Update on Drug Poisoning Data for 2015

The CDC just recently released the 2015 mortality data, so I thought it would be good to do an update of some previous posts that used the data through 2014. (See here and here and here for example; several other posts of mine from 2016 include this data.)

There were 52,623 drug poisoning deaths in 2015, which is up from 47,196 in 2014. But these are raw totals and aren’t very meaningful. They include accidental deaths, suicides, “undetermined intent”, and murder, and the CDC has separate codes for each of these. I don’t think it’s quite fair to add in suicides if we’re talking about “the drug problem,” because accidentally overdosing is very different from willfully ending your own life. These are different social problems, and the only reason to include suicides in the total is that you’re trying to inflate the number to make it look scarier. If we’re just looking at accidental drug poisonings, there were 44,335 in 2015 (up from 38,841 in 2014). (Accidental deaths are codes X40-X44, suicides are codes X60-X64, “undetermined intent” are codes Y10-Y14, and murder is X85, in the ICD10 coding system. Murdering someone with drugs is apparently rare; there are only 94 cases of it in 2015.)

I suspect you will see a lot of news stories starting with “There were 52,600 drug overdoses in 2015…” If you see such a story, scan it to see if it gives a breakdown by “accidental vs intentional.” If it doesn’t, that’s a big warning sign that the author didn’t do their homework. Or perhaps the author is trying to sell you a narrative using inflated and irrelevant numbers. Read such pieces carefully. If the piece starts with the 52k figure then shifts to talking about the “opioid epidemic,” the author has done a bait-and-switch. The relevant figure there is more like 15,000 to 19,000, depending on exactly how you count them.

Here are the notable things that happened between 2014 and 2015. Heroin deaths climbed from 10,754 deaths in 2014 to 13,069 deaths in 2015, taking the top spot. Other Opioids rose a little, from 12,657 to 12,927 deaths. Cocaine, which had been declining from 2006 to 2013, rose over the last two years (at 5495 deaths in 2013, 5992 deaths in 2014, and 6945 deaths in 2015). Other synthetic narcotics (which includes fentanyl and other very powerful synthetic opioids that are often a thousand times as potent as morphine) rose from 5695 deaths in 2014 to 9681 in 2015. This category shows the most dramatic increase. “Psychostimulants with abuse potential”, a category that includes methamphetamine and prescription analogs to it, has been increasing steadily over the past few years; it rose from 5114 deaths in 2014 to 6082 in 2015.

Here are the mortalities for each year for the most deadly substances, expressed as deaths per 100k population:

The chart above is from the CDC’s WONDER database, but it’s a little misleading. Don’t think you can add together the deaths from each substance to get a total, because there is significant overlap between categories. Single-substance overdoses are actually pretty rare. A typical heroin overdose actually involves two or three substances, so such a death might show up in, say, the heroin, cocaine, and benzodiazepine line. Such a death would be triple-counted if you made the mistake of summing the death totals for each substance. Also, some of these deaths aren’t even marked as “drug overdoses.” They are heart attacks or something else, but someone happened to write down “heroin” or “other opioids” on the death certificate. This is somewhat rare; usually if a drug is written down on the death certificate, the death is coded as a drug overdose. But this explains the discrepancy between the numbers in the above paragraph and the ones I will give below. Also, I’m excluding suicides, murder, and “undetermined intent” in the charts and figures below. The chart below is an update of a chart from an earlier post, illustrating the rarity of single-substance overdoses. I found some of these surprising when I first saw them (Click to enlarge, I realize the individual cells are small. Some rows are hidden, so the Total row might not match the sum of the column exactly in some instances):

To explore the details of what kinds of substances are found in combination, you need to dig into the raw data files here. There is a record for each death that happened in the United States in 2015, which lists all the contributing causes of that death. The CDC’s WONDER database is great, but it doesn’t allow you to determine what drugs *in combination* are contributing to these poisoning deaths. To see what kinds of substances are typically found in combination on a death certificate, see here:

(To read the cross table, go across by row and say, “This percent of [row label] poisonings also involves the substance [column label].” Thus you get 100% across the diagonal.)

