U.S. Overdose Deaths Fell 12% Last Year, but Rose in 430 Counties

The Centers for Disease Control and Prevention estimates there were nearly 68,000 drug overdose deaths in the United States in the 12 months ending March 2026, about 9,500 fewer than in the prior year — a decline of roughly 12%, according to an Axios analysis of the agency's provisional county-level data published Tuesday.

The national figure continues a trend now several years old. But the county map underneath it is not uniform. Among counties where a year-over-year change could be calculated, overdose deaths fell in 842, rose in 430 and held steady in 352.

What the provisional counts actually measure

The comparison runs between two rolling 12-month windows, ending March 2025 and March 2026, using death certificates that are still being finalized. Provisional counts are subject to revision as investigations close and toxicology results arrive, and the CDC warns that its provisional totals can understate final numbers.

The county breakdown carries an additional limitation that shapes any reading of it: for counties recording between one and nine deaths, the agency withholds the exact count, which leaves trends undeterminable in nearly half of U.S. counties. Those are, by definition, places where the toll is small enough to suppress — but also places where a jump from four deaths to nine would never appear in the map.

Where the decline stops

The counties that moved the wrong way are not randomly scattered. Several neighboring counties in Arizona and New Mexico recorded year-over-year increases, part of a regional pattern that has persisted while national totals dropped.

Rio Arriba County in northern New Mexico recorded 173 overdose deaths per 100,000 residents, the highest rate calculable from CDC counts and Census Bureau population estimates in the Axios analysis. Baltimore, which has spent years at the center of the epidemic's East Coast arc, recorded 82 deaths per 100,000.

New Mexico's trajectory has diverged from the national one before. State overdose deaths rose 21% in 2025 even as the country as a whole improved, a split that state health officials have attributed in part to the timing of the local supply transition and to treatment capacity in rural counties.

Arizona's story may follow a similar logic. Axios's Phoenix bureau has reported that the state's comparatively late shift from counterfeit opioid pills to powdered fentanyl could help explain why its crisis has outlasted the national peak — the drug supply changed later there, and the mortality curve followed.

What the CDC credits, and what it cannot quantify

The agency has cited several drivers for the national decline: wider distribution of naloxone, expanded access to addiction treatment, and shifts in the illicit drug supply itself. What the CDC has not done is assign a share of the improvement to any one of them.

That gap matters for policy. Naloxone distribution is measurable and cheap to expand; medication for opioid use disorder requires prescribers and durable funding; supply changes are largely outside any agency's control. Without attribution, a state cannot tell which of its investments is doing the work — and, more importantly, which one it could cut without stalling the decline.

Nationally, deaths involving fentanyl and other synthetic opioids, methamphetamine and cocaine have all fallen, according to CDC data released in May. That breadth is unusual for this epidemic, which for most of the past decade moved in one direction.

The number that hides the map

For agencies allocating opioid settlement dollars and state legislatures writing treatment budgets, the county table is the more useful document. A 12% national decline is compatible with dozens of counties where the toll rose, and those counties are where harm reduction programs, mobile units and bridge clinics face the sharper test.

The CDC's own framing has been cautious: provisional data, subject to revision, not yet a final accounting. The direction of the national curve is clear enough to be encouraging. The direction of any single county's curve is a separate question, and on this reading, 430 of them are still going the wrong way.