In a recent essay in The Atlantic, staff writer Michael Powell argued that some American and Canadian cities have allowed harm reduction to grow from a limited public-health strategy into an ideology, one that tolerates public drug use and disorder while failing to stem overdose deaths. Powell points to Seattle, Burlington, Vt., and Vancouver, among other places, and argues that their experiences show the negative consequences of placing too much emphasis on the autonomy of people who use drugs and too little on treatment, enforcement, and the interests of surrounding communities.
The problems Powell observes deserve attention. Open-air drug markets, discarded syringes, theft, public drug use, and neighborhoods overwhelmed by visible addiction should not be dismissed as the unavoidable price of compassionate drug policy. But establishing that these problems exist in cities that practice harm reduction is different from establishing that harm-reduction policies caused them. Powell frequently groups together very different policies, from syringe services and supervised consumption to safer supply, decriminalization, and decisions about policing, and then evaluates them against broad citywide outcomes. A more useful question is whether the evidence supports the specific causal claims he makes about each policy and place.
Harm reduction is not one uniform policy. Naloxone distribution, sterile syringe programs, drug-checking for contaminants, supervised consumption sites, and prescribed alternatives to the illicit drug supply all work in different ways and address different risks. Each should be assessed on its own terms, including when the evidence points to tradeoffs or unintended effects.
The conditions Powell describes should concern anyone interested in sensible drug policy. In Burlington, a city of roughly 45,000 people, 585 drug overdoses were recorded in 2024, 28 of them fatal. Powell argues that cities have placed too much emphasis on accommodating drug use and should instead restore enforcement against public drug use and dealing while putting greater pressure and resources toward treatment, including mandated treatment in some cases.
One of Powell’s central arguments is that overdose deaths rose sharply in cities that embraced strong harm-reduction policies. That claim deserves closer examination in the places he highlights.
In Burlington, the recent trend already complicates the picture. Powell notes that Vermont recorded 170 overdose deaths in 2025, a 37 percent decline from the state’s 2022 peak. Burlington’s experience with harm reduction is also more limited than Powell’s framing might suggest. The city approved an overdose prevention center only in 2024, meaning the facility Powell criticizes had not yet opened when many of the overdose and public-disorder problems described in his article occurred. Harm-reduction services existed before then, but the timing makes it difficult to attribute the city’s earlier deterioration to the supervised-consumption policy that features prominently in his critique.
Seattle offers a different test because harm-reduction services were already well-established there. Powell points to 7,089 overdose deaths in King County between 2016 and 2025 as evidence of the severity of the crisis. But that total obscures a significant reversal within the period he cites. King County’s overdose mortality rate more than tripled between 2018 and 2023, as fentanyl became pervasive in the local drug supply, then fell 32 percent between 2023 and 2025. Harm-reduction efforts continued during that decline, including widespread naloxone distribution and other overdose-prevention services. That does not establish that those policies caused the decline. But the rise and subsequent fall in deaths while harm reduction remained in place make the mortality trend considerably harder to attribute to the policies themselves.
Vancouver provides the longest test of Powell’s argument. North America’s first sanctioned supervised injection site opened there in 2003, and Powell acknowledges research finding fewer overdose deaths and infectious diseases associated with the program. Yet, he contrasts those findings with the dramatic rise in overdose deaths across the rest of British Columbia (a Canadian province geographically larger than the state of Texas) over the following two decades. Those are different levels of analysis. Evidence that a supervised consumption site reduces deaths among people who use it can coexist with rising population-level mortality driven by a rapidly changing illicit drug market. The relevant question is whether the intervention reduced the harms it was designed to address, not whether it single-handedly reversed a regional overdose crisis.
If harm-reduction policies themselves are driving higher overdose mortality, we should expect the same pattern elsewhere. But we don’t see that. New York City provides one useful counterexample. The city opened the nation’s first publicly recognized overdose prevention centers in November 2021 as overdose deaths were climbing. Deaths continued to rise the following year, increasing 12 percent from 2021 to 2022, before declining slightly in 2023 for the first time since 2018. That decline accelerated considerably in 2024, when overdose deaths fell 28 percent, from 3,056 to 2,192, the city’s first substantial decrease after nearly a decade of increases. The centers operated throughout this period as the city continued investing in syringe exchange services, overdose prevention, treatment, and other harm-reduction programs. The timeline does not establish that harm reduction caused the decline. But it does show that overdose deaths can fall substantially while harm-reduction policies remain in place, further weakening the argument that the presence of those policies explains citywide overdose trends.
The spread of illicit fentanyl makes that distinction even more important. Fentanyl reached different parts of the country at different times, rapidly changing the risk of overdose as it entered local drug supplies. A 2024 JAMA Network Open study of Oregon illustrates how easily that shift can be mistaken for the effect of a policy. Researchers initially observed an increase in overdose deaths following the implementation of Measure 110, which decriminalized possession of small amounts of drugs. After accounting for the timing of fentanyl’s spread into Oregon, however, they found no association between decriminalization and overdose mortality.
Another problem with evaluating harm reduction this broadly is that the term encompasses policies that work in very different ways. Among harm-reduction interventions, providing sterile syringes to reduce transmission of HIV and hepatitis C serves a different purpose than operating a supervised consumption site to prevent fatal overdoses or prescribing pharmaceutical opioids through safer-supply programs. Each should be evaluated based on its own objective evidence base.
For syringe service programs, the evidence is particularly strong. According to the Centers for Disease Control and Prevention, these programs are associated with an estimated 50 percent reduction in HIV and hepatitis C incidence. Nearly three decades of research also show that comprehensive syringe service programs do not increase illegal drug use or crime. In addition to providing sterile injection equipment, many offer infectious-disease testing, vaccination, and connections to medical care and substance use treatment.
Evidence that overdose prevention centers reduce deaths across an entire city is less settled. These facilities provide a supervised setting where trained staff can intervene when an overdose occurs, while also reducing risks associated with unsafe injection and connecting participants to other services. Powell acknowledges that studies have found people who consume drugs at supervised sites are far less likely to die from an overdose than those using elsewhere. A recent systematic review examining studies published in Canada between 2016 and 2024 found mixed results. Large provincial analyses generally found no statistically significant association between supervised consumption sites and overall overdose mortality, while some studies examining smaller geographic areas found reductions near the facilities themselves.
In Toronto, for example, neighborhoods within 500 meters of a supervised consumption site experienced 67 percent fewer overdose deaths per 100,000 residents after the sites opened. An Alberta analysis also found that greater use of supervised consumption sites was correlated with fewer fentanyl-related overdose deaths between 2017 and 2020. Both studies were observational, however, meaning they can identify associations but cannot establish that the sites alone caused the reductions.
The evidence therefore supports a narrower conclusion than either side of the debate often makes. Overdose prevention centers can prevent fatal overdoses among people who use them, and some evidence suggests benefits in the areas immediately surrounding them. The evidence that they reduce mortality across an entire city or province is less clear. That distinction matters when Powell cites rising citywide overdose deaths as evidence against the facilities themselves.
Powell raises legitimate concerns about diversion within safer-supply programs, meaning prescribed medications being sold, traded, or otherwise passed on to someone other than the person for whom they were prescribed. In British Columbia, he points to reports that prescribed hydromorphone was diverted and exchanged for fentanyl, prompting the province to tighten its program by requiring some recipients to consume prescribed drugs under pharmacist supervision. Diversion should factor into how these programs are evaluated, but it cannot be considered in isolation from the risks of the illicit supply they are meant to replace.
A more useful question when considering safer supply programs is how those risks compare with the alternative people already face. Safer supply is intended for people at high risk of overdose who would otherwise rely, at least in part, on an illicit drug market increasingly dominated by fentanyl and other adulterants that can lead to accidental overdose. Early research has associated participation in these programs with lower individual overdose risk, reduced use of fentanyl and other street-acquired drugs, and fewer hospital and emergency room visits, although Health Canada acknowledges that the evidence base is still developing.
That tradeoff is central to a harm-reduction approach. No drug policy eliminates every risk, and expecting one to do so sets an unrealistic standard for evaluating policy. The question is whether an intervention reduces overall harm compared with existing conditions, while identifying and addressing unintended consequences as they emerge.
Public drug use and disorder need to be addressed; safer-supply programs can create risks of diversion, and cities should not treat harm reduction as a reason to abandon enforcement against conduct that harms others. But those observations do not support his broader indictment of harm reduction. The evidence he cites from Seattle, Burlington, and Vancouver does not establish that syringe services, supervised consumption, naloxone distribution, or other harm-reduction interventions caused those cities’ overdose crises. In several cases, the evidence points in the opposite direction: Individual interventions have reduced the specific harms they were designed to address even as the broader overdose crisis worsened around them.
The lesson from these cities is not that harm reduction has failed. It is that no single approach can carry the weight of the overdose crisis. Policymakers should preserve interventions that demonstrably reduce death and disease, address diversion or other unintended consequences where the evidence identifies them, enforce laws against conduct that harms others, and expand access to evidence-based treatment. Abandoning effective harm-reduction tools because they did not solve every dimension of the drug crisis would repeat the mistake Powell warns against: allowing ideology to substitute for evidence.