A study of over 11 million Ontarians found that individuals who visited an emergency department or were hospitalized for cannabis use had more than five times the adjusted risk of deliberate self-harm and more than nine times the risk of death by suicide compared to matched controls. However, the study measured hospital-level severity of cannabis involvement, not cannabis use in general, so these findings apply to a small, clinically severe subset of users and should not be extrapolated to the broader cannabis-using population.

Cannabis-related emergency department visits and hospitalizations have risen substantially in North America alongside legalization and increasing potency. Clinicians need reliable data on whether these acute presentations mark downstream psychiatric risk, particularly self-harm and suicide. This study offers the largest population-level quantification to date, creating an empirical foundation for integrating structured self-harm screening into cannabis-related acute care pathways. If the signal is real, every cannabis-related hospital visit represents a potential intervention point that most emergency departments are not currently designed to capture.

Cannabis-related emergency department visits and hospitalizations have increased across North America alongside rising product potency and expanding legal access. Prior systematic reviews have identified associations between cannabis use and self-injurious behavior, with one meta-analysis of 37 studies reporting a pooled odds ratio of 2.57 for self-harm among cannabis users. However, most prior studies have been limited by modest sample sizes, cross-sectional designs, and reliance on self-reported cannabis use. Fabiano and colleagues addressed these gaps by leveraging Ontario’s linked administrative health databases, which cover approximately 97% of the province’s residents, to construct a matched cohort of over 861,000 individuals. The study defined its exposure as an incident cannabis-related emergency department visit or hospitalization using validated ICD-10 and DSM-5 diagnostic codes, providing a clinical anchor that is both specific and readily identifiable in practice.

At three years, 8.22% of individuals with a cannabis-related hospital contact experienced deliberate self-harm requiring acute care, compared to 0.64% of matched general population controls, yielding a fully adjusted hazard ratio of 5.35 (95% CI 5.04 to 5.67). Death by suicide was rare in absolute terms (0.20% versus 0.01%) but the adjusted hazard ratio reached 9.22 (95% CI 5.24 to 16.23), though the wide confidence interval reflects low event counts. A key sensitivity analysis excluding individuals with any prior recorded mental health or substance use treatment still showed a 7.18-fold elevation in self-harm risk, suggesting the association is not entirely attributable to pre-existing psychiatric comorbidity. The authors note that the exposure captures hospital-level severity and not cannabis use broadly, and they identify residual confounding from unmeasured variables, underascertainment of non-hospital self-harm, and the need for prospective studies with direct cannabis use measurement as essential next steps.

Cannabis in the Emergency Room: A Signal We Should Not Ignore, and Should Not Overread

If someone lands in the emergency room because of cannabis, what happens next? A study of more than 11 million Canadians followed for five years has a striking answer, and a set of important asterisks that most headlines will skip. Fabiano and colleagues have produced one of the most rigorous population-level analyses to date connecting cannabis-related hospital presentations to subsequent deliberate self-harm and suicide. The scale is impressive, the statistical methods are appropriate, and the sensitivity analyses are well conceived. What the paper genuinely contributes is a quantified, actionable clinical signal: when someone presents to acute care for cannabis, the subsequent three-year risk of self-harm is approximately 8%, which is a number high enough to change practice. The finding that this risk persists even among those with no recorded mental health or substance use treatment history is particularly important, because it means that psychiatric comorbidity flags alone will not catch all of the patients who need screening. This study transforms a vague clinical intuition into something that can be built into emergency department protocols.

The central methodological limitation, however, is fundamental and cannot be overstated. The exposure in this study is not cannabis use. It is a hospital contact for cannabis. The difference matters enormously. Studying people who were admitted to hospital for alcohol poisoning and then concluding that “drinking alcohol raises self-harm risk fivefold” would similarly conflate alcohol use with acute alcohol toxicity. The finding applies to a severe subset, not the full spectrum of users. The vast majority of people who consume cannabis never present to an emergency department for it. Those who do are clinically distinct in ways the administrative data cannot fully capture: they may have higher baseline trauma exposure, more chaotic social circumstances, concurrent polysubstance use, or undiagnosed psychiatric conditions that drove both the hospital visit and the subsequent self-harm. Residual confounding from these unmeasured factors is substantial. If you compare people who ended up in the ER for any reason to those who never did, the ER group will have worse health outcomes on nearly every measure, not necessarily because of what brought them to the ER, but because of the underlying circumstances that made an ER visit necessary in the first place.

What I would tell a patient is direct: if cannabis use has brought you to the hospital, we need to talk about self-harm risk and make a follow-up plan before you leave. What I would tell a colleague is that this study justifies universal self-harm screening at cannabis-related acute contacts, but not the causal claim that cannabis itself is driving the risk. And what I would tell a policymaker is that these data support investment in ED-based screening pathways, not blanket restrictions on cannabis that would disproportionately affect the millions of users who will never see the inside of an emergency room for it. Large relative risks in observational studies of severity-selected exposures are scientifically informative and clinically actionable, but they are not causal estimates for the broader exposed population. The more extreme the selection criterion, the larger the relative risk, and the more cautiously it should be extrapolated.

