Defining Cannabis Use Disorder in Administrative Health Data: A Systematic Review of Case Definitions and Validation Approaches

Abstract

Introduction:

Reliable case definitions (CDs) for cannabis use disorder (CUD) are essential for epidemiologic surveillance, health services research, and policy evaluation. As reliance on health administrative data increases, variation in diagnostic coding practices and limited validation of CDs may undermine the comparability and accuracy of CUD estimates. This systematic review aimed to identify, describe, and critically appraise how CUD has been operationalized within administrative health data sources, with particular attention to coding strategies and validation practices.

Methods:

Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, we systematically searched PubMed, EMBASE, and related databases for studies using International Classification of Diseases (ICD) codes to define CUD in administrative health data. Two reviewers independently screened studies, extracted data on CD components and validation methods, and assessed methodological quality using the Newcastle–Ottawa Scale.

Results:

A total of 56 studies met the inclusion criteria. Most relied on ICD-9 or ICD-10 diagnostic codes to identify CUD, typically using a one-or-more-code rule, although operational details varied by jurisdiction, coding framework, and observation window. No included study explicitly reported internal or external validation of its CD. Reported prevalence estimates ranged widely, from 0.06% in large administrative cohorts to 76.9% in highly selected clinical populations, reflecting differences in CD construction and study populations.

Conclusions:

CDs for CUD in administrative data vary substantially and lack empirical validation, limiting their reliability for surveillance and comparative research. The development and validation of standardized, transparent CDs are needed to strengthen cannabis epidemiology and to support reproducible research, health-system planning, and policy decision-making.

Keywords

cannabis use disorder administrative health data case definitions ICD coding validation surveillance

Introduction

Accurate surveillance of cannabis use disorder (CUD) is increasingly important in the postlegalization era, yet the tools used to identify cases vary considerably across health systems. Traditional epidemiologic data—such as national surveys and chart reviews—have advanced understanding of cannabis use patterns and associated harms, but they are resource-intensive, cross-sectional, and vulnerable to recall and reporting biases.^1–3 In contrast, health administrative data offer a cost-efficient and population-wide means of monitoring trends in substance use and related disorders.⁴ These data are routinely collected across hospitals, emergency departments, and outpatient programs for administrative purposes such as billing, performance monitoring, and service planning.⁵ When used for research and surveillance, they provide an opportunity to identify individuals with specific conditions at scale, using diagnostic and procedural codes.⁶

In administrative data research, it is important to distinguish between diagnostic codes, case definitions (CDs), and validated CDs. Diagnostic codes (e.g., International Classification of Diseases [ICD] codes) represent administrative translations of clinician-documented diagnoses and therefore carry some degree of face validity. However, the presence of a diagnostic code alone does not constitute a CD. A CD refers to an operational algorithm that specifies how diagnostic codes are combined with criteria such as care setting, frequency of encounters, and observation windows to classify individuals as having a given condition.⁷ For example, a CD for CUD might require at least one inpatient ICD-10 F12.x code or alternatively two outpatient F12.x codes within a 12-month period. Different operational choices can substantially affect who is identified as a case.

A validated CD goes a step further by empirically evaluating the performance of a CD against an external or internal reference standard, such as structured diagnostic interviews, chart review, registry linkage, or cross-database comparisons.⁸ Validation quantifies measurement properties such as sensitivity, specificity, and positive predictive value, allowing researchers to assess trade-offs between under- and overidentification. Importantly, increasing the number of diagnostic codes required for a CD may improve specificity but reduce sensitivity, whereas more permissive definitions may inflate prevalence through misclassification.⁹ Without validation, these trade-offs remain unknown, limiting confidence in epidemiologic estimates derived from administrative data.

Validated CDs have been developed for several psychiatric and neurologic conditions, including depression, anxiety disorders, autism spectrum disorder, and dementia, enabling reliable surveillance and cross-jurisdictional comparisons.^6,10,11 By contrast, comparable standards for CUD are lacking. Most studies relying on administrative data continue to use unvalidated or ad hoc combinations of ICD codes to identify CUD, often without justification for coding thresholds or time windows. This lack of standardization introduces substantial uncertainty and undermines comparability across studies.

Measurement challenges are further compounded by variability in clinical documentation and coding practices. Diagnostic assignment may differ by provider type (e.g., psychiatrist vs emergency physician), clinical context (e.g., acute intoxication vs chronic use), and health-system incentives, contributing to both under- and overidentification of CUD.^12–14 As a result, administrative data may systematically undercount milder or unrecognized cases while disproportionately capturing individuals with more severe presentations or comorbidities. These sources of misclassification bias can distort prevalence estimates and obscure associations with outcomes of interest.⁹

Despite the growing use of administrative data in cannabis research, the field lacks a comprehensive synthesis of how CUD has been operationalized and whether existing CDs have been validated.¹⁵ Without transparent and empirically evaluated definitions, estimates of CUD prevalence, comorbidity, and outcomes remain difficult to interpret, compare, or aggregate across jurisdictions. Establishing a clear understanding of current practices is therefore a necessary step toward improving surveillance, informing health-system planning, and strengthening the evidence base for cannabis policy.¹⁶

The objective of this systematic review was to identify, describe, and critically appraise CDs for CUD used in health administrative data. Specifically, we examined the diagnostic coding systems used, the operational criteria used to construct CDs, and whether internal or external validation methods were reported.

Methods

Study design

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA; see checklist in Appendix Table A2).¹⁷ The objective was to identify and evaluate CDs for CUD used in health administrative data. The review protocol was prospectively registered with the International Prospective Register of Systematic Reviews (CRD42023485042).

Eligibility criteria

Studies were eligible for inclusion if they met the following criteria:

Population: Individuals identified as having CUD or an equivalent diagnostic construct using health administrative data. Equivalent constructs included legacy diagnostic categories such as cannabis abuse or cannabis dependence (e.g., ICD-9 codes 305.2 or 304.3), which map onto the Diagnostic and Statistical Manual of Mental Disorders (DSM)-5 diagnosis of CUD.

Design: Observational or epidemiologic studies utilizing health administrative databases, including hospital discharge abstracts, insurance claims, electronic health records, or national registries.

CD criteria: Studies that described or applied diagnostic criteria for CUD using ICD codes or equivalent administrative classification systems. Studies were included regardless of whether the CD was validated. Because health administrative databases primarily rely on ICD coding systems for billing and surveillance, studies that identified CUD exclusively using DSM-based diagnostic criteria without an explicit ICD crosswalk were excluded. This exclusion was applied to ensure methodological consistency and relevance to administrative-data research, acknowledging that some DSM-based studies may have evaluated diagnostic validity outside the scope of this review.

Language: English.

Time frame: No publication date restrictions were applied; all records were considered from database inception to the date of the final search.

Data sources and search strategy

A comprehensive literature search was conducted on November 3, 2023, across PubMed, MEDLINE, EMBASE, PsycINFO, CINAHL, Web of Science, Scopus, and Google Scholar. Search terms combined keywords and medical subject headings related to CUD, CDs, administrative health data, and ICD coding systems. Reference lists of included studies were manually screened to identify additional relevant publications. Only peer-reviewed studies published in English were included. The search was updated on October 16, 2025, to capture newly indexed records; no additional eligible studies were identified (Appendix Table A1).

Study selection

All retrieved records were imported into Covidence (Veritas Health Innovation, Melbourne, Australia) for deduplication and screening.¹⁸ Two reviewers independently screened titles and abstracts, followed by a full-text review of potentially eligible studies. Discrepancies were resolved through discussion and consensus. Inter-rater agreement was moderate at both screening stages (κ ≈ 0.45), reflecting the heterogeneity of study designs and reporting practices.

Data extraction

A standardized data extraction form was developed and implemented in Covidence. Extracted variables included the following: •

Study characteristics: Authors, publication year, country, study design, and data source.

•

Population details: Sample size, demographic characteristics, and clinical or administrative setting.

•

CD components: ICD codes or other administrative classification systems used, inclusion and exclusion criteria, number of diagnostic codes required, and specified observation or look-back periods.

