Loading today’s stories
Science1 distinct publisher3 min readPublished
Scotland and Canada recorded declines in the same year, and the Commonwealth Fund's canvass of two dozen state health authorities found nobody willing to name a single cause, which leaves 2026 budgets ranking levers on plausibility.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
Virginia is the cleanest-looking number in the report, and it still settles nothing. Deaths there fell 40% from 2023 to 2024 after the state implemented a naloxone saturation plan [13], the kind the federal Substance Abuse and Mental Health Services Administration began requiring in 2022 of states taking opioid response money [12]. The nation fell 27% over 2024 [1]. The difference is 13 percentage points [1], and that residual is the most a careful reader can assign to Virginia's plan before even asking whether the relationship is causal. Everything underneath it was also happening in states that did something else.
Maryland shows the other half of the measurement problem. Its 25% decline is measured from 2019 to 2024 [14], a five-year window that includes years when deaths were still climbing, while the national 27% is a single year [1]. Line the two up and Maryland reads as a laggard [5], which is an artifact of the windows rather than a verdict on vending machines.
This is what an interview study can and cannot do. Evan Gumas of the Commonwealth Fund asked authorities in at least two dozen states, plus counterparts abroad, what they had been doing [8], and the uniform answer was that no one thing did it [9]. That is honest, and it is also the fingerprint of a design with no control group: everyone scaled something, nobody abstained, and the outcome moved everywhere at once. Naloxone saturation, medications for opioid use disorder, treatment in and around prisons, and better data infrastructure [10] are all plausible. New York's prisons and jails increased use of those medications sevenfold between 2022 and 2025 [21], in a system where substance use disorder is twelve times more common than in the general population [20]. Plausible is not ranked.
Two mechanisms in the report carry no budget line at all. Sarah Wakeman, who directs the substance use disorder initiative at Mass General and did not contribute to the report [24], notes that the years of accelerating death shrank the population still at risk [22]. Separately, the potency of fentanyl in the supply has been declining [23]. Both would lower mortality without any program touching it, and both would operate across borders, which is why the foreign numbers do more work than they appear to. Scotland fell 13% in 2024 and Canada 9% [4], roughly half and a third of the US figure [2]. Program scale-up in the three countries was not coordinated; a change in the drug supply is one of the few candidate explanations that would be shared by default.
The thing this does not tell you is whether the United States has moved to a new level or a new slope. Preliminary 2025 data show a further meaningful drop from a still-high base [3], which fits either reading. Brandon Marshall of Brown University calls the speed of the decline unprecedented and says in the same breath that the crisis is not over [5]; both halves are load-bearing. Against a baseline above 105,000 deaths a year from 2021 through 2023 [2], 27% is on the order of 28,000 people [3], so the attribution question is not academic.
Where it gets expensive: Gumas's report calls Medicaid expansion the single most influential lever for coverage and access to opioid use disorder medication [17], and Medicaid cuts are looming while the Trump administration, supportive of those medications, has positioned itself against some proven harm reduction measures [11]. Methadone capacity grew where coverage did, with Colorado going from 25 clinics to 54 in five years [18]. Contraction would run that experiment backwards, on people, and it is the one design likely to produce an attributable effect.
Ranked by verification strength, evidence, and original report placement.
Scotland saw a 13% decrease in overdose deaths in 2024 and Canada saw a 9% decrease.
Brandon Marshall, an epidemiologist at Brown University who studies overdose trends, said it is unprecedented to see overdose deaths come down so dramatically and so quickly, and added that we are nowhere near out of this crisis.
It is unclear what caused the decline; state health authorities who have fought the opioid crisis for decades are still surprised by the drop, according to Evan Gumas of the Commonwealth Fund.
Gumas said the thing every single person he spoke with said was that there is no one thing that led to the decrease.
The initiatives repeatedly mentioned to Gumas were naloxone distribution and harm reduction responses, expanded access to medications for opioid use disorder, better treatment in and around prisons, and enhanced data infrastructure.
Medicaid cuts loom, and the Trump administration, while generally supportive of medications to treat opioid use disorder, has positioned itself against some proven harm reduction measures.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 2, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
science
US life expectancy hit 79 in 2024, and the cushion over 2019 is about 73 days1 distinct publisher
invest
New York's K has a headcount: 40,700 people took 53% of the city's income growth1 distinct publisher
science
A 24% dementia gap, and 1.7% uptake: the shingles number nursing homes will be asked about1 distinct publisher
invest
Gig platforms top the SNAP employer table, and that ranking is now a regulatory input1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One report, carefully handled, never independently checked
The mortality figures are almost certainly right - a 27% national drop is not the kind of number a foundation invents - but nothing in our coverage lets a reader verify it. No CDC or provisional dataset is cited, the Commonwealth Fund report is described rather than shown, and the epidemiologist supplying the outside voice reviewed that report before it published, which STAT tells you. Where the reporting is genuinely strong is in refusing to launder the causal question: the phrase 'no one thing' is quoted, not paraphrased away.
The programs are real and countable; their effect is not
Unusually for a story about an unexplained trend, the interventions themselves are documented in units: naloxone distribution more than doubled in Maryland, methadone clinics went 25 to 54 in Colorado, carceral treatment rose sevenfold in New York, and every federally funded state has had to write a saturation plan since 2022. That is broad, multi-jurisdiction deployment across two countries and a decade in Sweden's case. The gap is attribution, not uptake.
Careful text, suggestive layout
STAT hedges where it matters and hedges twice. What tilts slightly toward overstatement is arrangement rather than assertion: a program is described, a percentage follows, and Virginia's 40% lands next to its saturation plan with no seam. Marshall's blunter view - that the dominant hypothesis is mostly drug supply - sits below four paragraphs of policy levers, and Wakeman's observation that the at-risk population has been thinned by death is filed under 'more sobering contributing factors'. Read fast and you come away thinking naloxone did it.
Aligned interests, openly labeled
The report's most quotable line - Medicaid expansion is the single most influential lever - is also the policy its publisher has long championed, arriving precisely as Medicaid cuts loom. Meanwhile the state officials who supplied the four levers were being asked what they had been doing, a question that reliably produces a list of one's own programs. Two things keep this from being worse: STAT says outright that Marshall reviewed the report, and says outright that Wakeman did not contribute - and Wakeman is the one who offers the least flattering explanation.
Confident about the fall, unsure about everything under it
Two claims here would survive almost any scrutiny: deaths fell sharply in 2024, and no one can yet say why. Between them sits a layer we would not lean on - which lever mattered, how much of the drop is thinner ranks of survivors, whether the potency of the supply is doing most of the work, and whether the turn is holding into 2026. Scotland's provisional rebound is the warning shot in this reporting, and it is one line long.