This study sits at an important juncture in the research arc connecting cannabis use to psychiatric harm. Prior systematic reviews have demonstrated modest but consistent associations between cannabis use and suicidal ideation, self-harm, and completed suicide, with pooled estimates in the range of two to threefold increased odds. The present study substantially advances this literature by using a validated, clinically anchored exposure definition at unprecedented population scale. However, it occupies a position below prospective cohort studies with directly measured cannabis exposure and far below any interventional evidence, meaning it generates hypotheses rather than confirming causal pathways.

From a pharmacological standpoint, the study does not distinguish between acute cannabis intoxication, cannabis use disorder, and cannabis-induced psychosis, each of which may carry distinct risk profiles and interact differently with co-occurring substance use and prescribed medications. Clinicians should be aware that patients presenting acutely for cannabis may be on concurrent psychotropic medications, including SSRIs, benzodiazepines, or antipsychotics, where pharmacokinetic interactions with cannabinoids are incompletely characterized. The most concrete actionable recommendation is straightforward: any cannabis-related emergency department or hospital presentation should trigger a structured self-harm risk assessment and a documented safety plan, irrespective of whether the patient carries a formal psychiatric diagnosis.

This is a retrospective population-level matched cohort study using linked administrative health databases. It sits in the upper tier of observational evidence, below prospective cohort designs with direct exposure measurement and far below randomized trials. The single most important inference constraint is that the exposure is defined by hospital presentation for cannabis, not cannabis use itself, which introduces selection bias toward high-acuity cases and limits the generalizability of any risk estimate to the broader population of cannabis users.

These findings are broadly consistent with, but substantially larger in magnitude than, prior meta-analytic estimates. A 2021 systematic review of 37 studies reported a pooled odds ratio of 2.57 for self-injurious behavior among cannabis users, while a Swedish population-level study of nearly 7 million adults found an adjusted hazard ratio of 3.10 for suicide in individuals with cannabis use disorder. The present study’s higher point estimates (aHR 5.35 for self-harm, 9.22 for suicide) likely reflect its more severe exposure definition, as hospital-level cannabis presentations select for a higher-acuity population than self-reported cannabis use or outpatient diagnoses. This pattern confirms the dose-severity relationship observed across the literature: the more clinically severe the cannabis exposure, the stronger the association with adverse psychiatric outcomes.

The most consequential analytic choice was defining the exposure as a hospital contact rather than attempting to capture a broader spectrum of cannabis use through outpatient diagnostic codes or prescription records. Had the investigators broadened the exposure definition to include outpatient cannabis use disorder diagnoses, the absolute event rates in the exposed group would likely have been lower and the relative risk attenuated, because a less severely selected population would dilute the association. Conversely, the unmatched comparison with alcohol-contact individuals introduces systematic demographic and clinical differences between the two populations that matching could have partially addressed. A propensity-score matched cannabis-alcohol comparison might have yielded a different relative risk estimate for that secondary analysis.

The most likely overinterpretation is the claim that “cannabis use causes a fivefold increase in self-harm risk.” This exceeds the evidence on two counts. First, the study is observational and cannot establish causation. Second, and more importantly, the exposure is not cannabis use but a cannabis-related hospital contact, which captures a small, high-acuity minority of all cannabis users. Extrapolating these risk estimates to recreational, moderate, or medical cannabis populations is not supported by this design. Similarly, the sensitivity analysis excluding prior mental health treatment does not prove that cannabis itself drives self-harm independent of psychiatric illness, because unmeasured and undiagnosed psychopathology may still account for part of the residual association.

This study provides strong, population-level evidence that a cannabis-related hospital contact marks substantially elevated subsequent risk of deliberate self-harm and suicide, even among individuals without a recorded psychiatric history. It does not establish that cannabis use itself causes self-harm, and it does not apply to the broader population of cannabis users. For clinical practice today, the finding supports implementing universal self-harm screening at cannabis-related acute care presentations as a high-yield, identifiable intervention point.

Does this study prove that cannabis causes self-harm or suicide?

No. This is an observational study that demonstrates an association, not a causal relationship. The people studied had reached a hospital because of cannabis, which means many other factors, including trauma, social instability, and undiagnosed mental illness, could explain part or all of the elevated risk. A causal claim would require a different type of study design entirely.

Should I be worried about self-harm risk if I use cannabis occasionally?

This study examined people whose cannabis use was severe enough to result in a hospital visit, which represents a very small fraction of all cannabis users. The findings do not indicate that occasional or moderate cannabis use carries this level of risk. However, if cannabis use has ever prompted you to seek emergency medical care, it would be prudent to discuss self-harm risk with your physician.

Does this mean cannabis is more dangerous than alcohol for mental health?

The study actually found that individuals with alcohol-related hospital contacts had modestly higher self-harm risk than those with cannabis-related contacts. The two populations differ in many ways that were not fully accounted for, so direct comparisons between the two substances based on this study alone are unreliable.

What should emergency departments do differently based on this research?

The study suggests that any cannabis-related emergency department presentation should be treated as an opportunity for structured self-harm risk screening and safety planning, regardless of whether the patient has a documented psychiatric history. This is a practical, low-cost intervention that could help identify at-risk individuals at a moment when they are already receiving clinical attention.

Have thoughts on this? Share it: