Cannabis and Psychosis: 5 Reasons This Study Doesn’t Say What You Think

A striking new paper claims psychosis hospitalizations rose after legalization. But context, nuance—and real-world sanity—are missing.

☕ What You’ll Learn in This Post

✔︎ Why a rise in cannabis-use-disorder (CUD) codes doesn’t automatically equal a rise in harm

✔︎ What’s missing from the popular “5x increase in psychosis” narrative

✔︎ Why countries with high cannabis use don’t show spikes in psychotic illness

✔︎ How public health systems are failing—not because cannabis is legal, but because it’s poorly integrated into care

✔︎ What smarter, stigma-free cannabis policy and clinical care would actually look like

Cannabis and Psychosis: The Headlines Are Loud. But Are They Honest?

It’s hard to argue with a 5x increase. Especially when it’s printed on a graph in a peer-reviewed journal and shows up in your LinkedIn feed next to phrases like “public health concern” and “unintended consequences.” A newly published study in Psychological Medicine tracked over 15 years of psychiatric hospitalizations in Colorado and found that the number of young adults admitted for psychosis alongside a cannabis use disorder diagnosis increased significantly—particularly after the state legalized recreational cannabis.

On the surface, it’s clean, compelling, and seemingly conclusive: legal cannabis, more psychosis. End of story?

Not quite.

Let’s be clear: the authors of the study deserve credit. The timeline is long. The dataset is large. The statistical model—interrupted time series—is appropriate for the question. And most importantly, the paper doesn’t overclaim. It stops short of calling cannabis the cause of psychosis. Instead, it draws attention to a troubling trend, and calls for further inquiry.

But what happens next—how media, policymakers, and even well-meaning clinicians interpret that trend—is where the real risk lies. Not a psychiatric risk. A scientific one. Because when it comes to cannabis and psychosis, context isn’t just helpful. It’s everything.

What the Colorado Study Got Right—and Why That’s Only Half the Picture

Let’s start with what deserves respect.

This study, led by investigators at Denver Health, looked at over 9,700 hospitalizations involving psychosis in people ages 10–29 across a 15-year span. It pinpointed a steady increase in psychosis-related hospital visits, with an especially sharp rise in those also coded with cannabis use disorder (CUD) after recreational legalization took effect.

The methodology was sound: the team used interrupted time series regression, a tool designed to detect changes in outcome trends after a policy shift. They didn’t just compare before-and-after snapshots—they measured the slope of change, and it was statistically significant. Notably, the most pronounced increase occurred in 21–29-year-olds—the age group legally permitted to purchase cannabis products.

So yes—there’s a signal here.

But a signal doesn’t mean we’ve heard the full message. And while the authors were careful in their language, the cultural interpretation that followed their graphs has been… less restrained. In many places, the signal has been amplified into an alarm, stripped of context, and translated into a simplistic narrative: cannabis use → psychosis.

The problem? That narrative ignores what’s missing—and what that missing data might reveal.

What if CUD isn’t a stable or meaningful diagnosis in this setting?

What if the rise in hospitalizations says more about documentation trends than actual incidence?

And what if the real story here isn’t cannabis causing psychosis—but a health system failing to guide cannabis use safely, especially for those most at risk?

These are the questions that deserve to be asked before we turn data into doctrine.

Why CUD Is a Blunt Tool: Diagnosing More Doesn’t Mean Harming More

Let’s talk about the acronym doing the heavy lifting in this study: CUD, or cannabis use disorder. At first glance, it sounds serious—maybe even self-evident. You wouldn’t be admitted to the hospital with “a disorder” unless something had gone terribly wrong… right?

Not necessarily.

CUD is not the same cannabis dependence, even though they’re often treated as synonymous. Here’s what CUD actually means in practice: a person has met two or more of 11 possible DSM-5 criteria within the past year. Those criteria range from the familiar (“developed tolerance”) to the almost laughably vague (“spent a lot of time getting cannabis”). That last one, in a legal state with cannabis delivery apps and 1,000 dispensaries, might just mean you got stuck in traffic on the way to your gummies.

To be fair, some CUD cases are truly disruptive. Daily use that derails relationships, impairs functioning, or exacerbates pre-existing psychiatric illness is a serious concern. But here’s the thing: none of that is confirmed in this study. The diagnosis is derived from a billing code, not a structured assessment. There’s no report of how it was made, by whom, or based on what information.

And this matters. Because after legalization, people are more likely to disclose cannabis use, clinicians are more likely to ask, and systems are more likely to flag it in the chart. So CUD codes go up—not necessarily because more people are “addicted,” but because more people are visible.

This is what’s known as detection bias, and it’s rampant in post-legalization environments. When you shine a brighter light, you find more dust. That doesn’t mean the room got dirtier.

Here’s a second problem: CUD isn’t necessarily synonymous with harm. Many people with chronic conditions—pain, insomnia, PTSD—use cannabis daily. They may build tolerance, prefer high doses, and feel lousy if they miss a dose. That technically qualifies them for a CUD diagnosis. But does it mean they’re disordered? Or just… treating themselves with the only thing that works?

If you’re looking for psychosis risk signals, these distinctions matter. A daily, balanced tincture user managing fibromyalgia isn’t the same as a teen dabbing high-THC concentrates 10 times a day in isolation. Yet this study treats both—at least on paper—as equally significant.

And in doing so, it builds its case for rising psychiatric harm on a foundation that is statistically clean but clinically muddy.

Want to know if CUD codes are meaningful? We’d need to know:

1️⃣ What products were used

2️⃣ At what doses and frequencies

3️⃣ Whether patients were self-medicating distress

4️⃣ If their use preceded or followed psychiatric symptoms

We get none of that here.

Instead, we get a simple box: CUD present or not. And the hospital data, like all big data, is only as useful as the definitions behind its digits.

Understanding CUD – my response to a patient’s question about it.

Another in-depth review of CUD

Cleveland Clinic’s view of CUD

Read Carefully: American Addiction Centers has a very clear view too

The Missing Pieces in the Study: No Dose, No Direction, No Control

It’s not that the study is poorly done. It’s that it’s incompletely read—by the headlines, the comment sections, and even some professionals who should know better.

Because beneath the clean graphs and p-values lies a quietly glaring problem: we’re trying to draw major conclusions from missing context. And in medicine, context isn’t window dressing—it’s the diagnostic lens.

Let’s walk through what’s not in this paper.

First: dose. The study tells us someone had psychosis and a CUD diagnosis. But was that person using 2mg of THC nightly to sleep, or 250mg a day via high-potency dabs? Were they smoking flower? Eating edibles? Using CBD alongside it? We don’t know. And yet, we’re asked to view all cannabis use—and all users—as functionally the same.

That’s not science. That’s a Rorschach test for your bias.

Second: directionality. Did cannabis use precede the psychosis? Was it part of a self-medicating attempt in the early stages of a developing condition? Or did the cannabis use begin after psychotic symptoms emerged? Again—we don’t know. The data don’t say. But the conclusions imply.

Third: no control group. This is key. The study shows psychosis-related hospitalizations rose in Colorado after legalization. But what about other states during the same timeframe? Were there parallel increases driven by post-COVID mental health collapses, rising homelessness, fentanyl adulteration, or economic stress? If so, then Colorado’s signal might not be about cannabis—it might be about everything else happening in the lives of vulnerable young people.

A well-matched control state could have helped us know. But that’s absent too.

Finally, there’s no assessment of why these hospitalizations happened in the first place. Was cannabis the trigger? Or simply present? Was the psychosis brief and substance-induced, or persistent and part of a first-break schizophrenia spectrum diagnosis? Were patients given a urine tox screen? Were they using other substances too?

You won’t find those answers in this paper. Which is fine—this was a population-level trend analysis, not a psychiatric deep dive. But what’s not fine is when people treat it like definitive evidence of cannabis harm, without ever asking what else it might mean.

This is the problem with overly enthusiastic interpretations of under-examined data: they skip past the complexity, and land squarely in the realm of moral panic.

What Public Health Data Actually Shows: The Global Reality Check

If cannabis really caused psychosis in anything more than a narrow slice of vulnerable users, we wouldn’t need a regression model from a single hospital to tell us. We’d feel it in the data. In the admissions. In the overwhelmed clinics. In the epidemiology.

We don’t.

Start global. The World Health Organization pegs the lifetime prevalence of schizophrenia spectrum disorders at about 0.4%—and that number has held steady for decades, across countries, across continents, and crucially, across eras of cannabis liberalization. Whether you’re looking at Portugal (where all drugs were decriminalized in 2001), the Netherlands (where cannabis tolerance began in the 1970s), or Canada (which federally legalized in 2018), there’s no psychosis wave. No public health surge. No cannabis apocalypse.

Zoom into the U.S. Look at California—the birthplace of legal medical cannabis and one of the first to legalize recreational use. Or Washington, D.C., which has some of the highest cannabis use rates in the country. Psychosis hospitalization rates haven’t ballooned in those places either. If anything, they’ve remained relatively stable.

In Ontario, Canada, researchers documented a rise in cannabis-related ED visits post-legalization. But when they zoomed in, the culprit wasn’t widespread psychosis. It was accidental overconsumption, panic reactions, and disorientation—especially in novice or unprepared users. Manageable stuff. Not a schizophrenia surge.

So what explains the dissonance?

Simple: the denominator is enormous. Millions of people are using cannabis regularly, even daily. If it reliably triggered psychosis in average users, we’d be seeing a mental health disaster right now. We’re not. Because most users are fine—especially those who are informed, intentional, and not combining high-potency THC with untreated trauma, genetic vulnerability, or other destabilizing forces.

That doesn’t mean cannabis is harmless. It means the harm is contextual, rare, and concentrated in predictable patterns—not randomly distributed like a pharmacologic lightning bolt.

