Friday, 13 September 2013

Population approaches to alcohol

I’ve seen some discussion about the total consumption model of alcohol harm in the past few weeks, mostly from Chris Snowdon – both in a blog post for the IEA and on his own blog.  This discussion prompted me to think a bit more about the various ways consumption of alcohol is understood, and the various interventions that might be employed to address it.  I’m not going to add anything particular original here, but maybe point to some issues that sometimes get overlooked – and certainly have done in the recent discussions I’ve seen.

I’ve written before about how public health approaches tend to focus on population-wide problems, and population-level consumption of alcohol might be one of these issues.  Chris Snowdon suggests that the total consumption model is all about targeting heavy drinkers with something of a blunderbuss approach.  The theory, he states, runs that the consumption of the heaviest - and therefore problematic - drinkers moves in step with overall population consumption.  Policies designed to reduce population-wide consumption are in fact aimed at targeting these heaviest drinkers.

As John Holmes has pointed out, this needn’t be in quite the way that Chris Snowdon criticises; the Sheffield MUP modelling isn’t actually based on population averages, but rather targeted groups.

But there’s something more too.  Snowdon’s point is that this is a theory that isn’t held up by evidence, but as he’d probably be the first to point out, evidence and policymaking isn’t as straightforward as all that.  There’s a suggestion that Kettil Bruun, one of the lead advocates of this kind of policy, proposed it on the basis of politics – that a universal policy would be less stigmatising than measures targeted at particular ‘problem’ drinkers.

And also the evidence Snowdon sees as undermining this model is questionable – at least in the way he understands and marshals it.  First, I’m always slightly questioning of hospital episode statistics, and how they have changed over time, as this relates to accuracies of coding.  Second, even if these were accepted as measures of alcohol-related harm, it shouldn’t be surprising if they don’t march in step with current consumption figures.  Apart from the fact that consumption figures themselves can be questioned*, one can expect a lag before the health effects are felt of any increased consumption.

There’s a more important point here, though, about the idea of population level interventions.  One is that such universal interventions don’t have to imply equal effects on all individuals or groups within society, as John Holmes points out.  (Of course, as I’ve pointed out before, in many of the arguments surrounding MUP these apparently targeted effects may be as much a political claim as a genuine aim.)

Another, often overlooked point is that in fact MUP can be understood as a genuinely universal measure that does not rely on the total consumption model as outlined by Snowdon.  I often reference James Morris’ article about MUP on this blog, and unsurprisingly I’m going to do so again here.  The point is: alcohol harm is all about risk.  Not all alcoholics necessarily get liver disease, for example.  Your risk is affected by the amount of alcohol you consume, but it’s hard to predict on an individual basis.  The population model doesn’t face this issue because it’s aggregating so many risks that the broad relationship holds.  That is, if we all reduce our drinking marginally, then the individual change in risk levels is barely noticeable.  However, when all these slight reductions in risk are aggregated at a population level, the difference becomes noticeable in terms of mortality figures.

The important point here is the contrast with Chris Snowdon’s position that alcohol problems are located in a small section of society – ‘those who have a genuine drinking problem’.  It’s certainly true that there are groups of people who have particular alcohol problems, but there’s a whole host of issues relating to alcohol consumption that could be deemed problematic.  It’s the very fact that there are genuine issues relating to the wider population that means I’m concerned about how resources will be targeted in the future to meet public health aims.

And I don’t think this idea that ‘problems’ are confined to a small section of society is helpful.

First, as Alastair Campbell has pointed out this week, the seriousness of someone’s alcohol issues might not be immediately apparent to those around them.  People can be very resourceful in hiding their problems.

Second, people are able to hide such problems from themselves.  I don’t just mean the ‘denial’ of addiction.  We like to think that there are safe (even beneficial) levels of alcohol consumption, but in fact the analogy is more like driving a car: there are safer ways to drive, and it can be beneficial to your mental and physical health as well as quality of life, through many knock-on effect such as sociability, but there are risks in any level of consumption/driving.  The question is about the management of those risks.

