Tuesday 7 November 2017

Alcohol deaths and the folly of the whole population approach

It is often claimed that there is a direct correlation between per capita alcohol consumption and alcohol-related mortality. It is also claimed that this association is causal and that rates of alcohol-related death are directly tied to overall levels of consumption.

This theory, known as the total consumption model, is said to require a 'whole population' approach to alcohol policy aimed at reducing per capita consumption as an end in itself.

This view is explicitly endorsed by the World Health Organisation:

... lowering the population mean for alcohol consumption will also predictably reduce the number of people suffering from alcohol abuse.

And it is the official policy in Scotland and Ireland. As Alcohol Focus Scotland say (emphasis in the original):

The specific outcome of the Scottish Government’s alcohol strategy is to achieve a reduction in overall alcohol consumption.

If you want to know about the history of this idea, read the IEA paper John Duffy and I wrote a few years ago. In short, the theory is only supported by correlations seen in some countries between overall consumption and alcohol-related harm. Here is a graph from Finland, for example...


And here is the USA in a paper published in 1967...



These correlations are pretty tight. Notice that there is hardly any time lag. This is true even of chronic diseases such as alcoholic liver cirrhosis because, as Terris says...

In many cases the cirrhotic process can be halted and decompensation prevented by avoiding further use of alcohol. Conversely, resumption of heavy alcohol use after a period of abstinence can decompensate a previously injured liver in a relatively short period of time.

Based on these correlations, neo-temperance campaigners such as Alcohol Focus Scotland claim that alcohol-related problems can be addressed by reducing per capita consumption. But this is like trying to make a dog happy by wagging its tail.

The fact is that alcoholics drink a disproportionately large share of the nation's alcohol. If you have fewer alcoholics, per capita consumption will decline. This alone can explain the correlation between falling alcohol consumption and falling alcohol related mortality.

But there is no reason to think that reducing per capita consumption by getting moderate drinkers to drink less or by having more teetotallers is going to have any effect on the behaviour of alcoholics. And yet, by the logic of the whole population approach, anything that reduces per capita consumption will inevitably reduce alcohol-related mortality.

Conveniently for the 'public health' lobby, this approach allows them to avoid having to get their hands dirty dealing with people who have alcohol problems. Instead, they can tinker with the guidelines and lobby for tax rises and advertising bans, none of which are likely to have any impact on dependent drinkers. Alcoholism is a complex problem and their cretinously simple 'solutions' misdiagnose the problem and are therefore doomed to failure.

One only needs to look at the relationship between alcohol consumption and alcohol deaths in different countries to see that things are far more complex than the total consumption model assumes. Here is what happened in the USA after the study mentioned above was published...

Notice how the rise in cirrhosis peaked ten years before the rise in alcohol consumption. By the time alcohol consumption started falling in the mid 1980s, rates of cirrhosis had already fallen by around 20% from their mid-70s peak.

The United Kingdom is another example. Alcohol consumption peaked in 2004 and has since fallen by 18 per cent, as this graph from the IAS shows...
By the logic of the whole population approach, alcohol-related deaths should have fallen by roughly 18 per cent since 2004 but, as data published by the ONS today show, they simply haven't. The rate of alcohol-related mortality was 11.7 per 100,000 in 2016. In 2004, it was 11.5 per 100,000. Twelve years after the peak in alcohol consumption, mortality rates are essentially unchanged.

If you take the NHS's hospital admissions data at face value (which you shouldn't), there has also been a very large increase in the number of alcohol-related hospital admissions since 2004.

Leaving the hospital data to one side, the number of deaths is essentially the same as it was before consumption starting falling. As mentioned, this lack of association cannot be put down to a time lag. There is no time lag for acute alcohol-related harm (eg. drink driving deaths) and the time lag for alcohol-related diseases is remarkably short (within a few years, as most). It is not like smoking.

There has been no relationship between consumption and mortality in Britain since 2004. Death rates went down slightly and then went up slightly while consumption fell consistently.

The correlation that is sometimes observed between these two variables is not causal. Both consumption and mortality can be dictated by a third variable. That third variable is heavy drinking and/or alcoholism. But other things can affect per capita consumption and there is no reason to believe that per capita consumption (ie. other people's drinking) will affect alcoholism.

The story in Britain since 2004 seems to be that per capita consumption has fallen without levels of alcoholism falling. There are more teetotallers, young people are drinking much less, and non-heavy drinkers seem to be drinking less. Moreover, the fall in alcohol consumption has been largely driven by a fall in beer consumption. This is potentially significant because beer is less associated with alcoholism (when epidemiologists looked at the 'paradigmatic' data for the USA shown above, they found a somewhat more convincing association when they looked only at spirits).

All of this has caused per capita consumption to fall sharply, but unfortunately it has not led to a fall in the number of alcohol-related deaths. The most obvious explanation for this is that the small minority of people who are genuinely at risk of alcohol-related mortality has not got any smaller.

Targeted interventions, rather than policies aimed at the entire population, are what is required.

No comments: