Everyone’s familiar with this graph:
Scary, although it was actually down slightly in 2018. Looking at this, one would assume that huge numbers of Americans are now addicted to opioids. However, although the numbers are annoyingly tricky to find, this doesn’t seem to be the case.
- 2 million Americans had an opioid use disorder in 2018, about 0.8% of the adult population. (A much larger number “misused prescription opioids”, but this is defined extremely broadly.) The average American high school has 750 students; in a group this size, on average, six would be opioid addicts.
- 800,000 Americans used heroin, about 0.3% of adults. In the 750-person high school, between two and three people would use heroin. This is the lowest figure since 2013.
- About 5 million Americans have ever used heroin during their lives. This is about 2% of adults, or 15 people in a high school.
Opioid prescriptions per capita peaked in 2010, and have since been going down:
The dramatic spike in overdose deaths, therefore, seems to mainly come from drugs getting much more dangerous, rather than more people using them. Fentanyl and other synthetic opioids, which have dangerously unknown doses and potencies, were initially used to make heroin stronger; they were then found in counterfeit prescription pills; and recently, they have appeared even in non-opioid drugs, like meth and cocaine. This suggests that one way to reduce the number of deaths would be distributing fentanyl test strips, which are widely available online.
The Rationalist Conspiracy 
Yes, but there are a lot more details you can see from the second graph.
The number of deaths among heroin users is the product of the number of users and the death rate. If one goes up and then the other goes up, the second has a larger absolute effect. It might get your attention if you didn’t notice the first rise. But causally the two are equally important. (Also, if you aren’t restricting your attention to heroin users, but to total opioid deaths or all drug deaths a big increase in their deaths might not be noticeable, as in the first graph.)
People have been complaining about an opioid epidemic for at least 2 decades. They weren’t completely wrong. There were three opioid epidemics. In the third, fentanyl caused the death rate of heroin users to rise from 1% per year to 3%. In the second, heroin use rose 3x or more from 2010-2015, with the death rate steady at 1% per year.
The first opioid epidemic was rising use of pills. They have held steady at 1 death per 10k patient-years. That’s probably worth it. Supposedly there was a conscious decision to increase the use of opioids. Since the death rate was steady, the total number of deaths was a predictable consequence of the decision. If the cost made sense in 1990, then it still made sense in 2010, when medical opinion turned against opioids. Of course, at new margin, the marginal patient is different. But if the slowdown in pills in 2010 caused the heroin epidemic, that complicates the picture.
I think I get my numbers like 1 death per 100 heroin-years and 10k pill-years from this blog post (which says a lot more, such as interesting skepticism about a lot of details about pills), but most of my points can be deduced from your second graph.
Thanks for providing all this additional detail! Greatly appreciated.