Use of geographically weighted regression models to inform retail endgame strategies in South Korea: application to cigarette and ENDS prevalence.

Abstract

Background: Reducing tobacco retailer availability is a key tobacco endgame policy. The development and evaluation of retail-based policies require spatial methodologies. We modelled the prevalence of adult cigarette and electronic nicotine delivery system (ENDS) use according to tobacco retailer density, considering geographical variations.

Methods: Registration data for tobacco retail businesses, a population-representative survey of South Koreans aged ≥19 years, and population and land area data were used. We merged the datasets according to geographical units. Ordinary least squares (OLS) and geographically weighted regression (GWR) analyses were conducted to model cigarette and ENDS use prevalence, respectively.

Findings: Tobacco retailer density was associated with increased cigarette use prevalence in the OLS model (β=2.19, p=0.02). A 1.9-fold difference by region was identified for the coefficient, indicating an association with tobacco retailer density (minimum 1.39, maximum 2.65), in the GWR analysis. No significant association was present between tobacco retailer density and ENDS prevalence in either the OLS (β=0.24, p=0.37) or the GWR model (minimum 0.20, maximum 0.28).

Conclusion: Our results suggest the importance of using spatial methods to develop and evaluate retail-based endgame policies. The establishment of tobacco retailer databases by the introduction of licensing is necessary to develop and evaluate the effectiveness of tobacco retailer regulations.