Slippery slopes, statistics and the telephone game

Leah Libresco of Unequally Yoked recently told her readers:

In statistical news, social scientists have not found any evidence of doctor assisted suicide launching us down a slippery slope.  The rates of people electing medically-assisted suicide has stayed flat in Oregon and the Netherlands (and in the Netherlands, there’s about 15 years worth of data).

These data aren’t a rebuttal to anyone who opposes doctor-assisted suicide for all people, but, if I’ve got any readers who were concerned the procedure would pressure the elderly to choose death rather than be a burden, how long would this trend need to last for you to consider it persuasive?

Now I’m one of the people she refers to in the second paragraph and would be opposed to Euthanasia even if there was no slippery slope. But I still think the slippery slope argument is very important. Not only does it add a practical problem, I also think that leading to further evil is a good heuristic for dubious act being evil in the first place.  So while not the only reason, the slippery slope is  one reason we know euthanasia is evil in the first place.

But does that slippery slope actually exist? Leah says no and she has a source to back her up.  Her reference is a post at TYWKIWDBI which in turn quotes an article by Peter Singer^[1], which in turn relies on the report[PDF] of a Royal Society of Canada expert panel on end-of-life decision making. That report in turn relies on various official data sourced of which we will talk in due course. My headline promised you a telephone game and now you know the players, but is the data transmission actually faulty? And if yes, where did the transmission errors actually happen?

Well, the relevant part of what Leah read at TYWKIWDBI is this:

In The Netherlands, voluntary euthanasia accounted for 1.7% of all deaths in 2005 – exactly the same level as in 1990…

In Oregon, where the Death with Dignity Act has been in effect for 13 years, the annual number of physician-assisted deaths has yet to reach 100 per year, and the annual total in Washington is even lower…

[I have deleted their emphasis because I quote only the emphasized part.] “The rates are flat” certainly seems like a good summary of this. But note, that the snippet doesn’t say that explicitly. The Netherlands rate in 2005 was the same as in 1990 but it doesn’t tell us anything about what happened in between. And on Oregon and Washington we get an absolute number not yet surpassed but nothing about what the actual numbers were. As we’ll see in a moment both these silences are telling.  But Leah’s interpretation clearly is the one the text naturally suggests, so I’ll say we have no reason to blame her.

TYWKIWDBI, too, is innocent. The quote is taken directly and faithfully from Peter Singer’s article. The shortening is correctly marked and doesn’t change the meaning. There is no omitted context that would change the message. In short they didn’t add anything and therefore didn’t add anything wrong.

Peter Singer is where it gets interesting. Let’s start with the Netherlands.

If we could rely on the numbers reported through official channels (prosecutors for part of the period and later commissions set up for that purpose) there would be no discussion. These numbers go up, period. Problem is, reporting practices used to be abysmal and have gradually improved. So how much of the increase is in actual euthanasias^[2] and how much is just improved reporting of what always happened? Answering this question is part of why the Dutch government commissioned five anonymous polls of doctors in 1990, 1995, 2001, 2005, and 2010. The results of these polls are extrapolated to the national rates most people rely on. ^[3] The results of the 2010 poll don’t seem to be out yet [Update: They are now, look at my follow-up post], so the Canadian panel had four data points, of which Peter Singer quotes the first and the last, noting they give the same rate.  To end the suspense, the report says

In 1990, 1.7% of all deaths were preceded by voluntary euthanasia, as compared to 2.4% in 1995, 2.6% in 2001 […]. This trend reversed in 2005, when again 1.7% of all deaths were the result of voluntary euthanasia[…]

This is interesting. Yes, the 2005 rate is back to its 1990 value. but that’s a decrease after it had been going up for 10 years.  And yes, that reversal follows the 2001 change from a non-prosecuting regime to official legalization. So slippery slope believers like me need to explain why the rate went down again after the law got worse, and I’ll be doing so infra. But the data  doesn’t quite fit with the other story either. After all, even the most ardent euthanasia supporter wouldn’t argue that legalization propels us up the slippery slope.^[4]  Basically the actual data should send us looking for other factors explaining the change.

So is Peter Singer the message-corrupting player of the telephone game? Well yes and no. No, because the Canadian panel actually argued against fear of the slippery slope and used this data in partial support of its argument. But their argument is more complex then “rate flat, therefore no slippery slope”. At least right now I won’t engage the real argument beyond saying it leaves me unconvinced. I couldn’t do it justice without derailing my train of thought.  But here we have the yes part of my last question. With a standard length column, of which he used one third on a touching lead-in story and another third to trash Christianity, Peter Singer didn’t have space do the real argument justice either. So he simplified it to a news bite of two data points and the conclusion.  Like most news bites it isn’t technically false. And like most news bites it stems from a desire to communicate something similar to a real argument. And as with most news bites the similarity turns out to be pretty remote. So his summary did set readers up for a serious misunderstanding.

