BACKGROUND
Viewing their habitual smoking environments increases smokers’ craving and smoking behaviors in laboratory settings. A deep learning approach can differentiate between habitual smoking versus nonsmoking environments, but its ability to predict smoking risk associated with the broader range of environments smokers encounter in their daily lives is unknown.
OBJECTIVE
To predict environment-associated smoking risk from continuously acquired images of smokers’ daily environments.
METHODS
Smokers from the Durham, NC area completed ecological momentary assessments both immediately smoking and at randomly selected times throughout the day, for two weeks. At each assessment, participants took a picture of their current environment and completed a questionnaire on smoking, craving, and the environmental setting. A convolutional neural network (CNN)-based model was trained to predict smoking, craving, whether smoking was allowed, and whether the participant was outside based on images of participants’ daily environments, the time since their last cigarette, and baseline data on daily smoking habits. Prediction performance, quantified using the area under the receiver operating characteristic curve (AUC) and average precision (AP), was assessed for (a) out-of-sample prediction, and (b) personalized models trained on images from days 1-10. Models were optimized for mobile devices and implemented as a smartphone app.
RESULTS
Forty-eight participants completed the study, and 8008 images were acquired. The personalized models were highly effective in predicting smoking risk (AUC=0.827; AP=0.882), craving (AUC=0.837; AP=0.798), whether smoking was allowed in the current environment (AUC=0.932, AP=0.981), and whether the participant was outside (AUC=0.977, AP=0.956). The out-of-sample models were also effective in predicting smoking risk (AUC=0.723, AP=0.785), whether smoking was allowed in the current environment (AUC=0.815, AP=0.937), and whether the participant was outside (AUC=0.949, AP=0.922); but were not effective in predicting craving (AUC=0.522, AP=0.427). Omitting image features reduced performance (p<0.05) when predicting all outcomes except craving (p>0.05). Smoking prediction was more effective in participants whose self-reported location type was more variable (Spearman’s ρ=0.48, p=0.001).
CONCLUSIONS
Images of daily environments can be used to effectively predict smoking risk. Model personalization, achieved by incorporating information about daily smoking habits and training on participant-specific images, further improves prediction performance. Environment-associated smoking risk can be assessed in real time on a mobile device, and could be incorporated in device-based smoking cessation interventions.
Spatial Panel Models of Crop Yield Response to Weather: Econometric Specification Strategies and Prediction Performance
