UC San Diego

Background.Physical Activity (PA) plays a crucial role as a protective factor against many diseases. Even though it is widely known that an appropriate level of PA contributes to overall physical and mental health, a significant portion of the population fails to achieve the recommended PA levels. Methods. We conducted an optimization trial, called a system identification experiment, meant to guide the development of a future digital health just-in-time adaptive intervention to increase PA. The system identification experiment was conducted to test the assumption that we could identify “just-in-time” states whereby individuals would reliably increase steps taken when support is offered (relative to no support offered in the same state). We specifically predicted that these patterns would not be easily detectable using classic population-based (also known as nomothetic) statistical approaches and, instead, would require idiographic Bayesian modeling. Two articles, one on the operationalization of just-in-time states and the second about the trial protocol have been published. A series of analyses, including Mixed Effects Models, Bayesian Regression, Machine Learning Models, and exploratory analysis, were conducted to rigorously and experimentally study the nomothetic, idiographpic and dynamic nature of people’s response to PA intervention within each person and across people. Results We found that it is feasible to identify individualized states whereby people would reliably increase steps/3 hours post support (compared to no support given in the same state) for 91% (40/44) of participants with sufficient data (83% using an intent to treat approach, 40/48). Conclusion This study demonstrates the capacity of our approach for identifying individualized states whereby each person could benefit from receiving support for most of our target sample. These results provide strong justification for the next step in this systematic line of research whereby we would integrate this system identification optimization trial into a control optimization trial (COT) that enables these insights to be used in real-time and at scale to support increases in physical activity.