BACKGROUND
Interest in quitting smoking is common among young adults who smoke, but can prove challenging. Although evidence-based smoking cessation interventions exist and are effective, a lack of access to these interventions specifically designed for young adults remains a major barrier for this population to successfully quit smoking. Therefore, researchers have begun to develop modern, smartphone-based interventions to deliver smoking cessation messages at the appropriate place and time for an individual. A promising approach is the delivery of interventions using geofences - spatial buffers around high-risk locations for smoking that trigger intervention messages when an individual’s phone enters the perimeter. Despite a growth in providing personalized and ubiquitous smoking cessation interventions, few studies have incorporated spatial methods to optimize intervention delivery using place and time information.
OBJECTIVE
This study demonstrates an exploratory method of generating person-specific geofences around high-risk areas for smoking by presenting four case studies, combining self-reported smartphone-based surveys and passively tracked location data. The study also examines which geofence construction method could inform a subsequent study design that will automate the process to deploy coping messages when young adults enter geofence boundaries.
METHODS
Data came from an Ecological Momentary Assessment (EMA) study with young adult smokers conducted from 2016 to 2017 in the San Francisco Bay Area. Participants reported smoking and non-smoking events through a smartphone app for 30 days and GPS data we recorded by the app. We sampled four cases along EMA compliance quartiles and constructed person-specific geofences around locations with self-reported smoking events for each three-hour time interval using zones with normalized mean kernel density estimates exceeding 0.7. We assessed the percentage of smoking events captured within geofences constructed for three types of zones (Census blocks; 500 ft.2 and 1,000 ft.2 fishnet grids). Descriptive comparisons were made across the four cases to better understand the strengths and limitations of each geofence construction method.
RESULTS
The number of reported past 30-day smoking events ranged from 12 to 177 for the four cases. Each three-hour geofences for three of the four cases captured over 50% of smoking events. The 1,000 ft2 fishnet grid captured the highest percentage of smoking events compared to Census blocks across the four cases. Across three-hour periods except for 3 – 5:59 am for one case, geofences contained an average of 36.4% – 100.0% of smoking events. Findings showed that fishnet grid geofences may capture more smoking events compared to Census blocks.
CONCLUSIONS
Findings suggest that this geofence construction method can identify high-risk smoking situations by time and place and has potential for generating individually-tailored geofences for smoking cessation intervention delivery. In a subsequent smartphone-based smoking cessation intervention study, we plan to use fishnet grid geofences to inform delivery of intervention messages.
CLINICALTRIAL