•

Validation methods: Internal validation approaches (e.g., chart review, cross-database comparison) or external validation approaches (e.g., linkage to registries or structured diagnostic assessments), if reported.

•

Outcomes: Reported prevalence estimates, diagnostic performance metrics (e.g., sensitivity or specificity, where available), and methodological limitations noted by the study authors.

Quality assessment

Methodological quality of included studies was assessed using the Newcastle–Ottawa Scale (NOS) for nonrandomized studies.^19–21 The NOS evaluates three domains: (1) selection of study groups, including representativeness and exposure ascertainment; (2) comparability of cohorts based on control of key confounders (e.g., age, sex, socioeconomic status); and (3) outcome assessment and adequacy of follow-up. Each study could receive up to nine stars, with higher scores indicating lower risk of bias. Two reviewers independently conducted quality assessments, with discrepancies resolved by consensus. The NOS was selected to evaluate the overall study design and internal validity; however, it does not assess the validity or performance of administrative-data CDs themselves. As such, NOS scores were interpreted separately from assessments of CD construction and validation.

Data synthesis

Given the substantial heterogeneity in study populations, designs, and CD structures, a narrative synthesis approach was used. Findings were summarized according to the following: 1.

Coding system (e.g., ICD-9 vs. ICD-10).

Operational criteria (e.g., number and combination of diagnostic codes, care settings, and observation windows).

Validation approach (internal validation, external validation, or none reported).

Where available, reported prevalence estimates and diagnostic performance metrics were summarized descriptively.

Use of artificial intelligence tools

Artificial intelligence (AI)-assisted software was used solely to support the administrative aspects of the review process. Covidence’s machine learning (ML) prioritization feature was used to assist with abstract screening. ChatGPT (OpenAI, San Francisco, USA) was used for grammar and clarity review during article preparation; no text generation, data extraction, or analytic decisions relied on AI tools. All methodological and interpretive decisions were made by the authors.

Results

Study selection

The database search identified 518 unique records. After title and abstract screening, 233 records were excluded as irrelevant. Following full-text review, 194 additional studies were excluded, resulting in 56 studies meeting the inclusion criteria (Fig. 1).

FIG. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram.

Table 1.

