Cannabis use and suicide: a case-control study based on integrative data analysis

Article type: Research Article

Keywords: Cannabis, Case-control, Data fusion, Integrative data analysis, Multiple imputation, Suicide.

Authors: Qiongyu Shi, Guohua Li

Affiliations: https://ror.org/00hj8s172grid.21729.3f0000 0004 1936 8729Department of Biostatistics, Columbia University Mailman School of Public Health, 722 West 168th St, New York, NY 10032 USA; https://ror.org/00hj8s172grid.21729.3f0000 0004 1936 8729Department of Anesthesiology, Columbia University Vagelos College of Physicians and Surgeons, 622 West 168th St, New York, NY 10032 USA; https://ror.org/00hj8s172grid.21729.3f0000 0004 1936 8729Department of Epidemiology, Columbia University Mailman School of Public Health, 722 West 168th St, New York, NY 10032 USA

License: © The Author(s) 2025 CC BY 4.0 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Article links: DOI: 10.1186/s42238-025-00352-1  |  PubMed: 41257848  |  PMC: PMC12629057

Relevance: Moderate: mentioned 3+ times in text

Full text: PDF (881 KB)

Background

Suicide is a major source of mortality worldwide. Each year over 720,000 individuals die by suicide globally (WHO ref. 2025). In the United States, suicide rates have increased by 37% over the past two decades, claiming 49,476 lives in 2022 (CDC ref. 2025). Use of alcohol and drugs has been recognized as an important risk factor for suicidality (Wong et al. ref. 2013; Artenie et al. ref. 2015; Poorolajal et al. ref. 2016; Borges et al. ref. 2017; Lange et al. ref. 2024). Specifically, cannabis use as a contributing factor has been implicated in suicidal ideation and attempts among young adults, military veterans, and individuals with certain psychological conditions (Borges et al. ref. 2016; Gobbi et al. ref. 2019; Han et al. ref. 2021; Daneshmend et al. ref. 2022; Shamabadi et al. ref. 2023). With more states legalizing cannabis for medical and recreational use, the prevalence of cannabis use among US adults aged 18 and older continues to rise (Hasin and Walsh ref. 2021; Yang et al. ref. 2023). Toxicological testing data indicate that the prevalence of cannabinoids detected in suicide fatalities increased from 8% in 2006 to 23% in 2017 (National Center for Drug Abuse Statistics ref. 2025).

Previous studies assessing the role of cannabis use in suicidality relied on small sample sizes and self-reported data, limiting the validity and reliability of their findings. A major barrier to epidemiologic research on cannabis use and suicide has been the lack of toxicological testing data from representative samples of the general population. Although many studies examined the association between chronic cannabis use and suicidality, few have investigated the relationship between acute cannabis use and completed suicide. To assess the direct causal relationship between cannabis use and suicide in the US general population, we conducted a case-control study using data from multiple nationally representative samples.

Methods

This research does not meet the definition of human subjects research under 45 CFR 46.

Data sources

National Violent Death Reporting System (NVDRS)

The NVDRS is a comprehensive surveillance system of deaths from suicide and homicide in the United States. Each death record includes victim information, circumstances surrounding the death, and toxicological testing results wherever available (NCIPC ref. 2022). In 2013, there were 17 states (Alaska, Colorado, Georgia, Kentucky, Maryland, Massachusetts, North Carolina, New Jersey, New Mexico, Ohio, Oklahoma, Oregon, Rhode Island, South Carolina, Utah, Virginia, and Wisconsin) participating in the NVDRS, representing approximately 30% of the US population (Lyons et al. ref. 2016). In this study, we used the 2013 NVDRS as the data source of cases, operationally defined as individuals who died from suicide at age 16 years and older. Data on cannabis use and alcohol use for cases were based on toxicological testing results recorded in the NVDRS.

National Roadside Survey of Alcohol and Drug Use by Drivers (NRS)

The NRS is a field survey designed to monitor the prevalence of alcohol and drug use in the US driver population, which is sponsored by the National Highway Traffic Safety Administration and conducted every 7 to 10 years since 1973 (Kelley-Baker et al. ref. 2016). The 2013 NRS is the most recently available data. In the 2013 NRS, drivers were randomly selected at 300 locations from 60 sites across the contiguous United States, and data for verbally consented drivers were collected during a single 2-hour Friday daytime session (either 9:30–11:30 AM or 1:30–3:30 PM) and four 2-hour nighttime periods (10:00 PM–midnight and 1:00 AM–3:00 AM on Fridays and Saturdays). Data collected included self-reported demographic information, breath samples, oral fluid samples, and blood samples. Survey data were weighted according to the multistage sampling scheme and the response rate (71.0%) to obtain unbiased estimates (Kelley-Baker et al. ref. 2016). Data on self-reported quantity and frequency of cannabis use and alcohol use in the past year were included. In addition to self-reported cannabis and alcohol use, toxicological testing data were collected in the 2013 NRS.

National Survey on Drug Use and Health (NSDUH)

The NSDUH is an annual survey that collects comprehensive information about illicit drug use, alcohol consumption, and tobacco use among the US civilian, noninstitutionalized population aged 12 years and older. The 2013 NSDUH used a state-based sampling plan and a combination of computer-assisted personal interviewing and audio-computer-assisted self-interviewing techniques, with a response rate of 71.7%. The self-reported survey data were weighted to obtain unbiased national estimates for survey outcomes in the population represented by the 2013 NSDUH (SAMHSA ref. 2014). The NSDUH data included respondent demographic characteristics, mental health conditions, and self-reported cannabis use and alcohol use in the past month. In this study, the 2013 NSDUH was used as the data source of substantiative controls with multiply imputed toxicological testing data on cannabis use and alcohol use through the algorithms developed based on the 2013 NRS dataset.

Outcome measurements

The binary outcome measure was case-control status. Cases (n = 12,750) were individuals who died from suicide at age 16 years and older and who were recorded in the 2013 NVDRS, and controls (n = 43,465) were participants aged 16 years and older in the 2013 NSDUH.

Exposure measurements

In the NVDRS, cannabis use was defined as a binary exposure, with a blood delta 9-tetrahydrocannabinol (THC) level above the detection limit (0.01 µg/mL) of the tests indicating positive. Alcohol use was also defined as a binary exposure, with a blood alcohol concentration (BAC) level ≥ 0.01 g/dL indicating positive.

In the NRS, cannabis use and alcohol use were measured in both blood and oral fluid samples. Similarly, cannabis use positivity was defined as testing positive for THC. Alcohol positivity was determined when a BAC level was ≥ 0.01 g/dL. Data on self-reported last-time cannabis use (past 24 h, past 2 days, past month, over a month, beyond a year/never) and the frequency of weekly alcohol use (0, 1–2, 3–4, 5–7, 8–14, more than 14 drinks) were also collected in the NRS. We converted the categorical self-reported data on cannabis use and alcohol use to be comparable with the self-reported past-month cannabis use and alcohol use in the NSDUH.

Because NRS participants were not a representative sample of the general population, it would be inappropriate to use them as controls. However, the NRS dataset contained both self-reported and toxicological testing data on cannabis and alcohol use, which allowed us to model the relationship between self-reported cannabis and alcohol use and toxicological testing results. In order to obtain toxicological testing data for the NSDUH respondents, we improvised integrative data analysis techniques by fusing toxicological testing data in the NRS into the NSDUH. Specifically, we developed prediction models using machine learning (Lasso logistic regression) to estimate toxicology-confirmed cannabis and alcohol use from self-reported data and demographic characteristics. These prediction models allowed us to combine the strengths of two datasets – NRS (with toxicological testing data) and NSDUH (with representative controls) – to generate toxicological testing data for appropriate controls to study the relationship between cannabis exposure and suicide. In addition, to address concerns related to missing data in the NRS, NVDRS, and the combined NRS-NSDUH dataset, we used multiple imputation techniques to create 20 imputed datasets with the Chained Equations Multiple Imputation algorithm. This data fusion with multiple imputation approach has been applied to assessing the association between cannabis use and homicide (Lee et al. ref. 2024) and described in detail elsewhere (Yu et al. ref. 2024).

Covariates

Covariates included age (16–20, 21–34, 35–49, 50–64, and ≥65 years), sex (male vs. female), race (non-Hispanic White, non-Hispanic Black, Hispanic, and Others), and education (less than high school, high school graduate, some college, and college graduate or more). There were slight differences in categorizing age and race between datasets. We adopted the NSDUH categories by re-arranging some individuals into each category using raw data. For race/ethnicity, we re-grouped participants as Hispanic, non-Hispanic White, non-Hispanic Black, and all other non-Hispanic individuals as ‘Others’. We included these covariates because they are included in all three datasets using similar questionnaires, providing sufficient information for correct imputation. Additionally, they are well-established risk factors for suicide and potential confounders for the causal relationship between cannabis use and suicide.

