clock menu more-arrow no yes mobile

Filed under:

A big study found spiking death rates for middle-aged white Americans. Critics say it's more complicated.

This week, a couple of researchers posted follow-up on a recent blockbuster study that documented a rise in the mortality rate among middle-aged white Americans.

These observers have been pointing out that the study's authors, Anne Case and Angus Deaton, did not adjust for age or sex in their work, and so they missed critical nuance about what's really going on with the rising death rates. Most interestingly, hidden in their data is a story about women of all races, who appear to be dying at faster rates than ever before. Here's a rundown of the controversy.

What Case and Deaton found

All-cause mortality, ages 45 to 54 for US White non-Hispanics (USW), US Hispanics (USH), and six comparison countries: France (FRA), Germany (GER), the United Kingdom (UK), Canada (CAN), Australia (AUS), and Sweden (SWE).
All-cause mortality, ages 45 to 54 for US white non-Hispanics (USW), US Hispanics (USH), and six comparison countries: France (FRA), Germany (GER), the United Kingdom (UK), Canada (CAN), Australia (AUS), and Sweden (SWE).
PNAS

Mortality rates in the United States have been declining for roughly a century. But, in their recent paper, Case and Deaton found that one group is now seeing a stark uptick: middle-aged white people.

As you can see in their chart above, the all-cause mortality rate for white men and women ages 45 to 54 (the red line in the graph) started climbing about half a percentage point per year in 1999 and has continued to rise ever since.

The turnaround is unique to this group: Other age, racial, and ethnic groups in the US have continued to see declines in their mortality rates. And middle-aged white people in other high-income countries have seen similar declines, too.

Case and Deaton couldn't say, exactly, what explains the turnaround, but they suggested that it might be linked to "economic insecurity," as well as to other problems that have recently been on the rise among this demographic group, including drug and alcohol poisonings (particularly from opioid painkillers), suicide, and chronic liver disease and cirrhosis.

Critics say that Case and Deaton didn't adjust their data properly

Not everyone is convinced by this analysis, though. Over the past week, other researchers have written pieces online arguing that Case and Deaton didn't account for sex and age differences in their analysis. Once you do that, these critics say, you get a far more nuanced story.

1) When you adjust for sex, it seems the real story is about women

Columbia statistician Andrew Gelman, who published a rebuttal to the paper on his blog and at Slate, says that the real story here isn't about all middle-aged white Americans. It's mainly about women.

By email, he explained to me that when you break up Case and Deaton's data by sex, you find that men saw a slight uptick in mortality in 1999-2005, followed by a decrease from 2005-2013.

But, he said, "you find a steady increase in mortality from 1999 to 2013 among non-Hispanic 45-54-year-old women." This is the group where we're really seeing a mysterious uptick in deaths, he added.

Andrewgelman.com

It's not just Gelman. In a separate blog post, researchers at the Urban Institute came to a similar conclusion: "By not looking at men and women separately, Case and Deaton failed to see that rising mortality is especially pronounced among women."

These researchers cited past studies documenting a similar rise in deaths among women. See, for instance, work by the National Academy of Sciences and the Institute of Medicine, which shows that "Americans are slipping behind other high-income countries when it comes to mortality and survival, and that this 'US health disadvantage' has been growing particularly among women." And women of all ages seem to be affected.

Health Affairs

So what's driving the uptick in deaths among women?

The Urban Institute researchers point out (as Deaton and Case did) that accidental poisonings — linked to prescription painkillers and heroin use — are a problem. But they say there's more to the story, too:

Accidental poisonings increased more than all the other causes combined, but they still explain only half of the total increase in deaths among white women at these ages.

In addition to suicide, obesity- and smoking-related diseases are driving these mortality increases. Our findings clearly point to the need for a stronger public health focus on the misuse of prescription opioid drugs, as well as more prevention and treatment of tobacco, alcohol, and other drug use; mental illness; and obesity-related illnesses.

2) When you adjust for age, subtle differences appear

The researchers also point out that Case and Deaton did not properly adjust for age. The study authors looked at mortality rates for the 45-54 group over more than 10 years — a group that hit their senior years during the study period and who will inevitably die at faster rates into their 60s and beyond.

There's a similar trend happening in the 35-44 set, Gelman said, noting that this rise in mortality is happening among people of other ages. He also said that if the economists had adjusted for the change in age over the 10 years, the mortality increase would have looked less dramatic.

"Case and Deaton did not age adjust," Gelman said. "This introduces a minor bias of about 5 percent in the change in middle-aged mortality rate, but fixing this minor bias is enough to change their US story about middle-aged whites from steady-rise-in-mortality to rise-from-1999-to-2005-followed-by-stasis-thereafter."

How Deaton and Case respond to their critics

I reached out to Deaton, who recently won the Nobel Prize in Economics, to see how he responds to the critics.

"Nothing [these critics] say contradicts what is in our paper," Deaton said, noting that mortality rates generally are different from mortality rates adjusted to account for various factors.

Still, Deaton added, the trend among women is real: "One thing that is happening here, which we know, is that women’s death rates have not done so well recently because they took up smoking later than men and gave it up later than men, so that their death rates from lung cancer have been rising."

He also noted that there may be something to the age-related issues: "It is true that at least some of the pattern of death rates can be explained by the population aged 45-54 getting older within the group. Exactly how much awaits further work."

But as Gelman noted, the major trend in the paper — that there's an international discrepancy in mortality trends among middle-age white people, as well as a discrepancy compared with other ethnic groups — remains true even after accounting for age and sex. "Case and Deaton's big story remains unchanged, and beyond this they get credit for bringing up the topic in the first place." 

The criticisms that have emerged do, however, suggest there's more work needed to unpack this trend.