Study Characteristics

Reference	Study type	Sample	Diagnoses codes used*	Primary outcomes	Key findings
Campbell et al. 2018²⁵	Retrospective Cohort Study	United States, 2012–2014, new KPNC enrollees, n = 13,165	ICD-9 codes 291, 292, 303–305 (excluding 305.1) and ICD-10 codes F10-F19	SUD prevalence, comorbidities, insurance coverage	Post-ACA, there was an increase in the prevalence of amphetamine and cannabis use disorders (CUDs), alongside more comorbidities and financial barriers to care.
Alhamad et al. 2020²⁶	Retrospective Cohort Study	United States, 2007–2015, kidney transplant recipients, n = 52,689	ICD-9: 304.3x, 305.2x	Death, graft failure, psychosocial issues, cardiovascular and pulmonary complications	CUD pre- and post-transplant was linked to higher rates of graft failure, graft loss, and post-transplant psychosocial issues.
Pavarin et al. 2023²⁷	Retrospective Cohort Study	Italy, 2009–2019, patients with AUD from emergency departments (EDs), n = 3,649	ICD-9: 304.0, 304.2, 304.3, 305.2, 305.5, 305.6	Suicide attempts	Females and individuals with cannabis or benzodiazepine abuse had higher suicide attempt rates.
Anona et al. 2023²⁸	Retrospective Cohort Study	United States, 2016–2020, hospitalizations for bipolar disorder, n = 993,000	ICD-10: F12.1	Suicidality, substance abuse	Bipolar disorder with coexisting personality disorder showed high suicidality and substance abuse rates, especially for cannabis.
Winhusen et al. 2019²⁹	Case–Control Study	United States, not specified, cannabis-using patients at MetroHealth System, n = 8,932	ICD-9: 304.3, 304.330, 305.2, 305.230; ICD-10: F12.1, F12.199, F12.2-F12.299, F12.1-F12.999	Respiratory diseases	Cannabis use was significantly associated with respiratory diseases, particularly among nontobacco users.
Campbell et al. 2020³⁰	Retrospective Cohort Study	United States, 2012–2016, 101,405 patients from community health centers	ICD-9: F12.x, 304.x, 303.x	CUDs, psychiatric disorders	CUDs were highly correlated with various psychiatric conditions.
Padwa et al. 2022³¹	Case-Control Study	United States, 2013–2017, 1,047,463 primary care patients in a medical cannabis state	ICD-10: F12	Medical conditions associated with cannabis use and CUD	CUD and cannabis use were significantly associated with multiple medical and behavioral diagnoses.
Glickman et al. 2023³²	Retrospective Cohort Study	USA, 2016, 5,075 adults with IBD	ICD-10: F12	ED visits, hospitalization, corticosteroid and opioid use, surgery, mortality	Cannabis use in inflammatory bowel disease patients increased the risk of negative health outcomes.
Hall et al. 2020³³	Retrospective Cohort Study	USA, 2016, 942 ER visits to University of Colorado Hospital	ICD-10: All F12 codes + T40.7X1-.7X5	Cannabis-related ER utilization	ICD-10 codes varied in their accuracy in predicting cannabis-related emergency room visits.
Mannes et al. 2023³⁴	Retrospective Cohort Study	United States, 2005–2019, 4.3–5.6 million yearly VHA patients, focused on veterans with chronic pain	ICD-9: 305.2, 305.20-22; ICD-10: F12.1X, F12.2X	CUD	There was an increasing prevalence of CUD among veterans with chronic pain.
Patel et al. 2023³⁵	Retrospective Cohort Study	United States, pre-2019, 13,634 adults with schizophrenia	ICD-10: F12, F12.10–99	Clinical outcomes in schizophrenia with comorbid SUD	Schizophrenia patients with comorbid CUD had worse clinical outcomes.
Ecker et al. 2020³⁶	Retrospective Cohort Study	United States, 2010–2016, 258,055 U.S. military veterans with CUD	ICD-9: 305.20, 304.30; ICD-10: F12.10, F12.20	Mental health treatment utilization	Veterans with CUD exhibited high mental health comorbidities and increased health care utilization.
Pavarin 2022³⁷	Retrospective Cohort Study	Italy, 2009–2019, 1136 residents	ICD-9: 304.3, 305.2	Mortality	Elevated mortality rates were observed, especially in individuals with concurrent alcohol use disorders.
Phillips et al. 2022³⁸	Retrospective Cohort Study	United States, 2016–2020, Adults older than 50	ICD-10: F12.x	Chronic health conditions and health care utilization	Cannabis users faced greater risks of various health conditions and increased health care utilization.
Jain et al. 2022³⁹	Retrospective Cohort Study	United States, 2007–2018, 22,815 adults undergoing lumbar fusion	ICD-9: 304.31, 304.32, 305.21, 305.22	Postsurgical outcomes	CUD patients experienced longer hospital stays and higher rates of medical complications.
McGuinness et al. 2021⁴⁰	Retrospective Cohort Study	United States, 2006–2015, 510,007 patients undergoing vascular surgery	ICD-9: 304.30, 304.31, 304.32, 305.20, 305.21, 305.22	Perioperative complications	Patients with CUD had a higher incidence of perioperative myocardial infarction and stroke.
Williams et al. 2022⁴¹	Cross-Sectional	United States, 2018–2019, 5,995 VHA outpatients	ICD-10: F12.x	Substance use disorders	Substance use disorders were underrecognized, particularly among specific demographic subgroups.
Bechard et al. 2022⁴²	Retrospective Cohort Study	Canada, 2003–2017, 14,697,778 ED visits by youths	ICD-10AM: F12	Cannabis-related ER visits	There was a significant increase in cannabis-related ED visits among youths.
Schriver et al. 2020⁴³	Retrospective Cohort Study	United States, 2017–2019, 9,811 patients in Penn Medicine Health System	ICD-10: F12.10, F12.20, F12.90	Suicidal ideation	CUD was linked to suicidal ideation in univariate but not multivariate analyses.
Lähteenvuo et al. 2021⁴⁴	Retrospective Cohort Study	Finland and Sweden, 1996–2017, 45,476 individuals with schizophrenia and comorbid substance use disorders	ICD-9/10: F12.1-F12.2, F12.4-F12.9	Morbidity and mortality	Cannabis users within schizophrenia cohorts had increased all-cause mortality and hospitalization.
DiGiovanni et al. 2020⁴⁵	Retrospective Cohort Study	Italy, 2006–2015, 2,159 hospital patients with addictive disorders	ICD-9: 304, 305	Hospital admissions due to mental and physical disorders	High incidences of mental disorder admissions were associated with cannabis abuse.
Burns et al. 2006⁴⁶	Retrospective Cohort Study	United States, 1998–2002, Pregnant women with drug-related hospital admission	ICD-10AM: F11.0–11.9, F12.0–12.9, F14.0–15.9	Adverse obstetric and neonatal outcomes	Pregnancies with drug-related diagnoses had a higher risk of adverse outcomes, including preterm birth and complications.
Frost et al. 2021⁴⁷	Retrospective Cohort Study	United States, 2009–2017, transgender and cisgender patients with drug use disorders in VHA	ICD-9: 304.30–304.32, 305.20–305.22; ICD-10: F12.1, F12.2 excluding certain subcategories	Disparities in diagnosis and treatment of drug use disorders	Transgender patients in the Veterans Health Administration system had a higher incidence of drug use disorders than cisgender patients.
Garg et al. 2018⁴⁸	Cross-Sectional	32,625 U.S. patients with hidradenitis suppurativa, 18–64 years, 24% male, identified using ICD-9 code 705.83	ICD-9: 304.3; 305.2	Opioid, alcohol, and CUDs	Patients with hidradenitis suppurativa had higher odds of substance use disorders, particularly among specific demographic groups.
Nguyen et al. 2022⁴⁹	Retrospective Cohort Study	TriNetX network patients, n = several thousand, incident diagnoses of CU-RD, age and gender not specified, captured using ICD-10 codes F12.x	ICD-10: F12.x	Cannabis-related disorder	Cannabis-related disorder prevalence in inflammatory bowel disease patients rose significantly over the decade.
Lapham et al. 2019⁵⁰	Retrospective Cohort Study	15,202 U.S. adults with an index CUD diagnosis, majority 18–29 years, 63.1% male	ICD-9: 305.2, 304.3	Treatment initiation and engagement	Low rates of treatment initiation and engagement were observed among patients with a new CUD diagnosis.
Campbell et al. 2017⁵¹	Retrospective Cohort Study	5,504 U.S. patients with CUD and matched controls, average 37 years, 64% male, matched on gender, age, and medical home	ICD-9: 305.2; 304.3	Health care service utilization	CUD patients had increased risks of ED and inpatient visits.
Coughlin et al. 2021⁵²	Retrospective Cohort Study	53,034 Veterans Health Administration patients with new depression diagnoses in 2017, mostly male, excluded prior depression treatment	ICD-10: F12.X	Guideline-concordant depression treatment	Patients with substance use disorders received less guideline-concordant depression treatment.
Reynolds et al. 2023⁵³	Retrospective Cohort Study	167,095 ICU patients with pneumonia in the Premier Health care Database, mean age 67.4, 50% male, included if they received antibiotics or antivirals within 1 day of admission	ICD-9 and ICD-10: 304.30, 304.33, F12.20	Pneumonia outcomes	CUD was not associated with altered outcomes in pneumonia patients after adjustments.
Campbell et al. 2018²⁵	Retrospective Cohort Study	396,452 U.S. adults using prescription opioids in 2011, mean age 51.83 years, 41% male, excluded cancer patients	ICD-10: F12.X	Opioid misuse and overdose events	Younger age, chronic conditions, and benzodiazepine use were linked to higher risks of opioid misuse and overdose.
Courchesne-Krak et al. 2021⁵⁴	Retrospective Cohort Study	10,125 U.S. women aged 18–44 delivering at ≥20 weeks of gestation, 0% male, matched on key demographic characteristics	ICD-10: F12.xx	Severe maternal morbidity (SMM) and blood transfusion	Substance-related diagnoses were linked to SMM, with alcohol-related diagnoses showing the highest odds.
Winhusen et al. 2020⁵⁵	Case-Control	410,484 patients in the MetroHealth System, about 57% male, with recorded medical assessment	ICD-10: Various F12 codes	Adverse cardiovascular outcomes	Regular cannabis users, especially nontobacco users, had higher cardiovascular risks and mortality.
Fortuna et al. 2018⁵⁶	Retrospective Cohort Study	83,920 adults in a Boston area safety net health care system, excluded if less than three primary care visits between 2013 and 2015	ICD-10: Various F12 codes	Time to drug & alcohol use treatment	The analysis focused on sleep disturbances rather than CUD in its relationship with substance use and mental health disorders.
Courchesne-Krak et al. 2023⁵⁷	Retrospective Cohort Study	13,769 patients aged 18–90 with an SRD, 65.8% male, included based on having an SRD as the primary reason for a clinical encounter	ICD-10: F.12xx	Factors associated with SRDs	Cannabis- and sedative-related diagnoses were not significantly associated with having a substance-related diagnosis as the primary reason for a clinical encounter.
Graupensperger et al. 2019⁵⁸	Retrospective Cohort Study	1,466,985 patients at Penn State Hershey Medical Center, age 16+ years	ICD-10: Various F12 codes	Substance use disorder	Individuals with spinal cord injuries had increased odds of various substance use disorders, including CUD.
Armoon et al. 2021⁵⁹	Retrospective Cohort Study	22,484 patients in Quebec addiction treatment centers, 12+ years, 66% male, enrolled in treatment during specified time frame	ICD-10: Various F12 codes	ED use and hospitalization	CUD-related hospitalizations and emergency room visits were more frequent compared with noncannabis substance use disorders.
Courchesne-Krak et al. 2022⁶⁰	Retrospective Cohort Study	16,770 pregnant women aged 18–44 with a single live birth or stillbirth at ≥20 weeks of gestation, 0% male, substance-related diagnosis during antepartum period	ICD-10: F.12xx	Health care utilization	Substance-related diagnoses were associated with more ED visits and longer inpatient stays among pregnant women.
Bahorik et al. 2017⁶¹	Retrospective Cohort Study	90,922 patients with SUD and matched patients without SUD in KPNC facilities, average age 44, 62.9% male	ICD-10: Various F12 codes	Major medical conditions	Patients with CUDs had a higher prevalence of major medical conditions compared with nonsubstance use disorder patients.
Courchesne-Krak et al. 2023⁶²	Retrospective Cohort Study	2,158 women aged 18–44 who delivered a single live or stillbirth at ≥20 weeks of gestation from 2012 to 2019, 0% male	ICD-10: F12.xx	Preterm and cesarian delivery	Cannabis-related diagnoses were not significantly associated with preterm or cesarian delivery.
Bahorik et al. 2018⁶³	Retrospective Cohort Study	35,148 patients with SUD in KPNC facilities, mean age 37 years, 64% male	ICD-10: Various F12 codes	ED admission trends	High ED usage was consistently reported among patients with substance use disorders across multiple years.
Hasin et al. 2022⁶⁴	Retrospective Cohort Study	5,556,315 Veterans Health Administration Patients, 90% male, having COVID-19+ during the study time frame	ICD-10: F12.1X, F12.2X	COVID-19 outcomes	CUD was associated with increased hospitalization among COVID-19+ patients but not with other COVID-19 outcomes.
Winhusen et al. 2019⁶⁵	Case-Control	50,897 patients in the MetroHealth System with T2DM and hypertension, mean age 61, 47% male	ICD-10: Various F12 codes	Medical complications of T2DM	CUD was associated with higher risks for myocardial infarction and mortality in patients with type 2 diabetes mellitus and hypertension.
Gan et al. 2021⁶⁶	Retrospective Cohort Study	Cross-sectional and longitudinal analysis of 1,396,634 Canadians with various SUDs, aged 45 on average, 50% male, assessing both those with and without prevalent CVDs.	ICD-9: 304.3, 305.2; ICD-10: F12, T40.7	CVDs (prevalent vs incident)	CUD was associated with an increased prevalence of heart failure and hypertension but did not significantly affect other CVD subtypes.
Winhusen et al. 2020⁶⁷	Case–Control	109,090 adults in Cleveland, OH, average age 58.4, 48.7% male, with type 2 diabetes mellitus and hypertension, undergoing medical assessment.	Various ICD-9 and ICD-10 codes for CUD	Morbidity and mortality	The CUD group experienced significantly worse outcomes for cerebrovascular accidents and all-cause mortality compared with non-CUD patients.
Gates et al. 2017⁶⁸	Cross-Sectional	10,988 offenders in the United States, mean age 37.5, 90.1% male, with valid records in ERD and OMS.	ICD-9: 304.2–305.6: Cannabis dependence, cannabis abuse	Suicide attempts & self-harm	Offenders with comorbid mental health conditions such as bipolar and anxiety, or those combining alcohol with depression, had the highest odds of suicide attempts, particularly when cannabis was also involved.
John et al. 2020⁶⁹	Retrospective Cohort Study	951,533 patients in the Duke University Health System, mean age 48.4, 57.4% female, with SUD, aged 18–90.	Extensive list of ICD-10 and ICD-9 codes for CUD	Prevalence of chronic noncancer pain (CNCP), overdose, ED utilization, inpatient hospitalization	Only 0.73% of the study population had CUD, but this group showed a significantly higher prevalence of chronic noncancer pain.
Hasin et al. 2023²³	Cross-Sectional	Veterans Health Administration patients, 18–75 years, mostly male (89.0% in 2019), sampled from primary care, emergency, or mental health visits.	ICD-9-CM and ICD-10-CM codes for CUD	CUD diagnosis	CUD prevalence rose significantly from 2005 to 2019, with the most considerable increases observed in states with medical cannabis laws.
Hasin et al. 2022⁷⁰	Retrospective Cohort Study	VHA patients from 2005 to 2019, ages 28–29 on average, mostly male (77.0%−98.2%), diagnosed with CUD.	ICD-9-CM and ICD-10-CM codes for CUD	CUD	There were notable increases in CUD diagnoses across different age groups and races, with the most significant rises among young Black males and females.
Hoggatt et al. 2023²⁴	Retrospective Cohort Study	Veterans’ Health Administration patients, aged 18–100, >90% male, documented encounters from 2010–2019.	ICD-9: 304.3, 305.2; ICD-10: F12	SUD diagnoses	Over a decade, CUD diagnoses gradually increased among Veterans Health Administration patients.
Liu et al. 2020⁷¹	Retrospective Cohort Study	120,706 ED visits across 18 states, all ages, 61.4% male, involving drug poisoning codes.	F12; T40.7X1A; T40.7X4A	Type of drug poisoning	6.4% of all ED visits analyzed were due to cannabis intoxication.
Livne et al. 2023⁷²	Retrospective Cohort Study	Veterans’ Health Administration patients, large sample with age-group-specific analysis, primarily male.	ICD-9-CM and ICD-10-CM codes for CUD	CUD	Among patients with psychiatric disorders, the prevalence of CUD diagnoses rose significantly.
Loyal et al. 2023⁷³	Retrospective Cohort Study	Hospitalizations in British Columbia, Canada; data spans 2008/2009–2017/2018, majority aged 15–44, slightly over 50% male.	Cannabis-related disorders: F12	Psychiatric hospitalizations	The study did not specifically examine individual outcomes but focused on trends in hospitalizations for mental and substance use disorders.
Tolan et al. 2022⁷⁴	Retrospective Cohort Study	United States, not specified, patient ED visits at two academic medical centers in Boston, MA, n = 180,656	Extensive list of ICD codes for cannabis-related disorders	Cannabis-related ED visits	Both THC immunoassay positivity and cannabis-related diagnostic codes in EDs increased as cannabis legalization progressed.
Hughto et al. 2021⁷⁵	Retrospective Cohort Study	United States, 2017, transgender and cisgender adults enrolled in commercial or Medicare plans, n = 62,548	ICD-10 F12.x	SUD prevalence	Transgender adults showed more than five times the prevalence of CUD compared with matched cisgender adults.
Jennings et al. 2021⁷⁵^*	Retrospective Cohort Study	United States, 2005–2018, Patients across three health systems with varying exposure to opioids, n = over 2.6 million	ICD-9 and ICD-10: 304X, 305X, F12.X	OUD co-occurring diagnoses	Substance use disorders were notably more prevalent in the opioid use disorder group than in other groups, with high rates of co-occurring alcohol, tobacco, and CUDs.

