Thursday, 7 October 2021

Alcohol-related deaths fall by 23% (sort of)

The Office for Health Promotion has got off to a flying start by announcing a 23% fall in alcohol-related deaths in England. This was achieved by changing the methodology behind the estimate and it was a piece of work carried out by Public Health England, but I'll take it. 

During the pandemic, there has been a lot of confusion about how many people have died 'with Covid' as opposed to 'of Covid', despite us having the death certificates showing how many died with Covid as the underlying cause (which is about 90% of all Covid-related deaths). 

I wonder how many of the people arguing the toss about this realise how shaky the figures are for obesity, smoking and alcohol-related mortality. For these diseases, academics tend to the use the system of attributable fractions. Put simply, they work out how many people are exposed to a risk factor (eg. alcohol) and then work backwards from epidemiological studies which show the increase in risk to estimate what proportion of deaths are caused by the risk factor.

Death certificates are not involved when it comes to establishing causation. The academics just look at the number of deaths from, for example, heart disease and state that x% were caused by smoking, alcohol, obesity etc. 

And so we decide that 50% of drownings are due to alcohol, for example, and then see how many drownings take place each year. It's a rough and ready estimate that has the virtue of being cheap, but it is prone to all kinds of flaws and uncertainties. Observational epidemiology is not an exact science and it is very difficult to establish the baseline risk (ie. the risk to someone who has no obvious risk factors).

Every few years, the academics update their assumptions. That is what has just been done. They also update their estimates of exposure. Previously, PHE was using alcohol consumption data from 2005 when the good people of England were drinking more than they do today. 

The result is a substantial reduction in the apparent harm done by excessive drinking. The number of alcohol-related hospitalisations has fallen below one million for the first time in years (a figure that has been inflated by the inclusion of more secondary diagnoses) and alcohol-related mortality is 23% lower than previously estimated.

I emphasise excessive drinking because the authors acknowledge that there is little to be gained by getting the average drinker to reduce his consumption.

As alcohol can be so damaging to health, wellbeing and society, it’s obviously a positive thing that England’s overall consumption of alcohol has fallen. However, before we become complacent, we should consider in more detail the alcohol consumption patterns across the population. If we look beneath the population-level trend, we can see data on consumption to suggest that many of the people who have chosen to drink less (or not at all) are those who were not at greatest risk of harm. It appears that many heavier drinkers, who are at most risk, have not reduced their alcohol consumption and may even have increased it.

This is another blow to the single distribution theory of drinking and the whole population approach to alcohol.

The new estimates cover 2019 but do not yet extend to 2020 when we saw a big rise in alcohol-specific deaths despite a significant fall in consumption. This, too, shows that the whole population approach is misguided.

No comments: