Observations Yankee Doodling

An unexpected headline: more US white people are dying

BMJ 2015; 351 doi: https://doi.org/10.1136/bmj.h6180 (Published 17 November 2015) Cite this as: BMJ 2015;351:h6180
  1. Douglas Kamerow, senior scholar, Robert Graham Center for policy studies in primary care, professor of family medicine at Georgetown University, and associate editor, The BMJ
  1. dkamerow{at}aafp.org

Middle aged white people’s death rates rise, while those of black and Hispanic people fall, new study finds

As recently as late October it seemed that all was well in public health in the United States, at least as far as mortality rates go. JAMA had just published a review of mortality data based on death certificates,1 showing that overall US mortality rates decreased 43% from 1969 to 2013. Most of the leading causes of death in that analysis had declining rates: heart disease, cancer, stroke, unintentional injuries, and diabetes.

So it came as quite a jolt to pick up the newspapers on 3 November and see front page headlines such as “Death rates rise among whites in middle age.”2 3 It turns out that there may be an unexpected exception to the blissfully falling mortality rates: white men and women aged 45-54.

Until now analyses always found that it was racial and ethnic minorities that got less healthcare and died sooner. This new analysis shocked the public health world, and perhaps the public as well, with its startling finding: mortality rates had actually been increasing among middle aged white people since 1999, while among other groups—yes, black and Hispanic people—they had been declining.4

Criticisms of the paper

Some criticisms of the paper have already surfaced, and it has been reported that it had been rejected by two leading medical journals before being accepted by the Proceedings of the National Academy of Sciences.5 It is true that the paper’s methods are only briefly described and that its signal graphic, reproduced in the newspapers, is a bit misleading. It compares death rates over time among middle aged US white and Hispanic people with death rates in other developed countries, to dramatic effect. Only the US white rate is rising, starting in the late 1990s, while all the others decrease. But it is a bit of an apples and oranges comparison, as none of the other countries’ rates are broken down by ethnicity. Perhaps Canadian middle aged white death rates are increasing too? We can’t tell.

Statisticians have argued that adjustment for age accounts for some of the dramatic difference in the affected age group, but the authors have shown that this is likely a minor effect and that middle aged white Americans are still worse off than others.6 Finally, the paper glosses over the fact that overall (age adjusted) death rates among US black people continue to be higher than among white people. It is only the changes in mortality rates that are going in opposite directions for the two groups.

But if the findings are largely accurate, and even if they are similar to previously unreported trends in other countries, they tell a shocking story. At some time around 1999 a decades long decline in death rates of about 2% a year reversed only among US white people aged 45-54. Other ethnic groups and ages continued to experience the 2% annual decline, while middle aged white mortality rates actually rose 0.5% a year from 1999 to 2013.

The findings applied equally to men and women in this group. Although middle aged people with all levels of education had increased death rates, the bulk of the increase was seen in those with only a high school education or less. The authors calculated that if the middle aged white mortality rate had held at its 1998 level, almost 100 000 deaths could have been avoided from 1999 to 2013.4

The two big questions that immediately emerge are why the increase and what can be done about it. The authors tried to answer the first question. Firstly, they showed that the increased mortality was largely accounted for by increasing death rates in only a few categories: drug and alcohol poisoning, suicide, and chronic liver diseases and cirrhosis. Though deaths in white people of all ages increased in these categories, only in the 45-54 age group was the increase not erased by larger declines in other categories.

Then, using completely different data from a number of US national surveys, they showed that self reported distress in this population was also a problem, citing declines in health, mental health, and ability to conduct activities of daily living and increases in chronic pain and inability to work.

Speculative hypothesis

Combining these findings, the authors wove a story of a generation that is in pain, misusing drugs and alcohol, and dying prematurely. It is a speculative but interesting hypothesis that awaits proof that is more definitive than simple associations. Still, if the finding is accurate it creates a concern about the future of this mini-generation. If they don’t all kill themselves first, they will be consuming lots of Medicare dollars when they get into their mid-60s and are depressed, in pain, and continuing to misuse drugs and alcohol (makes me feel grateful I’m too old for this group).

What we should do about all this is, of course, an entirely different question. Assuming that the findings hold up, I wonder what the next steps should be in clinical settings. Focus screening more on the middle aged? Ask clinicians not to formally screen but to “remain aware” of the potential for these problems? I’m not sure that either of these is practical or would achieve much.

What about changes in policy? The findings underscore the seriousness of drug and alcohol misuse in populations not generally thought to be at risk. Prescription drug misuse and the increasing availability and purity of heroin need to be tackled much more aggressively, but what does that mean? New laws? Better enforcement of existing laws? Better treatment?

Here’s hoping that the ruckus raised by this study leads to further confirmatory analyses and thoughtful responses.

Notes

Cite this as: BMJ 2015;351:h6180

Footnotes

References

View Abstract