STUMP » Articles » Mortality Monday: Cause of Death, 2014 » 19 June 2017, 22:13

Where Stu & MP spout off about everything.

Mortality Monday: Cause of Death, 2014  


19 June 2017, 22:13

I’m going to do something non-actuarial for a moment (or not-all-that-actuarial) and take a year’s worth of death records and slice them in odd ways.

As actuaries, we generally want to look at meaningful rates — mortality rate by age, sex, etc. — and sometimes incidence of cause tied to those as well.

For trends, we like age-adjusted mortality rates, but that can be a bit difficult for people to understand.

We often smooth the data to make tenable analyses — again, I would rather not explain that.

So for this post, I’m taking some data that have already been adjusted somewhat, but no smoothing has occurred.


Here is my data set I’m pulling from. It originally comes from the CDC, and this is the death records data set, which is derived from death certificates issued in the U.S. I linked to the kaggle page for this dataset, so you can see what other people have done with it.

This is not to be confused with the Social Security Death Master File, which only has death records for people with SSNs. The CDC file includes foreigners who died while in the U.S. (I’ll check that out in a bit)

There are 38 fields of record, including items such as age, sex, and an extremely detailed cause of death. I’m looking at the single cause file, because let’s start out with something easy.

I imported it all into Access, because I’m lazy (i.e., I have Access on my most-used computer), and I’ll share the query results in a spreadsheet at the end.

In 2014, there were 2,631,033 deaths recorded in this database, with 3,685 unique ICD10 codes for cause of death. There are different levels of categories I will investigate, so let’s have some fun!


The Death Records use ICD-10 codes to record cause of death. Perhaps you would be interested in seeing the top 20 causes of death at this level of detail.

The top one is heart disease, related to hardening of the arteries. The second one is lung cancer.

(Those docs and their fancy terminology).

In any case, not too many surprises, except perhaps for sepsis (unspecified) – I wouldn’t have guessed bacterial infection kills so many. But that could be hospital acquired infections, while patients for a different condition are treated for something else.

I will come back to this, but the main thing to note about all of these are that these are causes of death primarily for older people.

Especially pneumonia.

At this point, I’m warning you that there are some really disturbing items in the data. For example, “Legal execution” is one of the causes of death — (34 people in 2014). Some of the level of detail gets super-specific, and I’m going to be pulling out certain info in specific.

I am going to be “having fun” with this data for a while, so I’m mainly going to be looking only at different levels of cause/manner of death for this post, and not looking at demographic info (like age).


So here’s a thing.

There’s “Manner of Death”:

If I re-do top causes of death, also including the manner of death – the top 10 are all natural causes:

Well, that’s a bit boring. And not terribly surprising:

At least 78% of deaths are from natural causes, and likely much higher than that. Accidents are only 5% of total deaths; suicides and homicides less.

Makes me think of a fatuous image going around facebook & twitter comparing different causes of death. Well, homicides are only 1% of all deaths in the U.S., and we’re supposedly a really violent culture. So… we should just ignore homicide?

Speaking of….


Here we go!

Note that the first homicide category is “assault by unspecified firearm discharge”. Look, most coroners don’t really care whether the person was killed from a handgun vs. a rifle vs. a shotgun. Some states don’t make a distinction there. (There are some other records where the cause must be natural… like artherosclerosis… but it says the manner of death is “unspecified”. I assume those states are not making the distinction of manner of death.)

This is cause of death/manner of death combo ranked at #58.

You may be interested to know that cause #57 is the natural cause of multiple myeloma (aka a type of cancer). You may also not be interested.

Not getting how “by unspecified person” is relevant if we want to know how. These records are supposed to be how not who.

Which goes to show that death certificates need to be nationalized. (I’m not serious about that, fwiw.)

If you’re interested in the rarer causes of homicide, I will just pull out a few. There were only 169 causes under homicide, some of which seemed unusual, but I can understand. A few were homicides by heart attack — and I can see how that works.

Some things that caught my eye in the homicide stats:

  • 34 by legal execution
  • 32 by drowning/submersion
  • 12 by assault by gases/vapors
  • 6 by parent (I imagine it was more than that… but cf the “unspecified person” above… and gunshot, etc)
  • 6 by sequelae of war operations
  • 4 by assault by pushing from a high place
  • 1 by anemia, unspecified (hell if I know)
  • 1 by sepsis, unspecified
  • 1 by “Bus occupant injured in collision with car, pick-up truck or van: Driver injured in traffic accident”

Yes, it gets that specific. And yes, that can be a homicide, as opposed to an accident.

I’m not going to dig through the suicide stats right now, but you get some similarly disturbing specifics. Evidently, one can commit suicide via pneumonia. Or, at least, one person managed it in 2014.

Or, just perhaps, some coroner screwed up in entering the death certificate.

Seriously, suicide by pneumonia?


Suicides are more likely than homicides, but accidents are even more likely (if I ignore that whole “unspecified” category)

Here are the top ten causes of accidental death:

You need to know that “poisoning” encompasses overdoses of drugs, in addition to other meanings of “poisoning”.

But let’s go down that list — I’ve not combined traffic accidents, but we can see that accidental overdoses looks quite common … but look at all those deaths from falls.

I was skeptical re: this ‘deaths from falls’ item and I had this interaction:

So, there ya go.

Anyway, falls… watch out! (But yes, most fatal falls are of old people)

Let’s check that:

The first number in each row is the number of people who died that way, the second number is the average age at death.

Now again, these death numbers pale in comparison to those who die via heart disease or cancer. But death by falls are preventable, especially for the elderly. And those preventatives are easier than trying to cure cancer. Here’s a list of tips of preventing elderly people from falling in their homes.

Just because more old people die from cancer as opposed to falls doesn’t mean you should not try to prevent those falls. Mmkay?


I only concentrated on causes of death – I have yet to exploit many of the demographic details, like sex, detailed age, and race/ethnicity, that are in the data set. Don’t worry – I’ll get there.

But here are those top causes of death again, with average age at death:

So, those with lung cancer die younger. But in general, the top causes of death are for old people.

Spreadsheet with query results — has the whole list, not just the top 10.

Related Posts
Actuarial Stuff: Mergers, Bootings, and Secrecy (oh my!)
Happy Bobby Bonilla Day! In Praise of Valuable Annuities
Mornings with Meep: Happy Bobby Bonilla (and Bruce Sutter) Day!