Thursday, September 27, 2012

Fine-Grained vs. Indiscriminate Allocation

Reading the American Medical Association's Council on Ethical and Judicial Affairs,* I'm struck by their opposition to making fine-grained discriminations, and their general preference for more indiscriminate -- or "equal opportunity" -- allocation methods.  For example, they recognize that expected differences in magnitude of change in quality of life (approximated by the amount of functional improvement afforded by a treatment) between patients is morally relevant, but "only when they are very substantial" (30).  Due to the imprecision of the measure, they conclude that "considering small differences in functional improvement would often fail to maximize overall benefit to patients" and "would be unfair" to those for whom quality of life diverged significantly from functional status.

Of course, given limited information we are doomed to "often fail to maximize overall benefit".  But isn't a semi-reliable (albeit imperfect) guide better than none at all?  How does ignoring a piece of (even weak) evidence increase our chances of maximizing overall benefit?  Surely it just means that we'll end up making sub-optimal allocation decisions more often: Whereas tracking fine-grained difference in functional status allows us to identify the greatest improvement in quality of life with reliability at least slightly better than chance, now it is entirely down to chance.  Why is such an indiscriminate policy not just as "unfair" to the majority for whom quality of life does correlate with functional status?
In spelling out the implications of their view, the Council claims that it lessens the chances of discrimination against people with disabilities: "[S]ince only very substantial differences in the change in quality of life may be considered ethically relevant, a disabled patient will be given a lower priority only when doing so allows others to receive a much greater improvement in quality of life."

But why shouldn't we also give priority to one patient over another for the sake of a merely moderately greater improvement in quality of life?  If much more of a good thing is ethically relevant, why isn't some more also morally relevant (just weakened proportionally)?  If it isn't objectionably discriminatory to take large differences in functional status into account for determining quality of life, then I don't see how taking small differences of this same measure into account could be objectionably discriminatory either.  In either case we are attempting to track, as best we can, genuine differences in the magnitude of benefit that different patients can expect from a treatment.  And there's nothing objectionably discriminatory about prioritizing greater benefits over lesser ones, nor expected greater benefits over expected lesser ones.  Even when these judgments are highly fallible, an allocation method that's better than chance is, after all, better than an allocation method that uses chance alone.

Perhaps the most bizarre feature of the Council's paper is its (non-unique) endorsement of weighted lotteries as a means of applying allocation criteria.  This way, they explain, "the patients who are the most deserving of treatment according to the ethical criteria would have the best chance of receiving treatment." (34)  But why would we ever want to introduce a chance that the most deserving patients would not be the ones to receive treatment?  They excuse the inefficiency on the grounds that "this approach would have the advantage of non-abandonment": every patient would at least have some chance of treatment.  But why is it any advantage at all, in this context, that we might end up allocating our scarce medical resources inefficiently?  I mean, I see that it's pro tanto nice for the low-priority patients to have a chance at treatment, but this is clearly and massively outweighed by the pro tanto awfulness that the high-priority patients risk being denied the treatment that they need so much more.

A less appalling allocation method that the Council also endorses is a "three-tier approach", where patients are divided into "those who are clearly good candidates for treatment, those who are clearly poor candidates, and those who do not fall into either group."  (They even endorse a simpler, two-tier, "minimum threshold" approach.)  But they then insist on using an indiscriminate method for prioritizing patients within each category.  This seems strictly inferior to a weighted ordering approach (at least for the middle group, or near the threshold between patients who will receive treatment and those who won't) -- i.e., like the weighted lottery only without the "lottery" part.  A weighted formula combines the various ethically relevant factors into our best "rough guess" of each patient's moral priority for treatment, and then we simply go down the list as far as we can until we run out of resources.  Again, this is what would seem to maximize our chances of getting the allocation right -- or at least "closer to being right" than we otherwise would get.

The fundamental difference between the Council's perspective and mine seems to be that the Council implicitly sees an important asymmetry between mistakes of allocation due to incorrect discriminations vs. mistakes of allocation due to excessive indiscriminacy.  As I see it, the aim should be to allocate resources as best we can, and the particular cause of any given failure (of fixed magnitude) is not of intrinsic moral significance.  But the Council is very concerned about when we try and fail to "maximize overall benefit to patients", and much less concerned about failures to "maximize overall benefit to patients" that are due to our not even trying.  This inconsistency strikes me as odd and unprincipled.  Would anyone care to defend it?  Am I missing something?

