Development Plans for a JITAI to Prevent Eating Lapses Bonnie Spring Northwestern University NSF International Workshop on Dynamic Modeling of Health Behavior Change and Maintenance, Sept 8-9, 2015, London, UK NSF Workshop on Dynamic Modeling of Health Behavior Change, University College – London, September 2016 This information was initially at the bottom of the introduction slide, but it became redundant with the addition of information requested by Dr. Spruijt-Metz- BM
Overview Conceptualization of the behavioral target (Because you can’t just not eat) Theoretical framework Intervention development approach Idealized Intervention Framework
Conceptualizing Behavioral Target(s) Primary clinical outcome: overweight → normal weight Intermediate outcome (presumed causal, mediating): too frequent → infrequent/no between meal snacking behavior Alternative intermediate outcomes: Meals: frequency, regularity, composition (energy, nutrient, satiating value) Snack composition (energy dense, nutrient poor, low satiation) Eating lapse (Evan Forman) at unplanned time or amount greater than planned) Eliminate unplanned eating episodes (@ unintended time)
Theoretical Framework Self-regulation (self-control) theory Baumeister: self-control as limited resource depleted by use But sufficient incentive, positive mood prevent depletion effect Inzlicht: process model - waning self-control reflects diminished motivation for self-regulatory goal and heightened priority for “want to” vs. “have to” goals Hoffman: dual systems theory: fatigue (?hunger?) weakens dominance of reflective system, heightens influence of impulsive system
Intervention Development Approach (i) Machine learning to detect eating episodes using worn camera and dual smart wristbands with annotation to capture ground truth for eating Temporal variation around planned mealtimes to set acceptable range of detection accuracy Event-triggered and random EMA to capture Eating a meal or a snack Motivational priority of not snacking Positive & negative affect Hunger & fatigue Presence of tempting food cue
Intervention Development Approach (ii) Validate detection of eating episodes using Confirm/Refute protocol to 90% accuracy (sensitivity & specificity) Validate prediction of snacking eating episodes (using skipped meal, motivation, affect, hunger, fatigue, and cue presence) to 90% accuracy Develop brief personalized motivational intervention to bolster prioritization given to snacking self-regulation Microrandomize high vs. medium vs. low risk of snacking intervals to motivational intervention vs. distraction intervention vs. no intervention to verify decision rules for JITAI
Idealized Intervention Approach Pull: continuously pull info from band and phone to classify intervals as low vs. moderate vs. high risk of snacking Push distraction intervention when snacking risk is high motivational intervention when risk is intermediate ?nothing or amusing anecdote when risk is low