HHS Analysis: Drug Overdoses Drove 22% of the 2025 Mortality Drop

The Department of Health and Human Services' internal policy office published an analysis Tuesday that does two things at once: it confirms 2025 produced the lowest age-adjusted death rate ever recorded in the United States, and it shows that roughly half of the celebrated improvement is an artifact of how the federal government counts the population.

Once the Census Bureau's Vintage 2025 population estimates are applied to both years, the decline from 2024 to 2025 is 14.1 deaths per 100,000 people — about 2 percent — rather than the 4.6 percent in the headline comparison. What survives the correction is still the best year on record among the years the current population series covers, 2020 through 2025. And within that smaller, cleaner decline, drug overdoses account for nearly 22 percent of the total.

The brief, written by Stephen Murphy, Nicholas Holtkamp and Eric Sun at the Office of the Assistant Secretary for Planning and Evaluation, works from the National Center for Health Statistics' July 2026 report: 3,094,593 U.S. deaths in 2025 and a provisional age-adjusted rate of 689.2 per 100,000, below the final 2024 figure of 722.1.

Where the 4.6 percent goes

The arithmetic problem is a denominator problem. The final 2024 rate was calculated with Vintage 2024 population estimates, a blended base built from the 2020 census, the 2020 Demographic Analysis estimates and Vintage 2020 estimates. The provisional 2025 rate was calculated with Vintage 2025, which uses the enumerated 2020 census and the Census Bureau's Modified Age and Race Census file. The Census Bureau also revised its 2024 population estimates under the new methodology.

NCHS flagged the mismatch itself as a limitation. ASPE went further and rebuilt both years on the same footing. Holding 2024 deaths fixed but substituting the revised 2024 populations lowers the reconstructed 2024 rate from 722.1 to 703.9. Applying Vintage 2025 populations to the August 2026 provisional death counts puts 2025 at 689.8 — slightly above the official 689.2 because the agency's snapshot is three months fresher and has picked up a few more deaths.

The gap between those two exercises is the story: 14.1 deaths per 100,000 of real improvement, not the 32 implied by the published comparison. The report is careful to say neither official rate is wrong. Each was correct for the population estimates available when it was calculated.

What actually moved

On the common-vintage basis, unintentional injuries contributed the most to the decline, subtracting 4.23 deaths per 100,000 — about 30 percent of the net improvement. COVID-19 followed at 4.01 (28 percent), cancer at 2.55 (18 percent) and heart disease at 1.57 (11 percent). Those four causes account for roughly 87 percent of the total. Homicide and Alzheimer's disease also contributed, while rising influenza and pneumonia mortality offset 1.57 deaths per 100,000.

Inside the unintentional-injury category, the age-adjusted rate fell from 53.02 to 48.80 per 100,000, and drug overdose — underlying-cause codes X40 to X44 in CDC WONDER — accounted for 3.09 of the 4.23-point drop, or 73 percent of it. That 3.09 is the figure that becomes 22 percent of the all-cause decline.

Nearly all of the overdose improvement traces to one category: synthetic opioids other than methadone, the T40.4 multiple-cause code that covers fentanyl, fentanyl analogs and tramadol. That rate fell 2.98 per 100,000, roughly 96 percent of the net overdose decline, and the raw count of T40.4-involved unintentional overdose deaths dropped from 45,917 in 2024 to 36,406 in 2025, a decline of about 20.7 percent. Deaths involving another opioid but not T40.4 rose slightly. Non-opioid overdose deaths fell modestly.

Why the vintage matters for the rest of the table

The choice of denominator does not merely shrink the headline. It rearranges the causes. Under the mixed-vintage comparison, heart disease and cancer appear to be the two largest contributors to the decline, with a large residual left over for everything else. Under the common-vintage decomposition, unintentional injuries and COVID-19 take the top two slots.

ASPE notes one reason the overdose conclusion is more durable than the rest: overdose deaths cluster at younger and middle ages, where the Census revisions were smallest. The causes that moved most in the re-estimation were concentrated in the oldest age groups.

What the report declines to say

The brief separates description from explanation, and says so in a sentence that reads like a warning to anyone preparing to cite it: the analysis "helps identify specific causes driving the observed mortality decline between 2024 and 2025, but it does not explicitly identify why the decline occurred."

It then lists the candidate mechanisms without ranking them — changes in illicit drug availability, potency, contamination and market composition, alongside treatment-side factors such as overdose reversal agents, treatment initiation and retention, and recovery support. That is an honest inventory of hypotheses, not a finding.

HHS's press office nonetheless promoted the brief Tuesday under the headline "Trump Administration Drives Opioid Deaths to Lowest in Nearly a Decade." The analysis underneath attributes deaths to causes and age groups. It contains no policy variable, no comparison between states with different policies, and no test of any administration's actions. The claim in the headline is a political assertion sitting on top of a demographic decomposition that cannot support or refute it.

The same caution applies to the more modest version of the story. A 2 percent decline in age-adjusted mortality is a real improvement, and the fact that a single drug class explains almost all of the overdose component is a real finding about the fentanyl era. It is also compatible with the explanation that the synthetic opioid supply changed — which is not the same as saying the response worked, and not the same as saying it failed.

What the brief does establish is a measurement standard. Anyone citing the record-low death rate for 2025 now has to say which denominator they mean. The 4.6 percent number and the 2 percent number are both in circulation, and only one of them is a comparison between like things.

Deaths involving synthetic opioids remain concentrated among people who use drugs alone, and the treatment system that reaches them is still the same one described in earlier coverage of the 2025 overdose decline. Naloxone distribution and medication-assisted treatment for opioid use disorder are the two interventions the ASPE brief places on its list of unranked mechanisms. Programs and families looking for opioid rehab and treatment options can search by state through the site's directory.