America’s Drug-Death Crisis Began With Chronic-Pain Prescribing

The American drug-death crisis, as Steven Davis framed it, is not a marginal deterioration. The death rate from drug misuse in the United States rose tenfold from 1979 to 2022, with the sharpest increase concentrated in the 21st century. William Evans described a long, fairly linear rise from the mid-1970s through the late 1990s, followed by a much steeper escalation after 2000.

The recent history has three phases. The first was the rise in abuse of prescription opioids, typified by OxyContin. Evans emphasized that OxyContin was a prescription opiate with an extended-release formulation and high milligram content. Its abuse potential came partly from the fact that users could defeat the extended-release feature by crushing the drug and snorting it. The second phase was a shift to heroin. The third was the emergence of fentanyl as the dominant driver of deaths.

In 1980, Evans said, roughly 6,700 people were dying each year from drug poisonings. By 2021, the United States had reached 100,000 deaths in a year associated with drug poisonings. The number peaked, in Evans’s correction, at about 107,000 deaths in 2022, then declined somewhat in 2024 and appeared likely to be lower again in 2025. Davis treated 100,000 deaths per year as the round number to keep in mind for the recent scale of the problem.

The change in substances matters because it changes what a policy response can plausibly do. In the first phase, the supply was tied to the licit medical system: physicians, prescriptions, pharmaceutical promotion, regulatory boards, and patients who trusted medical authority. After 2010, the crisis moved increasingly into illicit markets. Ethan Lieber pointed to the reformulation of OxyContin in 2010, which made the drug harder to abuse. That reformulation had an unintended consequence: some users substituted toward heroin, which was relatively cheap and available.

Davis called this a classic case of unintended consequences from a well-meaning policy change. Lieber agreed, adding that a large part of the supply before then had come from the licit market; after the reformulation, the problem moved much more decisively to the illicit side.

Fentanyl changed the economics again. Evans said fentanyl had been available on the dark web, but Mexican transnational drug organizations were able to import precursor chemicals, manufacture it in large quantities, and ship it across the border. Lieber emphasized the potency: fentanyl is far stronger than OxyContin and many prescription opioids, so a small physical quantity can represent a large effective supply. It is cheap to produce and easier to traffic than bulkier drugs.

That potency makes fentanyl a different supply-control problem from older agricultural drug markets. Davis contrasted it with opium: destroying poppy fields can at least temporarily reduce supply. Evans agreed that fentanyl is “a very different animal” because of the relative ease with which precursors can be obtained and turned into the drug.

The story therefore begins with prescription practices, but it does not end there. Evans and Lieber’s central account is that American medicine helped ignite the crisis, but the resulting market is now much harder to extinguish because the deaths are driven largely by illicit fentanyl and by fentanyl contamination in other drugs.

The existing explanations for the drug-death rise include both demand-side and supply-side accounts. Evans began with the economists’ basic distinction: demand and supply.

On the demand side, the geography and demographics of early opioid deaths seem to fit a “deaths of despair” interpretation. Drug deaths were concentrated in places such as West Virginia, western Pennsylvania, eastern Ohio, and northern Kentucky. They were also concentrated among people with low education and among groups that had lost ground in the labor market. Evans acknowledged that these stories have validity: damaged labor markets and weakened institutions can encourage greater use of drugs or alcohol.

But he did not think demand-side stories alone could explain the magnitude of the increase. Lieber identified the relevant framework as the Case and Deaton “deaths of despair” story. Davis clarified the economic meaning of demand in this setting: demand could mean demand for pain management, or demand for the effects of opiates more generally. He distinguished that from a narrower claim that the underlying need for pain treatment had risen because work had become more physically damaging. Davis was skeptical of that version, noting that the economy had moved away from many physically demanding manual labor jobs that might create severe injuries requiring pain management. The more plausible demand-side story was despair, not a sudden national increase in injury-related need.

Evans and Lieber’s research does not deny demand. It gives demand a more specific role. Their account is that some places had more latent demand for pain relief before the prescribing shift, and those places were especially vulnerable once the supply-side environment changed.

