National overdose deaths keep falling in 2025, but Arizona, New Mexico and Colorado counties are surging against the trend

Summary

Yes, the national decline is holding, and even accelerating in relative terms — but it is not holding everywhere. CDC's own provisional data show overdose deaths fell nationally for a third straight year: from about 81,313 in the 12 months ending in 2024 to about 69,973 in the 12 months ending in 2025, a roughly 14% drop, on top of the historic ~27% drop reported the year before. A parallel check against the CDC county-level file (summing the same 949 counties that report every year) shows the identical pattern: 70,415 → 61,021 deaths, a 13.3% decline — independent confirmation the national number is not an artifact of reporting changes. But the decline is not uniform. CDC's own release names Arizona, New Mexico and Colorado as the only states with double-digit increases in the latest year, and the county data pinpoint exactly where: Maricopa County, AZ (Phoenix) added 342 more deaths year-over-year — the largest absolute increase of any county in the country — while a cluster of New Mexico counties (Santa Fe, Rio Arriba, Sandoval, Taos) and Colorado's Front Range counties (El Paso/Colorado Springs, Denver, Arapahoe, Douglas, Weld) are all rising sharply, several by 45-90%.

The national trend: is last year's decline holding up?

CDC's National Center for Health Statistics reported that U.S. overdose deaths fell almost 27% in 2024 (to roughly 80,000, the largest one-year decline ever recorded), after peaking near 110,000 in 2022. The most recent CDC release (May 2026, covering 12-month-ending 2025 data) confirms the decline is continuing into a third year: an estimated 69,973 deaths in the 12 months ending in 2025, down about 14% from 81,313 the year before, with opioid-involved deaths falling from about 55,296 to 44,564. A January 2026 CDC update (data through August 2025) had already shown the same direction: about 73,000 deaths in the 12 months ending August 2025, down 21% from roughly 92,000 the prior year, with deaths down in all states except Arizona, Hawaii, Kansas, New Mexico and North Dakota.

As an independent check, summing CDC's own county-level VSRR file (health.cdc_county_overdose_deaths) across the 949 counties that report a non-suppressed count in every year from 2020 through 2025 reproduces the same shape: totals rose from 81,689 (2020) to a peak of 95,551 (2022), stayed elevated through 2023 (93,291), then fell sharply to 70,415 (2024) and again to 61,021 (2025) — a 13.3% year-over-year decline, closely matching CDC's own 14% national figure. This is a useful sanity check but not a substitute for CDC's total: the county file suppresses any county-month with 1-9 deaths, so this sum systematically excludes small, mostly rural counties and undercounts the true national total; it should be read as confirming the DIRECTION and rough MAGNITUDE of the trend, not as a replacement national figure.

Where it is not holding: the sharpest county-level increases

Two different questions hide inside "sharpest rise": which counties added the most deaths in absolute terms, and which counties rose fastest in percentage terms. Both cuts point to the same region.

Largest absolute increase, Dec-2024-ending vs Dec-2025-ending 12-month totals: Maricopa County, AZ (1,564 → 1,906, +342 deaths) is far ahead of any other county nationally — more than 4x the next-largest increase. It is followed by El Paso County, CO (+71), Denver County, CO (+52), Santa Fe County, NM (+42), Pima County, AZ (+35), Rio Arriba County, NM (+30), Sandoval County, NM (+24), Pinal County, AZ (+22), Tarrant County, TX (+21), and Arapahoe County, CO (+20).

Sharpest percentage increase (restricted to counties with at least 20 deaths in the prior year, to avoid extreme percentages driven by tiny base counts): Taos County, NM (20→38, +90%), Sandoval County, NM (28→52, +86%), Rio Arriba County, NM and Santa Fe County, NM (both +67%), Cass County, ND (+63%), Linn County, IA (+62%), Grayson County, TX (+59%), and Beaufort County, SC (+55%). New Mexico alone supplies four of the top eight fastest-rising counties.

This lines up exactly with CDC's own state-level finding that Arizona, New Mexico and Colorado are the three states with 10%+ increases against the national decline — the county data show that within those states, it is specific metro counties (Phoenix/Maricopa, Denver, Colorado Springs/El Paso) and a cluster of northern New Mexico counties (Santa Fe, Rio Arriba, Sandoval, Taos) driving it, not a uniform statewide rise.

