• This stuff about life expectancy is old news

    I’m getting a lot of tweets and emails concerning this post by Avik Roy on life expectancy. I already addressed those points. In 2010. Here’s that post:

    Life expectancy as a metric is so controversial that I felt it deserved further exploration.  People don’t like it because they think lots of things are getting in the way besides the health care system.  Some of the most commonly brought up examples are how we score miscarriages, or the fact that we have more homicides, car accidents, or suicides.

    Yes, I know Betsey McCaughey made this argument on the Daily Show. She got it originally from an analysis done by Robert L. Ohsfeldt and John E. Schneider (published by the American Enterprise Institute ) for their book, The Business of Health: The Role of Competition, Markets, and Regulation:

    The abnormally high child mortality rate obviously contributes to the abnormally low life expectancy at birth in the United States. But death rates among adolescents and youth can also have a dramatic impact on estimated life expectancy. In that light, it is important to note that some specific cultural aspects of American society largely outside the purview of the health care system contribute to rates of death from injury, both unintentional (accidents) and intentional (homicide and suicide). Rates of death from injury are usually high in the United States compared to other developed countries, which affects the apparent underperformance of the U.S. health system (as measured by life expectancy at birth), because deaths from injury disproportionately affect adolescents and young adults.

    Those are fair points.  But do they alter things slightly or change how we perform greatly?

    First let’s look at homicides:

    Yes, we have more, but the number of actual deaths pales in comparison to total deaths per year (almost 700 per 100,000 population).

    We also have more deaths by accidents, but again, not in huge numbers.

    Our suicide rate is actually such that it should help us.  But again, not a huge factor.

    So let’s think this through.  What Ohsfeldt and Schneider did was calculate a regression to remove the effects from infant mortality, accidents, assault, etc. from the equation and recalculate the life expectancy (making us first).  But there are problems with this:

    Carl Haub, a demographer at the Population Reference Bureau in Washington, D.C., said the method was incomplete. A more-precise analysis would have removed those who died from these causes from overall mortality stats, and then recalculated life expectancy. (For more on how life expectancy is calculated, see this earlier blog post.) “Just because another method is a lot of work, does not mean regression will yield a correct result,” Mr. Haub told me.

    Prof. Ohsfeldt acknowledges that regression was chosen for its relative simplicity for what he called his “little book project.” And he agrees that some deaths that his book attempted to remove from the life-expectancy tables might be dependent on health-care systems.

    If it’s the case that the miscarriages, accidents, and homicides are screwing us, because we lose many more people at a young age, then if we look at the life expectancy of people who make it out of the dangerous young years, things should look better.  In fact, in their book, Ohsfeldt and Schneider make this exact argument:

    In contrast to life expectancy at birth, cross-national comparisons of healthy life expectancy at age sixty are relatively unaffected by differential death rates from unintentional injury and homicide.

    So let’s look at the life expectancy for 65 year olds.  According to the people who did the analysis for AEI, that’s the fairest way to assess life expectancy as a health care system population metric.

    Still the worst in recent years.

    Accept it.  Even if we look at life expectancy for sub-populations relatively less affected by the reasons people use to try and discredit the metric as a quality measure, we still look pretty bad.  Especially when you remember that people in the US over age 65 have much better access to the health care system overall, due to the universality of Medicare.  Moreover, even if you argue that the US is more dangerous, then don’t we really need a much better health system to keep us from dying?

    You can’t blame all deaths on the health system, but that doesn’t mean we couldn’t use a much better one.


    • I think a key point is that there are numerous variables that go into life expectancy, which makes it relatively unhelpful as a measure of the quality of healthcare in America. Throw in income, ethnicity and culture (try comparing the life expectancy of Asian women in the U.S. with Native American men sometime), and it’s easy to see that there are numerous factors at work.

      And I’d also refer you to the following observation:

      “It should be noted that the subject is the medical-care industry, not health. The causal factors in health are many, and the provision of medical care is only one. Particularly at low levels of income, other commodities such as nutrition, shelter, clothing, and sanitation may be much more significant. It is the complex of services that center about the physician, private and group practice, hospitals, and public health, which I propose to discuss.”

      That would be the second paragraph of Kenneth Arrow’s “Uncertainty and the welfare economics of medical care.”

    • I was shocked at how much worse our Homicide rates are! American Exceptionalism indeed! But also surprised at the conclusion of our post.

      Once Americans reach 65, we move from the barbaric private health insurance system into a socialized single-payer medical system (Medicare) which is not as exceptionally different from those of the peer nations you plot here. There are no longer the problems with lack of access that plague us younger folks. Our system may be uniquely expensive even for Medicare users, but the Government bears that cost, not the users, so nothing prevents them from taking advantage of everything modern medicine has to offer. Why are our outcomes so poor?

      Is it residual effects of poor health care before 65?

      Is it some difference in how we treat Medicare patients?

      I’d love to see a follow-up post addressing this.


      Also, there are several statistics where we ought to be doing much better than out peer countries. Smoking rates are much lower here than in Europe or Japan. I was trying to make a graph to show this, and playing with Gapminder, I found that even our Lung Cancer death rates are at the worse edge of our peer groups.

    • Please allow me to repeat a request regarding color-coded charts. Would it be possible to link to tables with numerical displays of data when these charts are used? Not everyone can tell colors apart. Thank you.

      • I was just about to make the same request.

        And it doesn’t help that KDE has decided to make all their color themes a WinDoze Vista/7 lookalike

    • While I am firmly in favor of improving our health care system — I am convinced that it under-performs in key respects and is far too costly — I am skeptical that doing so will make much difference to population-level health outcomes. My impression is that evidence is currently overwhelming that the social determinants of health simply swamp anything that a health care system, narrowly construed, can do. So I would suggest that the real failures of our health care system make no more than a trivial difference to our health outcomes, and fixing those real failures, without addressing the real drivers of our terrible outcomes, won’t make much of a difference.

      Does this mean that we shouldn’t strive for universal access to a reasonable level of health care? No, of course not. Does it mean that we shouldn’t be looking at ways to avoid over- and mis-utilizations of health care resources. No, obviously not. But it does mean that until, as a society, we address those social factors that are associated with bad health outcomes, we shouldn’t expect much from our efforts in terms of improved population-level health. (The failure to improve population-level health outcomes enough to “matter” won’t matter, of course, to those people who are saved from early, unnecessary deaths by access to health care, or who don’t end up in poverty because they lacked insurance — those are, I think, more than good enough reasons to keep pushing for better access, etc., even if doing so has no real impact on population level health outcomes.)

      Why do our health outcomes suck? Well, let’s see. Crazy levels of income inequality, lack of meaningful social and economic opportunities for too much of our population, entire regions of our country stuck in grinding, unrelenting poverty, a profound under-investment in social services and human capital more generally, etc etc. The list goes on, but it all comes down to the same damned thing, doesn’t it? And until we address that, I fully expect our population health outcomes to stay stuck at the bottom, no matter how much money we throw at fancy new medical treatments.

      • Why do our health outcomes suck? Well, let’s see. Crazy levels of income inequality, lack of meaningful social and economic opportunities for too much of our population, entire regions of our country stuck in grinding, unrelenting poverty, a profound under-investment in social services and human capital more generally, etc etc. The list goes on…

        Unlike Italy, one of the leaders, which has non of these problems. (I am of course being factious).

        BTW no on the list above but check out how long people of “first nations” live in Canada and aboriginals live in Australia.

        • Floccina — not to beat a dead horse, and I frankly don’t care all that much, but if you are going to pick some “dramatic counter example” to the substantial literature and impressive research on the social determinants of health broadly, and the strong correlation between life-expectancy (and health outcomes more generally) and income inequality in particular, can you at least take the time to pick one that is in fact a counter example? Italy isn’t even a statistical outlier for the most trivial straightforward correlation of the Gini coefficient and life-expectancy (nor is the U.S.). (No, I don’t endorse some naive straightforward causal pathway from inequality to health outcomes, but yes, I do think it is part of the story.)

    • The thing obamacare deals with and that everyone seems to care about is the stuff insurance pays for and hospitals do: treating sick people. Most breakdowns show we do pretty well and saying we suck at other obviously more consequential but unrelated aspects of wellness doesn’t change that. This is a terrible response.

      I agree the obsession with drugs and insurance is insane but hey those are things people want. They don’t want schools to teach meditation or nudges to help them exercise more.

    • Is it that hard to estimate our life expectancy-without-excess-homicide/infant-mortailty/car-crashes? I got curious and took a swag at it once, assuming that

      LE_observed = LE_adjusted * 0.9775 +
      25 * 0.006 +
      0 * 0.003 +
      30 * 0.0135

      25 = swag at average age of death from murder
      0.006 = fraction of our deaths due to murder

      0 = age of death from infant mortality (I know the average is larger).
      0.003 = fraction of our deaths due to “excess” infant mortality (vs Canada)

      30 = swag at average age of death from car crash
      0.0135 = fraction of our deaths due to car crash (100% of crashes)

      LE_observed = 78.37, this gives LE_adjusted = 79.6, moving us from #49 in the rankings to #37 (NOT applying a similar massage to the data for any other countries, e.g., Belgium, South Korea, and Portugal).

      But am I oversimplfying? (“Yes” with no explanation is too simple a reply).

      I also heard a claim that we classify “births” differently from other countries, and what they call a “miscarriage” we might call an “infant death” but did not see any data anywhere to support this, only a paper (from a conservative source, naturally) using uncertainty about this to throw sand in the general direction of our abominable infant mortality statistics.

    • @dr2chase:

      One of many.

      Paediatr Perinat Epidemiol. 2002 Jan;16(1):16-22.
      Registration artifacts in international comparisons of infant mortality.
      Kramer MS, Platt RW, Yang H, Haglund B, Cnattingius S, Bergsjo P.
      Department of Pediatrics, McGill University Faculty of Medicine, Montreal, Canada. michael.kramer@mcgill.ca

      Large differences in infant mortality are reported among and within industrialised countries. We hypothesised that these differences are at least partly the result of intercountry differences in registration of infants near the borderline of viability (<750 g birthweight) and/or their classification as stillbirths vs. live births. We used the database of the International Collaborative Effort (ICE) on Perinatal and Infant Mortality to compare infant mortality rates and registration practices in Norway (n = 112484), Sweden (n = 215 908), Israeli Jews (n = 148123), Israeli non-Jews (n = 52 606), US Whites (n = 6 074 222) and US Blacks (n = 1328332). To avoid confounding by strong secular trends in these outcomes, we restricted our analysis to 1987-88, the most recent years for which data are available in the ICE database for all six groups. Compared with Norway (with an infant mortality rate of 8.5 per 1000), the crude relative risks [95% confidence intervals] were 0.75 [0.69,0.81] in Sweden, 0.97 [0.90,1.06] in Israeli Jews, 1.98 [1.81,2.17] in Israeli non-Jews, 0.95 [0.89,1.01] in US Whites and 2.05 [1.95,2.19] in US Blacks. For borderline-viable infants, fetal deaths varied twofold as a proportion of perinatal deaths, with Norway reporting the highest (83.9% for births <500 g and 61.8% for births 500-749 g) and US Blacks the lowest (40.3% and 37.6% respectively) proportions. Reported proportions of live births <500 g varied 50-fold from 0.6 and 0.7 per 10000 in Sweden and Israeli Jews and non-Jews to 9.1 and 33.8 per 10000 in US Whites and Blacks respectively. Reported proportions 500-749 g varied sevenfold from 7.5 per 10000 in Sweden to 16.2 and 55.4 in US Whites and Blacks respectively. After eliminating births <750 g, the relative risks (again with Norway as the reference) of infant mortality changed drastically for US Whites and Blacks: 0.82 [0.76,0.87] and 1.42 [1.33,1.53] respectively. The huge disparities in the ratio of fetal to infant deaths <750 g and in the proportion of live births <750 g among these developed countries probably result from differences in birth and death registration practices. International comparisons and rankings of infant mortality should be interpreted with caution

      • Thanks — that is substantially more credible and authoritative than the data I was able to find.

        I’m still curious about whether my method for adjusting mortality to “remove causes” is 5% crap or 95% crap, or where it would lie in between.

    • Another point that could be made is: health care in Japan and Italy is far better than in Germany and the UK. To me it is easier to believe that Italians live longer than Germans because they are less likely than Germans to have alcohol problems (for whatever reason) than that German healthcare is that much worse than Italian healthcare.

      Otherwise you attempt to mimic Italian healthcare and say why follow German healthcare because though it is better than the USA system Italy is a far better model.

      Further you might say why even look out side of the USA we should do what North Dakota and Utah do!

      So Aaron if you were in Germany or the UK would you be arguing to scrap the German system and do what the Italians do or would you say at least we beat the USA by about as much as Italy beats us?

      Also I have never seen you post an argument that other states should follow ND’s lead.

    • Comments on this here:

    • How on Earth are deaths from accidents, homicide and suicide “outside the purview of the health care system”?