Washington Just Made Sobriety the Price of a Lease. The Evidence Points the Other Way.

You’re in a shelter bed tonight. You’ve been using — maybe you never stopped, maybe you slipped after ninety clean days. Under the model the federal government endorsed this year, that single fact can end your bed. Not your behavior toward other residents, not a safety violation — the fact of a positive test, or evidence you used, at any point before you’ve logged twelve to eighteen months of proof otherwise. Housing sits at the far end of that road, reserved for people who’ve already stopped. This isn’t a story about one shelter on one night. It’s the mechanism, as written into federal guidance and described in a formal objection filed against it: if you use, you lose your bed.

The federal government is now calling that mechanism a best practice.

A policy that brands itself “evidence-based” is running directly against the housing-outcomes evidence, and choosing faith in abstinence over the data.

On August 12, 2026, the Department of Health and Human Services, the Office of National Drug Control Policy, and the Department of Housing and Urban Development jointly released the Best Practices Toolkit: Addressing Homelessness and Addiction through Treatment First, a document tied to President Trump’s executive order on “Ending Crime and Disorder on America’s Streets.” It explicitly rejects Housing First, the model that has guided federal homelessness policy for two decades, in which someone gets a lease with no sobriety requirement and no obligation to accept treatment. In its place, the toolkit builds what it calls a phased path: crisis, stabilization, treatment, early recovery, and finally “thriving” — defined in the document as long-term abstinence, mental health stability, full-time employment, and living in housing that isn’t subsidized. Independent housing, in other words, isn’t the starting point. It’s the reward at the end.

“Treatment must come first,” HHS Secretary Robert F. Kennedy Jr. said at the launch. “Recovery begins when we connect people with effective care, not leave them trapped in addiction and homelessness. This toolkit gives communities practical, evidence-based strategies to expand treatment, support lasting recovery, and help people reclaim their independence.” HUD Secretary Scott Turner, standing alongside him, put it more bluntly: “I have met men and women whose lives have been transformed by treatment and recovery. Housing is essential, but housing alone is not enough.” ONDCP Director Sara Carter joined both officials for the announcement, though the agencies did not attribute a direct quote to her.

The toolkit itself traces back to a national summit ONDCP convened April 12–14, 2026, bringing together federal, state, and local officials, law enforcement, clinicians, and program leaders to build the “best practices” the document now codifies. That’s worth sitting with: four months from closed-door summit to signed guidance to a $44.3 million funding stream behind it — a fast track for a model that discards two decades of federal housing policy and the research built around it.

Then, on September 18, the money followed the rhetoric. SAMHSA awarded more than $42 million — $42.3 million — in supplemental funding to states and territories through two federal block grants, specifically earmarked to help them adopt the Treatment First model. Another $2 million went to the national office and all ten regional centers of the Addiction Technology Transfer Center (ATTC) Network — a SAMHSA-funded training system that helps state agencies retrain frontline addiction counselors and case managers on new clinical models — to build out training and implementation support. Five weeks after the toolkit rejected Housing First on paper, the federal government put real dollars behind making sure states actually switch.

The case the toolkit doesn’t make

Here’s what’s strange about calling this “evidence-based”: the strongest recent evidence on housing and addiction outcomes points at the opposite model.

A 2025 study in JAMA Network Open, led by researchers modeling outcomes for homeless adults with opioid use disorder — addiction to prescription painkillers, heroin, or fentanyl — found that providing stable housing without a treatment requirement was more cost-effective than treatment-contingent alternatives, at roughly $26,800 per quality-adjusted life year gained. Stanford’s own reporting on the study described supportive housing without conditions as a “high-impact, cost-effective” intervention — the kind of language HHS uses for its own preferred programs, applied here to the model HHS just rejected.

Disability Rights Education and Defense Fund attorney Michelle Uzeta laid out the broader research record in DREDF’s formal statement condemning the toolkit: “Housing First produces substantially better housing outcomes than treatment-contingent models: faster exits from homelessness, higher retention.” She goes further on the specific fear driving Treatment First — that housing without a sobriety requirement enables unchecked drug use. It doesn’t hold up, she argues: “Housing First participants use formal treatment services less often than people in treatment first programs…and they do so without worse substance use outcomes.” People get stable, and their substance use doesn’t get worse for having skipped the sobriety test at the door.

