The Overdose Numbers Are Falling. That Is Not the Same as the Crisis Ending.

Something genuinely good has been happening in American overdose data, and almost nobody has figured out how to talk about it without either overclaiming or immediately undercutting it.

The numbers first. CDC provisional data predicts 68,641 drug overdose deaths for the twelve months ending February 2026, a 12.1 percent decline from the prior year. That follows a 26 percent drop the year before. Three consecutive years of decline from the 2022 peak of roughly 108,000 deaths is the steepest sustained fall since the country started tracking this systematically.

That is not a rounding error or a reporting artifact. Tens of thousands of people are alive who would not have been on the previous trajectory.

So why does everyone working in this field sound so hedged about it?

Because we are not entirely sure why it happened

The comfortable explanation is that public health worked. Naloxone got distributed widely, treatment access expanded, surveillance improved, and the interventions built over a decade finally hit scale. There is real evidence for all of that.

The less comfortable explanation, which has been gaining ground in the peer-reviewed literature, is that the drug supply changed. Researchers have pointed to a drop in fentanyl potency linked to controls on precursor chemical exports, essentially a supply-side shock that no American policy caused and no American policy controls.

If that is a substantial share of the story, then the decline is partly weather rather than climate. Weather changes back.

The honest position is that both are probably true in some proportion, and the proportion matters enormously for what we do next. A decline driven by naloxone and treatment access is durable and worth funding harder. A decline driven by the chemistry of an unregulated supply chain is borrowed time.

The composition is shifting underneath the total

Here is the finding that has stuck with me most.

Research published in Addiction in 2026 tracked the decline by substance and demographic group. Deaths involving illicit fentanyl fell sharply between 2023 and 2024. Fentanyl-involved deaths without stimulant co-involvement dropped from 31,193 to 19,673. Fentanyl with stimulants dropped from 41,583 to 28,062.

But deaths involving stimulants without fentanyl went the other way, rising from 18,142 to 18,907.

The absolute increase is modest. The share is not. Stimulant-only deaths went from 17.3 percent of all overdose deaths to 23.8 percent in a single year.

That is what a changing epidemic looks like from the inside. The total falls while the shape reorganizes. And the tools built for the opioid crisis do not transfer cleanly. There is no naloxone for methamphetamine. There is no buprenorphine equivalent for cocaine. The pharmacological toolkit that drove a lot of the opioid progress simply does not exist for stimulants, where the evidence base points toward contingency management and behavioral approaches that are harder to scale and less well funded.

Xylazine tells a similar story. Its co-involvement in fentanyl deaths rose from 8.3 percent to 12.6 percent over the same period, and medetomidine has since displaced it in several major markets. The supply keeps adding non-opioid sedatives that naloxone does not reverse.

The decline is not distributed evenly

Averaged across racial groups, overdose death rates fell 24.4 percent between 2023 and 2024.

Non-Hispanic Black Americans saw the largest decline of any group examined, 29.3 percent. That sounds like the disparity narrowing, and in one sense it is.

It also leaves that group at 36.0 deaths per 100,000, roughly 1.51 times the national average.

A larger percentage decline from a much higher baseline can still leave the gap in absolute terms wide open. Both facts are true simultaneously, and reporting that picks one is misleading in either direction.

What has not moved at all

The number I keep returning to is not in the mortality data.

SAMHSA’s 2024 National Survey on Drug Use and Health found that of people who needed substance use treatment in the past year, about one in five received it. Roughly 80 percent did not.

Overdose deaths are falling. The treatment gap is not meaningfully closing. Those two things can coexist because they measure different failures. Mortality responds to naloxone, to supply chemistry, to emergency response. Treatment entry responds to something slower and more human: whether someone has concluded they need help, and whether the people around them know what to do about it.

That second thing is not a public health infrastructure problem in the usual sense. You cannot fund your way past it in a single budget cycle. It happens in kitchens and living rooms, one family at a time, usually badly, usually after years of trying the same approaches that have already failed.

Holding two ideas at once

The instinct in public health communication is to pick a register. Either the good news, to sustain political support for what is working, or the crisis framing, to prevent complacency.

The overdose data does not cooperate with either.

Sixty-nine thousand deaths in a year is still more Americans than died in combat in the entire Vietnam War. Overdose remains a leading cause of death for adults aged 18 to 44. And the conditions that produced the improvement, including Medicaid coverage and federal overdose prevention funding, are under active budgetary pressure in ways that could reverse the trend before anyone agrees on what caused it.

Progress and emergency are not opposites. We are in both.

The useful posture is neither celebration nor despair. It is asking, with some urgency, which parts of this decline we actually built and which parts we were handed. Because we only get to keep the first kind.