Visualizing 10 unusual causes of death in the CDC mortality database

Let me make two things clear right up front:

  1. The metrics I used to decide what causes of death are unusual are purely subjective, i.e. which of the thousands of causes I skimmed through caught my eye and made me go, “Huh.”
  2. It is in no way my intention to make fun of anyone’s death. I find these causes of death unusual, not amusing.

Also: I am a great believer in reproducible data science, so as always, I’ve made available everything anyone would need to reproduce (or extend!) my results in an IPython notebook (nbviewer version or faster-loading html version) and this GitHub repo folder.

The U.S. Centers for Disease Control maintains a data service called WONDER (Wide-ranging OnLine Data for Epidemiologic Research); among its databases is the Compressed Mortality File tracking underlying cause of death from 1968 to 2012.

The causes of death are taken from the International Classification of Diseases (which contains an enormous number of causes of death that are not what I would call diseases, such as being struck by a train). It went through revisions in 1979 and 1999, so the categories do not match up cleanly through every year. For example, after 1978 “transvestitism” is no longer listed as a possible cause of death. (I’m not making this up. There are no deaths attributed to transvestitism in this database, but it’s there in the schema, so perhaps it was assigned to someone before 1968)
Tools used: Python (with pandas and plotly) and Photoshop.

1. Dental caries

If my dentist had told me cavities could result in death, I might have flossed more often. We can see the change in cause of death definitions, as 1999 and on has slightly different wording.

2. Weather or storm

Here we see even more clearly the divide between cause of death classifications. I don’t know what they called these deaths before 1979, or what the big event was in 1980. I wonder about 2005; could it be Hurricane Katrina? The ICD-10 lists ‘hurricanes’ as a separate cause of death, but you always have to allow for human error in assigning these categories.

3. Migraine

A good friend of mine has her life encompassed by her migraines; I had no idea they could result in death. You can see that after the 1979 revision, increased knowledge of this condition led to parsing into further categories.

4. Spacecraft

This one I find somewhat puzzling. Since the database starts in 1968, we just miss the Apollo 1 fire the year before, but what about the seven deaths aboard the space shuttle Challenger in 1986?

5. Conjunctivitis

I had conjunctivitis, or ‘pinkeye’, at least once as a child, and it was no big deal, so I was bemused at the first season South Park episode in which the entire town is so afraid of pinkeye, they confuse it with zombification. (As an aside, this was the first time I’d ever heard of pirated media, as a coworker of mine downloaded it off Usenet in 1997.) Turns out it can be deadly, and there are many, many categories of conjunctivitis deaths. (The graph’s defaults don’t have enough different colors to differentiate them all, but I think the forest matters more than the trees here.)

6. Cleft palate or cleft lip

I find it encouraging that the death rate for this condition appears to have gone down. My dad grew up in the ’50s with a girl with a “hare lip”, as he called it, and hearing stories about it as a kid I felt so bad for her. Had I known it was a cause of death (and quite a more substantial one than many others on this list), my secondhand suffering would have been even worse.

7. Elbow

I tried to think of the most unlikely part of the body to result in death. Here it is. If you’re wondering why there are six causes of death in the legend and four in the graph, it means two had no entries during the time period. Also, the cause ‘Of elbow’ means it was a subgroup of a supergroup that does not appear in the database I downloaded (I could have downloaded the supergroup fields, but I didn’t, the file was huge enough already).
I’m assuming there was no outbreak of elbow deaths in the ’80s and ’90s, and that the higher bars are due to differences in criteria of classification. ‘Enthesopathy’ (a disorder of bone attachments) only appears in 1979, and its diagnosis drops down for all bones in 1999. If you’re curious, you can see the graph in my gist notebook.

8. Animal

This one was a little tricky. There are 13 categories of vehicle occupants in collision with ‘pedestrian or animal’ to remove, and then I thought to check specific animals like dogs and bees (cat fanciers will be happy to know there is no category devoted to death by feline, and yes, bees are animals.)

9. Ingrown nail

I actually had a pretty badly infected ingrown toenail as a kid. Still, it appears my odds were pretty good, as there’s less than two deaths per decade attributed to it.

10. War

You wouldn’t think war would be an unusual cause of death, the world being what it is, but I find the low numbers attributed to it unusual. There’s absolutely no increase when the Iraq war starts in 2003. Make of it what you will.