Cannabis, Tobacco Use, and COVID-19 Outcomes

Key Points Question Is cannabis use associated with COVID-19 outcomes? Findings In this cohort study of 72 501 patients diagnosed with COVID-19 in a large medical center, individuals who used cannabis had a higher risk of hospitalization and intensive care unit admission compared with those not using cannabis after controlling for other risk factors. Meaning These findings suggest the need to evaluate the potential impact of cannabis use on COVID-19 outcomes given the growing legalized use of cannabis.


Introduction
COVID-19 continues to be a public health concern, leading to morbidity and mortality.While nearly 76% of US adults have received at least 1 dose of the COVID-19 vaccine, 1

+ Supplemental content
Author affiliations and article information are listed at the end of this article.
Cigarette smoking has been found to be associated with more severe COVID-19 infection, including higher rates of disease progression, 9 hospitalization, intensive care unit (ICU) admission, oxygen requirements during hospitalization, and mortality following COVID-19. 4,81][12] In a large study using electronic health record (EHR) data, a substance use disorder diagnosis was associated with increased risk of COVID-19 and adverse outcomes, such as mortality and hospitalization. 11Another study suggests that there is an increased risk of COVID-19 breakthrough infections for people with substance use disorders. 11In a study of college students, high-risk alcohol use was associated with greater SARS-CoV-2 infection incidence, but not with COVID-19 outcomes. 10spite the increasing availability of cannabis, research on cannabis and COVID-19 outcomes has been limited.With regards to cannabis use specifically, some evidence has suggested that people who use cannabis are more likely to contract COVID-19 and less likely to survive the virus than nonusers. 13Other evidence suggests a protective effect of cannabis use on COVID mortality. 14Given existing limited and conflicting findings, more evidence is needed on the association between substance use-particularly cannabis-and health outcomes following COVID-19 infection.
Clinical data available in EHRs can be a powerful tool for examining gaps in knowledge about the association of substance use with COVID-19 outcomes.The objective of this study is to examine a large sample of patients with COVID-19 to evaluate whether substance use (ie, tobacco smoking and cannabis use) is associated with several COVID-19-related outcomes, including hospitalization, ICU admission, and all-cause mortality.We hypothesized that tobacco smoking and cannabis use would be associated with worse outcomes following a COVID-19 infection.

Study Design
or (4) a positive result on a COVID-19 antigen test.eFigure 1 in Supplement 1 shows the sample size filters of age range and missing data for inclusion in this study.

Outcome Variables
The primary outcomes measured were hospitalization, ICU admission, and all-cause mortality.
Posthospital mortality and other mortality outside the period a patient was hospitalized were included.Overall survival was also assessed using time-to-event analyses for those patients who had a documented date of mortality within our study period.

Covariates
Demographic-and treatment-related covariates, including patient age, sex, race and ethnicity, health insurance coverage, and date of COVID-19 diagnosis were extracted from the EHR.Patient race and ethnicity were self-reported and documented by the rooming staff during all routine clinical encounters.The racial categories included American Indian or Alaska Native, Asian, Black, Other Pacific Islander, White, and other.Ethnic categories include Hispanic and non-Hispanic.Racial and ethnic categories other than Black and White were combined into the other category due to limited sample sizes.Race and ethnicity data were collected because they were included in routine clinical workflow and the EHR data extraction, and it was important to evaluate whether race and ethnicity were associated with COVID-19 outcomes along with other factors.Insurance status was coded yes if patients had Medicare, Medicaid, commercial insurance, or listed other as a form of insurance; it was coded no if patients were uninsured or self-pay.
Tobacco smoking and cannabis use were assessed using encounter-level data from the EHR.For smoking, patients self-reported current, former, or never smoking.Only patients with documented smoking status (current, former, or never) were included in the analyses (13% of patients were excluded due to missing smoking status).Current cannabis use was coded positive when positive marijuana use status was documented yes in the encounter medical record.Alcohol abuse in past 3 years and current vape use (e-cigarette or electronic nicotine delivery systems) were coded positive when such use was documented in the medical record.Vaping was considered an independent covariate and not part of tobacco or cannabis use.Discrete fields in the medical record, rather than free-text scanning using normal language processing, were extracted using the standardized extraction code created by the consortium. 18 also assessed various comorbidities (ICD-10-CM codes extracted from the discrete diagnoses documented within 3 years prior to COVID-19 infection) that have previously been shown to affect COVID-19 outcomes (Centers for Disease Control [CDC] Higher Risk Category), 19 including history of malignant neoplasm, chronic kidney disease, chronic obstructive pulmonary disease, diabetes (type 1 or type 2), cardiovascular disease, obesity, and pregnancy.Obesity was defined as a body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) greater than 30 for persons aged 20 years and older. 20Given that BMI is an unreliable measure of obesity in adolescents, the growth percentile classification scheme was used for individuals younger than 20 years, with those at the 95th percentile or greater classified as having obesity.