Davis used Evans’s metaphor of “embers.” Some counties had many more embers than others: more underlying pain, more people likely to receive or seek treatment if physicians became more willing to prescribe opioids. The later shift in prescribing practice was the fuel.

That distinction is central. In their account, the early crisis was not simply produced by worsening economic conditions that made people more likely to use drugs. It was produced by an interaction between latent pain demand and a national medical change that made opioids more available for chronic pain.

The key empirical move in Evans and Lieber’s paper, as Lieber explained it, is to use county-level Social Security Disability Insurance rates around 1990 as a pre-crisis measure that predicts later opioid death rates. SSDI, Davis clarified, means people in a county were collecting disability benefits through the Social Security Disability Insurance program, having been identified through medical or related processes as seriously disabled.

Lieber said the 1990 SSDI rate in a county is “extremely predictive” of future drug death rates. High-SSDI counties later saw an explosion in opioid death rates relative to low-SSDI counties.

The timing matters. Davis emphasized that 1990 predates the change in prescribing practices that came in the second half of the 1990s. That makes the 1990 SSDI rate useful as a baseline measure rather than an outcome of the opioid prescribing shift itself. Lieber agreed with Davis’s formulation: the SSDI rate is treated as an indicator of underlying baseline demand for pain management.

The next question is what the SSDI rate is actually capturing. Lieber said the paper examines multiple possible explanations for why counties with higher disability rates would later have higher opioid deaths. Local labor market conditions are one possibility. Access to health facilities is another. The authors consider a range of alternatives. Lieber’s summary was that the factor that stands out is pain: the SSDI rate appears to be a good proxy for how much pain exists in the county.

They also compare opioid deaths with other “deaths of despair” outcomes. Lieber said the paper looks at whether county differences in SSDI rates explain later non-drug suicides and alcohol deaths. The same pattern does not appear. SSDI differences are tightly connected to future opioid problems, but not to those other outcomes in the same way.

That comparison matters because it weakens a pure despair interpretation. If high-SSDI counties were simply places of broader despair, one might expect the same divergence in alcohol deaths or non-drug suicides. Lieber’s claim is that the strong divergence is specific to opioids.

The research then connects the county-level pattern to changes in treatment. Lieber described a patient-level thought experiment: take a patient with chronic back pain in 1980 and ask the probability that the patient leaves the doctor’s office with an opioid prescription. Then ask the same question in 1985, 1990, and later years. From 1980 through the late 1990s, he said, the probability is roughly flat: a small fraction of patients with chronic pain get an opioid prescription. Then, in the late 1990s, the rate takes off, roughly doubling or tripling over the following decade.

Davis interpreted this as evidence that professional guidance and medical reevaluation were actually manifested in prescribing behavior. Lieber confirmed that the analysis controls for underlying health conditions and demographic characteristics. The claim is not merely that more opioids were prescribed over time, but that similarly situated patients with chronic pain became substantially more likely to receive opioids after the shift in medical practice.

The empirical chain, as presented, has two major links. First, counties with more baseline disability — interpreted after testing alternatives as more pain — later experienced larger opioid death increases. Second, doctors changed how they treated pain patients, especially in the late 1990s, making opioid prescriptions much more common for chronic pain. Together, these findings support the conclusion that the medical shift in prescribing practices played a key role in the early trajectory of the crisis.

Evans did not present the change in prescribing practices as the entire story. He described it as additive to the earlier work he and others had done on OxyContin’s role. Doctors were prescribing more opioids, and in some areas they had a more abuse-prone product to prescribe.

The underlying medical reevaluation preceded or accompanied OxyContin. In the late 1980s and early 1990s, Evans said, opioids were typically used for acute pain, severe burns, post-surgical pain, cancer patients, and people near the end of life. They were not commonly used for chronic back pain, neck pain, joint pain, or similar long-term conditions. But a view emerged that there was an “epidemic of untreated pain” and that prescription opioids could be an effective way to treat it.

