America Tour: Attribution, prediction, and the causal interpretation problem in epidemiology

Next week I’ll be visiting America to talk in Pittsburgh, Richmond, and twice at Tufts. I do not expect audience overlap so I’ll give the same talk in all venues, with adjustments for audience depending on whether it’s primarily philosophers or epidemiologists I’m talking to. The abstract is below. I haven’t got a written version of the paper that I can share yet but would of course welcome comments at this stage.


Attribution, prediction, and the causal interpretation problem in epidemiology

In contemporary epidemiology, there is a movement, part theoretical and part pedagogical, attempting to discipline and clarify causal thinking. I refer to this movement as the Potential Outcomes Aproach (POA). It draws inspiration from the work of Donald Ruben and, more recently, Judea Pearl, among others. It is most easily recognized by its use of Directed Acycylic Graphs (DAGs) to describe causal situations, but DAGs are not the conceptual basis of the POA in epidemiology. The conceptual basis (as I have argued elsewhere) is a commitment to the view that the hallmark of a meaningful causal claim is that they can be used to make predictions about hypothetical scenarios. Elsewhere I have argued that this commitment is problematic (notwithstanding the clear connections with counterfactual, contrastive and interventionist views in philosophy). In this paper I take a more constructive approach, seeking to address the problem that troubles advocates of the POA. This is the causal interpretation problem (CIP). We can calculate various quantities that are supposed to be measures of causal strength, but it is not always clear how to interpret these quantities. Measures of attributability are most troublesome here, and these are the measures on which POA advocates focus. What does it mean, they ask, to say that a certain fraction of population risk of mortality is attributable to obesity? The pre-POA textbook answer is that, if obesity were reduced, mortality would be correspondingly lower. But this is not obviously true, because there are methods for reducing obesity (smoking, cholera infection) which will not reduce mortality. In general, say the POA advocates, a measure of attributability tells us next to nothing about the likely effect of any proposed public health intervention, rendering these measures useless, and so, for epidemiological purposes, meaningless. In this paper I ask whether there is a way to address and resolve the causal interpretation problem without resorting to the extreme view that a meaningful causal claim must always support predictions in hypothetical scenarios. I also seek connections with the notorious debates about heritability.

JOB: UJ, prediction and/or philosophy of epidemiology

The Department of Philosophy at the University of Johannesburg has a fixed term vacancy expiring 30 April 2020, at level Senior Lecturer/Associate Professor/Professor, for someone willing to work on projects related to prediction or philosophy of epidemiology. The vacancy is created by Alex Broadbent’s 5-year appointment as a Dean.

Deadline: 2 August 2015 [NOTE EARLY DEADLINE]

Applicants must hold a PhD in philosophy and must demonstrate strong publication record and potential commensurate with level.

Responsibilities include teaching undergraduate and postgraduate modules, postgraduate supervision, research leading to publications, and assisting with administrative duties as determined in consultation with the Head of Department. Typical teaching load in the Department requires undergraduate teaching for no more than three our of the four terms, and usually the Department aims to organise teaching so as to leave two out of four terms free of undergraduate teaching for each of its staff.

To apply, go to

For further details go to or contact Prof Hennie Lotter

Why would an American medical professional help a tobacco company defend a lawsuit in Korea?

I find myself reading a document by a senior American epidemiologist and medical doctor, commissioned by Philip Morris Korea, in which this individual explains at great length why, in his view, the case of the National Health Insurance Service (NHIS) has brought against the tobacco companies must fail. The NHIS has asked me for an expert opinion, responding to this one.

I cannot help but wonder what would lead an epidemiologist and doctor to deliver a written expert opinion of this nature.

I was offered a fee for my opinion but I refused, since I felt credibility issues are at stake. I am spending my time on this because, first, I believe that the prevailing opinion that epidemiological evidence cannot be applied to the proof of specific causation is incorrect, and second, because I believe that it is worth the effort in this case to argue the point, given the public health burden of tobacco in Korea (where nearly half the male population smokes, and where cancer incidence is the highest in all Asian countries).

I have no idea what motivates my counterpart, but I am surprised that there remain any American epidemiologists who would do anything to help a tobacco company, given the history of the relationship between the profession of epidemiology and tobacco litigation in the USA.

I will double check there is no embargo, and if not, I will publish both the other side’s expert opinion and my own on this blog.

Psychiatric Disorders, Multifactorialism, and Kinds

Today I attended and talked at a great symposium called “What Kinds of Things Are Psychiatric Disorders?” organised by the University Center of Psychiatry at Groningen. The guest of honour was Kenneth Kendler, who gave a long, wide-ranging, and thoroughly excellent talk about the reality or otherwise of the psychiatric disorders captured by the DSM. It was one of the best talks I have heard in a long time, and philosophically more astute than many talks by self-identifying philosophers (including myself). The position he outlined got me thinking about a number of issues in the philosophy of epidemiology and of medicine.