Government discussions of alcohol policy haven’t always been helpful in this respect, as concepts such as ‘binge’ and ‘responsible’ are very malleable, and allow people to bend them to fit their own definitions of normal or acceptable – as in this research study which showed that older people tend to define acceptable drinking with reference to ‘propriety’ rather than ‘health’.  I found the same thing in my own research with younger drinkers, where people were perhaps surprisingly able to draw distinctions between themselves and the ‘other’, ‘irresponsible’ drinkers.  This isn’t to say that health should always be the number one priority in someone’s decision whether or not to drink; but it is illuminating to think about how it might be being disregarded.

This doesn’t mean I’m advocating MUP, or other policies that address total consumption across a population.  It simply means that the arguments for these interventions can’t be undermined by a narrow consideration of ideas around total consumption, and it may be unhelpful to characterise alcohol-related ‘problems’ as confined to a small group within society.

*The Mark Bellis et al research I link to is particularly relevant for Chris Snowdon’s point about harm not following consumption across socio-economic groupings.  This research, though not yet concluded, does suggest that there are all sorts of reasons for this pattern – not least that the consumption estimates may be out by differing amounts for different socio-economic groups.


  1. I agree with you about hospital admissions data. In the UK, at least, the criteria for 'alcohol-related' admissions have been expanded so vastly and so often that the stats for the last decade are almost entirely useless. I think my point still holds if you look at figures that are defined consistently, such as death and liver cirrhosis, and I hope that the figures from Australia which also undermine Ledermann are more reliable.

    If someone believes that alcohol is hazardous at almost any level (eg. WCTU, David Nutt, Ian Gilmore), then it is almost a tautology to say that harm will rise in line with consumption. The claims about alcohol and breast cancer would suggest that is true for women, but even if there is a causal relationship there (I'm not convinced), the relative risks are so low that they are simply not worth worrying about. Public health folk will use the line about small individual risks adding up to significant effects at the population level, blah, blah, blah, but if a risk is trivial to the individual it deserves a trivial response from government.

    I'm prepared to accept that MUP does not rely on the Ledermann hypothesis. There are plausible reasons to think of it as a slightly more targeted approach, so long as we assume that problem drinkers drink cheaper booze. In practice, it is more likely to reduce per capita consumption than problem drinking because demand is more elastic amongst moderate drinkers (the Sheffield model assumes the opposite, but that flies in the face of the evidence).

    At a very basic level, the Total Consumption Model is plausible. If fewer people ever drink, there will be fewer alcoholics. Similarly, if fewer people ever gamble, there will be fewer compulsive gamblers and if fewer people start smoking there will be fewer smokers. That is trivially true, but until the public health lobby starts preaching abstinence, it is not very helpful.

    It is also true that consumption and heavy drinking will often go up and down together. As John Duffy has been saying for years, this is reverse causation: the Pareto principle means that the consumption of the heaviest drinkers is bound to have an effect on overall consumption because they are drinking such a large share of the alcohol. This explains why there are plenty of correlations between consumption and harm, and it also explains why policies aimed at reducing per capita consumption often fail to reduce heavy drinking. They're wagging the tail to make the dog happy.

    Ultimately, the Total Consumption Model is popular because the policies it justifies are simple, easy and political. These policies are taken virtually word-for-word from the anti-tobacco playbook (in the case of smoking, per capita consumption and harm are clearly closely linked).

  2. In your previous post of 17th July, you say

    "That is, reduce everyone's consumption by a tiny average amount, and the difference to any specific individual in terms of risk might be almost negligible, but when you add all this up across a population, you get a noticeable fall in the total harm."

    I say that "total harm" is a ridiculous concept and smacks of fascism. If the population of the UK was ten times the size, the total harm would be ten times as big. But this is of no relevance to the individual. What is relevant is "excess absolute risk". My current alcohol consumption is about 35 units a week. I have a certain probability x, say, of developing liver disease. If I increase my consumption to 40 units, I have a new probability y. What is relevant to me is y-x, which is likely to be very small. If we take the example of passive smoking and accept the claims (which I don't) that 4 hours of daily exposure in a smoky pub results in a 20% increase in lung cancer risk, this equates to an increase in absolute risk of around 1 in 750. Putting it this way gives the invidual the requisite information to make a choice. A good way of getting a feel for this is to imagine a bag of 750 balls, 749 white and 1 black. How worried are you that if you pick a ball at random, it will be the black one? Probably not very. With breast there would be only 10 white balls.