OxyContin then arrived into that changed environment. Evans said it promised less risk of addiction and the convenience of a twice-daily pill because of its extended-release design. In his phrasing, the changed prescribing environment set “embers,” and OxyContin threw “gasoline” on the problem.

The promise of lower addiction risk proved badly wrong. Davis summarized the sequence this way: there was an underlying unmet demand for chronic pain management, OxyContin and similar extended-release drugs appeared to offer a less addictive way to manage that pain, and the second part of the claim was “seriously wrong.” Evans agreed.

The mechanism of persuasion involved pharmaceutical detailing and physician trust in presented evidence. Lieber said pharmaceutical firms conducted studies — the quality of which could be debated — and presented them to doctors as evidence that opioids were not nearly as risky as previously thought. The message to doctors was direct: patients were in pain, and the firms had drugs they said could help.

Evans gave a specific example involving Purdue Pharma and OxyContin. Detail representatives cited a New England Journal of Medicine item by Porter and Jick as evidence that the addiction risk from receiving opioids was low. But Evans said the cited item was not a full study of chronic pain patients. It was a roughly 100-word letter to the editor about post-surgical patients in a hospital setting who received opioids during their hospital stay. That, Evans emphasized, was not analogous to prescribing opioids to people with chronic pain.

Davis identified two failures in that example. Purdue was selling the product in a deceptive way by using evidence not suited to the claim. Doctors also failed, in many cases, to investigate the cited evidence and recognize that it was not a real chronic-pain study. Evans agreed that many physicians said they had been provided irrelevant information, but many also did not follow up to evaluate it.

Evans also pointed to two large medical organizations devoted to pain that helped lead the reevaluation of opioids for chronic pain. Much of their funding, he said, came from pharmaceutical companies, and they later became defendants in opioid-related court cases. In some cases, doctors were duped; in others, Evans suggested, they failed to read the science being put in front of them.

Davis stressed the broader trust problem. Many patients trust their doctors precisely because they are not experts in pharmacology, addiction, or chronic pain treatment. The argument was not that doctors were necessarily malicious. It was that misplaced trust in a professional system contributed to a grave failure whose consequences have lasted for decades.

The prescribing shift required more than physician persuasion. Evans said state medical boards were an important background condition. In the early 1990s, doctors who did not prescribe opioids for chronic pain could reasonably say they feared their state medical board would take away their license because such prescribing was not an approved treatment for those patients.

That means the change in prescribing practices required a change in the regulatory environment. State medical boards had to relax their view of how chronic pain could be treated. Evans said the “epidemic of untreated pain” was in part driven by doctors’ fear of medical boards; once boards became more accepting of opioid treatment for chronic pain, the practical constraint weakened.

Davis asked whether medical boards were influenced by the same professional associations and pharmaceutical narratives that affected doctors — especially the claim that drugs such as OxyContin were not as addictive as previously feared. Evans said surveys of board members suggest they were heavily influenced by the notion that addiction risks had been overstated.

Davis connected this to a separate discussion he had previously had with Rebecca Allensworth about medical boards and occupational licensing. The relevant point, as Davis summarized it, is that medical boards are often staffed by practicing physicians who have full-time jobs elsewhere. Board service is not their main occupation, and they may not be experts in addiction or in assessing the addictive properties of a particular drug. Davis characterized that as another weakness in the American regulatory apparatus.

A book cover for Rebecca Haw Allensworth’s The Licensing Racket: How We Decide Who Is Allowed to Work, & Why It Goes Wrong was shown while Davis made this point.

The regulatory piece matters because it complicates any simple villain story. Pharmaceutical firms had commercial incentives. Some evidence was misleadingly presented. Some professional organizations had funding conflicts. Some doctors failed to scrutinize evidence. But state medical boards also altered the boundary of acceptable practice, allowing chronic-pain opioid prescribing to become normal rather than professionally dangerous.

Evans said there has been recognition within medicine that the profession contributed to the problem. He cited the closing line of their paper, which refers to the Journal of the American Medical Association saying, in effect, that medicine contributed to the problem and has to be part of the solution. Evans said doctors need better training, especially for patients who will use opioids for long periods.