* = Specifically, their 1995 publication 'Ethical Considerations in the Allocation of Organs and Other Scarce Medical Resources among Patients', Archives of Internal Medicine 155: 29-40.

9 comments:

  1. "If it isn't objectionably discriminatory to take large differences in functional status into account for determining quality of life..."

    I believe it is.

    My partner Marsha had M.S., and at the end of her life could move her head and her left arm, and had some use of her left hand. (A combination of a shiny dot on her glasses and a camera on her computer allowed her to use the mouse and a virtual keyboard.) There were cognitive deficits as well.

    After her death I wrote "Marsha was the most pigheaded, tactless, warm, courageous, open-hearted, /alive/ human being I have ever known." She was that when I first met her; and she was still that in the nursing home.

    Now Marsha was a great risk-taker. If someone had offered her a gamble -- 10% chance, say, of being restored to perfect health, and 90% chance of death -- she'd probably have taken it. But she'd have been quite indignant to hear someone use her choice to argue that it would be better to save the life of one TAB (Temporarily Able-Bodied person) than nine people like her.

    Nor, I believe, would the odds she'd have accepted changed much over the course of her disease, though when I met her she could walk a city block if she had to. There'd have been a big drop when she went into the nursing home -- maybe from 20% to 10% -- but that's because being in a nursing home is in some ways like being in jail.

    Many a lifer would probably take a 90% chance of death in return for a 10% chance of liberty. Is it therefore right to save one free person rather than nine (let's say wrongly convicted) people imprisoned for life?

    ReplyDelete
    Replies
    1. Hi Angelweed, sorry for your loss. I hope you'll excuse the impersonal/academic nature of the rest of my response.

      I agree with your closing point that social factors (like imprisonment) can affect quality of life just as drastically as medical conditions. I would draw the opposite conclusion from this, though: Since life-saving treatment yields a greater benefit for a free person than a permanently imprisoned one, it would be right (in principle) to give priority to the free person if only one of the two can be saved. (Here I focus on 1-on-1 relative priority for simplicity; multiple person cases introduce further complications.)

      In practice, however, it might be wise for the medical establishment to stick to its general rule of ignoring "non-medical factors" (such as social conditions) when it comes to making medical decisions. That practical question is a complicated one, beyond my scope to address here.

      While the specific discussion in my post doesn't depend on any claims about multiple-person cases (as, again, I was primarily thinking of 1-on-1 tradeoffs), I think it is difficult to consistently accept the gamble of full health (with 90% chance of death) whilst denying the implication that saving one life at full health has moral priority (at least in principle) over saving nine at the reduced level.

      After all, this implication is just what we get when we imagine offering the gamble to ten people: if they all accept it, then their own choices would yield the expected result of one person surviving at full health, and nine of the ten dying. If it would be better to have everyone survive at reduced health, then they should not be willing to take the gamble. Otherwise you have an inconsistency between individual and aggregated evaluations. (How can it be better for each person to accept the gamble, and yet not better to achieve the result that's expected if all do exactly that?)

      Put another way: Imagine deciding on a policy from behind a "Veil of Ignorance", where you don't know which particular person you'll end up being. This method ensures impartial decision-making, as justice requires. But the Veil of Ignorance converts the policy decision into a gamble: If you have an equal chance of ending up as any individual in the scenario, then to choose to save 1 fully-healthy person over 9 in reduced health just is to accept a 90% chance of death and 10% chance of full health over an alternative gamble of a 10% chance of death and 90% chance of living with reduced health.

      Delete
    2. As I said, Marsha was a great risk-taker. It's easy for me to imagine her risking a 75% chance of death in circumstances where most people would only risk, say, a 25% chance. Your calculus then equates the life of one person in perfect health with the lives of four risk-takers with M.S., and also equates the life of the healthy person with one and one-third lives of cautious people with M.S. Doing the arithmetic then tells us that the life of one cautious person with M.S. is worth three times as much as the life of a risk-taker with M.S., which is absurd.

      The best reply I can think of is to say "Well, that's just measurement error. One life really is worth more than the other; we just don't know by how much." And perhaps this escapes assuming the conclusion since we can be sure of the qualitative result that being person X in perfect health is better than being person X with M.S.

      On the other hand, if person X were offered a complete cure from M.S., with a zero percent chance of death, but a one hundred percent chance of being person Y -- if Marsha had been offered a cure for M.S. on the condition that, cured, she would not be Marsha Metcalfe but Richard Chappell -- it's no reflection on you to say that I'm sure she would have turned the offer down.