What the county-level rankings above exclude

Both county rankings filter out CDC's privacy-suppressed cells (provisional_drug_overdose IS NOT NULL), and this materially shrinks the population being ranked. Of the 3,144 U.S. counties, CDC suppresses any county-month with a 12-month-ending count of 1-9 deaths: for the December 2025 period, 1,630 of 3,144 counties (52%) are suppressed and excluded from the ranking; for December 2024, 1,536 of 3,144 (49%) are suppressed. The counties dropped are overwhelmingly small, low-population, and rural — exactly the places where a swing from, say, 8 to 12 deaths would be a real 50% increase that this analysis cannot see or rank, because one or both endpoints falls in the suppressed range. This means the rankings above can only ever surface increases among counties large enough to consistently clear 10 deaths in a year; a genuine sharp percentage rise in a small rural county is systematically invisible in this data and is not reflected in either ranking. The 949-county national trend line is affected the same way: it is restricted to the ~30% of counties that clear suppression in every single year 2020-2025, so it undercounts the true national total (it does not attempt to estimate the suppressed deaths) even though it reproduces CDC's percentage trend closely.

Caveats

• All 2025 county and national figures are provisional and will be revised as more death records are finalized; CDC and STAT both note not all state records were fully reported as of the latest release, so any single state's or county's count could move as backfill continues.
• The percentage-rise ranking above is further restricted to counties with ≥20 deaths in the base year specifically to avoid noise from small numbers; a raw, unfiltered ranking of the already-suppression-filtered set would be dominated by small counties moving from, say, 11 to 22 deaths and registering as a huge but statistically fragile percentage jump.
• The 949-county national sum used here as a warehouse cross-check is a partial, suppression-filtered proxy for CDC's own official national estimate, not a replacement for it; it is presented only to confirm the direction and rough scale of the trend independently.

Every query behind this report

6 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT year, month, COUNT(*) as n, MAX(data_as_of) as data_as_of FROM health.cdc_county_overdose_deaths GROUP BY year, month ORDER BY year DESC, CAST(month AS INTEGER) DESC LIMIT 20

SELECT DISTINCT state_name FROM health.cdc_county_overdose_deaths WHERE state_name ILIKE '%united%' OR countyname ILIKE '%united%' FETCH FIRST 10 ROWS ONLY

WITH d2025 AS (
  SELECT fips, countyname, state_name, CAST(provisional_drug_overdose AS DOUBLE) AS cnt2025
  FROM health.cdc_county_overdose_deaths
  WHERE year='2025' AND month='12' AND provisional_drug_overdose IS NOT NULL
),
d2024 AS (
  SELECT fips, CAST(provisional_drug_overdose AS DOUBLE) AS cnt2024
  FROM health.cdc_county_overdose_deaths
  WHERE year='2024' AND month='12' AND provisional_drug_overdose IS NOT NULL
)
SELECT a.fips, a.countyname, a.state_name, b.cnt2024, a.cnt2025,
       (a.cnt2025 - b.cnt2024) AS abs_change,
       ROUND(100.0*(a.cnt2025-b.cnt2024)/b.cnt2024,1) AS pct_change
FROM d2025 a JOIN d2024 b ON a.fips=b.fips
WHERE b.cnt2024 >= 20
ORDER BY pct_change DESC
FETCH FIRST 15 ROWS ONLY

WITH d2025 AS (
  SELECT fips, countyname, state_name, CAST(provisional_drug_overdose AS DOUBLE) AS cnt2025
  FROM health.cdc_county_overdose_deaths
  WHERE year='2025' AND month='12' AND provisional_drug_overdose IS NOT NULL
),
d2024 AS (
  SELECT fips, CAST(provisional_drug_overdose AS DOUBLE) AS cnt2024
  FROM health.cdc_county_overdose_deaths
  WHERE year='2024' AND month='12' AND provisional_drug_overdose IS NOT NULL
)
SELECT a.fips, a.countyname, a.state_name, b.cnt2024, a.cnt2025,
       (a.cnt2025 - b.cnt2024) AS abs_change
FROM d2025 a JOIN d2024 b ON a.fips=b.fips
ORDER BY abs_change DESC
FETCH FIRST 15 ROWS ONLY

WITH base AS (
  SELECT fips, year, CAST(provisional_drug_overdose AS DOUBLE) AS cnt
  FROM health.cdc_county_overdose_deaths
  WHERE month='12' AND provisional_drug_overdose IS NOT NULL
),
counties_all_years AS (
  SELECT fips FROM base GROUP BY fips HAVING COUNT(DISTINCT year)=6
)
SELECT b.year, COUNT(*) AS n_counties, SUM(b.cnt) AS total_deaths
FROM base b JOIN counties_all_years c ON b.fips=c.fips
GROUP BY b.year ORDER BY b.year

SELECT year, COUNT(*) AS total, SUM(CASE WHEN provisional_drug_overdose IS NULL THEN 1 ELSE 0 END) AS suppressed
FROM health.cdc_county_overdose_deaths WHERE month='12' AND year IN ('2024','2025') GROUP BY year