Monday, November 9, 2015

Case and Deaton wrong; mortality rates increased for younger non-Hispanic white age groups as well [UPDATED]

Many have probably heard by now about the Case and Deaton publication, which documents an increase in all-cause mortality among non-Hispanic white aged 45-54, which supposedly goes against the trend for other racial groups and other countries, within that age group.

However, this is incorrect. The fact is that all-cause mortality rates have increased for age groups 25-34 and 35-44 as well, for non-Hispanic whites. How do I know this? By checking the same data that Case and Deaton cited in their paper, the CDC WONDER. The CDC WONDER is stated to be the source used to document the changes in white midlife mortality in their publication. Yet in the same publication, they state:

"The focus of this paper is on changes in mortality and morbidity for those aged 45–54. However, as Fig. 4 makes clear, all 5-y age groups between 30–34 and 60–64 have witnessed marked and similar increases in mortality from the sum of drug and alcohol poisoning, suicide, and chronic liver disease and cirrhosis over the period 1999–2013; the midlife group is different only in that the sum of these deaths is large enough that the common growth rate changes the direction of all-cause mortality."

Their publication thus suggests that mortality rates among non-Hispanic whites did not increase for any age group other than the 45-54 age group. However, this is what I found in the CDC WONDER database. Here is the change in mortality rates per 100,000 people for non-Hispanic whites, among the 25-34 age group:

Year
Results are sorted in by-variable order
Move this column one place to the rightDeathsClick to sort by Deaths ascendingClick to sort by Deaths descending
Move this column one place to the rightMove this column one place to the leftPopulationClick to sort by Population ascendingClick to sort by Population descending
Move this column one place to the leftCrude Rate Per 100,000Click to sort by Crude Rate Per 100,000 ascendingClick to sort by Crude Rate Per 100,000 descending
199923,98626,264,71391.3
200023,19125,735,24490.1
200124,19125,009,12896.7
200223,75824,581,49896.6
200323,55724,227,13797.2
200423,39723,954,18597.7
200523,82623,668,541100.7
200624,60623,508,119104.7
200724,75723,533,841105.2
200824,85223,740,364104.7
200925,20323,983,625105.1
201025,48624,143,320105.6
201126,75424,519,007109.1
201227,26624,744,491110.2
201327,58324,969,763110.5
The relative increase since 1999 is 21%. Yes that's right, 21%. Quite a lot more than the 8.9% relative increase observed for the age group 45-54, during the same period. Yet no one talks about the sharp increase in mortality rates for this age group. Probably because no one knows about it.

For the age group 35-44 it's not quite as noteworthy:

Year
Results are sorted in by-variable order
Move this column one place to the rightDeathsClick to sort by Deaths ascendingClick to sort by Deaths descending
Move this column one place to the rightMove this column one place to the leftPopulationClick to sort by Population ascendingClick to sort by Population descending
Move this column one place to the leftCrude Rate Per 100,000Click to sort by Crude Rate Per 100,000 ascendingClick to sort by Crude Rate Per 100,000 descending
199957,63032,411,498177.8
200058,23132,176,646181.0
200159,54631,661,084188.1
200259,52931,003,630192.0
200358,09530,270,056191.9
200455,54329,661,763187.3
200554,78629,094,648188.3
200653,19828,513,036186.6
200750,96627,790,388183.4
200849,09126,954,561182.1
200947,73626,085,420183.0
201044,99925,531,379176.2
201144,89124,880,878180.4
201243,94024,537,314179.1
201343,93124,265,361181.0
Still, a small increase in all-cause mortality.

Yeah I know, Deaton and Case didn't explicitly state that all-cause mortality rates for non-Hispanic whites increased for the age groups 25-34 or 35-44. They did, however, state that all-cause mortality did not increase for the age group 30-34 (see the aforementioned quote I provided).

But here is the 30-34 age group for non-Hispanic whites:

Year
Results are sorted in by-variable order
Move this column one place to the rightDeathsClick to sort by Deaths ascendingClick to sort by Deaths descending
Move this column one place to the rightMove this column one place to the leftPopulationClick to sort by Population ascendingClick to sort by Population descending
Move this column one place to the leftCrude Rate Per 100,000Click to sort by Crude Rate Per 100,000 ascendingClick to sort by Crude Rate Per 100,000 descending
199913,93813,756,381101.3
200013,36913,550,61798.7
200114,06113,465,904104.4
200213,85813,265,133104.5
200313,62012,937,899105.3
200412,94512,515,728103.4
200512,94312,026,446107.6
200612,66411,538,341109.8
200712,61911,340,965111.3
200812,69811,362,841111.8
200913,07811,529,602113.4
201013,27911,690,434113.6
201114,20312,008,437118.3
201214,48912,217,876118.6
201314,94912,400,050120.6

In conclusion, the part in the Case and Deaton publication which states that all-cause mortality rates for non-Hispanic whites did not increase for any other age group than 45-54 is, to put it bluntly, WRONG. At least according to the data they used.

If you don't believe me, check the data yourself on CDC WONDER. Also, check Case and Deaton's publication and you will see that they refer to the CDC WONDER data (among others) in their paper.

Also, mortality rates continued to decrease for Hispanics, non-Hispanic blacks, and Asian and Pacific Islanders for all the age groups. This you can check for yourself as well, if you wish to.

P.S. Could someone point this out to Deaton and Case?

Update: Finally got an email response from Deaton. Here is part of the email:

"You are right about the rising mortality rate in the younger group, and our statement that you quote should have been more precise and made it clear that we were referring to the older groups.  “Deaton and Case are wrong” seems a little strong though, especially as people will read it to mean that the headline result is wrong, rather than that we made an insufficiently qualified statement. But that is your choice.

The reason we did not want to say much about the younger groups is that their mortality rates are (a) extremely low, so that it is pretty easy for even a small number of deaths to increase the overall mortality rate for them, and (b) their mortality rates have indeed increased not infrequently in the past, for example from AIDS. So it is not such a big deal when their rate goes up. But, as you say, we should have been more precise."

And he notes:

"BTW: the study is Case and Deaton, not Deaton and Case, though you are far from the only one to have made that mistake!"

That's certainly not a mistake I'll make again.

I would still argue that the increase for the 25-34 age group is significant, given that even the absolute increase of 21 out of 100,000 people is not much lower than the 34 out of 100,000 people increase for the 45-54 age group (the relative increase is more than twice as large). In addition, almost all of the increase appears to come from poisonings (all of this can be confirmed on CDC WONDER).



No comments:

Post a Comment