Denotes that case definitions were based on the presence of ≥1 qualifying ICD code unless otherwise specified.

SUD, substance use disorder; case definition, unless otherwise stated, was any one of these codes.

Study characteristics

Study characteristics are summarized in Table 1. The included studies spanned multiple countries, predominantly the United States (n = 47, 84%), followed by Canada (n = 4), Italy (n = 3), and one joint study from Finland and Sweden. Most studies used retrospective cohort designs using health administrative data sources, including billing claims, hospital discharge abstracts, or electronic health records (EHRs).

Study populations were heterogeneous, ranging from narrowly defined clinical samples (e.g., veterans, pregnant women, or individuals with chronic medical conditions such as diabetes or hypertension) to large population-based datasets encompassing millions of individuals. Sample sizes ranged from fewer than 1,000 participants in smaller cohort or case–control studies to more than 5 million in national or system-wide analyses.

Participant demographics varied across cohorts. Mean ages generally fell between the mid-30s and early 50s, although some studies focused on older adults (≥65 years) or young adults (18–29 years). Across most studies, male participants predominated, often exceeding 50% and reaching over 90% in veteran-based cohorts.

Study periods ranged from the early 2000s through 2019. Data sources included inpatient and outpatient hospital settings, primary care, emergency departments, and community health systems. Several studies linked administrative records to registry, laboratory, or genomic data to enhance analytic scope.

CDs and diagnostic codes

All included studies defined CUD using ICD-9 or ICD-10 diagnostic codes. The most commonly used ICD-9 codes were 304.3 (cannabis dependence) and 305.2 (cannabis abuse), while ICD-10-based definitions primarily relied on the F12 category for cannabis-related disorders (e.g., F12.1 for abuse and F12.2 for dependence). Some studies additionally incorporated T40.7 codes related to cannabis poisoning or adverse effects. No studies used ICD-11 diagnostic codes.

Terminology across studies reflected legacy diagnostic frameworks, with cannabis abuse and cannabis dependence frequently used as distinct categories corresponding to severity levels now unified under the DSM-5 diagnosis of CUD. Inclusion of both categories was therefore necessary to capture the full spectrum of clinically documented cannabis-related disorders in administrative data.

Most studies used a permissive “one-or-more-code” rule, identifying CUD if any qualifying diagnostic code was recorded within a specified observation window. Fewer studies required multiple diagnostic codes or applied additional confirmatory criteria, such as repeated positive urine drug screens, EHR text mentions of cannabis use, or exclusion of remission codes. Some studies supplemented ICD coding with Current Procedural Terminology codes or SNOMED-CT mappings, while others incorporated text-mining approaches to identify cannabis-related terms within clinical notes.

A gradual transition from ICD-9 to ICD-10 coding was observed over time, with ICD-10 offering greater granularity through expanded subcodes (e.g., uncomplicated vs. psychotic-related cannabis disorders). Despite these differences, no study reported the use of ICD-11 codes, which may reflect the relatively recent adoption of ICD-11 and its limited implementation in administrative health databases during the study periods examined.

Validation of CDs

None of the 56 included studies explicitly reported internal or external validation of their CDs. All relied on administrative diagnostic coding alone to identify CUD, without verification against clinical records, structured diagnostic assessments, or linked registry data. The absence of reported validation limits assessment of both specificity and sensitivity and constrains confidence in epidemiological estimates derived from these definitions.

