Overdose deaths fell for a third straight year. The treatment access behind the fall is narrowing

U.S. overdose deaths fell for a third straight year, to an estimated 69,973 in the 12 months ending December 2025. On October 27, 2026, the Health and Human Services Department announced the count in a press release titled "Trump Administration Drives Opioid Deaths to Lowest in Nearly a Decade." A second document from the same department, published the same day, says its numbers do not explain why deaths fell.

The counts are provisional, and the two documents measure the change differently. The Centers for Disease Control and Prevention's May 2026 release put the 12-month total ending December 2025 at 69,973, about 14% below the 81,313 estimated for 2024. The Assistant Secretary for Planning and Evaluation recalculated the change in the age-adjusted death rate, adjusted for a population's age structure, and got about 2%, or 14.1 deaths per 100,000. CDC compared 2025 with 2024 using two different vintages of the population estimates that turn raw death counts into rates. On one vintage, the comparison shrinks: a provisional 2025 rate of 689.2 per 100,000, 4.6% below the final 2024 rate of 722.1.

The fullest available account of causes comes from Nabarun Dasgupta, an epidemiologist. His October 2026 working paper is not peer reviewed, he labels his causes as potential causes, and he declines to settle the credit question. The decline "deserves credit from Trump 1, Biden, and Trump 2 admins," he told STAT. The fall began before the current administration took office, after a 27% drop in 2024.

His map has three layers: the drug supply, how people who use drugs behave, and the clinical and public health response. The supply layer carries most of the arithmetic. In the health-policy office's breakdown, the death-certificate category for synthetic opioids other than methadone, which covers fentanyl and its analogues, accounts for about 96% of the net fall in unintentional overdose deaths, from 45,917 in 2024 to 36,406 in 2025. Unintentional injuries made the largest single contribution to the country's age-adjusted death rate decline, 30%, and overdose accounted for 73% of that reduction. The office says the concentration is consistent with a changed illicit supply, citing purity shifts, precursor controls and interdiction with Mexican counterparts. Dasgupta's series runs longer, from an annualized peak of 112,418 deaths in August 2023 to an estimated 66,937 in the 12 months ending April 2026, a 40.5% fall, with the rate at 19.5 per 100,000. His window ends four months after the federal counts do, so the two series do not line up. Deaths fell 25% or more in Rhode Island, New York, North Carolina, Alabama and Vermont, and rose 10% or more in New Mexico, Arizona and Colorado, according to CDC.

The administration's release credits border closure, the 2025 SUPPORT reauthorization, the law that funds opioid treatment programs, Drug Enforcement Administration fentanyl interdiction and a January 2026 "Great American Recovery" executive order. Dasgupta rejects most of that list: border closure, military strikes on drug boats, on the argument that DEA analyses show cocaine flows unchanged, street-level seizures, drug-induced homicide laws, Good Samaritan laws and GLP-1 drugs, the diabetes and weight-loss injectables. He also rejects the idea that the epidemic is ending because there is nobody left to die. The counterfeit-pill explanation fares worst. More than 95% of fentanyl seizures in Arizona, Colorado and New Mexico are pills, and those are three of the few states where overdose deaths rose 10% or more in the 12 months ending December 2025.

The administration's strongest numbers are about treatment, and the underlying commentary could not be obtained in full. HHS, summarizing a Health Affairs Forefront article published the same day, reports that opioid prescriptions dispensed fell to 45.3% of their March 2015 level, total opioid dosage dispensed to 35.2% of baseline and retail prescription-opioid patients to 49.6% of baseline, while retail buprenorphine and naltrexone, the two medications approved to treat opioid use disorder, rose to 218% of baseline between March 2015 and March 2026. Dasgupta credits the spread of those medications through telehealth, with long-acting injectable versions, naloxone, the medicine that reverses an overdose, medication in jails and opioid settlement money, with much of the third layer, and credits durability partly to generational turnover and to a post-pandemic recovery.

Two rules dated April 24, 2026 from the Substance Abuse and Mental Health Services Administration, the federal agency that funds addiction and mental health services, do separate work. One letter on medication for opioid use disorder says funding should support comprehensive care rather than medication-only models, supports individualized tapering, and requires at least annual medication reviews. A second letter on harm reduction funding says SAMHSA money may not buy fentanyl, xylazine or medetomidine test strips, overdose hotlines, sterile water and saline, or syringes, while naloxone and nalmefene, lock boxes, overdose education, wound care and HIV and hepatitis C prevention remain fundable, according to the National Association of Counties, which summarized the guidance.

Medicaid, the joint federal and state health insurance program for low-income and disabled people, carries the larger exposure. The Urban Institute estimates, on an explicitly approximate basis, that about 59% of Medicaid enrollees with opioid use disorder received buprenorphine in 2025, up from about 36% in 2019. The 2025 tax-and-spending law known as the One Big Beautiful Bill Act is expected to cut federal Medicaid spending by more than $900 billion over a decade, and work requirements plus six-month redeterminations are projected to reduce Medicaid expansion enrollment by 4.9 million to 10.1 million people in 2028.

Dasgupta warns in STAT that the coming Medicaid cuts and the paperwork they add could reverse addiction-treatment progress, and he notes that 2025 brought a slowing of the deceleration. He calls the recovery fragile. The machinery behind the decline is unglamorous: a medication that can be prescribed through telehealth, a bystander carrying naloxone, a test strip that shows what is in a sample. Two pressures now narrow it. SAMHSA funds can no longer buy the test strips and syringes that keep other drugs out of the supply, and Medicaid churn decides who keeps a prescription. The next CDC provisional release will show whether the 2025 slowdown continued. If the 19.5 per 100,000 in Dasgupta's paper stops falling while deaths from fentanyl and its analogues keep dropping, the drug supply was carrying less of this decline than the counts suggest, and the treatment programs were carrying more.