Overdose Deaths Fell 12.3%. The Government Is Cutting the Thing That Probably Did It.

On International Overdose Awareness Day this year, the CDC launched something almost boring on purpose: a data channel, a consolidated home for overdose numbers, prevention guidance, and treatment resources. Buried in the launch is the actual news: provisional counts for the twelve months ending March 2026 show 67,798 predicted drug overdose deaths nationwide, a 12.3% drop from the same period a year earlier. Nonfatal overdoses tracked in emergency departments fell too, down 2.6%.

A 12.3% year-over-year drop in overdose deaths is one of the best national numbers this crisis has produced in a decade, and almost nobody is treating it like headline news.

CDC Director Dr. Erica Schwartz framed the new channel as a transparency move: “CDC wants to ensure maximum transparency to empower patients, providers and other stakeholders as we continue our battle against the opioid epidemic.” Grant Baldwin, who directs the agency’s Division of Overdose Prevention, put it more usefully: “The most useful data does more than show us what has happened; it helps us decide what to do next.” That’s the right question, and the honest answer is uncomfortable for the moment we’re in. The interventions researchers most consistently credit with bending the overdose curve down — wider naloxone access, expanded syringe services, and fentanyl test strips that let someone know what they’re about to use — are the same interventions SAMHSA moved to defund in April this year, on the theory that they sent the wrong message. NIDA’s own research puts a number on why that theory doesn’t hold up under evidence: people who use syringe service programs are roughly five times more likely to enter drug treatment than people who don’t use them.

None of this means the drop is fake or that it will reverse overnight — provisional data lags, and a lot of the decline predates April’s funding letters. It does mean the timing is brutal for anyone trying to argue harm reduction was expendable: the numbers got better while the tools were still funded, and the funding got pulled anyway, on a bet that the improvement will hold without them. If you’re the person deciding whether to keep a test-strip line item in next year’s budget, this is the data to put next to that decision — not the political framing that came with the letter cutting it.

The next data channel update will say whether that bet paid off. This crisis has been wrong about worse bets before.