Quality assessment

Using the NOS, the overall methodological quality of included studies was high. In the selection domain, all studies received full scores for representativeness and exposure ascertainment. In the comparability domain, 50 studies adjusted for major confounders such as age, sex, and socioeconomic status, while 6 did not. In the outcome domain, 52 studies relied on validated administrative data sources or independent outcome assessments, whereas 4 relied on self-reported measures or had limited follow-up.

Most studies achieved scores of 8–9 out of 9, indicating strong internal validity and sound cohort design. However, high NOS scores primarily reflected study design and data source quality rather than the rigor of CD construction. Consequently, even methodologically strong studies frequently relied on unvalidated and inconsistently operationalized definitions of CUD.

Prevalence and epidemiological findings

Reported prevalence estimates of CUD varied widely, ranging from 0.06% in large administrative cohorts to 76.9% in highly selected cannabis-using clinical populations. This variation corresponded to differences in coding practices, study populations, and data sources.

Across multiple datasets, temporal analyses indicated increasing rates of recorded CUD diagnoses over time, such as a rise from 1.3 to 10.3 per 1,000 admissions in one U.S. cohort, coinciding with broader cannabis availability and heightened diagnostic recognition.

Comorbidity patterns were broadly consistent across studies. Individuals identified with CUD exhibited higher rates of psychiatric disorders (including depression, post-traumatic stress disorder, and bipolar disorder), chronic medical conditions (such as chronic obstructive pulmonary disease, cardiovascular disease, and chronic pain), and other substance use disorders. In specialized clinical populations, including transplant recipients, CUD diagnoses were associated with poorer clinical outcomes and increased psychosocial burden.

Methodological challenges identified

Several methodological challenges were consistently reported across studies: 1.

Coding variability: Diagnostic practices differed across providers, care settings, and health systems, increasing the risk of misclassification.

Incomplete capture: Mild or unrecognized cases of CUD were likely underrepresented due to underdiagnosis in administrative data.

Coding transitions: Shifts from ICD-9 to ICD-10 introduced discontinuities affecting temporal comparability of prevalence estimates.

Limited generalizability: Many studies focused on narrowly defined or institutional populations, limiting applicability to broader cannabis-using groups.

Data constraints: Administrative databases generally lacked detailed information on cannabis exposure characteristics (e.g., frequency, potency, route of administration) and contextual factors (medical vs. nonmedical use).

Despite generally strong methodological quality, these findings indicate that CD validity, rather than study design, represents the principal limitation in the existing literature. Without standardized and validated algorithms, results from administrative-data studies of CUD remain difficult to interpret, compare, and synthesize.

Discussion

Summary of findings

This systematic review identified substantial heterogeneity in how CUD has been operationalized within health administrative data. Across 56 studies, diagnostic codes and operational criteria varied considerably, and no study explicitly reported internal or external validation of its CD. Reported prevalence estimates ranged widely, from 0.06% in large administrative cohorts to 76.9% in highly selected cannabis-using clinical populations. This variability likely reflects a combination of methodological heterogeneity in CD construction and meaningful differences in underlying study populations. Collectively, these findings indicate that current approaches to identifying CUD in administrative datasets lack standardization and empirical validation, limiting the reliability and comparability of existing evidence.

Methodological considerations and risk of bias

Most included studies relied on retrospective cohort designs using EHRs or claims data, which are inherently vulnerable to misclassification and under-ascertainment. Variation in clinical documentation practices, provider coding behavior, and health-system incentives likely contributed to both false-positive and false-negative identification of CUD. Although the overall methodological quality, as assessed by the NOS, was high, this primarily reflected study design and data source quality rather than the rigor of CD construction. The absence of validated CDs therefore represents a critical methodological weakness that is largely independent of conventional risk-of-bias assessments.

In addition, observational designs limit causal inference, and generalizability was often constrained by the use of institutionally specific datasets, such as those derived from the U.S. Veterans Health Administration.²² These limitations underscore the need to distinguish between internal study validity and the validity of the measurement tools used to define CUD in administrative research.

Implications for research

The absence of validated CDs for CUD in administrative data has important implications for epidemiological research, surveillance, and cross-study synthesis. Without validation, it is difficult to determine whether observed differences in prevalence, comorbidity, or outcomes reflect true variation or artifacts of measurement. Developing standardized, empirically validated CDs would improve comparability across studies, enable meaningful meta-analytic synthesis, and support linkage of administrative data with survey-based or clinical datasets.

Future research should prioritize formal validation studies that compare administrative-data CDs against reference standards such as structured diagnostic interviews, clinician-confirmed diagnoses via chart review, or registry-based ascertainment. Validation should explicitly report diagnostic performance metrics, including sensitivity, specificity, and predictive values, to allow transparent evaluation of trade-offs inherent in different operational choices.

Observed demographic and clinical disparities across studies—including higher recorded rates of CUD among men, younger adults, and certain racialized groups—warrant further investigation. Prior work has documented the rising CUD diagnoses among younger Black veterans²³ and women,²⁴ but the extent to which these patterns reflect true epidemiologic trends versus differential diagnosis, documentation, or access to care remains unclear. Research incorporating more granular measures of cannabis exposure (e.g., frequency, potency, route of administration, and medical vs. nonmedical use) will be essential to contextualize administrative-data findings.

Emerging analytic approaches, including ML, may offer additional tools for refining CDs, although these were not evaluated in the studies included in this review. Interpretable ML models, such as decision trees or rule-based classifiers, may help identify systematic patterns in coding behavior across providers and settings, potentially improving consistency and reproducibility. However, such approaches should complement—rather than replace—transparent, rule-based definitions and should themselves be subject to rigorous validation.¹⁶

Implications for practice and policy

Reliable surveillance of CUD depends on the use of standardized and validated CDs. In the absence of such definitions, administrative data may under- or overestimate disease burden, potentially distorting resource allocation, program evaluation, and policy decisions. Harmonized CD algorithms applied consistently across jurisdictions would improve the accuracy of monitoring trends in CUD prevalence, treatment engagement, comorbidity patterns, and health outcomes.

From a clinical and health-system perspective, integrating consistent approaches to identifying CUD within administrative datasets may also facilitate earlier recognition and intervention, particularly when aligned with screening and referral pathways in primary care and specialty settings. Attention to social determinants of health—including socioeconomic status, structural inequities, and access to care—is essential, as these factors influence both diagnosis and outcomes and may contribute to observed disparities in administrative-data studies.

Limitations of the review

Several limitations should be considered when interpreting these findings. The included studies were heterogeneous in population, setting, and design, and most relied on retrospective administrative data subject to incomplete capture and misclassification. The exclusion of studies relying solely on DSM-based diagnoses without ICD crosswalks may have omitted validation studies conducted outside administrative-data contexts, although this decision was necessary to maintain methodological relevance. Publication bias may also have favored studies reporting higher prevalence estimates or stronger associations.

Future directions

Future work should focus on the development, validation, and dissemination of standardized CD algorithms for CUD in administrative data. Prospective studies integrating administrative records with clinical assessments and self-reported data would enable robust calibration of these algorithms. As ICD-11 becomes more widely implemented, opportunities may arise to align administrative definitions more closely with contemporary diagnostic frameworks. Establishing consensus CDs—analogous to those developed for depression and other psychiatric conditions—would provide a foundation for reproducible epidemiology, health services research, and surveillance in cannabis science.

Conclusion

This systematic review demonstrates that, despite the growing use of health administrative data to study CUD, the absence of standardized and validated CDs substantially undermines the reliability of the current evidence base. Reported prevalence estimates varied by orders of magnitude, reflecting differences in CD construction and study populations rather than clear epidemiological patterns. Advancing the field will require transparent, validated, and interoperable definitions of CUD that can be applied consistently across health systems. Accurate case identification is essential to quantify disease burden, evaluate interventions, and inform evidence-based public health and policy responses to CUD.

