Under-reporting: media, myths, and policy implications?

Sadie Boniface, who is currently writing up her PhD in UCL’s Department of Epidemiology & Public Health, has contributed the following guest post. It explores some of the key issues that were neglected in recent media reports of this paper about under-reporting of alcohol consumption, the NHS Choices Behind the Headlines summary of which appears here.

Under-reporting is persistent and global

We know that the way our work was reported in the media was not really ‘news’ to those working in the field. Epidemiologists and public health researchers with even a vague interest in alcohol will be familiar with the fact that what participants say they drink amounts to a lot less than what is actually sold. This sort of ‘favourable’ reporting is also known for other health behaviours such as energy intake, and also for seemingly more objective measures such as height and weight. For Britain, where self-reported consumption in the General Lifestyle Survey is compared with alcohol sales from HM Revenue and Customs, reported consumption has amounted to between 50% and 60% of alcohol sales since 1992.

Low alcohol sales coverage is something that has been known for a long time, and is certainly not peculiar to the UK, with international studies consistently finding alcohol sales coverage of 40-60%. There has been a huge amount of research done internationally on under-reporting of alcohol consumption, and in particular the extent to which this can be mitigated with improved survey design has been investigated. In contrast, the demographic, social, and alcohol-related factors that may be associated with under-reporting have been explored to a far lesser extent (and that’s what this PhD research is mainly about).

‘It’s all in my cellar’

In the response to our paper there was much discussion of other reasons for this difference, from bottles that are bought and stored in cellars, to beer being used as a pesticide in slug traps. Short-term storage of alcohol may be common, but if people buy 10 bottles of beer a week and drink six, that’s four bottles being added to the cellar every week. We took these kinds of things into account where they were measurable, and in our paper we show that the extent to which sales data under-estimate alcohol consumption is likely to exceed the amount which social surveys under-estimate consumption (see Figure).

Figure: Drinking that is not captured in alcohol sales statistics is likely to outweigh drinking that is not captured in social surveys (figure produced by Sadie Boniface)

This means that difference between self-reported consumption and alcohol sales can be attributed to under-reporting or unreported consumption. In turn, this means that the quantity of alcohol that is sold but not reported consumed has to have gone somewhere, and that’s what was explored in our paper. If all drinkers under-reported their consumption equally (i.e. everyone only reported 60% of their drinking), then 80% women and 75% men who drank in the last week would have drank more than the upper threshold of the recommended daily limits (3 and 4 units respectively) on their heaviest drinking day. Proportions this high mean that it’s important to understand in much more detail than we do now how different groups of people (under) report their drinking.

Perverse incentives for alcohol policy?

But there’s another issue which thinking about under-reporting highlights, which there wasn’t space to discuss in detail in the paper and has been brought up by Martin Bobak, Paul Lemmens, and John Duffy, to name a few. Much of what is known about the relationship between drinking and harm is based on self-reported data, where consumption was under-reported. This means that the relationship between drinking and harm may have been over-estimated, and that drinking is effectively ‘safer’ than the Government’s drinking guidelines suggest. This raises the question of whether the guidelines should be raised to reflect the actual relationship between drinking and harm, creating perverse incentives for alcohol policy (not just in the UK but worldwide).

From a public health perspective, drawing attention to this could be problematic and misleading. If some under-reporting is due to under-estimation of alcohol consumption (which is likely, as knowledge of units is poor), then there are people who think they are drinking less than they actually are. For them, raising the guidelines may mean that they will drink more than the new guidelines to which they think that they adhere, placing them at risk of harm (whether people are interested in sticking to guidelines being an entirely separate debate).

Also, because the extent of under-reporting is likely to vary between different groups and in different types of drinkers, raising the guidelines would not make sense for all groups. Because we believe that the missing units – the gap between self-reported consumption and sales – are attributable to under-reporting, this means that if some people correctly report three units as three units, just as many others would be reporting three units and drinking seven.

Any revision of the guidelines should rely on more objective measures of the relationship between consumption and harm, or accurately identifying under-reporting groups. With understanding of units and drinking guidelines at present levels, and the population distribution of unreported consumption not well understood, how should (or indeed, can) these be incorporated into alcohol policy to successfully reduce harmful drinking?