But back to the main question: Why did the rate go back down? Here it helps to look at the last step of the telephone game. The Canadian panel takes its data from the report[PDF] of the dutch panel that collected it. The report is, unsurprisingly, in Dutch, but it has an English summary.  And that summary actually has a plausible explanation. I’ll number the reasons for easier reference:

One of the most remarkable findings in the practical investigation is the decrease in euthanasia and assistance in suicide in 2005 compared to 2001 and 1995. […] This decline is linked to a number of other developments. First of all,[1.] the absolute number of deaths in 2005 was less than in 2001, whereas the proportion of persons aged 80 and over (euthanasia and assistance in suicide occur relatively infrequently in this age group) was actually greater. Secondly, [2.] an increase was found in other methods of controlling the symptoms of patients in the terminal phase of their lives, such as continuous deep sedation. In addition, the majority of the physicians thought [3.] there is a clear connection between improvements in palliative care and the decrease in life-terminating action taken by physicians. And finally, [4.] the decrease in the number of cases of euthanasia can probably partly be explained in part by changes in knowledge and opinions on the effects of morphine, which means that physicians are probably less inclined to attribute a life-shortening result to morphine. This has resulted in a decrease in the number of cases of life termination using morphine. We note here that in such cases, it is rather a question of a different appreciation of physicians of their own actions than an actual change in behaviour.

Let’s look at these reasons. #1 is a change in the eligible demographic. If anything it gives cause for the frightening suspicion that euthanasia declined because most of the eligible candidates have already been killed. But it could also be simple demographic change and I haven’t checked the data to see which story is more likely. Either way it isn’t an argument against a slippery slope. #2 has a dark side I’ll talk about later, but for now note that #2&#3 both talk of better alternatives having become available.  Great thing, but not an argument against slippery slopes. On a somewhat bitter note those aren’t newly invented options, just options now more widely available. Perhaps they should have, like,  been made available before, you know, people needing them got killed instead. And finally #4 is, as they say, better knowledge and not a real change.  So all in all none of these reasons contradict the slippery slope story.

Now one might hope these advances have reversed the trend permanently, but a look at the official statistics shows it isn’t so.  They have been rising again, and this time better reporting can’t be blamed alone. 2910 Euthanasias were officially reported in the year 2010, compared to the estimated 2325 for 2005 of which ca. 80% had been reported. Even in the best case, where reporting is now perfect, the euthanasia rate is up by a roughly a quarter  since the 2005 result on which the idea of it having stabilized rested.^[5]

This isn’t all. As I already mentioned, the second reason quoted for the 2005 decrease has a dark side.  Continuous deep sedation, one of the alternatives that has increased, means keeping the patient unconsciously sedated until death. In itself, this is not a moral problem. But at least in the Netherlands it is more often than not combined with the decision to forgo hydrating and feeding the patient as unnecessary medical treatment. And if we know they will die within a day anyway, as about half of them do, that might even be legitimate. But otherwise it is simply a more distanced way of killing. Looking at it from this angle it is very disturbing that 7.1% of Dutchmen dieing in 2005 were under continuous deep sedation, up from 6.0% in 2001.  It’s hard to figure out how much of this is killing and how much legitimate palliative care.  But there is at least a very real possibility that the killings just moved to another name and protocol.

So all in all we have seen that the rise of euthanasia in the Netherlands has at best been temporary interrupted by factors unrelated to legalization.

And in Oregon and Washington? Well both Oregon[pdf] and Washington[pdf] publish official reports of their acts’s implementation. In Oregon both dispensations and deaths were higher than ever before. The Washington regime only took effect in 2009 so the rate increases in the second year don’t prove any trend yet. If the point is supposed to be that the numbers are smaller than in the Netherlands, that would be an apples and oranges comparison. For one, the American regimes are much stricter. But more importantly, assisted suicide (patient takes prescribed poison) isn’t as common as euthanasia (doctor directly kills patient) in the Netherlands either.

So on Oregon and Washington too, Peter Singer cited statistics that are technically true but give a very wrong impression.

To sum up my message, euthanasia rates have a clear upwards trend, we shouldn’t rely on Peter Singer to summarize complex circumstances and it is easy to fall for “facts” produced by telephone games.