Drug-Linked Heart Infections Fell 18 Percent. Stimulants Now Drive Nearly Half of Them.

The most dangerous infection a person can get from injecting drugs has become less common in the United States since 2020. Read that sentence alone and the last six years look like progress. Read the paper it comes from and the picture is closer to a trade.

A national analysis published Oct. 27 in Clinical Infectious Diseases estimates 118,190 hospitalizations for drug use-associated infective endocarditis between 2016 and 2022. Cases peaked at 18,735 in 2020 and then fell in each of the two following years, reaching 15,345 in 2022 — a decline of about 18 percent. The authors, a group from the University of Colorado Anschutz, Washington University in St. Louis, UCSF and Boston University, present that drop as consistent with wider access to addiction medication and harm reduction, and they are careful to say they cannot prove it.

Inside the same dataset, though, the drug mix behind those infections moved almost point for point in the opposite direction.

A Decline That Is Really a Substitution

Opioid-only cases fell from 49.4 percent of the total in 2016 to 39.9 percent in 2022, a loss of 9.5 percentage points. Over the identical window, the share involving stimulants — methamphetamine or cocaine, alone or alongside opioids — rose from 37.3 percent to 46.9 percent, a gain of 9.6 points.

That symmetry is the finding. The overall number came down because one category shrank, and nearly all of the room it vacated was filled by another. This is not a rising tide lifting every boat; it is a current changing direction.

It also lands hardest on the group with the thinnest treatment options. For someone who uses methamphetamine or cocaine and wants to stop, stimulant addiction treatment offers no equivalent to the medication shelf built for opioids — no methadone, no buprenorphine, no naltrexone equivalent with a comparable evidence base. The clinical infrastructure exists mostly as behavioral care, and behavioral care is the part of the addiction system that insurers reimburse worst.

What the Numbers Actually Moved

The shares tell one story; the raw counts tell another, and the two are worth reading side by side.

Drug category in the discharge record	2016	2022
Opioids only	49.4%	39.9%
Opioids and stimulants together	22.6%	28.6%
Stimulants only	14.7%	18.3%
Neither coded	13.3%	13.2%

The co-use row deserves attention. Polysubstance presentations — a person using both — grew faster than stimulant-only ones, and now account for more than a quarter of all these hospitalizations. That combination is what the fentanyl-era supply makes easy: stimulants to stay upright, opioids to come down, often in the same session and sometimes in the same bag.

The "neither" row is flat to within a tenth of a point across seven years, which is the quiet proof that the shift is not a general drift toward some third category. Everything that changed, changed along the opioid-stimulant axis.

Where the West Diverged

Regionally, the country is not one country. Using logistic regression that adjusted for patient age, sex, race and ethnicity, insurance, and hospital teaching status and bed size, the researchers modelled the probability that a given endocarditis hospitalization involved stimulants. In the Pacific census division it was 0.59, with a 95 percent confidence interval of 0.56 to 0.62. In New England it was 0.34, interval 0.31 to 0.38.

That is a 25-point spread between the two coasts for the same diagnosis. In several Western divisions stimulants overtook opioids outright as the leading drug category.

The authors do not explain the geography, and the design does not let them. Census division is the exposure variable in that model, so the comparison is between regional averages, not between local markets. A Western state with an unusual stimulant supply and a Western state with unusual coding habits both show up as "Pacific" here.

Why the Authors Hedge on the Cause

Two mechanisms are offered, and the paper tests neither.

The first is the hopeful one: more naloxone and syringe access, more education on safer use, more people connected to medication for opioid use disorder. If that is what happened, the decline is a policy return on investment, arriving years before anyone expected to see one in a hospital discharge record.

The second is less flattering to policy. Because injecting is what pushes bacteria into the bloodstream, a population that shifts from injecting to smoking carries less endocarditis risk for reasons that have nothing to do with treatment access. A change in how a drug is sold and consumed would produce the same curve.

There is a third possibility the paper's data cannot exclude at all: fewer people injecting opioids because fewer people are alive to inject them. Overdose mortality was climbing through most of this window.

The paper's own language — that current efforts "may be reducing" these hospitalizations — is the right register for what is available. Anyone citing this study as proof that harm reduction worked is going beyond it, and anyone citing it as proof that it failed is going equally far in the other direction.

The Medication Asymmetry

Paul Christine, the paper's first author and an assistant professor of medicine at Colorado, put the asymmetry plainly in the university's announcement. The problem, he said, "isn't simply that stimulants are becoming more prominent. It's that the healthcare system currently has fewer effective tools to treat stimulant use disorder than opioid use disorder."

He is describing a gap that is structural rather than incidental. Opioid use disorder has three medications with decades of trial evidence behind them, and a treatment system — opioid treatment programs, pharmacy dispensing, telehealth induction — built specifically to deliver them. Stimulant use disorder has contingency management, an intervention with a genuinely strong evidence base, and then a long tail of approaches with weaker support.

Contingency management also has an implementation problem that medication does not. It pays people for verified abstinence, which means it needs a funding stream that most state Medicaid programs have been reluctant to authorize at scale, and its reach is limited by how few clinics run it.

So the disease burden has shifted toward the half of the addiction system that is least equipped, least reimbursed and least standardized. That is the policy consequence of this paper, and it does not depend on which causal explanation for the decline turns out to be right.

What a Four-Year Lag Hides

The data stop at the end of 2022, and that matters more than it first appears.

The national opioid settlements began distributing in 2022, with roughly $50 billion scheduled across nearly two decades. States were still standing up the spending infrastructure — councils, approved-use lists, procurement — through 2023 and 2024. Whatever that money has or has not achieved, it cannot be visible in a series that ends the same year the payments started.

The same window excludes the current adulterant picture, including xylazine's spread into the Eastern supply, and it excludes the Medicaid redetermination process that removed millions of people from coverage in 2023 and 2024. Those three forces — settlement spending, a changing adulterant mix, and coverage loss — are exactly the ones a reader would want to weigh against the 18 percent decline, and the dataset predates all of them.

The 2020 peak is also a COVID artifact worth naming. Hospital capacity, admission thresholds and discharge patterns all shifted violently that year, in ways that could push a chronic infection like endocarditis either up or down.

The Coding Problem

Endocarditis does not announce its cause. The researchers identified drug use-associated cases from ICD-coded discharge records, which means a case counts only if a clinician wrote down both the infection and the drug use.

Drug use is under-recorded in hospital charts for reasons that have been documented for years: stigma in the encounter, time pressure, and the fact that a patient who does not disclose gets coded as having no history. A downward trend in a coded series is therefore compatible with a genuine decline in disease and with a decline in documentation.

The drug category has a second layer of softness. "Opioid only" and "stimulant only" are assignments made from codes, and a person who used both but mentioned one will land in a single bin. That is one reason the co-use row is likely an undercount, and one reason regional comparisons of drug mix should be read as comparisons of recorded drug mix.

What This Means for Patients Leaving the Hospital

Infective endocarditis is not a diagnosis anyone walks out of quickly. Treatment typically means weeks of intravenous antibiotics, often a valve replacement, and a period of inpatient monitoring that gives the care team an unusual amount of contact with a person who uses drugs.

That contact is the opportunity, and it is routinely wasted. The gap is rarely the antibiotics — it is what happens on the day of discharge, when a patient with a new valve, no stable housing and an active stimulant habit is sent out with a follow-up appointment. Hospital-based addiction consult services and post-discharge levels of care that can actually accept someone within days are the two pieces that determine whether the surgery was worth doing.

For stimulant-involved cases specifically, there is also nothing to prescribe at the door — no equivalent of a buprenorphine bridge. What a discharging team can offer instead is contingency management enrollment, a warm handoff to a program that runs it, and a plan that names where the person is sleeping.

What Comes Next for the Researchers

Christine and his colleagues say their next work is on hospital-based interventions for people who use stimulants, with a focus on connecting them to treatment and to ongoing care after discharge. They also want to know whether the same declines are showing up in the other infections that follow drug use — skin, soft tissue and bone.

Those are the right questions, and they are the questions this dataset cannot answer. A discharge record tells you that an infection happened and roughly what caused it. It cannot tell you whether the patient got a bed, a prescription, or a call the following week.

What the paper does establish is that the shape of this particular harm has changed. The opioid share of drug-associated endocarditis is smaller than it was. The stimulant share is not, and the treatment system's answer to stimulants is the weakest part of the house.