But Davis pressed the crucial limitation: even if prescribing practices returned to their pre-mid-1990s pattern, that would not solve the current crisis. Lieber agreed. The crisis has moved into illicit markets. Changing what doctors do now may matter at the margin, especially over a long horizon by reducing the number of people who transition from medical opioid exposure to illicit use. But most current deaths are from illicit opioids, not from a doctor’s decision after a minor surgery or a back-pain visit.

Evans described the current crisis as a poisoning epidemic. Fentanyl is now driving deaths. Some people use fentanyl as their drug of choice, but a large share of deaths involve people taking cocaine or methamphetamines that are laced with fentanyl. Whether the contamination is intentional or unintentional is uncertain, he said. The result is that people may believe they are taking cocaine or meth and instead ingest a drug contaminated with fentanyl.

That means the policy problem has changed. The crisis started in legal prescription medicine and then morphed into the abuse, misuse, or accidental poisoning associated with illicit drugs. Fixing prescription practices can address the original channel, but it cannot by itself unwind the market and mortality pattern now in place.

Davis described the conclusion as depressing: the United States largely created the problem through domestic medical and regulatory choices, and even fixing those choices now does not offer an obvious way out. Evans agreed with that assessment.

The one positive note Evans offered was that if supply can be changed, deaths can fall. He pointed to the decline after the 2022 peak and referenced a Science article by long-time drug-crisis researchers arguing that the recent drop may reflect a reduction in fentanyl supply. Border seizures, he said, have fallen even with continued aggressive enforcement, suggesting some kind of supply disruption.

Davis was skeptical about relying on that as a long-term answer. Fentanyl precursors are not hard to obtain or manufacture, and fentanyl’s potency makes it easy to move large effective quantities in small volumes. He accepted that disrupting supply can reduce availability and raise price, but called indefinite suppression “an enormous challenge.” Evans did not dispute that. His narrower claim was that if supply can be stopped or reduced, deaths appear to fall. Davis replied that “if” is doing a great deal of work.

That skepticism led Davis toward demand-side policies of a different kind from the despair explanation: education and persuasion about risk, especially the risk that a person intending to take cocaine may instead take cocaine laced with fentanyl. He contrasted that with decades of supply-side drug-control strategies, from versions of the war on drugs to “just say no,” which he did not see as a viable long-term strategy by themselves.

Evans said other countries have had opioid problems, but the prescription-drug addiction problem did not appear in most other places in the same way. Davis drew out the implication: the U.S. mortality experience has been unfavorable relative to other rich countries in recent decades, and drug deaths are a material part of that story, especially for middle-aged men with low education.

Evans said that for middle-aged men, particularly low-educated men, deaths of despair — alcohol-related deaths, suicides, and drug overdoses — are a large component of the divergence from other countries. Davis added that some suicides may themselves be downstream of addiction, creating another connection between drug misuse and broader mortality outcomes.

The policy implication Davis emphasized is diagnostic. If the United States misidentifies the source of the crisis, it will likely choose ineffective responses. His summary was that the problem is largely homegrown and American. Evans agreed.

That does not mean there were no nefarious actors, especially in illicit supply. But Davis argued that illicit suppliers are not the main origin of the problem. Treating them as the main source misses the central sequence: a medical and regulatory system changed its understanding of opioids for chronic pain; pharmaceutical firms promoted drugs under claims that minimized addiction risk; professional and regulatory institutions accepted that reevaluation; doctors prescribed more; patients in high-pain areas were especially exposed; and the market later shifted into heroin and fentanyl.

Evans ended by pushing back on the idea that economics is merely “the dismal science.” In his view, the tools of economics — supply and demand, market incentives, and empirical variation across places — are powerful precisely because they can illuminate problems that might seem outside economics. Davis accepted the point: the value of the analysis is not that it makes the story less grim, but that it clarifies how the United States got into the crisis and what kinds of responses are unlikely to be enough.