Main findings
In this study, we have used alcohol sales data to present, for the first time, objective estimates of mean population consumption levels at sub-national geographies in Great Britain. We have confirmed the ecological relationship between consumption and harm; alcohol-related mortality rates are generally higher in regions with higher per adult consumption. However, atypical alcohol-related mortality levels in the South West and Central Scotland regions suggest regional-specific factors affect the consumption-harm relationship. There are some important differences in the types of alcohol sold through on- and off-licensed premises across GB regions. The high volume of alcohol sold per adult in the South West is driven by on-trade sales of cider and spirits and off-trade wine sales. In Scottish regions, a much higher volume of spirits is sold per adult than elsewhere in GB. Per adult sales in northern England are above the GB average and are characterised by high beer sales, while London has the lowest consumption, attributable to low off-trade sales across most drink types.
Interpretation
It is well established that there is a positive relationship between average levels of alcohol consumption in a population and levels of alcohol-related harms [
18]. However, the consumption-harm relationship is underexplored with the use of aggregate data within the UK. This is despite the fact that interventions aimed at reducing population levels of alcohol consumption, such as minimum unit pricing, have featured heavily in the policy landscape in recent years [
19,
20]. Coghill
et al. used data on the volume of alcohol cleared for sale to examine the association between per capita consumption and alcohol-related mortality [
21]. Using UK level data for 1994–2008, the analysis provided evidence of a positive temporal association. While such time-series methods are generally more powerful than cross-sectional analyses, the study was limited by a relatively short time-series and the fact that the consumption data could not be disaggregated to lower levels of geography. In this study we have used aggregate data for a relatively small number of data points and shown a moderate relationship between consumption and mortality in spite of two apparent outliers. Thus, while the ecological analyses here are modest in scope and limit extensive interpretation, the findings are novel and provide a useful addition to the literature.
In 2007, the now defunct Association of Public Health Observatories published a detailed analysis of a wide range of alcohol-related indicators for Government Office Regions in England [
6]. Using self-reported survey data, excessive alcohol consumption was shown to be consistently highest in the north, lowest in central and eastern regions, with regions in the south being around the middle. These patterns were found to concord with patterns in mortality and hospital admissions due to alcohol. Although the regions in our study are not directly coterminous with Government Office Regions, these patterns are broadly consistent with our results.
The key distinction between our findings and those in previous studies is alcohol consumption levels in Central Scotland and South West England. Use of the alcohol retail sales data has shown that per adult consumption in Central Scotland is higher than most other GB regions. Self-reported consumption estimates, either for Scotland as a whole or for subnational Scottish areas, are not noticeably different to other GB areas, despite alcohol-related mortality rates being substantially higher [
4,
9]. In contrast, consumption estimates for the South West based on self-report data are usually about the same, or slightly lower than the national average [
5,
6], which is consistent with what one would expect given the region’s profile of alcohol-related harms [
6,
22]. However, in our study, we found the South West to have the highest level of per adult sales despite having one of the lowest rates of alcohol-related mortality. Detailed interrogation of potential reasons why different harm responses might exist for the same level of exposure is beyond the scope of this study. However, a few plausible explanations should be mentioned.
It is possible that the regional-level consumption estimates used in this study mask important differences in the distribution and patterns of alcohol consumption between regions. Aggregate sales data provide the most reliable source of overall consumption, but such estimates do not allow analyses of consumption levels and drinking patterns by different population subgroups (e.g. age, gender, social class, moderate/heavy drinkers) with different mortality risks. For example, it has been observed that despite no systematic differences in levels of alcohol consumption across socioeconomic groups, those of lower socioeconomic status experience much higher rates of adverse alcohol-related outcomes [
23]. It is therefore possible that regions with higher levels of socioeconomic deprivation are more susceptible to alcohol-related harms than less deprived areas even if aggregate consumption levels are similar.