Footnotes

Author Disclosure Statement

A.B. is supported by the Canadian Institutes of Health Research (CIHR) and Alberta Innovates through doctoral awards. S.P. holds the Cuthbertson & Fischer Chair in Pediatric Mental Health at the University of Calgary. G.M. and M.I.A. declare no competing interests.

Funding Information

The authors confirm that this work did not receive any specific funding.

Abbreviations Used

Appendix

Table A2.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses Checklist

Section and topic	Item #	Checklist item	Location where item is reported
TITLE
Title	1	Identify the report as a systematic review.	1
ABSTRACT
Abstract	2	See the PRISMA 2020 for Abstract checklist.
INTRODUCTION
Rationale	3	Describe the rationale for the review in the context of existing knowledge.
Objectives	4	Provide an explicit statement of the objective(s) or question(s) the review addresses.
METHODS
Eligibility criteria	5	Specify the inclusion and exclusion criteria for the review and how studies were grouped for the syntheses.
Information sources	6	Specify all databases, registers, websites, organizations, reference lists, and other sources searched or consulted to identify studies. Specify the date when each source was last searched or consulted.
Search strategy	7	Present the full search strategies for all databases, registers, and websites, including any filters and limits used.
Selection process	8	Specify the methods used to decide whether a study met the inclusion criteria of the review, including how many reviewers screened each record and each report retrieved, whether they worked independently, and if applicable, details of automation tools used in the process.
Data collection process	9	Specify the methods used to collect data from reports, including how many reviewers collected data from each report, whether they worked independently, any processes for obtaining or confirming data from study investigators, and if applicable, details of automation tools used in the process.
Data items	10a	List and define all outcomes for which data were sought. Specify whether all results that were compatible with each outcome domain in each study were sought (e.g., for all measures, time points, analyses), and if not, the methods used to decide which results to collect.
Data items	10b	List and define all other variables for which data were sought (e.g., participant and intervention characteristics, funding sources). Describe any assumptions made about any missing or unclear information.
Study risk-of-bias assessment	11	Specify the methods used to assess risk of bias in the included studies, including details of the tool(s) used, how many reviewers assessed each study, and whether they worked independently, and if applicable, details of automation tools used in the process.
Effect measures	12	Specify for each outcome the effect measure(s) (e.g., risk ratio, mean difference) used in the synthesis or presentation of results.
Synthesis methods	13a	Describe the processes used to decide which studies were eligible for each synthesis (e.g., tabulating the study intervention characteristics and comparing against the planned groups for each synthesis (item #5)).
	13b	Describe any methods required to prepare the data for presentation or synthesis, such as handling of missing summary statistics, or data conversions.
	13c	Describe any methods used to tabulate or visually display results of individual studies and syntheses.
	13d	Describe any methods used to synthesize results and provide a rationale for the choice(s). If meta-analysis was performed, describe the model(s), method(s) to identify the presence and extent of statistical heterogeneity, and software package(s) used.
	13e	Describe any methods used to explore possible causes of heterogeneity among study results (e.g., subgroup analysis, meta-regression).
	13f	Describe any sensitivity analyses conducted to assess robustness of the synthesized results.
Reporting bias assessment	14	Describe any methods used to assess risk of bias due to missing results in a synthesis (arising from reporting biases).
Certainty assessment	15	Describe any methods used to assess certainty (or confidence) in the body of evidence for an outcome.
RESULTS
Study selection	16a	Describe the results of the search and selection process, from the number of records identified in the search to the number of studies included in the review, ideally using a flow diagram.
Study selection	16b	Cite studies that might appear to meet the inclusion criteria, but which were excluded, and explain why they were excluded.
Study characteristics	17	Cite each included study and present its characteristics.
Risk of bias in studies	18	Present assessments of risk of bias for each included study.
Results of individual studies	19	For all outcomes, present, for each study: (a) summary statistics for each group (where appropriate) and (b) an effect estimate and its precision (e.g., confidence/credible interval), ideally using structured tables or plots.
Results of syntheses	20a	For each synthesis, briefly summarize the characteristics and risk of bias among contributing studies.
	20b	Present results of all statistical syntheses conducted. If meta-analysis was done, present for each the summary estimate and its precision (e.g., confidence/credible interval) and measures of statistical heterogeneity. If comparing groups, describe the direction of the effect.
	20c	Present results of all investigations of possible causes of heterogeneity among study results.
	20d	Present results of all sensitivity analyses conducted to assess the robustness of the synthesized results.
Reporting biases	21	Present assessments of risk of bias due to missing results (arising from reporting biases) for each synthesis assessed.
Certainty of evidence	22	Present assessments of certainty (or confidence) in the body of evidence for each outcome assessed.
DISCUSSION
Discussion	23a	Provide a general interpretation of the results in the context of other evidence.
	23b	Discuss any limitations of the evidence included in the review.
	23c	Discuss any limitations of the review processes used.
	23d	Discuss implications of the results for practice, policy, and future research.
OTHER INFORMATION
Registration and protocol	24a	Provide registration information for the review, including register name and registration number, or state that the review was not registered.
	24b	Indicate where the review protocol can be accessed, or state that a protocol was not prepared.
	24c	Describe and explain any amendments to information provided at registration or in the protocol.
Support	25	Describe sources of financial or nonfinancial support for the review, and the role of the funders or sponsors in the review.
Competing interests	26	Declare any competing interests of review authors.
Availability of data, code, and other materials	27	Report which of the following are publicly available and where they can be found: template data collection forms; data extracted from included studies; data used for all analyses; analytic code; any other materials used in the review.

From: Page et al. 2021.¹⁷

PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

References

1. Connor

, Stjepanović

, Le Foll

, et al. Cannabis use and cannabis use Disorder. Nat Rev Dis Primers 2021;7(1):16; doi: 10.1038/s41572-021-00247-4

2. Rotermann

. Analysis of trends in the prevalence of cannabis use and related metrics in Canada. Health Rep 2019;30(6):3–13; doi: 10.25318/82-003-x201900600001-eng

3. Plana-Ripoll

, Musliner

, Dalsgaard

, et al. Nature and prevalence of combinations of mental disorders and their association with excess mortality in a population-based cohort study. World Psychiatry 2020;19(3):339–349; doi: 10.1002/wps.20802

4. Huỳnh

, Kisely

, Rochette

, et al. Measuring substance-related disorders using Canadian Administrative Health Databanks: Interprovincial comparisons of recorded diagnostic rates, incidence proportions and mortality rate ratios. Can J Psychiatry 2022;67(2):117–129; doi: 10.1177/07067437211043446

5. Emerson

, McLinden

, Sereda

, et al. Secondary use of routinely collected administrative health data for epidemiologic research: Answering research questions using data collected for a different purpose. Int J Popul Data Sci 2024;9(1):2407; doi: 10.23889/ijpds.v9i1.2407

6. McBrien

, Souri

, Symonds

, et al. Identification of validated case definitions for medical conditions used in primary care electronic medical record databases: A systematic review. J Am Med Inform Assoc 2018;25(11):1567–1578; doi: 10.1093/jamia/ocy094

7. Larvin

, Peckham

, Prady

. Case-finding for common mental disorders in primary care using routinely collected data: A systematic review. Soc Psychiatry Psychiatr Epidemiol 2019;54(10):1161–1175; doi: 10.1007/s00127-019-01744-4

8. DeYoung

, Chen

, Beum

, et al. Validation of a syndromic case definition for detecting emergency department visits potentially related to Marijuana. Public Health Rep 2017;132(4):471–479; doi: 10.1177/0033354917708987

9. Chubak

, Pocobelli

, Weiss

. Trade-offs between accuracy measures for electronic healthcare data algorithms. J Clin Epidemiol 2012;65(3):343–349.e2; doi: 10.1016/j.jclinepi.2011.09.002

10.

10. Fiest

, Jette

, Quan

, et al. Systematic review and assessment of validated case definitions for depression in administrative data. BMC Psychiatry 2014;14(1):289; doi: 10.1186/s12888-014-0289-5

11.

11. Coo

, Ouellette-Kuntz

, Brownell

, et al. Validating an administrative data-based case definition for identifying children and youth with autism spectrum disorder for surveillance purposes. Can J Public Health 2018;108(5–6):e530–e538; doi: 10.17269/CJPH.108.5963

12.