  3. Good points.

    I'd disagree about individual vs population risk, though. It could be argued that lots of risks are trivial to the individual (in terms of probability, not seriousness) but it's fair enough that the government takes action. Road safety might be one but you could equally count a huge amount of regulation more generally.

    And there are plenty of cases where we're not all at equal risk of suffering ill effects but there's a universally-applied policy on the issue (including alcoholism).

    I don't think that's an unfair approach per se. It's about balancing that attempt at central control with the individual's right to put themselves at risk.

  4. When I talk about trivial risks, I mean - roughly speaking - those reported in epidemiological studies that show less than a doubling of risk - ie. most of modern lifestyle epidemiology. Things like alcohol and breast cancer or red meat and bowel cancer, where a normal person would not consider the sacrifice of abstaining to be worth the very modest reduction in absolute risk from one disease.

    In each case, it could be argued that the elimination or reduction of the risk factor would 'save' a substantial number of lives if we multiply the risk over a population of millions, but there are several reasons why this does not justify government action (besides from the libertarian objection).

    Firstly, the epidemiology is rarely convincing at this level, particularly when the results show less than a 50% increase in relative risk, which is typical. Causation is not proven and the mechanisms are frequently biologically inexplicable. We also don't know what the substitution effects and unintended consequences will be.

    Secondly, even if the epidemiological evidence was better, there are THOUSANDS of relative risks that would require intervention. Take the World Cancer Research Fund's reports, for example. They find links between virtually every type of food and some cancer or other. Extrapolating from this, you would find that millions of lives could be 'saved' from lifestyle regulation, but it would require a near-totalitarian system. The risks we hear most about are the ones that are politically useful to certain interest groups (eg. alcohol and breast cancer, passive smoking and lung cancer), but there are many, many more.

    Thirdly - and most importantly, IMO - very few, if any, people who currently enjoy moderate drinking or eating red meat are inclined to give up or reduce their consumption of alcohol or steak on the basis of a modest increase in their personal risk of getting a couple of types of cancer. That is an entirely rational choice. These activities may not be 'safe' in absolute terms, but they are safe enough for them.

    From a population/'public health' point of view, these things are 'unsafe'. They 'kill' x number of people in the UK/Europe/worldwide* and SOMETHING MUST BE DONE. But on whose behalf? The individuals in the society aren't sufficiently bothered and the public health people cannot seriously believe they know these individuals' preferences better than they do (OK, they probably do believe that, but they're wrong). So what mandate does the government have to interfere?

    * Yes, when the risk is really small and the disease is really rare, they resort to extrapolating across the whole planet:

  5. Jonathan - I don't think 'total harm' is a problematic concept in itself, unless you were to take a complete anarchist view of the role of government. Decisions are made all the time on the basis of something that might also be called net benefit. NICE is the classic example, but every decision in engineering is made in this way. All policy decisions are to some extent a balancing of competing values and interests, and if decisions are to be made across a society (such as what sort of crash barriers we should put on motorways) we must make decisions about total harm, whether these are implicit or explicit.

    Chris - lots of that is persuasive, and I'm still undecided on MUP, partly because I'm unclear exactly what problem it's definitely aiming at (it can't possibly address all alcohol-related 'problems' people identify in the UK today).

    However, I'd still say it's all a matter of degree. Living isn't 'safe' in that I'm quite likely to die something of an 'avoidable death'. But that doesn't mean that we shouldn't address some of the issues on a societal basis. It might be rational to allow certain sorts of risks, and I think there are certainly problems with how public health in particular sees some risks as acceptable and others as irrational, but that doesn't mean that all risks should be disregarded by the state - and very few risks can be said to have effects that are isolated to that individual.

    I don't know if that's really an answer, but I'd just always want to stress that this isn't an area where there are definitive lines; it's about judgement as to what is reasonable to accept. We don't accept that all individuals' expressed preferences are rational, justifiable and should be given free rein, so we need to think carefully about what makes the case for intervention. My concern would be that we often duck out of these discussions (as in nudge approaches, where effectiveness of an intervention is often seen as a justification, when the two are quite distinct). But here we are having a discussion about when government should intervene - so that's encouraging!