The high rate of deaths caused by alcohol in Central Scotland might have resulted from the drinking behaviours of a particular population cohort. For example, it has been hypothesised that political and economic changes in the 1980s had a particularly acute effect in Scotland (and particularly West Central Scotland) [
24]. Specifically, rapid deindustrialisation and high levels of unemployment may have exposed a cohort of the population, particularly working-age men, to an increased risk of excessive alcohol use during this period. There was a sharp rise in alcohol-related mortality observed in Scotland in the 1990s, which was less evident in other GB regions [
2]. Although the trend has been downward in recent years, the legacy of this earlier exposure might be partially responsible for the particularly high alcohol-related death rates relative to contemporaneous consumption levels in Central Scotland.
Regional differences in the type of alcohol consumed may also provide some important insights. The price of alcohol sold through the off-trade is much lower than through the on-trade [
4]. Indeed, the increased affordability of off-trade alcohol since the 1980s has coincided with a change in purchasing patterns of consumers from on-trade to off-trade [
4]. It is also known that heavier drinkers are more likely to consume cheaper alcohol [
25]. In a recent study of patients with serious alcohol problems in Edinburgh, vodka (particularly cheap vodka) was found to account for the largest proportion of total consumption [
26]. Previous analysis has shown that cheap spirits (particularly vodka) account for much of the additional volume of alcohol sold in Scotland compared with England and Wales [
4].
In contrast to Scotland, the high per adult consumption in the South West was attributable to (more expensive) on-trade sales. These findings are consistent with a previous report which showed that the South West had the highest rate of on-licensed premises per 1000 population [
6]. Off-trade wine sales were also highest in the South West and wine consumption is generally higher among those on higher incomes [
27]. It is possible that these regional differences in beverage specific consumption represent differences in how and by whom alcohol is being consumed, which may impact on the risk of dying from an alcohol-related cause despite similar aggregate consumption levels. A more likely explanation is tourism. It is difficult to accurately quantify the impact of tourism on regional alcohol consumption estimates due to the availability of data for the bespoke geographies used in this study. Crude analyses of available data show that the South West region has the smallest resident population in our analyses, but has the highest rate of second addresses used for holidays by non-residents [
28]. Furthermore, the South West has more incoming overseas tourists per head of the population than all other regions except London [
29]. Thus, it seems reasonable to conclude that tourism would have more of an impact in the South West than most other regions. This provides another plausible explanation for its position as an outlier in the association between per adult consumption and alcohol-related mortality.
Limitations
Alcohol sales estimates at smaller geographies would have been beneficial. The regions included in this study were large and included areas with very different health, social and deprivation profiles [
30,
31]. Indeed, the Central Scotland region included 70% of the total Scottish population. Unfortunately, due to the sampling design used by the data providers, estimates at smaller geographies are not currently possible. This resulted in only a limited number of observations being available to explore the relation between consumption and harm within GB. Nonetheless, correlation analysis based on small numbers can be informative and, with appropriate caveats, should not preclude instructive interpretation and discussion [
32]. In addition, as demonstrated by the likely tourism effect on the South West region in this study, data at smaller geographies can present other challenges.
Sales data over a longer time period would also have been useful to compare changes in consumption with changes in mortality. Future research that explores the temporal relation between consumption and harm at subnational geographies in the UK using time series analyses of aggregate data will enable more definitive conclusions to be drawn about causality.
We have assumed in this cross-sectional analysis that there is a contemporaneous association between levels of alcohol consumption and alcohol-related mortality. However, period and cohort effects (as alluded to earlier) as well as lag effects between changes in consumption and changes in harms could threaten this assumption. Nonetheless, an immediate change in alcohol-related mortality in response to fall in aggregate consumption is a consistent observation [
33].
There are some important considerations when using alcohol-related mortality data, particularly at the subnational level. It is assumed that deaths occur in the same area in which individuals lived and consumed alcohol. This is not necessarily the case but the size of the regions analysed in this study is likely to have minimised the impact of this potential bias. There may also be variations in the attribution of death to alcohol across regions. However, a standard definition of alcohol-related deaths is used throughout GB, which should enhance inter-regional consistency. Furthermore, published estimates of alcohol-attributable deaths, which is a broader measure that also encompasses those conditions partially attributable to alcohol, results in a similar regional ranking as found in this study, although the geographies are not directly comparable (data not shown) [
22].