12. Tang

, Lucyk

, Quan

. Coder perspectives on physician-related barriers to producing high-quality administrative data: A qualitative study. CMAJ Open 2017;5(3):E617–E622; doi: 10.9778/cmajo.20170036

13.

13. Albagmi

, Shawaheen

, Asiri

, et al. Assessment of clinical miscoding errors and potential financial their implications on healthcare management – A case of local hospital in Najran, Saudi Arabia. Saudi Pharm J 2024;32(1):101894; doi: 10.1016/j.jsps.2023.101894

14.

14. Cheng

, Gilchrist

, Robinson

, et al. The risk and consequences of clinical miscoding due to inadequate medical documentation: A case study of the impact on health services funding. Health Inf Manag 2009;38(1):35–46; doi: 10.1177/183335830903800105

15.

15. Bahji

. The unseen addiction: Improving surveillance and governance of medical cannabis in Canada. Can J Addict 2025;16(3):10–12; doi: 10.1097/CXA.0000000000000244

16.

16. Ng

, Lad

, Patel

, et al. Applications of machine learning in cannabis research: A scoping review. Eur J Integr Med 2025;74:102434; doi: 10.1016/j.eujim.2025.102434

17.

17. Page

, McKenzie

, Bossuyt

, et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021;372:n71; doi: 10.1136/bmj.n71

18.

18.Veritas Health Innovation. Covidence Systematic Review Software. 2019.

19.

19. Wells

, Shea

, O’Connell

, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2019.

20.

20. Luchini

, Stubbs

, Solmi

, et al. Assessing the quality of studies in meta-analyses: Advantages and limitations of the Newcastle Ottawa Scale. WJMA 2017;5(4):80–84; doi: 10.13105/wjma.v5.i4.80

21.

21. Stang

. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 2010;25(9):603–605; doi: 10.1007/s10654-010-9491-z

22.

22. Boyko

. Observational research opportunities and limitations. J Diabetes Complications 2013;27(6):642–648; doi: 10.1016/j.jdiacomp.2013.07.007

23.

23. Hasin

, Wall

, Choi

, et al. State cannabis legalization and cannabis use disorder in the US Veterans Health Administration, 2005 to 2019. JAMA Psychiatry 2023;80(4):380–388; doi: 10.1001/jamapsychiatry.2023.0019

24.

24. Hoggatt

, Chawla

, Washington

, et al. Trends in substance use disorder diagnoses among Veterans, 2009-2019. Am J Addict 2023;32(4):393–401; doi: 10.1111/ajad.13413

25.

25. Campbell

, Bahorik

, VanVeldhuisen

, et al. Use of a prescription opioid registry to examine opioid misuse and overdose in an integrated health system. Prev Med 2018;110:31–37; doi: 10.1016/j.ypmed.2018.01.019

26.

26. Alhamad

, Koraishy

, Lam

, et al. Cannabis dependence or abuse in kidney transplantation: Implications for posttransplant outcomes. Transplantation 2019;103(11):2373–2382; doi: 10.1097/TP.0000000000002599

27.

27. Pavarin

, Fabbri

, Turino

, et al. Epidemiology and clinical-demographic characteristics of suicide attempts in alcohol use disorders in an Italian population. J Psychoactive Drugs 2023;55(4):456–463; doi: 10.1080/02791072.2022.2107464

28.

28. Anona

, Olaomi

, Udegbe

, et al. Co-occurrence of bipolar disorder and personality disorders in the United States: Prevalence, suicidality, and the impact of substance abuse. J Affect Disord 2024;345(h3v, 7906073):1–7; doi: 10.1016/j.jad.2023.10.087

29.

29. Winhusen

, Theobald

, Kaelber

, et al. Regular cannabis use, with and without tobacco co-use, is associated with respiratory disease. Drug Alcohol Depend 2019;204:107557; doi: 10.1016/j.drugalcdep.2019.107557

30.

30. Campbell

, Bailey

, Hoffman

, et al. Associations between psychiatric disorders and cannabis-related disorders documented in electronic health records. J Psychoactive Drugs 2020;52(3):228–236; doi: 10.1080/02791072.2020.1747665

31.

31. Padwa

, Huang

, Mooney

, et al. Medical conditions of primary care patients with documented cannabis use and cannabis use disorder in electronic health records: A case control study from an academic health system in a medical marijuana state. Subst Abuse Treat Prev Policy 2022;17(1):36; doi: 10.1186/s13011-022-00467-1

32.

32. Glickman

, Dalessio

, Raup-Konsavage

, et al. The impact of cannabis use on clinical outcomes in inflammatory bowel disease: A population-based longitudinal cohort study. Inflamm Bowel Dis 2024;30(7):1055–1061; doi: 10.1093/ibd/izad151

33.

33. Hall

. Evaluation of Cannabis ICD-10-Cm codes’ Performance to Predict Emergency Department Visits Partially Attributable to Cannabis. University of Colorado Denver, Anschutz Medical Campus; 2020.

34.

34. Mannes

, Malte

, Olfson

, et al. Increasing risk of cannabis use disorder among U.S. veterans with chronic pain: 2005–2019. Pain 2023;164(9):2093–2103; doi: 10.1097/j.pain.0000000000002920

35.

35. Patel

, Chan

KMY

, Palmer

EOC

, et al. Associations of comorbid substance use disorders with clinical outcomes in schizophrenia using electronic health record data. Schizophr Res 2023;260:191–197; doi: 10.1016/j.schres.2023.08.023

36.

36. Ecker

, Lang

, Hogan

, et al. Cannabis use disorder among veterans: Comorbidity and mental health treatment utilization. J Subst Abuse Treat 2020;109:46–49; doi: 10.1016/j.jsat.2019.11.003

37.

37. Pavarin

. Mortality risk for individuals with cannabis use disorders in relation to alcohol use disorders: Results of a follow-up study. Psychiatry Res 2022;316:114741; doi: 10.1016/j.psychres.2022.114741

38.

38. Phillips

, Pedula

, Choi

, et al. Chronic health conditions, acute health events, and healthcare utilization among adults over age 50 in Hawai’i who use cannabis: A matched cohort study. Drug Alcohol Depend 2022;234:109387; doi: 10.1016/j.drugalcdep.2022.109387

39.

39. Jain

, Cloud

, Gordon

, et al. Cannabis use disorder is associated with longer in-hospital lengths of stay, higher rates of medical complications, and costs of care following primary 1- to 2-level lumbar fusion. Global Spine J 2024;14(1):67–73; doi: 10.1177/21925682221093965

40.

40. McGuinness

, Goel

, Elias

, et al. Cannabis use disorder and perioperative outcomes in vascular surgery. J Vasc Surg 2021;73(4):1376–1387.e3; doi: 10.1016/j.jvs.2020.07.094

41.

41. Williams

, Fletcher

, Frost

, et al. Comparison of substance use disorder diagnosis rates from electronic health record data with substance use disorder prevalence rates reported in surveys across sociodemographic groups in the Veterans Health Administration. JAMA Netw Open 2022;5(6):e2219651; doi: 10.1001/jamanetworkopen.2022.19651

42.

42. Bechard

, Cloutier

, Lima

, et al. Cannabis-related emergency department visits by youths and their outcomes in Ontario: A trend analysis. CMAJ Open 2022;10(1):E100–E108; doi: 10.9778/cmajo.20210142

43.

43. Schriver

, Lieblich

, AlRabiah

, et al. Identifying risk factors for suicidal ideation across a large community healthcare system. J Affect Disord 2020;276:1038–1045; doi: 10.1016/j.jad.2020.07.047

44.

44. Lähteenvuo

, Batalla

, Luykx

, et al. Morbidity and mortality in schizophrenia with comorbid substance use disorders. Acta Psychiatr Scand 2021;144(1):42–49; doi: 10.1111/acps.13291

45.

