In May 2025, the Centers for Disease Control and Prevention announced that drug overdose deaths had fallen 27%, from roughly 110,000 in 2023 to about 80,000 in 2024. Provisional data through late 2025 projects a further drop to around 72,000.
As an addiction researcher at Stanford, and someone in long-term recovery, I felt the relief of that news personally. I also watched as my colleagues exhaled for the first time in years.
After two decades of relentless escalation, the overdose curve had finally bent downward. Policymakers are celebrating and calling it “unprecedented progress.” The CDC is framing it as saving “more than 81 lives every day.”
They are both right. More than 81 lives a day — that is real, and it matters. But it should give us reason to pause.
Here’s what we’re celebrating: a death toll that exceeds total American combat fatalities in Vietnam, every single year. In 2015, when overdose deaths first topped 50,000, the shameful milestone was treated as a national wake-up call. A decade later, 50,000 deaths is now an aspirational target. That shift in expectations is a warning sign: The crisis isn’t getting better — we’re just getting more used to it.
This exact pattern has shown up in historical trends of substance use before. I call it the “stable floor.” This is the moment when a death toll stops being treated as an emergency and starts being treated as the cost of doing business in modern society.
In 1982, when federal tracking began, roughly 21,000 Americans died in alcohol-impaired crashes. This tragic total was fueled by a combination of lowered state drinking ages, a culture that normalized driving while intoxicated, and a lack of modern vehicle safety features. When people heard the number for the first time, the response was swift, massive, and furious: Mothers Against Drunk Driving mobilized public outrage, Congress raised the drinking age to 21, states adopted 0.08 blood alcohol content laws, and sobriety checkpoints became standard. Deaths dropped nearly 50% by the mid-1990s.
Then something quietly happened that most people ignored: Progress on lowering drunk driving deaths stalled. Once the new policies were in place, pushing the death toll lower would have required deeply unpopular structural changes, such as massively expanding public transit, sharply raising alcohol taxes, or mandating ignition interlocks in every vehicle. For political reasons, these changes were never made even though there was significant evidence that they would work.
For the past three decades, between 10,000 and 13,000 Americans have died annually in drunk driving crashes. People don’t hold candlelight vigils for these deaths anymore. We’ve absorbed them into the background calculus of a society that drinks and drives.
That is exactly what could happen with drug overdose deaths if we allow it. In 2017, the president declared a public health emergency when the toll hit roughly 70,000. In 2021, when it surpassed 100,000, there was renewed outcry, congressional hearings, and emergency funding. Now that deaths have returned to the range that first triggered the emergency, the tone has overwhelmingly shifted from alarm to relief. We are now benchmarking our success against the worst years instead of against any principled standard of how many deaths are tolerable.
The early warning signs of a stable floor appearing, and normalization approaching, are flashing red. The rate of decline is decelerating. After the 27% drop in overdose deaths in 2024, provisional data for 2025 shows roughly a 19% year-over-year decline, with several states actually reporting increases.
Most alarmingly, policymakers are already dismantling the very infrastructure driving this decline. Last year, the White House withheld roughly $140 million in CDC grants dedicated to local overdose tracking and prevention, followed by staffing cuts at the agency’s injury prevention center.
Now, proposed federal budget cuts for fiscal year 2026 threaten to make deeper cuts to the CDC and the Substance Abuse and Mental Health Services Administration. That policymakers would defund overdose surveillance, treatment, and harm reduction at precisely the moment these programs are demonstrating effectiveness is itself evidence of normalization.
Some will argue that concerns about stagnation are premature when numbers are still falling. But this objection misunderstands how normalization works. Normalization and the emergence of a stable floor don’t begin when progress stops. They begin when progress becomes the justification for disengagement in funding and the refusal to make the difficult choices to put an end to the crisis. The drunk driving movement failed at precisely the moment when success was redefined downward until 10,000 annual deaths became the accepted price of mobility.
The technologies currently driving the overdose decline — such as widespread naloxone distribution, expanded access to medications for opioid use disorder, fentanyl test strips, and data-driven surveillance — are working. But they work only as long as they are funded, distributed, and politically defended. It remains to be seen whether this welcomed decline in overdose deaths prompts political complacency or sustained investment.
The question before us is not whether 72,000 deaths is better than 110,000. Of course it is. The question is whether 72,000 deaths is acceptable to us as a society. And if not 72,000, will 52,000 be acceptable? Because until we confront that question directly, the answer will be written for us, incrementally, in the same quiet arithmetic that has kept drunk driving fatalities at 10,000 a year for over a generation.
Wayne Kepner, Ph.D., M.P.H., is a postdoctoral fellow in the Department of Psychiatry and Behavioral Sciences at Stanford University School of Medicine. His research focuses on addiction health services and policy.
