A large federal surveillance study found that cannabis-related healthcare visits among commercially insured working-age adults rose 71% between 2017 and 2022. Notably, state cannabis legal status did not meaningfully predict encounter rates in this population, challenging assumptions that legalization alone drives healthcare burden in younger adults.

Cannabis use among American adults is evolving rapidly alongside shifting legal frameworks, changing product potency, and declining perceived risk. Clinicians, health systems, and policymakers urgently need reliable data on whether cannabis-related healthcare demand is actually growing and which populations bear the greatest burden. This study, drawing on 115 million commercially insured individuals from the FDA’s own surveillance infrastructure, provides the most comprehensive picture yet of how cannabis-related encounters are trending in working-age adults, a population that has been substantially underrepresented in prior analyses focused on Medicare and VA cohorts.

As cannabis legalization expands across U.S. states and cannabis product potency continues to increase, the clinical and public health consequences of cannabis use in working-age adults remain poorly characterized. Prior federal studies have primarily examined Medicare beneficiaries over 65 and Veterans Affairs populations, leaving a major gap in understanding cannabis-related healthcare burden among younger, commercially insured individuals. Perez-Vilar and colleagues used the FDA Sentinel Distributed Database, which aggregates standardized administrative claims from four major national commercial insurers, to identify and characterize all cannabis-related healthcare encounters among adults aged 18 to 64 from 2017 through 2022. Cannabis encounters were defined using ICD-10-CM codes for cannabis-related disorders (F12x) and cannabis poisoning (T40.7x) in any diagnosis position.

Among 115 million eligible individuals, 963,345 (0.8%) had at least one cannabis-related encounter, generating over 5.6 million total encounters. Annual rates rose from 44.0 per 10,000 person-years in 2017 to 75.1 per 10,000 person-years in 2022, a 71% increase that was statistically significant by Mann-Kendall trend test. Outpatient and emergency department encounters drove the increase, while inpatient encounter rates appeared to plateau after 2020. Adults aged 18 to 25 had consistently the highest rates. Notably, no meaningful difference in encounter rates was observed between states with adult-use legalization, medical-only legalization, or no legal cannabis. Among those with encounters, psychiatric comorbidity was common: 30.6% had anxiety disorders, 25.7% had depression, and 31.7% had prior pain diagnoses. The authors acknowledge that rising rates may partly reflect improved clinical documentation rather than true increases in cannabis-related harm, and that ICD-10-CM codes used to identify encounters have not been validated against chart review.

More Cannabis Visits, No Legal-Status Effect: What Federal Surveillance Data Tell Us About Working-Age Adults

Between 2017 and 2022, the rate of cannabis-related healthcare encounters among commercially insured American adults nearly doubled. That is a striking number, but before drawing conclusions about whether cannabis is becoming more dangerous or legalization is driving harm, it is worth asking a harder question: are we seeing more cannabis-related illness, or are we just getting better at writing it down? This study, funded and overseen by the FDA, deploys one of the most powerful surveillance tools available in U.S. healthcare research. The Sentinel Distributed Database covers over 115 million commercially insured adults from four national insurers, offering unmatched scale within this population. And the study’s authors deserve credit for intellectual honesty. They present a null finding on legal status that would have been easy to downplay or bury, and instead highlight it as noteworthy. They acknowledge their coding limitations transparently. These are the marks of careful science. The 71% increase in encounter rates and the high psychiatric comorbidity profile among affected individuals are genuine contributions to our understanding of who is presenting for cannabis-related care.

But here is where interpretation must be precise. The study’s conclusion leans toward framing rising encounter rates as “consistent with” increasing cannabis use and potency. That framing is not wrong, but it is incomplete in a way that matters. Counting more arrests for a crime does not tell you whether crime is rising or police are patrolling more; you need independent data on both to interpret the trend. Similarly, rising ICD-10 cannabis codes in claims data could equally reflect expanded screening protocols, payer incentives to document substance use, greater patient willingness to disclose cannabis use in an era of destigmatization, or genuine increases in clinical harm. Without time-trend data on coding practices themselves, these explanations cannot be separated. The study also does not distinguish between F12x codes (cannabis use disorder) and T40.7x codes (cannabis poisoning), which carry very different clinical and policy implications. Nor does it tell us whether more individuals are presenting for the first time or whether the same individuals are returning more frequently, a distinction that separates new case incidence from intensification of existing burden.

The most scientifically provocative finding is the null result on state cannabis legal status. In older Medicare and VA populations, legal status does appear to correlate with encounter rates. In this younger, commercially insured population, it does not. Think of it this way: speed limits do not change how fast teenagers drive as much as they change how fast grandparents drive. For working-age adults who likely have access to cannabis regardless of legal framework, legalization may not be the dominant variable shaping healthcare encounters. If I were speaking to a patient, I would say: if you use cannabis and have anxiety, depression, or chronic pain, please tell your provider openly so we can evaluate whether it is helping or interfering. To a colleague: this trend is real, but be cautious about attributing it entirely to rising harm rather than rising documentation. To a policymaker: do not assume that prohibition keeps younger adults out of the emergency room for cannabis-related reasons, because this data suggests it does not. In surveillance studies using administrative data, a rising rate is a hypothesis, not a conclusion. It prompts investigation into whether the signal reflects changing real-world phenomena or changing measurement practices, and honest science requires holding both possibilities simultaneously until the evidence can distinguish between them.

This study sits at the surveillance and hypothesis-generating level of the research arc. It establishes a directional trend with high confidence but cannot explain why the trend exists, which clinical subgroups are most affected by actual cannabis-related harm, or what interventions might alter the trajectory. It complements, rather than replaces, the companion Medicare and VA analyses by demonstrating that findings in older populations do not automatically generalize to younger, commercially insured adults. The divergent legal-status result between age groups is a critical clue for researchers designing next-generation studies: it suggests that age, access dynamics, and behavioral norms may matter more than legal status for healthcare utilization in working-age populations.

From a pharmacological and safety standpoint, the high prevalence of psychiatric comorbidities among those with cannabis-related encounters reinforces existing evidence of bidirectional associations between cannabis use and conditions such as anxiety, depression, and psychotic disorders. Clinicians should note that 31.7% of individuals with cannabis-coded encounters also carried pain diagnoses, consistent with evidence that chronic pain is a risk factor for cannabis use disorder. The study does not capture product type, cannabinoid composition, or route of administration, so no conclusions about specific cannabis formulations or therapeutic versus recreational use can be drawn. The most concrete actionable recommendation is this: regardless of state legal status, clinicians should implement routine, nonjudgmental cannabis screening in outpatient and emergency department settings, particularly for adults aged 18 to 25 and those with co-occurring mental health or pain conditions.

This is a repeated cross-sectional descriptive surveillance study using administrative claims data from the FDA Sentinel Distributed Database. It occupies the lower tier of the evidence hierarchy, below cohort studies, case-control studies, and clinical trials. While its scale is exceptional, the most important constraint on inference is that administrative claims data cannot confirm whether cannabis was the clinical cause of the encounter, cannot verify coding accuracy against a reference standard, and cannot measure critical confounders including race, socioeconomic status, and COVID-19-related healthcare disruptions.

This study extends a growing body of federal surveillance research on cannabis-related healthcare utilization but provides a distinctive contribution in the commercially insured working-age population. Companion analyses using the same Sentinel infrastructure in Medicare and VA populations reported that cannabis legal status was associated with differential encounter rates among older adults, making the null legal-status finding in this younger cohort a meaningful point of divergence. Prior administrative studies in Medicaid populations and hospital discharge databases have documented rising cannabis-related hospitalizations, but none approach the scale of this analysis. The results confirm the directional trend of increasing cannabis-coded encounters seen across multiple data systems while raising a new question: whether the determinants of cannabis-related healthcare utilization fundamentally differ by age group and insurance status.

The most consequential analytic choice was the use of ICD-10-CM codes in any diagnosis position to define cannabis-related encounters, without distinguishing primary from secondary diagnoses and without disaggregating cannabis use disorder codes (F12x) from cannabis poisoning codes (T40.7x). If the analysis had been restricted to encounters where cannabis appeared as the primary diagnosis, the encounter rates and trends could look substantially different, potentially revealing whether the increase is driven by encounters specifically prompted by cannabis or by incidental documentation of cannabis use during visits for other reasons. Similarly, separate trend analyses for F12x and T40.7x codes would have allowed readers to distinguish between rising cannabis use disorder documentation and rising acute toxicity, which carry very different clinical and policy interpretations.

The most likely overinterpretation is that the 71% increase in encounter rates proves cannabis is becoming more dangerous or that cannabis use itself nearly doubled over this period. This exceeds what the evidence supports because administrative claims data capture coding events, not clinical realities. An increase in coded encounters can be driven by greater provider awareness, expanded screening, changed documentation incentives, or patient willingness to disclose cannabis use in addition to or instead of actual increases in cannabis-related harm. Equally, the null legal-status finding should not be read as proof that legalization is safe; it means the study could not detect a meaningful difference in this specific population and time frame, which may reflect unmeasured confounding, the dominance of illicit market access among younger adults, or limitations in how legal status was assigned.

This study establishes that cannabis-related healthcare encounter rates among commercially insured U.S. adults aged 18 to 64 rose substantially from 2017 to 2022, with younger adults and those with psychiatric comorbidities bearing the greatest burden. It does not establish whether the increase reflects growing harm or improved documentation, it does not demonstrate a causal relationship with legalization, and it does not apply to uninsured or publicly insured populations. For practice now, it supports routine cannabis screening in outpatient and emergency settings regardless of state legal status.

Does this study prove that cannabis is becoming more dangerous?

No. The study shows that more cannabis-related diagnoses are appearing in healthcare records, but it cannot distinguish whether this reflects increased harm, increased documentation of cannabis use by clinicians, or greater willingness by patients to report their use. Both possibilities are plausible, and the study design cannot separate them.

Does this mean legalizing cannabis has no effect on health?

Not exactly. The study found no meaningful difference in encounter rates between states with legal adult-use cannabis and states without it, but only in commercially insured adults aged 18 to 64. Studies of older populations have found legal-status effects. This null result may reflect the reality that younger adults can access cannabis through illicit channels regardless of legal status, but it does not prove legalization is without health consequences.

Should I be concerned about using cannabis if I have anxiety or depression?

This study found that about 30% of people with cannabis-related healthcare encounters also had anxiety disorders and about 26% had depression. While this does not prove cannabis caused those conditions, it underscores the importance of discussing your cannabis use openly with your healthcare provider, especially if you are managing a mental health condition, so that your provider can evaluate whether cannabis is helping or potentially interfering with your overall care.

Do these findings apply to everyone, or only certain populations?

These findings apply specifically to commercially insured adults aged 18 to 64. People who are uninsured, on Medicaid, or on Medicare were not included. The trends may look quite different in those populations, and readers should be cautious about applying these results broadly.

Have thoughts on this? Share it: