I. The compound noun
We speak of “the opioid crisis” as though it were one thing. It is at least four, and the habit of naming them with a single phrase has cost us more than any individual policy error of the last thirty years.
There is a crisis of dependence originating in medical treatment. There is a separate crisis of mortality in an illicit market that long ago stopped depending on prescriptions for its supply. There is a crisis of undertreatment among patients in durable pain who are now taxed for the sins of a marketing campaign they never saw. And there is a global crisis of absence, in which most of humanity cannot obtain the cheapest opioid in existence for the purpose of dying without screaming. These four have different causes, different victims, and different remedies. Policies built for one have been applied, with predictable results, to the others.
I want to attempt a disentangling here, and I want to be honest at the outset about the position I am writing from, because it determines both what I can see and what I cannot.
II. The observer’s position
I have been offered opioid analgesics twice in my life: after wisdom teeth extraction as a teenager, and after elbow surgery last year. In both cases I used a few of them, briefly, and stopped without difficulty or regret. This is, statistically, the ordinary experience. It is also nearly useless as a credential.
It is useless in the permissive direction because it is precisely the intuition that made the expansion of the 1990s feel safe. If you reason outward from the median patient — who takes two tablets, dislikes the fog, and leaves the rest in a drawer — liberal prescribing looks harmless, and the warnings look like fussiness. The risk in opioid exposure is not distributed evenly across a population; it is concentrated in a minority who cannot be reliably picked out in advance. A policy calibrated to the median will produce a small percentage of catastrophes across an enormous denominator, and the median patient will never see one of them.
It is equally useless in the punitive direction, and this is the more common failure among people situated as I am. “I took two and stopped; why can’t they” is the engine of a great deal of contempt, and it rests on a false premise — that the person who cannot stop is exercising a weaker version of the same will I exercised. Sustained exposure reorganizes the systems that make stopping possible. My two prescriptions tell me nothing whatever about what it is like to be four months into a regimen a body has rebuilt itself around. I am in the position of a man who has crossed a stream twice at low water offering an opinion about drowning.
What moderate and infrequent use does confer is a kind of distance, and distance permits observation. I have no stake in vindicating my own consumption and no wound driving me to defend or condemn anyone else’s. That is a modest advantage, and I will try to spend it carefully.
III. The thesis: three offices in one act
The central structural fact about opioids in modern medicine is this. A single act — the writing of a prescription — is simultaneously the discharge of three incompatible offices.
The first is clinical: the physician relieves suffering, believing the patient’s report, because pain has no biomarker and the report is the whole of the evidence.
The second is allocative: the physician rations a scarce and hazardous good, weighing this patient’s present relief against that patient’s future harm, and against the harm to third parties into whose hands the tablets may pass.
The third is custodial, and it is a police function: the physician stands as the state’s checkpoint against diversion, evaluated by licensing boards, monitoring databases, and federal enforcement on how skeptically she treats the very testimony her first office obliges her to believe.
No other therapeutic decision is structured this way. Nobody audits an antibiotic prescription for evidence of criminal intent. And these three offices cannot be optimized together, because the first requires trust, the second requires prognosis, and the third requires suspicion. The oscillations of the past three decades are best read not as a series of mistakes but as a system alternately maximizing one office at the expense of the other two, discovering the cost, and reversing.
The four experiences that prompted this essay are useful precisely because each isolates one of the offices, in something close to laboratory purity.
IV. The clinical office, working invisibly
My own two prescriptions isolate the first office operating normally. A surgeon caused a controlled injury; the injury hurt; a drug was provided; the drug worked; the drug was discontinued. The transaction was unremarkable in every particular, which is exactly why it is analytically important.
The ordinary case generates no data. It produces no overdose record, no monitoring-program flag, no lawsuit, no newspaper account, no grieving family willing to speak on camera. It is invisible to the entire apparatus that generates our knowledge of the subject. Consequently, every public representation of opioid analgesia is drawn from a sample constructed entirely of its failures.
This is not a conspiracy; it is the structure of surveillance. But it means that when someone argues that opioids are, for most people most of the time, a boring and effective tool, the argument has almost no evidentiary constituency, while the argument that every prescription is the opening move of a tragedy has a documented body count behind it. Any fair analysis has to correct for that asymmetry deliberately, because nothing in the data will correct for it automatically.
V. The allocative office, suspended
My mother told me that a close friend of ours had died of terminal cancer in Florida, and that he had been in considerable pain until his release to hospice, where palliative dosing finally ended it.
This isolates the second office — and it isolates it by showing us what happens when the office is switched off. Rationing under uncertainty is justified by future harm: dependence, escalation, overdose, a life reorganized around a molecule. When the future shortens to weeks, the entire cost side of the ledger goes to zero. Relief becomes free, and is therefore granted without argument.
The logic is coherent. But notice what it implies about the rest of the system. We are not, in the ordinary case, rationing according to the severity of suffering. We are rationing according to how much future the sufferer has. A man with three weeks left receives whatever it takes. A woman with thirty years of degenerative spinal disease and comparable daily pain receives a lecture about physical therapy and a query against a state database. Her suffering is not smaller. It is longer — and length is precisely what disqualifies her.
The defense of that arrangement deserves a fair hearing, because it is not made in bad faith. The evidence for long-term opioid therapy in chronic non-malignant pain is genuinely thin; tolerance is real; dose escalation without functional gain is a common trajectory; and a treatment that is unambiguously good across three weeks may be net harmful across three decades. All of that can be argued honestly. What cannot be argued honestly is the sleight of hand by which a contested empirical claim about long-run efficacy gets delivered to the patient as a moral verdict about who deserves mercy. Chronic pain patients hear the moral verdict clearly, because it is the part that is actually communicated.
There is a further irony in the geography. Florida in 2010 was the national wholesale source of diverted oxycodone: by that year, 90 of the 100 physicians purchasing the most oxycodone in the United States were in Florida, dispensing largely for cash outside the scope of standard practice. The state responded with laws in 2010 and 2011 restricting dispensing at the site of care, alongside federal enforcement operations that arrested dozens of clinic operators and suspended scores of registrations. The crackdown worked on its own terms — researchers estimated overdose mortality rates 3.0 per 100,000 lower by 2012 than the counterfactual trend predicted — and it worked without producing the heroin substitution that a comparison state experienced, whose heroin overdose rate quadrupled from early 2011 to late 2012 while Florida’s rose far more slowly.
I record this because a fair-minded paper cannot pretend that restriction never works. It plainly did there. And yet the same state, fifteen years later, is where a dying man waited months for adequate relief. The tightening that shut the pill mills and the tightening that left him in pain were not separate policies. They were the same policy, encountering two very different patients, and the machinery had no way to tell them apart.
VI. The custodial office, unprotected
My friend Parisa, who worked as a nurse in South Tehran, describes something Americans rarely picture: patients with addictions becoming violent toward the clinical staff who refused them morphine.
This isolates the third office and strips it of every institutional cushion. When a controlled substance is scarce, legally rationed, and physically located behind a nurse, that nurse is the state’s refusal, embodied, at three in the morning, in a corridor. The custodial function has been assigned to a person whose training, vocation, and entire professional identity are organized around not refusing.
Iran is an instructive case rather than an exotic one. It borders the world’s principal opium-producing region, carries one of the higher burdens of opioid dependence anywhere, and — unusually for its neighborhood — adopted methadone maintenance and needle exchange relatively early on frankly pragmatic grounds. What Parisa describes is not the failure of an alien system. It is what the custodial office looks like when the buffer between refusal and desperation is one human body.
The American version is subtler and, in its way, more corrosive, because the violence is done to the clinical encounter rather than to the clinician. Once the patient understands that describing pain accurately is a strategic act, he begins to calibrate: overstating to clear the threshold, or understating to avoid the flag. Once the physician understands that she will be audited on her credulity, she begins reading the encounter for signs of performance. Both parties are now managing each other rather than the illness. The working material of medicine is testimony, and we have introduced a systematic incentive to corrupt it on both sides.
The profession’s whiplash follows from this and deserves more sympathy than it receives. Physicians spent one decade being graded on patient satisfaction scores that included pain management, and the following decade being investigated for the prescribing those scores rewarded. It is not obvious what a conscientious practitioner was supposed to have done.
VII. When all three offices are bought
The fourth window is the oldest and the strangest: a friend’s devotion, in the 1990s, to Blood on the Dance Floor, and particularly to “Morphine.”
The song is built on a hard alternation — an abrasive, percussive outer section against a soft, floating middle in which a person is administered Demerol. The arrangement makes the pharmacological case more precisely than any lyric could. The assault stops. The room goes quiet. Then the assault resumes. Jackson had publicly acknowledged dependence on analgesics in 1993, cancelling a tour and entering treatment. He released the song in 1997. He died in 2009 of an anesthetic administered in his bedroom by a physician he employed, who was afterward convicted of involuntary manslaughter.
This case is analytically valuable because it inverts the standard narrative completely. The public story about opioids is a story about access being too easy in the aggregate. Jackson’s case is a story about what happens when the patient is wealthy enough to purchase all three offices at once. The clinician’s independent judgment, the rationer’s arithmetic, and the constable’s suspicion were not overwhelmed; they were hired. Prince and Tom Petty are variations on the theme.
The gatekeeping model, then, fails at both ends of the distribution. The unimpressive and the poor get the custodial office and very little of the clinical one. The rich get servants and none of the custodial office at all. What almost nobody gets is a trained professional exercising judgment that is genuinely her own.
VIII. The evidence of the pendulum
The historical record supports the three-office reading better than it supports any villain narrative.
For most of the twentieth century, American medicine had a documented undertreatment problem severe enough to earn a name in the literature. Its least defensible expression was the undermedication of the dying, which is the specific abuse the modern hospice movement was founded to correct. Cicely Saunders’ contribution was not pharmacological but administrative: treat pain on a schedule rather than on demand, before its return rather than after, in amounts sufficient to stop it. The World Health Organization’s analgesic ladder codified the approach in 1986. My family friend’s release into hospice is that reform functioning as designed, four decades on.
The reform then generalized, and generalization is where it broke. If the dying are undertreated, the reasoning ran, everyone else must be too. Pain became “the fifth vital sign.” Accreditation bodies scored institutions on it. A single paragraph published as a letter in 1980 — concerning hospitalized inpatients, not outpatients on sustained therapy — was cited hundreds of times as evidence that iatrogenic addiction was negligible. OxyContin arrived in 1996 with a campaign resting heavily on that claim. Per capita opioid prescribing roughly tripled between 1999 and 2015; total prescriptions peaked in 2012 at about 255 million, or 81.3 per 100 persons.
The return swing came in 2016. The CDC’s prescribing guideline was written as guidance, its numerical thresholds intended as prompts for conversation. Within months they had been hardened into statute, insurer caps, pharmacy refusal policies, and software ceilings. Patients stable for a decade were tapered involuntarily or discharged from practices. The agency issued a clarifying letter in 2019 stating that its guideline was being misapplied, and rewrote it in 2022 specifically to strip out the numbers that had been converted into rules. Dispensing fell from 260.5 million prescriptions in 2012 to 125.7 million in 2024, with total morphine-milligram-equivalents down about 65%.
And here is the finding that should discipline everyone’s confidence, in both directions. Prescribing was cut in half, and overdose mortality rose through most of that period — from roughly 69,700 deaths in 2018 to about 110,000 in 2021 and 2022. The market had already relocated. The proportion of opioid deaths involving fentanyl went from 16.4% in 2012 to 72.9% in 2019. Pressure on the licit channel did not reduce demand; it moved demand into a channel with no dosing standard whatever.
The most recent news is better and deserves to be stated plainly, since pessimism on this subject has become a reflex. Provisional CDC figures put 2025 overdose deaths at approximately 69,973, down almost 14% from 81,313 in 2024, with opioid-involved deaths falling from about 55,300 to 44,600 — a third consecutive annual decline from the 2022 peak near 110,000. The causes are contested: wider naloxone distribution, expanded access to buprenorphine, shifts in the illicit supply, and the grim arithmetic of a shrinking exposed population all have advocates. What is clear is that the decline cannot be credited to further prescribing restriction, since 22 states saw dispensing rise between 2023 and 2024 while drug mortality fell.
IX. The global inversion
The fourth crisis is the one Americans almost never register, and it reverses the entire domestic frame.
