Drug Poisonings by Intent Over Time

I've stitched together a nice dataset of all drug poisonings for all years 1999-2016. The raw data files are here. Usually people who want mortality statistics over time can pull them from the CDC's Wonder database, but that is limited. If you want to do some complicated filtering and counting based on multiple criteria on the death record (like I'm trying to do), you need to go back to the raw data. I'll do a series of posts on what I find.

The ICD10 coding system has codes for drug overdoses and they vary by intent. There are separate codes for accidental poisonings, suicide, undetermined intent, and murder. Most media stories that report on the "opioid epidemic" lump them all together and report the total. I think this is misleading. If someone deliberately takes his own life using a bottle of painkillers, is that part of a drug epidemic? I don't think so. On the other hand, if recreational drug users are recklessly killing themselves by scarfing down too many Vicodin and chasing it with vodka, I think that counts. Intent matters.

Here is what I found. Accidental deaths have been rising dramatically. Suicides have increased, but by a much smaller factor. "Undetermined Intent" bounces around but has been roughly flat. Murder has always been small. "Other" means that it wasn't labeled as a drug poisoning but some kind of substance was mentioned on the death certificate (well, one of the ~dozen most common substances anyway). In this last category I see a spread of different underlying causes: vehicle accidents, drownings, various organ diseases, the effects of chronic drug use (these are not coded as drug poisonings but should plausible be counted as part of "the drug problem".) Mere mentions of drugs have been rising, but again not as dramatically as accidental poisonings.

I think the table is perfectly informative, but since I spent all morning toying with ggplot I might as well share the pretty graphs I made from this data.

A slightly different look at the same numbers is in the table below.

And the same numbers with a ggplot graph.

None of this is news to me, but this is the first time I've had the complete dataset to build this on.

One thing that makes me suspicious is the small fraction in the "other" category. I've written several previous posts suggesting that there might be a spurious trend in the death counts. I suspect many deaths from other causes are inappropriately labeled drug poisonings just because the decedent happened to have drugs in their system when they died. With the very large number of drug users in this country (tens of millions of illicit drug users and ~90 million prescription opioid users each year), I would expect to see a large number of death where opioids are mentioned on the death certificate but are incidental. A lot of people who happen to have drugs in their system die of totally unrelated causes. Are medical examiners scrupulously avoiding any mention of drugs when they aren't relevant to the cause of death? Or are those same medical examiners assuming that the presence of drugs contributed to the death, even in cases where it might be incidental? I'll follow up if I find anything that favors one story over the other. If the second thing is happening, that means the recent increase in drug poisonings is being overstated.

______________

Ugh. The text on the graphs looks blurry in the main post, but if you click directly on it the text gets sharper.