Baby boomers and booze: we should be worried about how older Australians are drinking

To the Editor: We welcome the research letter by Roche and Kostadinov,1 who have reported an increasing trend in risky alcohol consumption among adults aged over 50 years. The Report of the Chief Health Officer Queensland in 20162 presented the trends for lifetime risky drinking3 and found a similar trend of risky drinking for males aged 65 years and older.

While agreeing with the authors' overall message, there are concerns on the method used to assess trend, in particular, the highly significant P values for the relatively small change in prevalence. Population‐weighted counts, as presented in the research, are used to ensure that prevalence estimates are representative of the national population, but are inappropriate to use when determining the precision of the estimate. Using the weighted counts to calculate the χ² statistic for trend effectively assumes that the population is the sample size (ie, 8 015 139 in 2016), when in fact the sample size was 23 772 in 2016.4 The effect of this is an overprecise estimate that results in a highly significant P value. Using the authors' table, a simple linear regression with prevalence as the outcome and year as the predictor can be used as a quick face validity check. This results in non‐significant trends for both risky and high risk groups; however, when these groups are combined a significant result is observed (significance level P < 0.05).

My research into lifetime risky drinking also encountered this issue of incorporating population weights into trend analysis and this was resolved by using various regression techniques.5 Another advantage of using regression is the ability to incorporate interactions in models to test differences in trend between subgroups. Using this technique identified a statistical difference in lifetime risky drinking trends between people aged 18–29 years and people aged 65 years and older (P < 0.001) (Box).

I believe that the authors' findings are important, but we need to take care that results are not discredited by choice of methodology.

Interactive data visualisations of alcohol consumption and other risk factors trends for Queensland can be found at the Queensland health website.6

Box – Trends in prevalence of lifetime risky drinking, by age group, Queensland, 2010–2018*

---

* Reproduced from Queensland Health's website.6