17

An article titled The Myth of Americans' Poor Life Expectancy claims that the US leads the OECD in life expectancy once the numbers are adjusted to remove deaths due to fatal injuries.

Unfortunately, the link to the research by Ohsfeldt and Schneider is broken, and I haven't been able to find the PDF anywhere else. And I can't find any other primary sources for this claim.

Brian
  • 369
  • 1
  • 7
  • 2
    What they could have done would be to use the internationally accepted numbers for mortality amenable to healthcare (i.e. based on causes of death where good healthcare could, in principle, have affected the outcome). They made up something flattering to the US instead, lowering the credibility of the comparison. – matt_black Jul 20 '17 at 08:36
  • 4
    "You'd all live longer if we had no guns or cars". – Jamiec Jul 20 '17 at 08:47
  • 1
    @jamiec and probably even longer if you put tobacco and alcohol in that mix. – Mindwin Remember Monica Jul 20 '17 at 17:32
  • 2
    There are also major geographic factors involved, See e.g. this http://www.nbcnews.com/health/health-news/who-lives-longer-study-finds-colorado-wins-20-years-n756471 for a map of life expectancy by county. Of course some self-selection is involved. The highest life expectancy is in mountain areas. but people with healthy lifestyles tend to move to such areas. – jamesqf Jul 21 '17 at 05:22
  • @Jamiec: And no burgers. – Martin Schröder Jul 22 '17 at 11:22

1 Answers1

19

It is highly dubious.

This data comes from a 2006 book published by a conservative think tank, the American Enterprise Institute. The report is titled The Business of Health: The Role of Competition, Markets, and Regulation.

Sure enough, they ranked USA the highest after "standardizing" the (old) data:

Table 1-5 showing the USA at the top

One clue that this shouldn't be trusted is that the "standardized life expectancy" - described as removing deaths from fatal injuries from the life expectancy tables - actually reduces life expectancy for many of the countries at the top of the raw table.

The report shows the formula actually used:

Equation 1-1

Not only does it fail to remove actual deaths from unintentional or intentional injuries for each country (instead using some coefficients from a regression of death rates on deaths from injuries from across the OECD - why do this? - the update to the article misdescribes this with "the adjustment factor was based on fatal injury rates relative to the average"), it pushes up the "standardized life expectancy" for countries, notably the USA, with high GDP per capita.

To me, this seems to suggest that the USA should be doing better than other countries on raw life-expectancy given its high GDP per capita, but is failing to do so.

Oddthinking
  • 140,378
  • 46
  • 548
  • 638
Henry
  • 13,472
  • 1
  • 54
  • 62
  • 4
    I made a fairly sizeable edit because I think you nailed the source, but didn't provide enough context for people (like me) not familiar with the background. Please check I didn't over-step. – Oddthinking Jul 20 '17 at 01:53
  • 1
    It may be worth noting that they don't justify the inclusion or exclusion of any of the variables (beyond the implication that the deaths are not the hospitals' faults, without acknowledging or accounting for deaths from those causes that only occur after the victim makes it to the hospital) – GDPPC is particularly suspect; and they provide no insight into *how* they arrived at the coefficients (or why they use logs). And there is the "year-effects" fudge-factor that they explicitly did not release. – Kevin Jul 20 '17 at 16:39
  • @Kevin This looks like the result of a mixed effects regression with year as a random effect. It seems appropriate to consider year a random effect and one can't give one coefficient for this. One can give the variance of this effect or a coefficient for each year in the dataset. I can't judge if using this regression is appropriate or not but if they use it they should provide the uncertainty for each value in the table. Possibly some of these values shouldn't even be considered different. – Roland Jul 22 '17 at 21:00