So, for example, benzodiazepines are extremely likely to be found in conjunction with opioids. Alcohol is frequently found in conjunction with other drugs. You get the sense from looking at the cross table that many of these people would still be alive if they hadn’t been mixing drugs.

These drug interactions are a big deal, because a big increase in mortality due to heroin can cause an increase in mortality due to cocaine, even if the cocaine problem is declining. You can’t just look at the graph above and say, “The prescription painkiller problem (mostly the ‘other opioids’ curve) is getting worse, because it went up.” Looking at the cross table, for those deaths with “heroin” stamped on the death certificate, about 20% involve cocaine, ~15% benzodiazepines, and ~11% “other opioids.” So a rise of 3,000 heroin deaths per year (roughly what we saw from2014 to 2015) will result in 600 additional “cocaine” deaths, 450 “benzodiazepines” deaths, and 330 “other opioids” deaths. Likewise, the big spike in “Other synthetic narcotics” deaths is going to drive up the deaths attributed to other substances. Roughly 20% of these deaths involve other opioids, 18% involve benzodiazepines, and 17% involve cocaine.  So the roughly 4,000 additional “other synthetic narcotics” deaths will give us 800 additional “other opioids” deaths and about 700 additional benzodiazepines and cocaine deaths. This could be true even though there may be no additional users or perhaps even fewer users of these other drugs. The problem is that a heroin habit is so potentially deadly, and heroin users tend to use these other substances in combination. Anyone looking at the raw totals by substance will be misled if they aren’t thinking carefully about the poly-drug use problem.

Be cautious interpreting the big jump in “other synthetic narcotics.” I suspect that the year-to-year changes aren’t as big as the chart above implies, but rather the reporting is getting more accurate. There have been numerous media reports of multiple people overdosing at the same time in the same city (even the same neighborhood) because of a bad batch of “heroin.” Well, it’s not really heroin but actually something like fentanyl or something even stronger. The first responders don’t always know what drug is actually causing the drug overdose. The lab testing required to figure out the exact substance sometimes takes weeks (citation needed, but I’m speaking from memory based on an NPR story I heard months ago). I don’t expect that death certificates are always updated after the test results come back. I think the big jump in this "other synthetic narcotics" category shows a growing awareness that this is what’s really happening in these “heroin overdose” cases, most likely combined with a real increase in the number of overdoses. So there's a real trend with a reporting trend on top of it.

Some will surely draw the wrong lesson from this most recent year’s worth of CDC data. Some will say, ”The problem is bigger than ever, so we need to hammer it even harder!” In reality, the rising heroin and fentanyl overdoses are proving that drug prohibition isn’t working. Or in other words, prohibition is having *exactly* the effect that its critics predict. Dealers are attracted to drugs that are easiest to conceal, so they are opting for super-concentrated opioids that are sometimes thousands of times stronger than heroin. You can collapse a multi-ton shipment of heroin down to something the size of a briefcase. But when it reaches the user, it may not have been diluted properly. This problem is causing a lot of unnecessary overdoses. Ending prohibition would pretty much fix the problem, because strict quality control on the drug supply would be possible under a legalization regime. Buyers could ensure that they are getting pharmaceutical grade heroin that hasn't been spiked with a lethal dose of fentanyl, and producers would subject themselves to outside audits, the results of which they could show to their customers. Producers would subject themselves to the threat of lawsuits if they mislabeled something. Such quality controls simply aren't possible under a prohibition regime.

One final note here. The CDC's method for counting "prescription opioid overdoses" used to be to count all deaths involving methadone, other opioids, and other synthetic narcotics, which up until 2013 might have been a reasonable methodology. No longer. Unless the spike in other synthetic narcotics deaths is driven entirely by a recent dramatic spike in fentanyl prescriptions, we should suspect that the heroin and other synthetic narcotics categories are bleeding into each other. Be very careful if you're reading anything that says that 2015 saw a "new maximum" in opioid deaths. They are either using an old and now dubious methodology for counting prescription opioids or they are surreptitiously adding prescription opioids to heroin without explaining that these are two different problems.