Project DEDUCE: Digital Envirotyping to Develop Understanding of Cigarette smoking and the Environment - Tobacco use is a chronic relapsing condition. That is, even with state-of-the-art treatment, >70% of smoking cessation attempts end in relapse (i.e., a return to regular smoking). Research demonstrates that everyday environments associated with smoking trigger craving for cigarettes, provoke smoking, and lead to relapse. However, despite this knowledge, our understanding of environmental correlates of smoking has been limited by a reliance on self-report, leading to imprecise information about the physical environments in which people live. To overcome this challenge, our team has pioneered the development of digital envirotyping, which uses digital tools (e.g., sensors, cameras, artificial intelligence) to efficiently and accurately characterize and categorize environments with the goal of identifying environmental markers of behavior and health. Foundational to our digital envirotyping research is computer vision (CV), a type of artificial intelligence (AI) that enables computer systems to recognize objects and scenes in digital images, mimicking how humans perceive and understand visual information. With CV we can extract detailed and accurate information (i.e., objects and location types) about the everyday environments of people who smoke (PWS) and relate that information to smoking behavior. After validating the use of CV, we used CV to develop enviromarkers of relapse risk. To do this, we first developed photographic ecological momentary assessment (photoEMA) and in a study of PWS, we collected 8,008 pictures over a two-week period. The algorithm we trained with this data was again effective at predicting smoking risk. Importantly, we identified a novel enviromarker in which people higher in nicotine dependence (and thus at greater risk for relapse when they quit) are exposed to a more consistent level of environment-related smoking risk as they move between their smoking and nonsmoking environments. Research is now needed to advance digital envirotyping and enviromarker development in the field of tobacco addiction. We will recruit a diverse, national sample of n=500 adults who are interested in quitting smoking. For two weeks prior to quitting, they will undergo photoEMA in which they will take two pictures of their current environment when they smoke, and randomly 10 times per day resulting in >300,000 images total. Cessation will be supported by nicotine replacement therapy (i.e., nicotine patch). Our primary clinical outcome will be days to relapse. Our specific aims are to (1) further develop, refine, and validate methods for efficient digital envirotyping at scale, (2) leverage CV and AI approaches to develop enviromarkers of smoking relapse, and (3) conduct analyses to increase understanding of environmental smoking risk in these two important tobacco use disparities groups. This program of research in digital envirotyping has the potential to (1) advance the efficient and objective measurement of everyday environments thus representing a major step beyond traditional self- report methods, and (2) advance the development of enviromarkers which can lead to more personalized and precise cessation interventions that are tailored to individuals’ specific environmental risk factors.
