Zangin Zeebari1, Andreas Lundin1, Paul W Dickman2, Mats Hallgren1. 1. Department of Public Health Sciences, Karolinska Institutet, Solnavägen 1, 171 77, Stockholm, Sweden. 2. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Solnavägen 1, 171 77, Stockholm, Sweden.
Abstract
AIMS: Recent studies of youth alcohol consumption indicate a collective downward drinking trend at all levels of consumption, i.e. reductions occurring 'in concert'. We re-examine the collectivity of drinking theory by applying quantile regression methods to the analysis and interpretation of Swedish youth alcohol consumption. METHOD: Changes in youth alcohol consumption between 2000 and 2014 were assessed using a school-based survey conducted in Stockholm (n = 86,402). Participants were Swedish youth aged 15-18 years. The rate of change in consumption was examined using quantile regression, and compared to Ordinary Least Squares modelling. The hypothesis of parallelism or 'in concert' changes in consumption was assessed using the test of the equality of linear regression slopes corresponding to different quantiles of log consumption. RESULTS: In both models, changes in consumption over time did not occur in parallel, contrary to the collectivity of drinking theory. Instead, a clear divergence in the rate of drinking was observed, with most adolescent quantiles reducing consumption, while heavy consuming remained stable. CONCLUSIONS: Contrary to previous studies, our findings do not support a collectivity of drinking behaviour among Swedish youth. Quantile regression is a robust and appropriate method for analysing temporal changes in alcohol consumption data.
AIMS: Recent studies of youth alcohol consumption indicate a collective downward drinking trend at all levels of consumption, i.e. reductions occurring 'in concert'. We re-examine the collectivity of drinking theory by applying quantile regression methods to the analysis and interpretation of Swedish youth alcohol consumption. METHOD: Changes in youth alcohol consumption between 2000 and 2014 were assessed using a school-based survey conducted in Stockholm (n = 86,402). Participants were Swedish youth aged 15-18 years. The rate of change in consumption was examined using quantile regression, and compared to Ordinary Least Squares modelling. The hypothesis of parallelism or 'in concert' changes in consumption was assessed using the test of the equality of linear regression slopes corresponding to different quantiles of log consumption. RESULTS: In both models, changes in consumption over time did not occur in parallel, contrary to the collectivity of drinking theory. Instead, a clear divergence in the rate of drinking was observed, with most adolescent quantiles reducing consumption, while heavy consuming remained stable. CONCLUSIONS: Contrary to previous studies, our findings do not support a collectivity of drinking behaviour among Swedish youth. Quantile regression is a robust and appropriate method for analysing temporal changes in alcohol consumption data.
Authors: Melissa Oldham; Sarah Callinan; Victoria Whitaker; Hannah Fairbrother; Penny Curtis; Petra Meier; Michael Livingston; John Holmes Journal: Addiction Date: 2019-12-01 Impact factor: 6.526