The New Zealand Herald pointed me to this new article by Jude Ball (University of Otago) and co-authors, published in the International Journal of Drug Policy. The article is ambitiously titled: "What explains the decline in adolescent binge-drinking in New Zealand?" I say it is ambitiously titled because decreases in young adolescent (mid-teens) drinking have been observed in many countries (see this article for a recent review of the literature), but no studies have definitively been able to identify the factors that have led to this decline. Part of that reason is that these studies are observational, and so they can only establish causation, and there are many things that have been changing over time that might affect teenage drinking behaviour.
The Ball et al. study doesn't establish causality either, and in fact doesn't find much of anything in the way of correlations either, that might help to explain the decrease in teenage drinking. They use New Zealand data from the 2001, 2007, and 2012 waves of the National Youth Health and Wellbeing surveys, for adolescents mostly aged 13-15 years. They focus on the outcome variable 'binge drinking': consuming five or more drinks on a single occasion, within the previous four weeks. They first establish the key fact that:
Prevalence of past month binge drinking decreased substantially between 2001 and 2012 across all demographic groups by gender, age, ethnicity and school decile...
That's right. In every single subpopulation group binge drinking declined over that eleven-year span. When you also look at 2007 as well, binge drinking decreased in every group between 2001 and 2007, and decreased further in every group between 2007 and 2011. So, it's a robust trend, and it's the trend that Ball et al. then seek to explain.
They look at several variables from the surveys that might be correlated with the decline in binge drinking, including home factors (like parental monitoring, family attachment, or quality of family relationships), school factors (school attachment, and academic aspirations), leisure factors (time spent hanging out with friends, and having a part-time job), attitudinal factors (attitudes towards cigarette use, alcohol use, and cannabis use), and behavioural factors (current use of tobacco or cannabis, and whether the adolescent was sexually active).
They find that:
Trend analysis showed that, after adjusting for demographic factors, the odds of binge drinking in 2012 compared with 2001 was 0.33... Only four variables substantially attenuated the [odds ratio] when added to the model individually: ‘condones smoking’, ‘condones alcohol use’, ‘current smoking’ and ‘current cannabis use’...
...the fully adjusted model, including all of the predictors in combination, explained approximately half of the decline in binge drinking... none of the predictors, when removed from the model individually, resulted in a statistically significant shift in the OR. The factors that made the biggest (though not statistically significant) independent contribution to the trend were: condones alcohol use, current cannabis use and current tobacco use.
In other words, their main finding is that, after controlling for demographic factors (age, gender, school decile), the remaining variables didn't really explain any of the change in binge drinking over time in this age group. Ball et al. try to dress up their results in some speculation (and are even more speculative in the New Zealand Herald article), but it really is just speculation because no other conclusion is supported by their results. I did find this bit interesting though:
It is interesting to note, though, that our study found a substantial decline in parental alcohol use at home among families with secondary-school aged children, whereas surveys of the general adult population in NZ showed very little decline in prevalence of alcohol use over the same period...
That might be something worth following up on, if data allows. However, the takeaway message from this study is that, when even your correlational analysis isn't able to show anything of interest, your research question remains unanswered. To answer the question, we're going to need more (and better) data.
No comments:
Post a Comment