Statistical Analysis
Cohort descriptive data were presented as numbers (proportions) with appropriate χ 2 test statistics for categorical variables.To evaluate the association of substance use and COVID-19 outcomes (hospitalization, ICU admission, and mortality), we used both univariate and multivariable logistic regression models, adjusting for age, sex, race and ethnicity, health insurance status, comorbidities (composite), date of COVID-19 diagnosis, smoking status, and cannabis use.We also performed timeto-event analysis for overall survival using Cox proportional hazards regression models to evaluate the association of substance use and all-cause mortality.The proportional hazards assumption was checked using the scaled Shoenfeld residuals.Separate multivariable models were also constructed to assess comorbidities individually in sensitivity analyses.The association between alcohol abuse

JAMA Network Open | Substance Use and Addiction
Cannabis, Tobacco Use, and COVID-19 Outcomes

Comparing Current and Former Smoking
In addition to the previous comparisons with never smoking, we compared current vs former smoking.We found a significantly higher probability of hospitalization (OR, 1.28; 95% CI, 1.20-1.38;P < .001)but not ICU admission (OR, 0.98; 95% CI, 0.88-1.10;P = .78)or all-cause mortality (OR,

ICU Admission
Similarly, cannabis use was associated with an increased risk of ICU admission following COVID-19.

Mortality
However, cannabis use was not associated with an increase all-cause mortality following COVID-19.

Adjusting for Comorbidities
eTable 1 in Supplement 1 shows the frequency of comorbid conditions in this sample.Further analysis was done to see whether any of the 7 comorbidities defined by the CDC as a tier 1 comorbidity were individually associated with COVID-19 outcomes.These analyses found that these comorbid conditions increased the risk of COVID-19 outcomes.We reached similar results regarding tobacco and cannabis use when adjusting for each of the comorbid conditions (eTable 5 in Supplement 1).We have conducted additional analyses with refined categories for covariates (insurance types, ethnicity, age groups) and reached similar results (data not shown).

Tobacco, Cannabis Use, and Receipt of COVID-19 Vaccine
The

Other Substance Use and COVID-19 Outcomes
EHR data are limited for other substance use.Alcohol abuse in the past 3 years was documented in 250 patients (0.3%), and current vape use was documented in 1384 patients (1.9%) (eTable 1 in Supplement 1).These variables are known to be severely underrecorded in the EHR, therefore limiting statistical power significantly. 17Despite these limits, the OR for alcohol abuse and hospitalization was greater than 1, but the P value was greater than our prespecified level of statistical significance (OR, 3.34; 95% CI, 2.08-5.69;P = .01)(eTable 8 in Supplement 1).Likewise, the OR for  vaping and hospitalization was greater than 1, but the P value did not meet our prespecified level of statistical significance (OR, 1.20; 95% CI, 1.06-1.37;P = .006).There was not sufficient data on alcohol and vaping to evaluate their association with COVID-19-related ICU admission and mortality (eTable 8 in Supplement 1).

