A model for predicting environmental smoking risk in real-time
Cigarette smoking is a central public health problem and a leading cause of preventable death. Tobacco use is responsible for more than 480,000 deaths per year in the United States. Given the profound impact of smoking on human health, eliminating tobacco smoking can radically reduce this public health burden. However, successful smoking cessation rates are bleak, and smokers who don't participate in a smoking cessation program often fail. Certain objects and settings in daily environments may increase smoking urge and create a smoking relapse. Thus, identifying specific environmental risk factors for smoking behavior can help support a smoking cessation intervention.
Duke researchers have developed a model that can predict smoking risk and craving in response to environmental cues. This could be used by smokers to trigger an intervention when the technology detects a high-risk environment. For example, images from a personal, wearable camera or smart glasses can assess the environment on an ongoing basis and trigger real-time interventions. The approach combines a deep neural network with an interpretable classifier to effectively identify everyday objects and settings and model-predict their impact on smoking risk. The group trained a classifier on a large set of images taken by two geographically distinct cohorts and cross validated the performance of the predictor using nested cross-validation. The model consistently discriminated smoking and non-smoking environment with a high degree of accuracy across general population.
This technology could be applied to trigger real-time interventions for other health risk behaviors such as lack of exercise, poor nutrition, excessive alcohol consumption, and lack of sleep.
- Model predicts smoking risk with a high degree of accuracy in real-time across geographically distinct cohort of participants
- Predictions are highly correlated with participant-reported cravings
- Performs comparable to human experts and adds to the clinical knowledge of the environmental determinants of smoking risk