He begins by arguing that realism about psychiatric disorders cannot be modelled on realism about physical kinds: it cannot be like realism about the periodic table. At best it would need to be like realism about biological species – acknowledging the lack of essential properties, the fuzzy boundaries, and so forth. His own view, however, is that realism of even this moderate kind is not warranted about psychiatric disease kinds, at least not those in the DSM. Part of his argument is a pessimistic induction from past psychiatric diagnoses that are no longer recognised. The other part comes from the massive multifactorialism of psychiatric disorders.

One interesting thing about this view is that it also applies to multifactorial diseases outside psychiatry – cancer, for example. The same considerations apply – both the mutlifactorialism, and the pessimistic view of the durability of our current classifications. The two arguments interact, perhaps: massive multifactorialism means that we do not have a good general explanation of why a given set of symptoms occurs, and this lack of explanation indicates that we don’t understand the phenomenon very well and thus that we are quite likely to be wrong about it in some yet-to-be-discovered ways.

Kendler’s view of illness in general, however, is not constructivist. He combines a sceptical stance towards the kinds in the DSM with a realist-leaning pragmatism about psychiatric disorder, the overarching thing, as opposed to the ways we have cut it up. In other words, he thinks that psychiatric illness is a real thing; but he is sceptical about our way of individuating psychiatric diseases. So, at least, I understood his position.

This view also connects in interesting ways with the debate about health in the philosophy of medicine. In that debate, “realism” is not used – instead, “naturalism” is the closest correlate. It is not an equivalent, however, since a naturalist typically denies that there is any normative component to health facts, while Kendler – in response to an acute question from the audience – thinks that health is fundamentally about human goals and suffering, and thus fundamentally normative. The naturalism/normativist debate elides two dimensions of disagreement, one concerning the objectivity of health facts (on which Kendler is a pragmatist, but towards the objective end of the spectrum, opposed to so-called normativists in the health debate), the other concerning whether health is value-laden (on which Kendler appears to be with with the normativists). I have thought for a while that it should be possible to occupy a position like this – agreeing with naturalists on one dimension of disagreement, normativists on another – and I’m pleased to have found someone who does, and someone so credible.

The picture, then, is of a spectrum of disorder which is fairly independent of us, and also normative, but which is fairly arbitrarily cut up by us in our attempts to understand it. That’s a crude summary, of course. But it made me wonder if an exact reverse might also be attractive. Some disease kinds, especially infectious diseases, might suit realism of the kind we might have about biological species. After all, infectious diseases correspond to actual biological species, or something similar enough. Thus we might be (fairly) realist about cholera – it’s what happens when something we are fairly realist about (vibrio cholerae) gets into the small intestine of something else that we are fairly realist about (us). However, we might be irrealist, or less than realist, about health facts in general – not just psychiatric health and disease, but health and disease in general. We might think that health is not a fundamental category – not a kind, not even in the way that species are kinds – and yet think that certain diseases do count as kinds.

I am not sure where this leads, but my hope would be to develop a notion of health and illness – the mere absence of health – as secondary properties, like colour, which depend in non-arbitrary non-contingent ways on us (as opposed to a constructivist stance, which makes facts depend on us in contingent ways). Such a view may make room in the picture for the diagnoser as well as the diagnosed, and it may explain why a particular set of objective characteristics seem to us to belong together as “health” or as a given illness, despite the fact we often find little objective to bundle them together. Perhaps it also would answer to Kendler’s pragmatist intuitions, which I also find very compelling.

Whatever the prospects for Kendler’s view or for this hastily-sketched alternative, the lecture really helped me pull together a number of these issues. It left me thinking that the overarching goal for contemporary philosophical work on the nature of health and disease must be to link the debates about the status of health (naturalism, normativism, and all that) with debates about the status of disease kinds and with the literature on natural kinds more generally. Kendler’s excellent talk today embarked on this extremely complex task and went a remarkable way towards offering a comprehensive and plausible position.

Tobacco and epidemiology in Korea: old tricks, new answers?

Today I participated in a seminar hosted by the National Health Insurance Service (NHIS) of Korea, which is roughly the equivalent of the NHS in the UK, although the health systems differ. The seminar concerned a recent lawsuit in which tobacco companies were sued by the NHIS for the costs of treating lung cancer patients. The suit is part of a larger drive to get a grip on smoking in Korea, where over 40% of males smoke, and a packet of 20 cigarettes costs 4500 Korean Won (about USD 4.10 or UKP 2.80). The NHIS recently suffered a blow at the Supreme Court, where the ruling was somewhat luke-warm about a causal link between smoking and lung cancer in general, and moreover argued that such a link would anyway fail to prove anything about the two specific plaintiffs in the case at hand.

I was struck by the familiarity of some of the arguments that are apparently being used by the tobacco companies. For example, the Supreme Court has been convinced that diseases come in two kinds, specific and non-specific, and that since lung-cancer is a non-specific disease, it is wrong to seek to apply measures of attributability (excess/attributable fraction, population excess/attributable fraction) at all.