45. Di Giovanni

, Di Martino

, Zecca

IAL

, et al. Trend in hospital admissions of drug addicts and associated factors from 2006 to 2015: An observational study on the hospitals’ discharge registries from a region of Central Italy. Ann Ig 2020;32(4):376–384; doi: 10.7416/ai.2020.2361

46.

46. Burns

, Mattick

, Cooke

. The use of record linkage to examine illicit drug use in pregnancy. Addiction 2006;101(6):873–882.

47.

47. Frost

, Blosnich

, Lehavot

, et al. Disparities in documented drug use disorders between transgender and cisgender U.S. Veterans Health Administration patients. J Addict Med 2021;15(4):334–340; doi: 10.1097/ADM.0000000000000769

48.

48. Garg

, Papagermanos

, Midura

, et al. Opioid, alcohol, and cannabis misuse among patients with hidradenitis suppurativa: A population-based analysis in the United States. J Am Acad Dermatol 2018;79(3):495–500.e1; doi: 10.1016/j.jaad.2018.02.053

49.

49. Nguyen

, Perez

, Celio

, et al. S888 Epidemiology of documented cannabis use and/or cannabis-related disorders in a multi-center cohort of inflammatory bowel disease patients from 2012–2021. Am J Gastroenterol 2022;117(10S):e643–e644; doi: 10.14309/01.ajg.0000860192.20968.39

50.

50. Lapham

, Campbell

, Yarborough

BJH

, et al. The prevalence of healthcare effectiveness data and information set (HEDIS) initiation and engagement in treatment among patients with cannabis use disorders in 7 US Health Systems. Subst Abus 2019;40(3):268–277; doi: 10.1080/08897077.2018.1544964

51.

51. Campbell

, Bahorik

, Kline-Simon

, et al. The role of marijuana use disorder in predicting emergency department and inpatient encounters: A retrospective cohort study. Drug Alcohol Depend 2017;178:170–175; doi: 10.1016/j.drugalcdep.2017.04.017

52.

52. Coughlin

, Pfeiffer

, Ganoczy

, et al. Quality of outpatient depression treatment in patients with comorbid substance use disorder. Am J Psychiatry 2021;178(5):414–423; doi: 10.1176/appi.ajp.2020.20040454

53.

53. Reynolds

, Afshar

, Wright

, et al. Association between substance misuse and outcomes in critically III patients with pneumonia. Ann Am Thorac Soc 2023;20(4):556–565; doi: 10.1513/AnnalsATS.202206-532OC

54.

54. Courchesne

, Smith

, Zúñiga

, et al. Association of alcohol and other substance‐related diagnoses with severe maternal morbidity. Alcohol Clin Exp Res 2021;45(9):1829–1839; doi: 10.1111/acer.14671

55.

55. Winhusen

, Theobald

, Kaelber

, et al. Increased morbidity and mortality in hypertensive patients with substance use disorders: Electronic health record findings. J Stud Alcohol Drugs 2020;81(4):471–478; doi: 10.15288/jsad.2020.81.471

56.

56. Fortuna

, Cook

, Porche

, et al. Sleep disturbance as a predictor of time to drug and alcohol use treatment in primary care. Sleep Med 2018;42:31–37; doi: 10.1016/j.sleep.2017.12.009

57.

57. Courchesne-Krak

, Zúñiga

, Chambers

, et al. Substance-related diagnosis type predicts the likelihood and co-occurrence of preterm and cesarean delivery. J Addict Dis 2023;41(2):137–148; doi: 10.1080/10550887.2022.2082834

58.

58. Graupensperger

, Corey

, Turrisi

, et al. Individuals with spinal cord injury have greater odds of substance use disorders than non-sci comparisons. Drug Alcohol Depend 2019;205:107608; doi: 10.1016/j.drugalcdep.2019.107608

59.

59. Armoon

, Grenier

, Cao

, et al. Frequencies of emergency department use and hospitalization comparing patients with different types of substance or polysubstance-related disorders. Subst Abuse Treat Prev Policy 2021;16(1):89; doi: 10.1186/s13011-021-00421-7

60.

60. Courchesne-Krak

, Kepner

, Rubano

, et al. Differences in outpatient, emergency, and inpatient use among pregnant women with a substance-related diagnosis. Am J Obstet Gynecol MFM 2022;4(2):100559.

61.

61. Bahorik

, Satre

, Kline-Simon

, et al. Alcohol, cannabis, and opioid use disorders, and disease burden in an integrated healthcare system. J Addict Med 2017;11(1):3–9; doi: 10.1097/ADM.0000000000000260

62.

62. Courchesne-Krak

, Marienfeld

, Kepner

. What brings you in today? Sex, race, substance type, and other sociodemographic and health-related characteristics predict if substance use is the main reason for a clinical encounter. Yale J Biol Med 2023;96(3):277–291.

63.

63. Bahorik

, Satre

, Kline-Simon

, et al. Alcohol, marijuana, and opioid use disorders: 5-Year patterns and characteristics of emergency department encounters. Subst Abus 2018;39(1):59–68; doi: 10.1080/08897077.2017.1356789

64.

64. Hasin

, Saxon

, Malte

, et al. Trends in cannabis use disorder diagnoses in the U.S. Veterans Health Administration, 2005–2019. Am J Psychiatry 2022;179(10):748–757; doi: 10.1176/appi.ajp.22010034

65.

65. Winhusen

, Theobald

, Kaelber

, et al. Medical complications associated with substance use disorders in patients with type 2 diabetes and hypertension: Electronic health record findings. Addiction 2019;114(8):1462–1470; doi: 10.1111/add.14607

66.

66. Gan

, Buxton

, Scheuermeyer

, et al. Risk of cardiovascular diseases in relation to substance use disorders. Drug Alcohol Depend 2021;229(Pt A):109132; doi: 10.1016/j.drugalcdep.2021.109132

67.

67. Winhusen

, Theobald

, Kaelber

, et al. The association between regular cannabis use, with and without tobacco co-use, and adverse cardiovascular outcomes: Cannabis may have a greater impact in non-tobacco smokers. Am J Drug Alcohol Abuse 2020;46(4):454–461; doi: 10.1080/00952990.2019.1676433

68.

68. Gates

, Turney

, Ferguson

, et al. Associations among substance use, mental health disorders, and self-harm in a prison population: Examining group risk for suicide attempt. Int J Environ Res Public Health 2017;14(3):317.

69.

69. John

, Wu

L-T

. Chronic non-cancer pain among adults with substance use disorders: Prevalence, characteristics, and association with opioid overdose and healthcare utilization. Drug Alcohol Depend 2020;209:107902; doi: 10.1016/j.drugalcdep.2020.107902

70.

70. Hasin

, Fink

, Olfson

, et al. Substance use disorders and COVID-19: An analysis of nation-wide Veterans Health Administration electronic health records. Drug Alcohol Depend 2022;234:109383; doi: 10.1016/j.drugalcdep.2022.109383

71.

71. Liu

, Vivolo-Kantor

. A latent class analysis of drug and substance use patterns among patients treated in emergency departments for suspected drug overdose. Addict Behav 2020;101:106142.

72.

72. Livne

, Malte

, Olfson

, et al. Trends in prevalence of cannabis use disorders among U.S. Veterans with and without psychiatric disorders between 2005 and 2019. medRxiv 2023; doi: 10.1101/2023.05.24.23290358

73.

73. Loyal

, Lavergne

, Shirmaleki

, et al. Trends in involuntary psychiatric hospitalization in British Columbia: Descriptive analysis of population-based linked administrative data from 2008 to 2018. Can J Psychiatry 2023;68(4):257–268; doi: 10.1177/07067437221128477

74.

74. Tolan

, Terebo

, Chai

, et al. Impact of marijuana legalization on cannabis-related visits to the emergency department. Clin Toxicol (Phila) 2022;60(5):585–595; doi: 10.1080/15563650.2021.2012576

75.

75. Hughto

, Quinn

, Dunbar

, et al. Prevalence and co-occurrence of alcohol, nicotine, and other substance use disorder diagnoses among US transgender and cisgender adults. JAMA Netw Open 2021;4(2):e2036512; doi: 10.1001/jamanetworkopen.2020.36512