JAMA Network Open | Substance Use and Addiction
Cannabis, Tobacco Use, and COVID-19 Outcomes

Discussion
Given the continued risk of COVID-19, this study extends current evidence on the potential impact of substance use on COVID-19 outcomes.Using EHR data from a large medical center, we provide further evidence on the association of tobacco use with an increased risk of hospitalization, ICU admission, and all-cause mortality related to COVID-19 infection.Importantly, we present new evidence suggesting that cannabis use may be associated with an increased risk of hospitalization and ICU admission following COVID-19, while adjusting for other factors, such as tobacco smoking, comorbidities, and COVID-19 vaccination before diagnosis.
Our findings may help clarify the complex multidimensional impact of tobacco smoking on COVID-19 outcomes.While some research indicates a protective association between smoking and COVID-19 severity (referred to as a smoker's paradox 21 ), most research demonstrates that tobacco smoking is associated with an increased risk of symptomatic infection with SARS-CoV-2 as well as an increased risk of disease progression. 4,6,7Interestingly, some studies indicate increased severity of COVID-19 infection in individuals who formerly smoked, including higher rates of hospitalization, ICU admission, oxygen requirement during hospitalization, in-hospital mortality, but not in those who currently smoke. 4Here, we presented evidence using EHR data of more than 72 000 COVID-19 cases and showed that current and former smoking status were both associated with poor COVID-19 outcomes, characterized by an increased risk of hospitalization, ICU admission, and all-cause mortality following COVID-19, compared with those who have never smoked, after considering other risk factors.
Given the rising availability of cannabis, these findings also contribute to the existing limited research on potential effects of cannabis use on COVID-19 outcomes. 7,22A recent study shows that any substance use disorder was associated with worse COVID-19 outcomes; however, the design did not have large enough sample sizes to evaluate the association of specific substances, such as cannabis, with COVID-19 severity. 12Another study suggested a protective association of cannabis with COVID-19 mortality 23 ; however, their sample size was smaller and unidentified collider bias could be an important source of paradoxical associations. 23In this study, we provide evidence of an association between cannabis use and poorer COVID-19 outcomes characterized by both an increased risk of hospitalization and ICU admission.Furthermore, we presented preliminary data on the association between other forms of substance use, including alcohol abuse and vaping, and COVID-19 outcomes.There is currently very limited research examining the association of vaping (ie, e-cigarette use) and the severity of COVID-19 outcomes in patients. 24In our study, we presented preliminary findings that vaping may be associated with an increased risk of hospitalization, despite limited documentation regarding vape use in the EHR data.Similarly, there is research on increased alcohol consumption during the COVID-19 pandemic, 25 while little is known about the association between alcohol use and COVID-19 outcomes.One study found that alcohol use was associated with an increased risk of SARS-CoV-2 infection in a small cohort of college students. 10We present a potential association between alcohol abuse and increased risk of hospitalization following COVID-19 infection.Notably, further studies are needed, as our findings were limited by small sample sizes and limited documentation within our EHR database.

Limitations
This study has limitations.First, the study spanned

Conclusions
This cohort study found that both current and former smoking were associated with an increased risk among patients with COVID-19 for hospitalization, ICU admission, and all-cause mortality.These associations remained after adjusting for demographic and comorbidity factors.Specifically, older patients who reported current or former smoking showed a faster progression to all-cause mortality than those who reported never smoking.In addition, cannabis use was associated with an increased risk of hospitalization and ICU admission among patients with COVID-

Findings
many factors, including vaccine hesitancy and the emergence of new, more virulent strains of the SARS-CoV-2 virus, highlight the continued importance of identifying factors that contribute to poor outcomes from this viral Key Points Question Is cannabis use associated with COVID-19 outcomes?In this cohort study of 72 501 patients diagnosed with COVID-19 in a large medical center, individuals who used cannabis had a higher risk of hospitalization and intensive care unit admission compared with those not using cannabis after controlling for other risk factors.Meaning These findings suggest the need to evaluate the potential impact of cannabis use on COVID-19 outcomes given the growing legalized use of cannabis.
Cannabis, Tobacco Use, and COVID-19 Outcomes and vape use and risk of hospitalization was also assessed in a sensitivity analysis given the underdocumentation of these substance use risk factors.Given the hypotheses on the association of tobacco smoking and cannabis use on 3 major COVID-19 outcomes involving approximately 9 tests of association, we adjusted the significance thresholds from .05 to .005 for a Bonferroni correction to correct for multiple comparisons.All analyses were conducted using R version 3.5.3(R Project for Statistical Computing), SAS version 9.4 (SAS Institute), and SPSS version 27 (IBM Corp).