This is reminiscent of the use of non-specificity in the 1950s, when it was seen as a problem for the causal hypothesis that smoking causes lung cancer. It also gives rise to a strategy which is legally sound but dubious from a public health perspective, namely, first going for lung cancer, and leaving other health-risks of smoking for later. This is legally sound because lung cancer exhibits the highest relative risk of the smoking-related diseases, and perhaps it is good PR too because cancer of any kind catches the imagination. But the health burden of lung cancer is low, even in a population where smoking is relatively prevalent, since lung cancer is a rare disease even among smokers.

The health burden of heart disease, at the other end of the spectrum, is very large, and even though smoking less than doubles this risk (RR about 1.7), the base rate of heart disease is so high that this amounts to a very significant public health problem. I do not know what the right response to this complex of problems is: clearly, high-profile court cases are have an impact that extends far beyond their outcome, and also the reason that people stop smoking, or accept legislation, need not be an accurate reflection of the true risks in order for those risks to be mitigated. (If you stop smoking to avoid lung cancer, you also avoid heart disease, which is a much better reason to stop smoking from the perspective of a rational individual motivated to avoid fatal disease.) Nonetheless I am struck by the way that legal and health policy objectives interact here.

I was also interested to hear that the case of McTear was a significant blow to the Korean case because of its findings about causality, which indeed are exactly those of the Korean case. That case is not well regarded in the UK, and not authoritative (being first instance), so it is interesting – and unfortunate – that it has had an effect here.

The event was an extremely good-spirited affair, and the other speakers had some interesting things to say. My book, in Korean, received a significant plug, not least, I suspect, because the audience not understanding much of my talk, were repeatedly referred to it for more detail. The most shocking thing about the event was to hear the same obfuscatory strategies that are now history in Europe and America being used, to good effect, by the very same companies in this part of the world. It is one thing to defend a case on grounds that one believes, but there is not anyone who still reasonably believes that smoking does not cause lung cancer, which seems to be the initial burden that plaintiffs in this sort of case need to prove. It is a bit like being asked to begin your case against a scaffolder who dropped a metal bar on your head with a proof of the law of gravity, and then being asked to prove that the general evidence concerning gravity proves that gravity was the cause in this particular case, given that not all downward motions are caused by gravity. – Not exactly like that, of course, but not exactly unlike, either.

On the positive side, I am hoping that a clear explanation of the reasoning behind the PC Inequality that I favour might help with the next stage of the case, although I am unclear what that stage might be.

Is consistency trivial in randomized controlled trials?

Here are some more thoughts on Hernan and Taubman’s famous 2008 paper, from a chapter I am finalising for the epidemiology entry in a collection on the philosophy of medicine. I realise I have made a similar point in an earlier post on this blog, but I think I am getting closer to a crisp expression. The point concerns the claimed advantage of RCTs for ensuring consistency. Thoughts welcome!

Hernan and Taubman are surely right to warn against too-easy claims about “the effect of obesity on mortality”, when there are multiple ways to reduce obesity, each with different effects on mortality, and perhaps no ethically acceptable way to bring about a sudden change in body mass index from say 30 to 22 (Hernán and Taubman 2008, 22). To this extent, their insistence on assessing causal claims as contrasts to well-defined interventions is useful.

On the other hand, they imply some conclusions that are harder to accept. They suggest, for example, that observational studies are inherently more likely to suffer from this sort of difficulty, and that experimental studies (randomized controlled trials) will ensure that interventions are well-specified. They express their point using the technical term “consistency”:

consistency… can be thought of as the condition that the causal contrast involves two or more well-defined interventions. (Hernán and Taubman 2008, S10)

They go on:

…consistency is a trivial condition in randomized experiments. For example, consider a subject who was assigned to the intervention group … in your randomized trial. By definition, it is true that, had he been assigned to the intervention, his counterfactual out- come would have been equal to his observed outcome. But the condition is not so obvious in observational studies. (Hernán and Taubman 2008, s11)

This is a non-sequitur, however, unless we appeal to a background assumption that an intervention—something that an actual human investigator actually does—is necessarily well-defined. Without this assumption, there is nothing to underwrite the claim that “by definition”, if a subject actually assigned to the intervention had been assigned to the intervention, he would have had the outcome that he actually did have.

Consider the intervention in their paper, one hour of strenuous exercise per day. “Strenuous exercise” is not a well-defined intervention. Weightlifting? Karate? Swimming? The assumption behind their paper seems to be that if an investigator “does” an intervention, it is necessarily well-defined; but on reflection this is obviously not true. An investigator needs to have some knowledge of which features of the intervention might affect the outcome (such as what kind of exercise one performs), and thus need to be controlled, and which don’t (such as how far west of Beijing one lives). Even randomization will not protect against confounding arising from preference for a certain type of exercise (perhaps because people with healthy hearts are predisposed both to choose running and to live longer, for example), unless one knows to randomize the assignment of exercise-types and not to leave it to the subjects’ choice.

This is exactly the same kind of difficulty that Hernan and Taubman press against observational studies. So the contrast they wish to draw, between “trivial” consistency in randomized trials and a much more problematic situation in observational studies, is a mirage. Both can suffer from failure to define interventions.