The wealthiest tenth of countries hold roughly 90% of distributed morphine-equivalent opioids. Half the world’s population, in the poorest countries, receives under 1% of the morphine distributed worldwide. In 2023, about a fifth of humanity consumed 86% of global morphine production. More than 80% of the roughly 61 million people experiencing serious health-related suffering live in low- and middle-income countries with severely limited access to palliative care, including plain oral morphine. The Lancet Commission estimated the annual cost of closing the world’s unmet palliative morphine need at about $145 million. Morphine is off-patent, inexpensive, and has negligible street value; it is manufactured reluctantly because the margins are thin and the liability exposure is not.
So the world contains, at the same moment, a nation that saturated itself with high-potency synthetics and a majority of humanity that cannot obtain the cheapest opioid ever made for the purpose of easing a death. Whatever else is true about these drugs, the primary global fact about them is not excess. It is absence — and the absence is sustained partly by regulatory architecture exported from countries reacting to a problem the recipients do not have.
X. The moral register
The older tradition is markedly less squeamish about this than modern policy discourse.
Proverbs 31 warns rulers away from strong drink in verses four and five — kings must not drink, lest they forget the law and pervert the judgment of the afflicted — and then, two verses later, prescribes it: give strong drink to him that is ready to perish, and wine to those of heavy hearts, that he may drink and forget his poverty and remember his misery no more. The same passage, within a breath, forbids intoxication to the one who must judge and commands it for the one who is being crushed.
That is, structurally, the two-tier arrangement Western medicine reached by a century of trial and error. The illuminating difference is the criterion. Scripture does not sort by prognosis. It sorts by office and burden: the one who bears responsibility for others must keep a clear head, and the one who is perishing may be given relief. Applied honestly, that is a more generous standard than ours, because it does not require you to be dying in order to qualify for mercy — and a more demanding one, because it locates the danger of intoxication in the abandonment of duty rather than in the molecule.
The distinction worth preserving is between relieving suffering and manufacturing oblivion. The first is commanded. The second is what occurs when relief becomes an end rather than a means. And since the term circulates loosely in Christian writing on this subject, it is worth stating that the New Testament’s condemnations of pharmakeia concern sorcery and its associated drug practices; reading them as a blanket prohibition on analgesia is poor exegesis pressed into the service of a modern anxiety.
There is also a pastoral point that belongs in an analytical paper because leaving it out distorts the analysis. Nothing in the above licenses contempt for the person who could not stop. Whatever dependence is, it is not simply a smaller quantity of the virtue the rest of us happen to have more of. The person struggling here is not a failed version of me. He is someone whose circumstances, neurochemistry, and history intersected the same substance at a different angle. I hold that as a working premise rather than a sentiment, because the alternative premise — that the difference between us is moral quality — has been tried at scale, and it produced the enforcement architecture that now leaves dying men in pain.
XI. What disentangling buys
If the three offices are genuinely incompatible, the goal cannot be to balance them perfectly within one professional role. It can only be to stop pretending that any single instrument governs all four crises at once.
Separating them yields several conclusions that are otherwise obscured. Restriction of diversion channels — the Florida case — can succeed on its own terms without telling us anything about the wisdom of dose ceilings for stable patients. Falling prescription volume cannot be credited or blamed for illicit fentanyl mortality, which decoupled from prescribing more than a decade ago. Guidance written for population averages will be converted into individual ceilings by every actor with a liability interest, and must be drafted in the certain knowledge that this will happen. And the global picture indicates that the world’s dominant opioid problem is a distribution failure, not a consumption failure, which means the regulatory templates we export are frequently answers to the wrong question.
The four windows that opened for me in a single day were each showing one facet of a compound we have insisted on treating as an element. My own uneventful prescriptions, a dying man in Florida finally released into relief, a nurse in South Tehran standing between desperation and an empty answer, and a famous man who bought his way past every safeguard we had built — these are not four illustrations of one problem. They are four different problems that happen to involve the same class of molecule, and the persistent failure to say so has produced a policy history in which each correction generates the next catastrophe.
The song my friend loved in the 1990s was, it turns out, a fairly precise piece of testimony about the moment when medicine stops being medicine. The information was public, in plain form, for twelve years. What was missing was not knowledge. It was any arrangement capable of acting on it — because the one person positioned to refuse was on the payroll, which is only the most vivid instance of a more general truth: we keep asking a single office to perform three jobs, and then expressing surprise when it performs none of them well.
