VA Cannabis Misuse Diagnoses: Why the Data Isn't So Simple

October 7, 2026

Photo by RDNE Stock project via Pexels.

On October 5, 2026, Nate Landau published the fifth installment of an ongoing investigative series examining how the Department of Veterans Affairs handles cannabis-related diagnoses, this time taking direct aim at a statistic that's been circulating as proof of a crisis: a sharp rise in veterans diagnosed with cannabis use disorder. The op-ed, running via Marijuana Moment and built on Freedom of Information Act records obtained by the Veterans Action Council, doesn't deny that diagnosis numbers have climbed. It argues instead that the climb is a far messier story than the headline suggests, shaped as much by how the VA codes paperwork as by what veterans are actually doing with cannabis.

Landau isn't a disinterested observer here. He directs Contra Costa NORML and sits on the Veterans Action Council, which puts him squarely inside the advocacy apparatus pushing for expanded, better-documented veteran access to cannabis. That context matters for reading the piece, but it doesn't undercut the records he's working from, which span more than 17 years of VA data. His central thread is one that often gets lost in policy fights: a 2015 shift in how the VA codes diagnoses may be doing more to inflate the numbers than any actual change in veteran behavior. Whether that reframing holds up is worth walking through piece by piece.

What the FOIA Records Actually Show

Photo by https://kaboompics.com/ via Pexels.

The series Landau is contributing to has been working through VA records methodically, and this installment leans on internal correspondence rather than public-facing reports. A 2022 VA communication cited in the FOIA materials put the number of veterans with cannabis-positive urine drug screens that year at 134,790, with a nearly matching 139,336 veterans separately identified as carrying a cannabis use disorder diagnosis. Those two figures sitting so close together looks, at first glance, like confirmation that cannabis use among veterans translates almost one-to-one into a clinical disorder.

Landau's point is that this reading skips several steps. Part 5 follows four earlier installments in the Veterans Action Council's series, which have already surfaced a VA provider education packet on cannabis, an incident involving a dental-intake checkbox that appears to have fed into broader screening data in ways nobody anticipated, and an analysis of how federal rescheduling could affect veteran access to care without legal jeopardy. Each of those pieces has chipped away at the assumption that VA's cannabis-related numbers mean exactly what they appear to mean on their face.

What ties the series together is a simple methodological warning: raw counts, pulled out of administrative systems built for billing and compliance rather than clinical nuance, can be easily misread. A positive drug screen is not a diagnosis, and a diagnosis entered into a chart isn't automatically the product of a thorough clinical workup. Treating 134,790 and 139,336 as two sides of the same coin, Landau argues, requires ignoring how those numbers actually get generated inside VA's records systems.

The 2015 Coding Shift That Changed Everything

In 2022, the VA recorded over 134,000 cannabis-positive urine screens and more than 139,000 veterans diagnosed with cannabis use disorder (CUD), highlighting the growing scale of marijuana-related health issues among VA patients.

The mechanism Landau points to as the likely driver of inflated counts is a coding transition that has nothing to do with cannabis policy at all. In October 2015, the VA switched its diagnostic coding system from ICD-9 to ICD-10, bringing its recordkeeping in line with the DSM-5 framework that the American Psychiatric Association had published two years earlier, in 2013.

That alignment carried a quiet but significant consequence. Under the older DSM-IV framework, clinicians distinguished between cannabis abuse and cannabis dependence as two separate diagnostic categories, each with its own threshold and its own code. DSM-5 scrapped that split and merged both into a single combined diagnosis: cannabis use disorder, or CUD. On paper, that's a clinical refinement meant to better reflect how substance use problems actually present on a spectrum rather than as two distinct boxes. In practice, it also means every veteran who once might have been coded under either of two separate categories now gets funneled into one.

That's a mechanical effect, not a behavioral one. If 100 veterans were previously split between abuse and dependence codes, and all 100 now land under the single CUD code, a records system comparing pre-2015 and post-2015 numbers will show what looks like a surge in CUD diagnoses, even if not a single additional veteran actually developed a cannabis problem. Landau's core argument in this section of the op-ed is that nobody has cleanly separated how much of the reported increase reflects this coding consolidation versus how much reflects an actual rise in problematic use. Without that separation, a decade's worth of trend lines built on comparing ICD-9-era and ICD-10-era data may be measuring a definitional shift dressed up as an epidemiological one.

A Gap Between Use and Disorder

Photo by RDNE Stock project via Pexels.

Landau brings in national survey data to put the VA's internal diagnosis numbers in context, and the gap is striking. Citing NESARC-III figures included in the FOIA materials, the op-ed notes that 7.3 percent of U.S. veterans reported past-year cannabis use, while only 1.8 percent met full DSM criteria for cannabis use disorder. That's roughly a four-to-one ratio between people who use cannabis and people whose use rises to the level of a diagnosable disorder, which tracks with how clinicians generally understand substance use patterns: most people who use a substance, including alcohol, don't meet criteria for a use disorder tied to it.

The Veterans Action Council reads that gap as evidence that something inside the VA's own screening and coding pipeline is over-assigning the CUD label relative to what the broader clinical literature would predict. If national data suggests roughly one in four cannabis-using veterans might plausibly meet disorder criteria, but VA's internal numbers suggest something closer to a near match between positive screens and diagnoses, that discrepancy needs an explanation beyond genuine clinical severity.

One possible explanation shows up elsewhere in the same FOIA records: VA research reviewed as part of this reporting found that nearly a third of sampled patient charts contained no documented discussion at all between provider and patient about cannabis use. No notes on frequency, context, impact on daily functioning, or any of the criteria that actually distinguish casual use from a disorder. That's a meaningful documentation gap, and it raises an uncomfortable possibility: that some share of CUD diagnoses may be getting generated more or less automatically off a positive screen, without the conversation that's supposed to determine whether a disorder is actually present.

What Independent Research Adds to the Picture

Photo by https://kaboompics.com/ via Pexels.

Landau doesn't build his case in a vacuum, and the op-ed draws on independent academic research to show that at least part of the rise in CUD diagnoses looks real. A study led by Ofir Livne, conducted through Columbia University in partnership with VA Puget Sound, tracked diagnosis rates across the Veterans Health Administration system and found that CUD diagnoses more than doubled among VHA patients between 2005 and 2019, a period that spans the coding transition but extends well beyond it in both directions.

What makes that study useful here is its granularity. The sharpest increases in CUD diagnoses weren't spread evenly across the veteran population. They clustered among veterans with bipolar disorder or conditions on the psychotic spectrum, groups where cannabis use carries documented risks of exacerbating symptoms and where clinicians have reason to screen more closely and intervene more actively. That pattern suggests a genuine clinical signal sitting underneath the broader trend line, not just an artifact of how diagnoses get coded.

Landau's op-ed treats this as a complication rather than a contradiction. He isn't arguing that the entire rise in CUD diagnoses is a statistical mirage produced by the 2015 coding switch. His argument is narrower and, frankly, more defensible: that whatever real increase exists among vulnerable subpopulations is sitting on top of a measurement change that makes the overall magnitude of the trend unreliable. That distinction matters practically. If VA leadership is allocating screening resources or training based on a headline number that's partly a coding artifact, they may be over- or misdirecting attention away from the subgroups, like veterans with psychotic-spectrum conditions, where the Livne research suggests the real clinical risk is concentrated.

Why This Matters for Pending Legislation

Photo by Baldur93 via Pixabay.

None of this is happening in a policy vacuum. The Veterans Medical Marijuana Safe Harbor Act, H.R. 2435, sponsored by Rep. Earl Blumenauer, advanced out of the House Veterans' Affairs Committee on May 12, 2026, by an 18-14 vote. Its Senate companion, S. 1147, introduced by Sens. Cory Booker and Dan Sullivan, remains pending. Both bills aim, broadly, to protect VA providers who discuss cannabis with patients in legal-access states and to reduce the chilling effect that current policy can have on honest conversations between veterans and their doctors.

Rising CUD diagnosis numbers have been used in committee debate and in public commentary as evidence that expanding cannabis access for veterans carries real clinical costs. Landau's op-ed complicates that argument without fully dismantling it. If a meaningful share of the reported increase traces back to a 2013 DSM-5 reclassification and a 2015 VA coding transition rather than to an actual surge in problematic use, then citing the raw diagnosis trend as proof that access itself is driving harm is building an argument on uncertain ground.

The Veterans Action Council's position, as Landau frames it, isn't that the data should be ignored or that CUD isn't a real concern for some veterans. It's that cleaner, standardized data collection should be a parallel priority alongside any expansion of access, so that future debates over bills like H.R. 2435 and S. 1147 aren't fought with numbers nobody can fully vouch for. As always, cannabis legality and VA policy interact differently depending on the state a veteran lives in, and anyone navigating this terrain should confirm the current rules where they are rather than assume federal movement changes local realities overnight.

Strip away the advocacy framing on both sides and what's left is a narrower, more technical question than the headline numbers suggest. This was never really a debate about whether veterans can be trusted with cannabis access. It's a debate about whether the VA's own records infrastructure can be trusted to accurately describe what's happening inside it, given a coding transition, a diagnostic merger, and a documented gap in provider-patient conversations all sitting on top of one another.

Until the VA can show that a positive screen, a documented clinical conversation, and a final diagnostic code are consistently and transparently linked, any count of rising CUD cases ought to get a second look before it gets cited as settled fact in a committee hearing or a floor debate. That's true whether the number is being used to argue for tighter restrictions or for faster expansion of access.

Landau's op-ed won't be the last word here. The Veterans Action Council has signaled this is an ongoing series, and more FOIA records are likely to surface as H.R. 2435 and S. 1147 move, or stall, through Congress. Anyone following the veteran cannabis access debate would do well to keep watching this particular data fight as closely as the legislative one.