Statistical analysis

We first generated descriptive statistics to assess the distributions of the exposure variables and covariates. Then, we examined the associations of cannabis use and alcohol use with suicide using weighted multivariable logistic regression models to calculate adjusted odds ratios (aORs) and 95% confidence intervals (CIs). Bootstrap sampling was used to obtain valid estimates of aORs and 95% CIs (Lee et al. ref. 2024). We fitted three models to generate the estimated aORs and 95% CIs: Model 1 was based on the complete NVDRS cases and NRS controls; Model 2 was based on NVDRS cases and NRS controls with 20 multiple imputations; and Model 3 was based on NVDRS cases and NSDUH controls with the data fusion approach and 20 multiple imputations.

Results

There were considerable differences in demographic characteristics between cases (i.e., suicide deaths recorded in the 2013 NVDRS) and controls (i.e., respondents to the 2013 NSDUH) (Table 1). Specifically, cases were more likely than controls to be male (77.4% vs. 48.2, p < 0.001), be non-Hispanic White (84.9% vs. 65.5%, p < 0.001), and have received only a high school education or less (61.5% vs. 41.5%, p < 0.001). Age distribution, however, was similar between cases and controls (Table 1). As a select sample of drivers, participants in the 2013 NRS differed from both the NVDRS cases and the NSDUH controls and were disproportionately young adults aged 21–34 years (39.7% vs. 23.0% and 24.0%, p < 0.001), non-Hispanic Black (24.4% vs. 5.8% and 11.8%, p < 0.001), and attained education beyond high school (69.1% vs. 38.5% and 58.4%, p < 0.001) (Table 1).

Over one quarter (28.7%) of the cases in NVDRS were tested for THC; of them, 16.9% were positive (Table 2). Of the NRS participants, 44.7% were tested for THC, with 9.4% being positive (Table 2). Blood alcohol testing results showed that 39.5% of the cases in NVDRS and 2.4% of the NRS participants had elevated BACs (Table 2). Results from multiply imputed data were consistent with those from actual testing data. The data fusion approach with multiple imputation estimated that of the NSDUH respondents, 7.2% were positive for THC and 2.6% were positive for alcohol (Table 2).

Multivariable logistic regression modeling revealed that cannabis use and alcohol use were each significantly associated with increased odds of suicide (Table 3). The case-control analysis based on the data fusion approach yielded an aOR of 1.83 (95% CI: 1.17, 2.79) for the association between cannabis use and suicide and an aOR of 20.53 (95% CI: 11.83, 29.24) for the association between alcohol use and suicide (Table 3, Model 3). Results from the three models were similar. In all the models, significantly increased odds of suicide were found in males, non-Hispanic Whites,, and those with less education attainment (Table 3).

Discussion

Our analysis of multiple national data systems indicates that cannabis use is associated with 83% increased odds of suicide. This finding is consistent with the Han et al. study (Han, et al., ref. 2021), in which the prevalence of suicidality for cannabis users was significantly higher than for cannabis non-users among NSDUH respondents. The validity of our finding that there exists a robust association between cannabis use and increased suicide risk is further bolstered by the corroborative evidence that reaffirms several well-established risk factors for suicide, such as alcohol use, male sex, non-Hipanic White race, and less education attainment (Flensborg-Madsen et al. ref. 2009; Borges et al. ref. 2017; Peterson et al. ref. 2020).

Our study contributes empirical evidence to the growing body of research linking cannabis use to increased suicide risk. Numerous studies have demonstrated a strong association between cannabis use and the development of psychotic disorders, such as schizophrenia (Volkow et al. ref. 2014) and depression (Lev-Ran et al. ref. 2014). These mental health issues can directly influence suicidal behavior, exacerbated by additional factors such as relationship problems, lower career achievement, and reduced life satisfaction. Furthermore, chronic cannabis use has been linked to cognitive and motor impairments, which may disrupt daily functioning and independence. For instance, impaired motor skills and slowed reaction time can affect driving ability, sometimes resulting in driving cessation—a loss of autonomy that has been associated with increased depression and suicide risk (Chihuri et al. ref. 2016; Ko et al. ref. 2021).

This study has several notable strengths. First, we used the case-control design and integrative data analysis techniques, including a novel data fusion approach to handling missing toxicological testing data. Selecting cases and controls from nationally representative samples enhances the generalizability of our findings compared to studies conducted in specific geographic regions or population groups. Additionally, by incorporating toxicological testing results instead of relying on self-reported data to determine cannabis use exposure, we minimized misclassification bias.

