By David Collins and Helen Lapsley: Remarks on the Draft
Workshop on Guidelines for Estimating the Avoidable Costs of Substance Use and Abuse, sponsored by Health Canada, June 22-23, 2005, Ottawa.
Keywords: Health;
Thankyou for the invitation to attend what is proving to be a very interesting seminar in, if I may so, a pleasing and attractive city. As one would expect, David Collins and Helen Lapsley have contributed a valuable paper, albeit as they insist, a preliminary draft.
It behoves me, as the first speaker this morning, to go back to the first workshop in 1995 which led to the International Guidelines for Estimating the Costs of Substance Abuse published by the World Health Organisation in 2003. Those guidelines are the foundation for this paper. The progress reflects both new data bases and new analytical developments, together with addressing loose ends which had not been considered sufficiently in the earlier report.
It is worth recalling that the initial intention was a gold standard for the guidelines, an ideal grounded in standard economic theory, which would enable the social cost estimates to be extended to other areas of economics such as the System of National Accounts and the evaluation of policy options including substance abuse prevention and treatment. Systematising and measuring the notion of avoidable costs enables us to go a step further towards systematic policy evaluation. That it does, in part is a consequence of the gold standard decision. Had we not been so disciplined, this new work would be much less potentially valuable.
However, while the gold standard is the underlying principle, we have all found it difficult to apply the standard for various practical reasons, typically because of data deficiencies. So while keeping the gold standard in mind, we have had to use silver, copper and pewter. This is, of course, inevitable but sometimes we forget.
That is my first recommendation to David and Helen. I would like them to be more explicit in their paper as to where there is a gold standard, and where, because of the difficulties of implementing it, they are using a lower standard.
Let me illustrate this by reference to the paper’s advocacy of the Arcardian norm. A number of participants have argued that there are deficiencies with the measure. One problem was nicely illustrated by the comparison of the high suicide rate for Australia and the lower one for Greece. We know that not only are suicide rates culturally sensitive, but so are the recording of suicides. This problem of valid international comparisons because of different recording practices applies for other conditions, as have various contributions have instanced.
Even were there not these difficulties, economics have had problems using an arcadian type norm in at least two other areas that I know of. One is to identify the most productivity-efficient firm, the other is to identify the most cost-efficient health care provider. The parallel is Armstrong’s Arcadian norm attempts to identify the most health-efficient public health configuration.
Economists have put some effort into trying to improve the quality of their arcadian measures – especially in regard to production frontiers. (The first, and crudest, is to average the three most efficient observations.) No doubt those methods could be applied to the Armstrong arcadian too,. But while I would welcome such improvements in due course, my point is another.
At this stage, the gold standard for the minimum level of disease is totally impractical. So the paper recommends a lower standard. Whether it is the best practical one is initially the decision of this workshop, and later will be settled by the international research community. The point here is that the report has to recommend a procedure which, whatever it is, is not on the gold standard.
All I am asking is that the report clearly state when it is recommending a lower standard as a practical response to data inadequacies. In particular in a decade or so, people need to look at the report and recognise where in it there are (eternal) analytical truisms, and where there are proposals relevant for these times which have since been superceded.
The report also needs to say something about where better data already exists. Were I to be looking at avoidable costs for smoking cessation, I would go to the various longitudinal empirical studies which have followed the health record of those who have given up smoking. Ideally there would be a meta-study. Without one I would in effective create one, albeit a little differently from industry standard, because I would need slightly different information.
Of course I would check my conclusions against Armstrong’s Arcadian norm. If there was a major disagreement, I’m not sure what I would do – it would depend upon an assessment of why the inconsistency.
Although the epidemiological data for this strategy is available for tobacco and alcohol it is not generally available for illicit drugs. For them – and inter-country comparisons – I may need to go back to something like the Arcadian norm as a backstop, when the available research fails us.
Even so, I wonder if the norm can be avoided for illicit drugs. One argument is that anti-policies lead to substitution, perhaps as users move from heroin to cocain. The Arcadian norm is a way of dealing with these shifts of use. But would it not be better to directly estimate the degree of substitution. Indeed is that not what a policy maker really wants. The argument that we cant know precisely the degree of substitution is not valid, insofar as if the method implicitly assumes it anyway. The need is to be explicit.
The primary justification for my strategy is that it uses better quality data. Were I to use inferior data for, say, evaluating the impact of a hike in tobacco excise, you may be sure that someone subsequently doing, say, the impact of a public education campaign, would use another one – judged better – and the two results would be incomparable, which is exactly what policy makers do not want.
But I would also use this data because I need to a good time-response profile. As much as the conclusions may be sensitive to the choice of the Arcadian norm or whatever, I suspect they may be even more responsive to the profile. This is because the evaluation discounts events so that those near the asymptote have far less quantitative significance than those close to the policy’s initial impact.
That is, of course, depending on the size of the discount rate. Some one has to say that because DALYs already have a discount rate incorporated in them – 5 percent p.a. I think it is – that they cant be used for avoidable cost estimates, unless by coincidence the evaluation is using the same discount rate. (The NZ rate is 10 percent p.a.) To be useful in the context, a DALY has to be broken down into its component parts, becoming QALY like.
(DALYs also use an age adjustment. But public policy may not accept the validity of treating different people differently according to their age. Again the underlying data upon which DALYs are based may be more useful, again leading us back to the more neutral QALYs.)
If the response to the policy under consideration is rapid, then irrespective of the discount rate, the avoidable cost estimate becomes very much like the total cost estimate. That needs to be said doesnt it? Also that in the long run, the total cost configuration may look very similar to the avoidable cost configuration.
Actually there is much that needs to be added to this draft report to remind the neophyte reader that it is based on the International Guidelines. But here I am sure that I am counselling the writers what they have already have planned.
Finally, I have not mentioned the criminology. While the paper pays some attention to the social costs of drug induced crime, that seems to me to be and issue for the International Guidelines, and should be incorporated in the next edition. At this stage – I am open to persuasion – I dont see that it raises particular issues for avoidable cost estimates, which do not occur elsewhere. I may be wrong, in which case the final report will correct me. If I am correct we are most fortunate that this report is filling the gap until we get the third edition of the Guidelines.
Their paper opens up a new set of issues. One might say that it opens up too many issues: those who commissioned the paper must be well pleased with it, for it covers much more than they might have expected from their commission. But that only reflects the enthusiasm and commitment we have come to expect from David and Helen.