Figure 2 .
Figure 2. Association of Tobacco Smoking and COVID-19-Related Hazard of Mortality, Stratified by Age

Table 1 .
Distribution of Patient Characteristics and Outcomes Involving Hospitalization, ICU Admission, and Mortality

Table 2 .
Associations of Patient Characteristics With Outcomes of Hospitalization, ICU Admission, and Mortality Among 72 501 Individuals with COVID-19 a characteristics of patients who received vaccination before diagnosis are shown in eTable 6 in Supplement 1.A total of 19 410 patients (26.8%) received a COVID-19 vaccine before the diagnosis.
24 months (February 2020 to January 2022), which may have included significantly different SARS-CoV-2 disease manifestations due to the emergence of new variants, time-varying policies related to universal masking and lockdowns, and the introduction of the COVID-19 vaccine in December 2020.To reduce this concern, we included date of diagnosis and vaccination before diagnosis in our multivariable models to reduce the confounding effect of different outcomes related to time.However, caution should still be exercised when interpreting our results due to the potential for persistent confounding.Second, EHR data are limited by relying on patient self-report of substance use and subsequent documentation by medical staff.Therefore, substance use data quality within EHR often suffers from variable reporting and missing documentation.The best existing measure in the EHR data, current cannabis use, is a very crude measure without specific details on cannabis type, frequency, or recency.We have tried to reduce this bias by using data from all available hospital encounters.These findings should be viewed with caution because a detection bias is possible if heavier marijuana users were more likely to have that status documented and were at greater risk for morbidity.This study sample representing patients who required health care services for COVID-19 may not be representative of the general population regarding their substance use.Furthermore, there were insufficient data on the types of tobacco products (eg, cigars, pipe) because product types were often undocumented and nonmandatory in most medical encounters.We acknowledge that health care system-wide EHR-based data suffer from these biases despite the large set of clinical data they represent.Third, additional factors, such as mental health status, were not included in this investigation given their potential impacts on substance use and health outcomes.Additionally, these findings are based on information in our EHR data.Although our EHR tracks patient mortality outcomes beyond our health care system, it is possible that the data do not fully capture outcomes for patients who sought care or died at another institution.
19.Our preliminary data also suggest a need for further investigation into whether other forms of substance use, including nicotine and cannabis vaping and alcohol abuse, are associated with worse COVID-19 outcomes.Overall, this research calls for further investigation into the associations of tobacco and cannabis use with COVID-19 outcomes.Given the recent legalization of recreational marijuana use in more states, including the area served by this academic medical center, further research may aid in guiding interventions, such as substance use prevention and treatment, that would benefit patient outcomes moving forward in the COVID-19 pandemic and the associated heath consequences it will have in our communities.Comorbidity Details and Other Clinical Characteristics of Patients with COVID-19 and Outcomes Involving Hospitalization, ICU Admission, and Mortality eTable 2. Associations of Patient Characteristics and the Hazard of Mortality eTable 3. Association of Tobacco Smoking and COVID-19-Related Hazard of Mortality, Stratified by Age eTable 4. Specific Comparison Between Patients With Current Smoking vs Patients With Former Smoking eTable 5. Associations of Patient Characteristics (Including Comorbidity Details) and Outcomes of Hospitalization, ICU Admission, and Mortality eTable 6. Characteristics of Patients With COVID-19 and Outcomes of Receiving COVID-19 Vaccine Before Diagnosis eTable 7. Associations of Patient Characteristics and the Outcome of Receiving the COVID-19 Vaccine Before Diagnosis Using Logistic Regression eTable 8. Associations of Patient Characteristics (Including Vaping and Alcohol Use) With Outcomes of Hospitalization, ICU Admissions, and Mortality eFigure 1. Flowchart Demonstrating the Data Filter and Selection Process eFigure 2. Association of Tobacco Smoking and COVID-19-Related Hazard of Mortality, Stratified by Age, Adjusted for Other Patient Characteristics