Why it's important to wait before making up your mind ...

Parent Cafe
1

You’ve probably seen or heard about the study that says white Americans aged 45-54 are dying faster, along with the graph that shows a spike upwards. You may also have heard that the increase is directly related to drugs, alcohol, etc.

I had the study open on my desktop for a few days but put off reading past the abstract because the result seemed a bit far-fetched: that would be a lot of extra deaths. Turns out caution may have been the best response, and my point is that it often is.

The paper may not have done the math right: they didn’t account for the aging of the 45-54 age group, which seems to have increased from 49.1 years to 49.7 years in the period studied. As that group ages, more will die each year and this works out to about an 8% increase in the probability of dying for that whole year for this group. So you take the .6 increase and multiply that by the expected increase in mortality of 8%, and you get 4.8%. That eliminates most of the difference. Note this isn’t “statistics” in the sense of difficult calculation but really a logic error, meaning they didn’t think to do this adjustment. You can see how they only needed to check a few sources to figure this out [this post](Correcting statistical biases in “Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century”: We need to adjust for the increase in average age of people in the 45-54 category | Statistical Modeling, Causal Inference, and Social Science) by statistics professor Andrew Gelman. As Andrew points out, you could say they found for this group a leveling off of life expectancy but not a decrease and even that conclusion would need more thought.

Note you can see why stats people glommed on to this: like I noted, the effect is huge given the number of people involved and that’s a lot of alcohol death! Most claimed big effects disappear when you figure out the problem better.

On a related note, the Economist ran a piece about how earnings differences between men and women largely - but not entirely - disappear when you compare actual jobs versus broad categories and even broader groupings of jobs. There may be more inequity in the types of jobs women have than in whether they’re paid fairly versus men for the jobs each actually have. This is the same issue we see on the other side of the political spectrum: the claim that government workers make more also falls apart when you look at actual jobs instead of broad groups and actual skill levels, etc. Big effects tend to disappear when you look closer. I think this is a very useful thing to remember: whenever someone claims there’s a big difference, the odds are pretty good that person is a) simply wrong if you look more closer and/or b) manipulating you.

2

The linked post does mention that, even after the age adjustment showing that the US white death rate went flat (instead of rising by about 5%), there is still a big difference between that and the large (around 30%) declines in other rich country and US Hispanic death rates.

3

An update by Andrew Gelman: [url=<a href=“http://andrewgelman.com/2015/11/06/what-happened-to-mortality-among-45-54-year-old-white-non-hispanic-men-it-declined-from-1989-to-1999-increased-from-1999-to-2005-and-held-steady-after-that/%5Dhere%5B/url”>http://andrewgelman.com/2015/11/06/what-happened-to-mortality-among-45-54-year-old-white-non-hispanic-men-it-declined-from-1989-to-1999-increased-from-1999-to-2005-and-held-steady-after-that/]here[/url]. Gist is that mortality increased from 1999 up until 2005, though the data from before 1999 wasn’t examined, and has been generally level since.

4

If you read Gelman’s entire discussion, all three posts, you see he emphasizes that the general conclusions of the paper still remain: people in other other countries, and Hispanic people and non-white people in the US, are seeing drops in mortality rates. But white people in the US are not seeing a drop in mortality. There’s a big difference in the trend, even after Gelman’s correction. Also, the death rates from the “despair” causes (suicide, liver diseases, poisoning from drugs and alcohol) are, as Deaton and Case reported, increasing dramatically in middle-aged white people.

5

The scary part of that report, that is not statistical shifting or noise, is the rise in deaths from alcoholism, suicide and drugs. What also is not statistical noise is that much of this rise is among those at the lower end of the educational and economic spectrum. It mirrors something else,the epidemic of drug use, things like heroin and meth, in rural areas and some of the resulting side effects (like HIV infection rates). Not surprising to me, if you look at economic dislocation in this country, it is this very same group that has gotten pounded, and economic misery is directly related to these ills. It will be interesting to see what the response to this is, will it be looking at root causes or an expansion on the ‘war on drugs’? Will it be some real action, or ‘Just say No to drugs’ kind of thing?

6

I was mostly attracted to the extraordinary increase reported, not to the relative changes versus the world. Those are hard to analyze.

So for example, this is not the group which has suffered most in the Great Recession; that would be minorities. And one can track such things as incarcerations and arrests along with that.

I’m also unsure about the quality of data from other places.

My guess is some portion of this would, if it’s real, be attributable to the generally poor US health system. Our system is great if you have money, fairly lousy if you don’t. My “prior” on this is influenced by the recent maps of the uninsured, which show that every state which hasn’t expanded Medicaid has lots of uninsured. That says to me many people in the states which have expanded Medicaid only have insurance because of that expansion. This situation is pretty unique in the Western world; essentially all other countries have universal or near universal coverage. But I’m not familiar with research on the mortality effects of health insurance.

7

Update: [here[/url]. The gist remains, despite [url=<a href=“http://www.nytimes.com/2015/11/07/upshot/more-details-on-rising-mortality-among-middle-aged-whites.html?mabReward=CTM%5Dthis”>http://www.nytimes.com/2015/11/07/upshot/more-details-on-rising-mortality-among-middle-aged-whites.html?mabReward=CTM]this column](Age adjustment mortality update | Statistical Modeling, Causal Inference, and Social Science) in the NYT, that Gelman is correct: there has been no increase since 2005 but it appears other nations have had a decrease in death rates among this group. (In other words, the increase from 1999-2013 was all up to 2005.)

I have to note that Paul Krugman has picked up this study and used it in his column - as well as his blog - without any reference to the statistical discussion of the validity of the results. I think that’s something you could definitely fault him on.

8

A further update: Andrew downloaded the data used in the paper and finds that, contrary to what commentators are saying in the national press - see my note above - the actual increase is among women, not men. This completely blows away the storyline about white middle-aged men in the 2000’s. And the increase is in women younger than 52. [url=<a href=“http://andrewgelman.com/2015/11/10/death-rates-have-been-increasing-for-middle-aged-white-women-decreasing-for-men/%5DHere%5B/url”>http://andrewgelman.com/2015/11/10/death-rates-have-been-increasing-for-middle-aged-white-women-decreasing-for-men/]Here[/url]. Note that this doesn’t get into causes at all and isn’t a multi-level model that relates one issue to the next but rather a look at the gross numbers of deaths.

9

The biggest effect seems to have occurred in the early 2000s. I wonder if 9/11 had any effect on the suicide rates in this country. I personally have two friends who committed suicide in 2002 and 2003 and their decline into the depths of depression seemed to be influenced in at least a small way by 9/11. I’m not saying that it caused the depression but I believe in each case it contributed to their descent into total despair. (Neither had family members who were victims of 9/11.)

10

And apparently, the actual experts in this field not only agree with Andrew Gelman but specifically say it’s too bad the original study didn’t separate men and women because the issue is women of child-bearing age. [url=<a href=“http://healthaffairs.org/blog/2015/11/10/to-understand-climbing-death-rates-among-whites-look-to-women-of-childbearing-age/%5DHere%5B/url”>http://healthaffairs.org/blog/2015/11/10/to-understand-climbing-death-rates-among-whites-look-to-women-of-childbearing-age/]Here[/url].

As in: "Another study by researchers at the University of Wisconsin–Madison shows that in the decade between 1992-96 and 2002-06, female mortality rates increased in 42.8 percent of US counties. Only 3.4 percent of counties, by comparison, saw an increase in male mortality rates.

Furthermore, our own analysis of the same data used by Case and Deaton shows that the average increase in age-specific mortality rates for whites age 45-54 was more than three times higher for women than men. More specifically, between 1999 and 2013, age-specific mortality rates for US white women age 45-54 increased by 26.8 deaths per 100,000 population, while the corresponding increase for men was 7.7 deaths.

By lumping women and men together, the study also missed the important point that the increases in mortality are affecting women of reproductive and childrearing ages, a finding that has huge implications for children, families, and communities.

We blogged about this nearly two years ago in January 2014 and recently updated the original NAS cross-national comparisons. Here is our updated figure showing the probability of women surviving to age 50 in the United States and in other high-income countries, including western European countries, Canada, Australia, and Japan. There is simply no mistaking the reality that American women are currently dying much earlier than their counterparts in other advanced nations." [Note: by Laudan Aron, Lisa Dubay, Elaine Waxman, and Steven Martin.]

To focus on the reporting first, it’s pretty bad considering these studies date from 2012, 2014, etc. and thus should have been picked up by any reporter who can use Google. And the commentating on the Case/Deaton study has been significantly wrong, led by Krugman on the left and Ross Douhat on the right - both NYT, btw.

In terms of causation, these experts think more about health insurance and lack of care which affects poorer women, particularly as obesity has become a bigger issue. Accidental poisonings and suicides together still leave half the number to explain. Poisonings refers to both prescription drugs and illegal ones, which also gets at issues such as the increase in meth use and the abuse of prescription drugs as a substitute for better medical care.

But I think what really bothers me is that this work has been taken to mean “something’s wrong with white middle-aged American men” when it’s actually about women. And that’s typical, isn’t it? This is really a story about worsening health of American women.