Apparent Drug Overdose Spike In 2025 Driven By Flawed Modeling

Last June, respective mainstream media outlets reported a surge successful U.S. supplier overdose deaths successful early 2025 that was based connected information from nan U.S. Centers for Disease Control and Prevention (CDC). 

A caller Northwestern University study has recovered that surge was not real, and nan evident spike was a statistical modeling artifact, not a existent reversal successful overdose trends. In fact, nan study recovered overdose deaths person continued to diminution pursuing a highest successful August 2023, marking nan longest sustained alteration successful much than 4 decades. 

The findings straight reside speculation that nan CDC made information reporting errors aliases intentionally mischaracterized aliases concealed nationalist wellness information owed to governmental unit aliases interference, said lead author Lori Ann Post.

"Many group deliberation CDC supplier overdose information are being cooked, but they're not," said Post, head of nan Buehler Center for Health Policy and Economics astatine Northwestern University Feinberg School of Medicine. "We tin spot them because they're scientists trying to do nan champion occupation they tin pinch difficult circumstances. There was nary clear inducement for immoderate management to inflate these numbers. This was not politics."

The study was published coming (April 8) in the American Journal of Public Health.

What happened?

To hole nan June 2025 report, CDC scientists utilized information from anterior years (2022 and 2023) erstwhile location was a monolithic acceleration of supplier overdose deaths, Post said. Because of that, provisional national information suggested overdose deaths had risen sharply successful January 2025, prompting nationalist interest and media coverage. 

"That spike made headlines astatine nan time, but it didn't bespeak reality," Post said "What we recovered was a mismatch betwixt predictive models and a quickly changing epidemic. CDC scientists did nan champion occupation they could pinch less people, much constraints and much group watching them."

Subsequent revisions to national estimates corrected nan overestimate and confirmed that overdose deaths had continued to fall, Post said. 

The U.S. relies connected provisional mortality estimates from nan National Center for Health Statistics, which uses statistical models to relationship for reporting delays successful decease investigations. These models performed good for years but struggled erstwhile overdose deaths began declining aft a prolonged play of accelerated growth, Post said. 

The study authors stress nan section highlights a broader issue: surveillance systems are astir susceptible during turning points, erstwhile trends displacement direction.

Why it matters

Accurate information are basal for nationalist wellness response. Misinterpreting trends tin misdirect argumentation decisions, undermine nationalist spot and distort assets allocation, Post said. 

Despite this episode, nan researchers accent that national mortality information stay nan astir reliable adjacent real-time root for search overdose death.

Why nan models broke

The overdose situation has been reshaped by fentanyl, a synthetic opioid that dispersed unevenly crossed nan state and drove accelerated increases successful deaths done 2022 and 2023.

Models trained connected that explosive maturation were later applied to a play of decline. The result:

• Overestimation of deaths successful early 2024 and 2025
• False signals of a nationalist "spike"
• Confusion among policymakers, researchers and nan public

Looking ahead

The authors telephone for greater transparency successful national information systems, including beforehand announcement of methodological changes and clear archiving of revisions.

"When nan numbers change, group notice," Post said. "We request to make judge they understand why."

More astir nan study

The scientists utilized Northwestern's national, location and authorities dashboard, the OD Pulse, which is designed to way supplier overdose deaths betwixt January 1999 to April 2025, pinch unprecedented timeliness and precision. They analyzed nationalist overdose decease information from 2015 done early 2025, comparing observed counts pinch aggregate versions of national predictive estimates.

The insubstantial is titled, "The 2025 Drug Overdose Spike That Wasn't: Neither Politics nor Data Errors Explain nan Anomaly." Co-authors see researchers from University of California, San Francisco, University of Maryland, Baltimore, and University of North Carolina astatine Chapel Hill.