And here’s where the real public health lesson lies: instead of treating cannabis like a contagion, we should be treating it like any other powerful tool. Respect it, guide its use, educate users early, and intervene in the right places. Not with blanket suspicion, but with precision care.

Because if psychosis is on the rise, cannabis might not be the match—it might just be the smoke we noticed too late.

What’s Really Happening Here: Unguided Use in Vulnerable Populations

The most honest takeaway from the Colorado study isn’t that cannabis is making people psychotic—it’s that people who are already at risk are using cannabis in risky ways, without any support or direction. And then landing in hospitals.

Look closely at who’s showing up in these encounters. In the psychosis + CUD group, 68% were male, most were in their early 20s, and the majority were on Medicaid. That paints a picture—not of the average cannabis user, but of a very specific population: young, often underserved men navigating trauma, poverty, mental illness, or all three, and self-medicating with potent cannabis products they don’t fully understand.

These aren’t casual users. These are high-frequency, high-potency consumers often using without structure, without education, and without access to non-stigmatizing care. They’re not problems. They’re canaries. And the coal mine isn’t cannabis—it’s the gap between legalization and medical infrastructure.

If you legalize a substance with complex neurological and psychiatric effects—and then fail to provide clinical scaffolding for how to use it safely—you don’t get chaos. You get confusion, overcorrection, and preventable harm in the places that were already hurting.

This is a public health story, yes—but one about lack of guidance, not about chemical villainy. Because let’s be honest: if these same young people were taking prescription amphetamines, or alcohol, or even over-the-counter decongestants at the same doses and frequencies, we’d be seeing psychiatric side effects too. But we wouldn’t be racing to re-ban the substance. We’d be asking: where was the education? where were the supports?

The cannabis in these cases might be strong. But the system’s silence is stronger.

So if we’re going to interpret the hospitalization spike as a call to action, let’s make sure we’re acting on the right target. Not cannabis itself, but the lack of professional presence in cannabis use—especially for those already vulnerable to psychiatric destabilization.

What Smart Policy Would Do: Regulate Risk, Not the Plant

When studies like this hit the news cycle, the policy response often swings toward restriction. Stricter laws. Harsher warnings. More fear-based messaging. But if we want to make cannabis safer—and prevent the kind of psychiatric events highlighted in the Colorado data—then we don’t need a crackdown. We need a compass.

Because smart cannabis policy doesn’t start with banning a plant. It starts with recognizing that not all cannabis use is equal, and neither are the risks.

What if we approached cannabis the way we already do with alcohol or antidepressants?

We don’t call a glass of wine alcoholism.

We don’t assume that a 10mg SSRI will destabilize the brain the same way a 100mg dose might.

We don’t treat all users as ticking time bombs.

We look at the dose, the intent, the duration, the context—and the person.

Smart cannabis policy does the same. It:

♦️ Flags high-risk patterns—like high-potency THC concentrates in teens

♦️ Offers front-loaded education on how cannabis interacts with sleep, trauma, and underlying psychiatric risk

♦️ Incentivizes lower-risk products (like balanced THC/CBD formulations)

♦️  Encourages early screening—not for addiction, but for misuse with medical need

♦️ Funds cannabis-aware clinicians who know how to treat patients with curiosity, not condescension

And perhaps most importantly, it resists the urge to treat correlation as causation. Because when we confuse documentation trends with psychiatric epidemics, we start crafting policy for a fantasy problem—and ignoring the real one unfolding in our ERs and clinics.

Legalization was the beginning. Integration is what comes next.

What We Owe the Public: Context Over Panic, Precision Over Fear

The cannabis and psychosis conversation isn’t going away. Nor should it. But if we’re going to have it, we owe the public something better than quick conclusions wrapped in tidy headlines. We owe them context.

We owe them the truth that cannabis can destabilize people—but mostly those already standing on shaky ground.

We owe them the insight that a diagnosis code doesn’t always mean dysfunction, especially when that code is based on frequency, not harm.

We owe them the reminder that most people who use cannabis—daily, occasionally, medicinally—aren’t ending up in psych wards.

And we owe them care models that help prevent those who might from ever getting close.

The real risk isn’t cannabis. It’s the assumption that cannabis is always the risk.

It’s the lazy science that lumps all use into one category.

The policy impulse that treats one data signal like a universal danger.

The silence in too many exam rooms, where cannabis conversations are either brushed off or shut down entirely.

We can do better.

We can teach patients how to use cannabis with eyes open, not in fear.

We can differentiate between wellness, self-medication, and misuse.

And we can build a healthcare system that finally speaks fluently about a plant that millions already turn to—without judgment, without panic, and with an actual plan.

Because the real “public health response” isn’t to panic when a paper shows psychosis went up.

It’s to ask the deeper question: why did it happen, to whom, and what could we have done better?

The answer isn’t prohibition.

It’s precision.

And it starts with telling the whole story.

❓ 10 Optimized FAQ Questions + Answers