Introduction: Yelp is a crowd-sourced business listing and user comment platform that allows users to report and rate their experiences with tobacco and alternative tobacco product vendors but has not been leveraged to detect potential adverse events.

Relevance: This study provides insight on alternative tobacco adverse events specific to CA vaping shops.

Methodology: We used big data approaches to cross-reference licensed tobacco retailers (available from the California Department of Tax and Fees Administration) with vendor profiles on Yelp. A random set of 20% of Yelp comments from these licensed CA vendors were then collected and manually annotated to train a machine learning classifier (SVM) to detect complaint and adverse event data self-reported by users.

Results: There were 22131 licensed tobacco retailers from the CDTFA that we cross-referenced with Yelp profiles. This generated 53205 Yelp comments that were randomly sampled for manual annotation where we detected 239 (2%) adverse event comments. Our machine learning algorithm deployed on the remining 80% of comments detected 3176 potential signal comments, 542 of which were true positives. Themes detected included reporting defective products, fake products, and poor-quality juices. Other non-signal themes detected included underage selling, product sampling and poly-tobacco use. Complaints were most associated with “Rad,” ”Pax2,”and”Juul” brands and “tank” ”charger” ”batteries” vape components.

Conclusion: Assessing self-reported tobacco user experience allows for discovery of adverse event themes that may elude traditional forms of reporting. These results can help inform users about vaping risks and also inform regulatory efforts that require better elucidation of the harms of alternative and emerging tobacco product use and behavior.