How a Soft Diagnosis Becomes a Hard Headline

Two of eleven. That's the entire bar for a diagnosis now making headlines as "cannabis addiction." Someone who clears it by two mild symptoms gets the same label, the same statistic, and increasingly the same scary headline as someone who clears it by nine severe ones. Almost nothing in the public conversation about cannabis and addiction distinguishes the two.

A recent example shows how easily that gap gets papered over. Smart Approaches to Marijuana (SAM), a US advocacy group whose stated mission is to reduce marijuana use and oppose what it calls the commercialization and normalization of cannabis, published a release headlined "Marijuana Is on Track to Become America's #1 Addiction." The figure behind it: 19.3 million Americans meeting clinical criteria for cannabis use disorder (CUD). The release calls the entire group "addicted."

This isn't a new observation for us. When we covered a new Yale-led genetics study that drew a biological line between cannabis use and cannabis use disorder, we flagged that the diagnostic category doing that dividing work has never been well validated at its low end. This piece is that follow-up.

The figure is real. The word attached to it is doing more work than the diagnosis behind it can support, and that's true well beyond the US, including in Australia's own drug data.

What's inside the number

CUD, like every DSM-5 substance use disorder, is diagnosed on a sliding scale from that two-symptom minimum up through four or five criteria for "moderate" and six or more for "severe."

That structure matters because of where the diagnosis actually sits in the population. In the 2022 National Survey on Drug Use and Health, 7.0% of US adults met criteria for CUD, and of that group, mild CUD alone (3.9%) accounted for more than moderate and severe combined (3.1%). Longer-term data tells a similar story: between 2002 and 2017, mild CUD prevalence among all US adults roughly doubled, from 1.4% to 1.9%, even as moderate and severe CUD among cannabis users declined over the same period. Most of the diagnostic category's growth has come from its softest tier.

That would matter less if the mild tier were well understood clinically. It isn't. A 2022 study examining construct validity in the DSM-5 CUD framework found that severe CUD is strongly associated with clinical markers across multiple domains, but mild and moderate CUD are linked to cannabis-specific measures alone, meaning the diagnostic categories aren't yet well shown to track meaningfully different levels of real-world impairment. The same body of research shows the diagnosis is comparatively unstable: test-retest reliability for CUD sits at a modest kappa of 0.41 in a general-population sample, well below the near-perfect agreement (kappa 0.78) found for the much simpler question of whether someone used cannabis at all in the past year. Agreement between two different clinical interview methods assessing the same DSM-5 CUD diagnosis was only moderate (kappa 0.51 to 0.60). Because the threshold is just two criteria, researchers studying the framework note that a difference of a single symptom between assessments can be enough to flip someone in or out of a diagnosis entirely.

None of this means CUD isn't real, or that severe cases aren't serious. The literature is clear that they can be. It means a flat total blends a well-supported severe end with a comparatively unstable and thinly validated mild end under one word.

Australia isn't exempt, it just hides the seam differently

Australia's main population data source, the National Drug Strategy Household Survey, doesn't ask DSM-5 CUD questions at all. It uses a different WHO screening tool, ASSIST-Lite, which sorts recent cannabis users into "low/no risk," "moderate risk," and "high risk" categories rather than diagnosing disorder. The 2022 to 2023 survey found 71.6% low or no risk, 22.2% moderate risk, and 6.2% high risk among people who had used cannabis in the past three months.

That's a genuinely different instrument, but the flattening reappears anyway. The AIHW's own reporting still cites the global, DSM-based statistic that roughly one in ten people who use cannabis develop a disorder, without any severity breakdown. And when Australia's national mental health survey (the ABS) does produce a formal "substance use disorder" prevalence figure, 3.3% of Australians, or 647,900 people, it's generated via a structured interview, the CIDI, that operationalises the same DSM-5 polythetic criteria. Two different local instruments, neither well mapped onto the other in public reporting, both ultimately feeding the same "addiction" shorthand once a number reaches a headline.

Why the system allows this

This isn't unique to how SAM used its number. It's a structural feature of any polythetic, symptom-count diagnosis reported as a single topline figure. Nothing in a flat prevalence statistic distinguishes someone who ticked two mild boxes from someone who ticked nine severe ones. Anyone citing the raw total, whether a public health agency, a journalist, or an advocacy group of any persuasion, inherits that flattening. It's also worth noting that the largest funder of cannabis research, the US National Institute on Drug Abuse, has a statutory mandate to study drug abuse specifically. That shapes what gets funded and measured, independent of anyone's intent, and it means a large share of the available literature is built to produce disorder-framed findings in the first place.

SAM's release illustrates the mechanism cleanly. It takes the full CUD total and applies "addiction" to all of it, with no acknowledgment that a majority of that total sits in the tier the diagnostic literature itself says is least validated and least reliably measured. That's not a claim about SAM's motives. It's what the diagnostic architecture permits, used exactly as designed.

What a more honest number requires

Reporting severity tiers separately, noting that mild cases make up the largest share of the total, and flagging that the mild tier is the least stable on reassessment isn't complicated. The underlying data already supports it. Its absence in a headline is a choice, not a gap in available evidence.

For patients, doctors, and the legal cannabis sector, this isn't academic. An unexamined "X million addicted" figure shapes workplace policy, insurance underwriting, and how a patient's own prescribed use gets perceived by people around them, often against a number that was never built to distinguish their situation from an unrelated one.

Sources: Fink, D.S., Shmulewitz, D., Mannes, Z.L., Stohl, M., Livne, O., Wall, M., Hasin, D.S. Construct validity of DSM-5 cannabis use disorder diagnosis and severity levels in adults with problematic substance use. Journal of Psychiatric Research, 2022;155:387-394. Hasin, D.S., et al. Prevalence and Correlates of DSM-5 Cannabis Use Disorder, 2012-2013. American Journal of Psychiatry (2016). Compton, W.M., Han, B., Jones, C.M., Blanco, C. Cannabis use disorders among adults in the United States during a time of increasing use of cannabis. Drug and Alcohol Dependence, 2019;204:107468. NSDUH detailed tables, SAMHSA. AIHW, National Drug Strategy Household Survey 2022-2023 and Alcohol, tobacco & other drugs in Australia reports. Australian Bureau of Statistics, National Study of Mental Health and Wellbeing. SAM release: "Marijuana Is on Track to Become America's #1 Addiction," learnaboutsam.org, July 2026.

This article was researched and drafted with AI assistance and fact-checked against primary sources.

This article is published under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0. You may republish it for free, in print or online, provided you credit Marijuana.com.au, keep the article unchanged, and link back to the original. It may not be republished for commercial use or edited without permission. For syndication queries, contact syndicate@marijuana.com.au.