      What I'm arguing here is that lives are not commensurable in the quality-of-life sense -- there's no knowing if a year of being Marsha-with-M.S. is better or worse than a year of being Richard-in-good-health. And the person behind the Veil of Ignorance has no more clue than we do.

      Even putting aside those worries, concerns of practicality weigh against measuring quality of life too finely. People usually return to an equilibrium of happiness after life changes, whether disastrous or joyous. And we generally don't believe that, overestimating the misery or happiness that life events will bring us. So an able-bodied person contemplating a life in a wheelchair, or a life without eyesight, is likely to fear a great deal more unhappiness than will actually ensue; while a wheelchair user or a blind person, is likely to anticipate much more happiness from a return to health than would actually be the result.

      Delete
    3. Ok, I see I misunderstood your initial point. With the "veil of ignorance" thought experiment we're supposed to consider what a prudent choice (given the uncertainty) would be. I don't think one's actual dispositions, if imprudent (being either excessively reckless or excessively cautious), show anything about how different outcomes compare morally. It's only once we have a grasp of what we think a reasonable tradeoff would be in the intra-personal case that we can extend it via the VoI to inter-personal cases.

      It's true that people generally won't want to change into a different person, because that would be for them to die. This doesn't show that we can't compare their welfare levels. I wouldn't want to be replaced even by someone uncontroversially better-off, because I'm not impartial: I want me to live. Again, this doesn't show that we can't make reasonable inter-personal utility comparisons. (All such judgments are fallible, of course, but as I note in the main post, the crucial question is whether we can do better than chance, which I think we surely can.)

      I absolutely agree that we should take into account empirical evidence of the sort you mention in your final paragraph. Our untutored intuitions about what an unfamiliar life would be like (esp. in the long term) may be demonstrably mistaken. We should of course try to avoid such mistakes as much as possible when developing a measure of quality of life. But that's not a reason to refrain from making fine-grained discriminations unless the alternative is more likely to get it right, which I've never seen anyone argue for. Instead, you seem to be implicitly relying on the same asymmetry that I critique at the end of the main post.

      Delete
    4. Perhaps it's a touchstone of the difference between us that I would neither assume that risking a 75% chance of death was imprudently reckless nor that risking 20% was imprudently cautious. I don't think there's a fact of the matter, only dispositions.

      I can't, for instance, deem it imprudently cautious for someone to refuse any non-negligible risk of death, if their current life is worth living. (Perhaps refusing a 1/10**100 chance of death in return for perfect health is imprudent -- but then we might make the same argument for refusing that risk in return for $100. So a working definition of "negligible risk of death" might be "so small it's worth risking in return for $100.) Upper limits are perhaps easier -- but say we were asking people what chance of death they would risk in return for the power of unaided flight: Is it really a fact of the matter that, say, risking a 99% chance of death is imprudent? 90%? Surely there is some risk that most of us would reject, but that someone who longed for flight might accept without our censure.

      So that's one point -- if there is no actual fact of the matter as to whether it's better to be Pat with a disability or Robin without one, then trying to measure this non-existence is at best useless.

      The second point is that it tends toward the pernicious. If (as I believe) we as a society think that the lives of people with disabilities are barely worth living, and as a result make them even more difficult, then trying to measure fine differences is likely to amplify that problem. People in North America with dark skin tend to have more difficult lives than those with light skin. Should we therefore be more willing to allocate scare medical care to those with lighter skins?

      Delete
  2. I don't think it defensible, and it doesn't seem to have affected subsequent policy. For example, Colvin-Adams et al (2012) review US heart and lung transplantations allocation procedures and say:

    The revised allocation policies [of 2005] removed the emphasis on waiting time and replaced it with a combination of geographic priority and the LAS, a calculation of illness severity and projected posttransplant survival that was intended to place the sickest candidates with the best chance of survival at the top of the waiting list. This was the first time “utility” of the transplant was included as part of an organ allocation policy.

    ReplyDelete
    Replies
    1. Hmm, that's interesting, as the Council here does emphasize the moral significance of "utility" when there are large differences between patients, which seems consistent with the actual changes you describe as being implemented in 2005. Do you know if it is now standard practice to also attempt to make fine-grained discriminations of the sort I'm defending here?

      Delete
    2. The illness severity/predicted post-transplant survival calculation is
      fine grained, as I understand it. You might wade through some of the material at http://optn.transplant.hrsa.gov to see what they currently do, and what they are proposing eg for renal transplant:

      http://optn.transplant.hrsa.gov/SharedContentDocuments/KidneyConceptDocument.PDF

      Delete

Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.