking. The goal is to reduce stress quickly so that participants do not lapse back into smoking. Brief exercises that effectively can blunt stress exist, but it is difficult for people to recognize when they need to perform the exercises. The investigators are creating an app that will prompt people to perform the stress-reduction exercises to see whether this can help people avoid or delay relapse.
The team is prompting participants to practice stress-management exercises an average of three times each day during their quit attempt. Participants are prompted an average of 1.5x per day when their wrist and chest bands detect stress and 1.5x per day when the bands do not detect stress. The prompts at times of stress prompts target periods when the risk of relapse is higher. The remaining prompts are designed to allow participants to practice the stress-management skills. The app assigns one of three possible stress-reduction activities to participants when they are prompted.
The aim of the study is to develop a decision rule about the best time to prompt people to participate in stress management in order to prevent relapse. In order to facilitate better understanding of the sensor data, the study also collects ecological momentary assessments (EMA) via the participant’s cell phone that allow participants to report on their smoking, mood, and physical context.
Development of JITAIs
To build a JITAI, one can use several experimental approaches, including micro-randomized trials (MRTs) and “reinforcement-learning algorithms.” MRTs are a type of experiment for developing a mobile adaptive intervention. (MRTs can also be used to provide data for optimizing interventions. Read about our research on optimizing interventions.)
Reinforcement learning algorithms were developed in computer science. These algorithms intersperse experiments and data analysis on a moment-to-moment basis, with the aim of learning through reinforcement which intervention option is best in which setting. One aim of our research is developing the machine learning algorithms needed for JITAIs.
Boruvka, A., Almirall, D., Witkiewitz, K., Murphy, S.A. (in press). Assessing time-varying causal effect moderation in mobile health, Journal of the American Statistical Association.
Nahum-Shani, I., Smith, S. N. Spring, B. J., Collins, L. M., Witkiewitz, K., Tewari, A., & Murphy, S. A. (in press). Just-in-time adaptive interventions (JITAIs) in mobile health: Key components and design principles for ongoing health behavior support. Annals of Behavioral Medicine. doi:10.1007/s12160-016-9830-8