Several limitations should be taken into consideration when interpreting the study results. First, all toxicological testing data for NSDUH controls were imputed using prediction models trained on NRS data. This approach assumes that the relationship between self-reported and toxicology-confirmed use of cannabis and alcohol is consistent across populations. Differences in reporting behavior, such as underreporting in roadside survey settings, may limit calibration of the imputation model and underestimate uncertainty. Second, as NRS data cannot be disaggregated to state level, we could not restrict the analysis to the 17 states participating in the 2013 NVDRS. Differences in cannabis policy across states may therefore contribute to residual confounding. Also, our study may have limited applicability to certain racial and ethnic minorities due to inadequate representations in the study samples. Third, despite leveraging data fusion and multiple imputations to mitigate inaccuracies in self-reporting, selection bias might still be present due to the sampling methodology employed by NRS that targeted specific time windows on weekends. Fourth, the high prevalence of missing toxicological testing data on cannabis use from the NRS dataset may affect the accuracy of the multiply imputed data, as demonstrated in the discrepancies in the odds ratios derived from models before and after imputations using NRS data (Table 3). Fifth, our study results are susceptible to unmeasured confounding as some confounding factors, such as substance use disorder and history of depression and other mental health problems, are not accounted for due to the lack of data. Psychiatric conditions, such as major depression, schizophrenia, and anxiety disorder, have been recognized as important risk factors for suicide (Harris and Barrachlough ref. 1997; Reutfors et al. ref. 2009; Rivlin et al. ref. 2010), and have been found to be positively associated with cannabis use (Bovasso ref. 2001; Patton et al. ref. 2002). Finally, this study does not assess the dose-response relationship between cannabis use and suicide because it is not feasible to multiply impute quantitative cannabis data.

Conclusions

Results of this study indicate a positive association between cannabis use and suicide. Specifically, we found that cannabis use, as indicated by toxicological testing results, is associated with 83% increased odds of suicide with adjustment for demographic characteristics and alcohol use. If confirmed, the excess risk of suicide associated with cannabis use, along with other adverse health outcomes related to cannabis use, should be taken into consideration in the development and evaluation of cannabis policies.