Uzeta’s statement is where the toolkit’s mechanics get spelled out in plain terms. The document, she writes, “builds a system in which disabled people have to earn housing rather than have a right to it.” It requires “alcohol- and drug-free living environments,” with relocation — losing your placement, your bed, your progress — as the consequence for a lapse. Independent housing sits “at the far end of a five-phase ladder, reached only after twelve to eighteen months of documented abstinence.” DREDF represents disabled people nationally, and Uzeta’s statement is explicit that substance use disorder is itself a recognized disability under federal civil rights law — which means a system that conditions housing on managing it differently than any other chronic health condition raises exactly the kind of question her organization exists to raise.

Uzeta’s statement is where the toolkit’s mechanics get spelled out in plain terms.

She also names who gets hurt hardest by conditional-shelter systems generally: 58% of transgender homeless survey respondents report having been denied emergency shelter because of their gender. Add a documented-sobriety gate on top of an already-discretionary shelter system, and you’ve handed staff one more reason to turn someone away — one dressed up as clinical judgment instead of bias.

Not every recovery organization is treating this as a fight to the death. Faces & Voices of Recovery, a national recovery advocacy group, took a more measured line in its September policy update: “treatment and secure housing must go hand in hand, and neither need to be placed ‘first’ in front of the other.” They called the toolkit’s five-phase outcome definitions “a bit simplistic and convenient,” said they’d push to “broaden their scope,” and committed to keeping funding for the resources it references a “high priority” through congressional advocacy. It’s a both-housing-and-treatment position, not a rejection — but even their measured take treats the toolkit’s five-phase framework as something that needs fixing, not something that’s already right.

STAT News reported on launch day that many addiction-medicine and harm-reduction clinicians were “likely to be dismayed” by the toolkit, and that providers who run contingency management programs — treatment that uses small rewards, like gift cards, to reinforce someone showing up for care and staying negative on drug tests — feared their model would prove “ideologically incompatible” with an administration built around abstinence as the precondition for everything else, rather than a milestone within it. That distinction matters clinically. Contingency management is one of the best-evidenced treatments for stimulant use disorder — addiction to drugs like methamphetamine or cocaine, for which there’s no FDA-approved medication — precisely because it rewards small steps forward without punishing setbacks. A model that treats any documented use as a reason to pull someone’s bed is structurally at odds with a model built to expect, and work through, exactly that.

We’ve run this experiment before

This isn’t the country’s first attempt at treatment-first shelter policy, and that’s the part missing from the toolkit’s framing of itself as new evidence-based ground. In the early 1990s, most homeless services for people with serious mental illness and addiction ran on what was called the Continuum of Care model: sobriety, medication compliance, and program participation, in sequence, before anyone got a lease. Psychologist Sam Tsemberis watched that model fail the people it was built for — the ones whose addiction and mental illness made it functionally impossible to clear the qualifying hoops — and in 1992 founded Pathways to Housing in New York City to try the opposite: a lease first, with support services delivered afterward, no sobriety test at the door. It worked. Within Pathways’ first five years, 88% of participants remained housed, against 47% in the city’s existing treatment-first residential system. That gap is where Housing First came from. It wasn’t an ideological preference; it was a response to watching the treatment-first model leave people on the street. The 2026 toolkit doesn’t cite that history. It doesn’t have to grapple with it if it never mentions it.

What this actually changes, this week

None of this is abstract for the person it applies to. If you’re in a shelter system in a state that takes this September funding, your case manager’s incentives just shifted — toward documenting your abstinence, not toward getting you a key. The Treatment First toolkit doesn’t ban housing. It reorders it, and reordering is the whole policy: instead of a home as the floor everyone starts from, it’s the ceiling you reach after eighteen months of proof. For someone managing a chronic, relapsing condition — which is what addiction is, clinically, not a discipline problem — eighteen months of unbroken documentation is a very high bar, and one bad week can reset the clock.

Here’s what hasn’t changed: naloxone, the opioid-overdose reversal medication, remains legal, federally funded, and free to access in every state regardless of which housing model your local shelter system adopts. Treatment-first policy governs who gets a lease. It doesn’t govern who gets to survive long enough to have that argument at all.

It doesn’t govern who gets to survive long enough to have that argument at all.

The toolkit’s authors are right about one thing: housing alone doesn’t treat addiction. But the research doesn’t say housing alone is the goal — it says housing first is the precondition that lets treatment actually work, because you can’t stabilize a chronic illness from a shelter cot with a curfew and an exit interview. The evidence the administration cites to justify a stricter ladder is the same evidence that built the case against one, thirty years ago, in the same city, for the same reason. Washington just spent $44.3 million betting that this time, the ladder works.