A Little Something to Fight the Poison

Saturday, November 23, 2002
COMPARATIVE ADVANTAGE? WHAT'S THAT? Just thought I'd note that it's funny that the last couple of posts have involved Manu commenting political science and myself commenting on health care, with both of us prefacing our posts with "well, I don't know much about this subject, but..."

I guess that means we both have promising careers as public intellectuals ahead of us!

I'll get to the Median Voter Theorem sometime soon, though, I promise...
IT'S HARD FOR ME to accurately express the extreme irritation I felt deep within the cockles of my heart when I read this gawdawful David Frum National Review column attacking Al Gore's proposal to institute a Canadian-style single-payer health care system. There is already so much ignorance on this side of the border regarding other health care systems; the last thing anyone needs is for people to take seriously the opinions of a hack like Frum on the issue.

Frum argues that the main virtue of the Canadian health care system is that it is "hassle free"--nevermind that it costs half as much per capita than the U.S. system does; or that no one goes without health care coverage, as opposed to the 16% of Americans who aren't covered; or that virtually no Canadians went bankrupt because of medical bills last year, whereas 500,000 Americans did; or that HMOs can't pressure doctors into determining which procedures are necessary or not--but that this convenience isn't worth the long waiting times that occur in the Canadian system.

I'm not an expert on either statistics or health care. Fortunately, neither is Frum, which is why I'm going proceed to rip into him for his use of the Fraser Institute's report on "waiting times" to prove this point.

If one saw a report subtitled "Hospital Waiting Lists in Canada," one might expect that the report's data is based on either an aggregation or random selection of data compiled from the waiting lists of Canadian hospitals. One might also assume that these waiting times cover all procedures or a representative sample of procedures.

Well, one might have this expectation only if one was unfamiliar with the methodological standards of right-wing think tanks.

Switching to the more ornery second-person, what you instead find when you read the report is survey data based on the responses of individual specialists to mail-in surveys, who have been asked about their patients' waiting times for a selection of procedures.

Can you say "self-selection bias"? Clearly, the Fraser Institute can't.

The response rate to these mail-in surveys was 30%, which the report notes is "good quality for a mail-in survey." Great. And if I told you that a 30% survival rate is "good quality for open-heart surgery done with a screwdriver," that wouldn't give you any more confidence in that procedure's quality.

The report is nice and readable, though, because it doesn't bother to clutter up its text or its charts with those tough-to-read statistical tests for the median waiting times that it presents, or the standard deviations that would be required to calculate them. Statistical significance? Who needs it?!? So even if we accepted that the survey data was unbiased, we wouldn't be able to figure out whether any of the results mean a damn thing.

I am, however, going to go out on a limb and guess that neither the 3 British Columbia radiation oncologists (out of 45 possible respondants) nor the 1 Newfoundand otolaryngologist (out of 7 possible) constitute a sufficient sample from which one can determine a statistically representative median. And yet the "median" waiting times drawn from these samples are presented alongside all of the other data without an asterisk or qualifying footnote in sight. It requires, I have to say, a certain amount of pure chutzpah to use the word "median" to describe a number drawn from a sample of 1.

And then there's the selection of procedures. Based on the report, Frum says that mean waiting time for tonsillectomies in Saskatchewan is 80 weeks, which might lead someone to believe that there's a bunch of poor kids wandering Regina with horribly distended throats.

That is, you might believe that, if you didn't realize that the Fraser Institute's survey of surgical procedures only measures the waiting times for elective surgeries (except for cardiology), which are ranked on the triage list below "urgent" and "emergent" surgeries. As a revealing but well-buried passage in the report explains, there's a pretty good reason why the Fraser Institute's waiting times look a lot worse than the numbers issued by provincial govenrments:
Saskatchewan Health measures waiting times for all non-emergent surgeries while Waiting Your Turn measures waiting times for all elective surgeries. This means that urgent surgery wait times (which are significantly shorter than elective wait times [my emphasis]) are included in the medians and averages measured by Saskatchewan Health but not in those measured here.
This omission clearly limits the data's usefulness quite a bit--how useful is a statistic on median waiting times for measuring the overall state of a health care system when you exclude all of those cases for which waiting times would be most crucial? The Fraser Institute of course covers itself by placing a bunch of caveats on how these numbers should be interpreted--caveats that the Institute is completely aware hacks like Frum will ignore when using the numbers as ornaments for the ideological arguments they present to the general public.

That's because for a guy like Frum, statistics are wonderful playthings, but not worthy of real care or respect. If Frum actually did respect the integrity of facts and numbers, he'd at least bother to present us with comparative U.S. waiting times for the procedures that he serves up. Instead, the only comparative statistics Frum uses in his article are the survival rates of a couple of arbitrarily chosen diseases of a handful arbitrarily chosen subregions of each country. He doesn't mention, of course, that life expectency is several years higher in Canada than in the U.S. Nor does he mention that report after report shows that aggregate health care outcomes in Canada are the same or better than American outcomes, even though the U.S. spends twice as much money on health care. See, for example, this compilation of studies by the New England Journal of Medicine, which calls the U.S. health care system "at once the most expensive and the most inadequate system in the developed world."

Arg! OK, that's enough. I could go on for pages. I wonder how long I'll have to wait to have these inflamed cockles examined...
SORRY ABOUT THE SILENCE: Lots of work, some disorganization, and a bit of time thinking about the utility of this blogging thing. Oh, and those usual, vague grad school blues.

I'm also trying to figure out the purpose and focus of my doing this (independent of Manu, since I think I've given him carte blanche, except for that restriction on porn). No firm decisions yet, except that I'll try to keep posting regularly for the near future. Suggestions and criticisms (as long as they're not pure snark) welcome.
Friday, November 22, 2002
Hmmmm...its been a week and a half since the last post. Can't say that the break from blogging was "refreshing" per se (given that my time was spent in the lab), but I think I have the itch again.

Jacob T. Levy has a series of interesting posts -- here, here, and here -- on the Median Voter Theorem, which posits that in a two-party system, the postions of these two parties will generally cluster around the positions and inclinations of a so-called Median Voter. Without knowing anything specific about this theorem, I would guess that it is based on the premise that each party's ideological base (left or right) -- has nowhere else to go, and will (grudgingly) support their centrified "home" party.

(Of course, over the last ten years, ideological splinter groups led by Perot, Buchanan and Nader have had dramatic effects on otherwise evenly-matched presidential elections. But this phenomenon doesn't seem to have affected Senate or House races...)

Levy argues that while the Median Voter Theorem has some basis, voters are unlikely to be highly sensitive to changes in party platform. Besides, many people seem to routinely vote for a specific party, election after election, that they are comfortable with. He suggests that in a two party system, there will be relatively lengthy cycles of single-party dominance...until the dominant party seems to veer far enough from the political median to be smacked down. TNR seems to back this hypothesis in part -- though they assert that the frequency of political tectonic shifts will be much greater.

Interesting stuff...and I must say that Levy's blog is highly recommended. Scholarly, yet engaging and highly readable. Give it a visit...