References

1. AA Artenie, J Bruneau, É Roy, G Zang, F Lespérance, J Renaud. Licit and illicit substance use among people who inject drugs and the association with subsequent suicidal attempt. Addiction, 2015. [DOI | PubMed]
2. G Borges, CL Bagge, R Orozco. A literature review and meta-analyses of cannabis use and suicidality. J Affect Disord, 2016. [DOI | PubMed]
3. G Borges, CJ Cherpitel, R Orozco, Y Ye, M Monteiro, W Hao. A dose-response estimate for acute alcohol use and risk of suicide attempt. Addict Biol, 2017. [DOI | PubMed]
4. GB Bovasso. Cannabis abuse as a risk factor for depressive symptoms. Am J Psychiatry, 2001. [DOI | PubMed]
5. Centers for Disease Control and Prevention (CDC). Suicide Data and Statistics. https://web.archive.org/web/20250603013408/https://www.cdc.gov/suicide/facts/data.html?CDC_AAref_Val=https://www.cdc.gov/suicide/suicide-data-statistics.html. Accessed 3 June 2025.
6. S Chihuri, TJ Mielenz, CJ DiMaggio, ME Betz, C DiGuiseppi, VC Jones. Driving cessation and health outcomes in older adults. J Am Geriatr Soc, 2016. [DOI | PubMed]
7. AZB Daneshmend, J Stewart, DA Jarkas, SI Franklyn, RL Gabrys, ZR Patterson. Examining risk factors in the cannabis–suicide link: considering trauma and impulsivity among university students. Int J Environ Res Public Health, 2022. [DOI | PubMed]
8. T Flensborg-Madsen, J Knop, EL Mortensen, U Becker, L Sher, M Grønbaek. Alcohol use disorders increase the risk of completed suicide–irrespective of other psychiatric disorders. A longitudinal cohort study. Psychiatry Res, 2009. [DOI | PubMed]
9. G Gobbi, T Atkin, T Zytynski, S Wang, S Askari, J Boruff. Association of cannabis use in adolescence and risk of depression, anxiety, and suicidality in young adulthood: a systematic review and meta-analysis. JAMA Psychiatr, 2019. [DOI]
10. B Han, WM Compton, EB Einstein, ND Volkow. Associations of suicidality trends with cannabis use as a function of sex and depression status. JAMA Netw Open, 2021. [DOI | PubMed]
11. EC Harris, B Barraclough. Suicide as an outcome for mental disorders. A meta-analysis. Br J Psychiatry, 1997. [DOI | PubMed]
12. D Hasin, C Walsh. Trends over time in adult cannabis use: a review of recent findings. Curr Opin Psychol, 2021. [DOI | PubMed]
13. T Kelley-Baker, JH Lacey, A Berning, A Ramirez, C Moore, K Brainard. 2013–2014 National roadside study of alcohol and drug use by drivers: methodology (Report No), 2016
14. TM Ko, VA Kalesnikava, D Jurgens, B Mezuk. A data science approach to estimating the frequency of driving cessation associated suicide in the US: evidence from the National violent death reporting system. Front Public Health, 2021. [DOI | PubMed]
15. S Lange, L Llamosas-Falcón, KV Kim, AM Lasserre, H Orpana, CL Bagge. A dose-response meta-analysis on the relationship between average amount of alcohol consumed and death by suicide. Drug Alcohol Depend, 2024. [DOI | PubMed]
16. S Lee, G Li, S Chihuri, Y Yu, Q Chen. Using data fusion with multiple imputation to correct for misclassification in self-reported exposure: a case-control study of cannabis use and homicide victimization. Inj Epidemiol, 2024. [DOI | PubMed]
17. S Lev-Ran, M Roerecke, B Le Foll, TP George, K McKenzie, J Rehm. The association between cannabis use and depression: a systematic review and meta-analysis of longitudinal studies. Psychol Med, 2014. [DOI | PubMed]
18. BH Lyons, KA Fowler, SP Jack, CJ Betz, JM Blair. Surveillance for violent deaths: National violent death reporting system, 17 states, 2013. MMWR Surveill Summ, 2016. [DOI | PubMed]
19. National Center for Drug Abuse Statistics. Drug Abuse Statistics. https://web.archive.org/web/20250627133456/https://drugabusestatistics.org/. Accessed 27 June 2025.
20. National Center for Injury Prevention and Control (NCIPC)National violent death reporting system web coding manual version 62022Atlanta, GACenters for Disease Control and PreventionNational Center for Injury Prevention and Control (NCIPC). National violent death reporting system web coding manual version 6. Atlanta, GA: Centers for Disease Control and Prevention; 2022.
21. GC Patton, C Coffey, JB Carlin, L Degenhardt, M Lynskey, W Hall. Cannabis use and mental health in young people: cohort study. BMJ, 2002. [DOI | PubMed]
22. C Peterson, A Sussell, J Li, PK Schumacher, K Yeoman, DM Stone. Suicide rates by industry and occupation—National violent death reporting system, 32 states, 2016. MMWR Morb Mortal Wkly Rep, 2020. [DOI | PubMed]
23. J Poorolajal, T Haghtalab, M Farhadi, N Darvishi. Substance use disorder and risk of suicidal ideation, suicide attempt and suicide death: a meta-analysis. J Public Health (Oxf), 2016. [DOI | PubMed]
24. J Reutfors, L Brandt, EG Jönsson, A Ekbom, P Sparén, U Osby. Risk factors for suicide in schizophrenia: findings from a Swedish population-based case-control study. Schizophr Res, 2009. [DOI | PubMed]
25. A Rivlin, K Hawton, L Marzano, S Fazel. Psychiatric disorders in male prisoners who made near-lethal suicide attempts: case-control study. Br J Psychiatry, 2010. [DOI | PubMed]
26. A Shamabadi, A Ahmadzade, K Pirahesh, A Hasanzadeh, H Asadigandomani. Suicidality risk after using cannabis and cannabinoids: an umbrella review. Dialogues Clin Neurosci, 2023. [DOI | PubMed]
27. Substance Abuse and Mental Health Services Administration (SAMHSA). Results from the 2013 National Survey on Drug Use and Health: Summary of National Findings. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2014. NSDUH Series H-48, HHS Publication No. (SMA) 14-4863.
28. ND Volkow, JM Swanson. Effects of cannabis use on human behavior, including cognition, motivation, and psychosis: a review. JAMA Psychiatr, 2014. [DOI]
29. SS Wong, B Zhou, D Goebert, ES Hishinuma. The risk of adolescent suicide across patterns of drug use: a nationally representative study of high school students in the United States from 1999 to 2009. Soc Psychiatry Psychiatr Epidemiol, 2013. [DOI | PubMed]
30. World Health Organization (WHO). Suicide. https://web.archive.org/web/20250717081532/https://www.who.int/news-room/fact-sheets/detail/suicide. Accessed 17 July 2025.
31. KH Yang, RM Tam, N Satybaldiyeva, W Kepner, BH Han, AA Moore. Trends in past-month cannabis use among US adults across a range of disabilities and health conditions, 2015–2019. Prev Med, 2023. [DOI | PubMed]
32. Y Yu, R Little, M Perzanowski, Q Chen. Multiple imputation of more than one environment exposure with non-differential measurement error. Biostatistics, 2024. [DOI | PubMed]