Ryan S. Sultan, MD

Home → Research → Grants

Cannabis Use and New-Onset Psychosis

Title: Cannabis Use and New-Onset Psychosis: Research Background and Aims
Principal Investigator: Ryan S. Sultan, MD
Funding: $335,500
Period: 2025-2026
Type: Pharmacoepidemiological Study

1. Background and Significance

Epidemiological Evidence: Cannabis use has long been linked to psychosis in epidemiological studies. Chronic or heavy cannabis use correlates with a higher likelihood of developing psychotic symptoms, a concern relevant due to the 2x increase of adult cannabis use between 2001-2002 and 2012-2013 in the US. Notably, cannabis can induce transient psychotic-like experiences even in otherwise healthy individuals – roughly 1 in 200 people who use cannabis may experience short-lived psychotic symptoms.

Dose-Response Relationship: There is evidence of a dose–response relationship: more frequent cannabis use (e.g. weekly or daily) is associated with increased risk of psychosis. A meta-analysis estimated that approximately one-third (33.7%) of individuals experience cannabis use at the onset of a first psychotic episode, highlighting how common cannabis exposure is among new psychosis cases.

Key Finding - European Multi-Site Study: In a large multi-site European study, daily cannabis use was reported by 29.5% of patients with first-episode psychosis (versus only 6.8% of controls). After adjusting for other factors: Daily cannabis users had ~3x higher odds of developing a psychotic disorder compared to never-users. Daily use of high-potency cannabis (high THC content) was associated with ~5x higher odds.

Biological Mechanisms: Cannabis is believed to interact with the dopamine system in the brain – a key pathway implicated in psychosis. Recent neuroimaging evidence shows midbrain dopamine alterations (substantia-nigra/ventral tegmental area), mirroring the dopamine signaling abnormalities seen in people with untreated psychosis. This suggests cannabis may trigger or exacerbate psychosis via increased dopamine neurotransmission.

Epigenetic Evidence: Emerging epigenetic evidence suggests that recent or cumulative cannabis use induces genome-wide DNA methylation patterns in middle aged adults. These alterations were overrepresented in pathways related to schizophrenia and bipolar disorder, suggesting that cannabis exposure may imprint enduring epigenetic signatures which may serve as a mechanism for changes in neurodevelopmental and psychiatric pathways.

Critical Period - Adolescent Exposure: Initiating cannabis use during adolescence (while the brain is still developing) has been linked to lasting changes in brain structure and higher psychosis risk in adulthood. Frequent high-THC exposure has been associated with earlier onset of psychotic disorder.

Public Health Context: As cannabis becomes more accessible due to legalization, understanding its impact on psychosis incidence is critical. More permissive state cannabis policies are associated with increased adult cannabis use and CUD. The natural experiment created by staggered state-by-state legalization offers a unique opportunity to examine whether looser cannabis laws are linked to an uptick in psychotic disorders.

2. Research Aims and Hypotheses

Overall Objective: To determine the association between clinically documented cannabis use and incident psychotic disorder, and to test modification of this association by state cannabis legalization contexts. This project will use a large healthcare database to conduct a retrospective cohort analysis.

Aim 1: Association Between Cannabis Use and Incident Psychosis

Hypothesis: Individuals with a record of cannabis use will have an elevated risk of a subsequent psychotic disorder diagnosis 12 months after a CUD diagnosis compared to those with no cannabis use record.

Methodology: Identify cohort with evidence of cannabis use (ICD codes F12.x), comparison group without cannabis codes, focus on NEW episodes of cannabis use (first diagnosis during study period), ensure no prior history of psychotic disorder, follow forward in time (1-2 years) for onset of psychotic disorder.

Statistical Approach: Calculate incidence rate/cumulative incidence of psychosis in exposed vs. unexposed, regression modeling to estimate strength of association, adjust for confounders: age, sex, medications, psychiatric history, other substance use, Cox proportional hazards model or logistic regression for 1-year incidence, penalized estimation if needed for sparse outcomes, subgroup analyses by age group (adolescents, young adults, adults).

Expected Findings: Cannabis exposure associated with significantly increased risk of psychotic disorder onset, potentially mirroring the ~3x increased risk seen in daily users in previous studies.

Aim 2: Impact of Cannabis Legalization Context on Psychosis Risk

Hypothesis: The relationship between cannabis use and psychosis is moderated by the cannabis legalization environment. States or time periods with more permissive cannabis laws will exhibit higher rates of cannabis-associated psychosis.

Methodology: Stratify analyses by state and time period according to legalization status. Categorize exposure contexts: (a) Pre-legalization (neither medical nor recreational legal), (b) Medical-only legalization, (c) Recreational legalization in effect. Compare psychosis incidence among cannabis users across different contexts. Difference-in-differences analysis exploiting staggered timing of legalization.

Statistical Approach: Stratified cohort analysis by legalization context, test for interaction between individual cannabis use and state-level legalization status, group states by legalization timing (early-adopting vs. late-adopting vs. non-legalizing), Cox models within each policy category or unified model with interaction term, time-stratified analysis: compare rates before vs. after legalization.

Expected Findings: In more permissive environments, the cannabis–psychosis association will be stronger or absolute rates of psychosis will be higher. Higher attributable risk in states with legal recreational use.

3. Data Source and Analytical Approach

Data Source: Large-scale de-identified healthcare database (IBM MarketScan Commercial Claims and Encounters or similar claims/EHR dataset). Contains medical and pharmacy claims for millions of insured individuals across the U.S.

Timeframe: Approximately 2010–2025 (period encompassing significant cannabis policy changes)

Cannabis Use Identification: ICD-10 codes in F12 category (cannabis use, abuse, dependence), ICD-9 equivalents: 304.30, 305.20, etc., ancillary codes indicating cannabis-related clinical encounters, date of first observed cannabis-use diagnosis = index date for cohort entry.

Psychotic Disorder Outcomes: ICD-10 codes F20–F29 (schizophrenia, schizophreniform, schizoaffective, delusional disorder, acute/transient psychosis, unspecified psychosis), ICD-9 codes: 295.x, 297.x, 298.x, include mood disorders with psychotic features and substance-induced psychosis, classify appropriately (primary vs. substance-induced) in sensitivity analyses.

Exclusion Criteria: Patients with any psychotic disorder code prior to index date (focus on new-onset cases only).

4. Expected Outcomes and Implications

Scientific Impact: Quantify risk of psychotic illness attributable to cannabis use in large U.S. cohort, provide estimate of public health burden, shed light on whether cannabis legalization influences mental health outcomes at population level.

Clinical Implications: If Aim 1 confirms significantly elevated risk, reinforces need for targeted prevention (educating young cannabis users about psychosis), early intervention strategies, clinical guidance for high-risk individuals (adolescents, family history of psychosis).

Policy Implications: If Aim 2 shows higher psychosis incidence in permissive environments, policymakers should approach legalization with caution, implement safeguards: potency limits, youth access restrictions, public health campaigns. If no increase observed post-legalization, alleviates some concerns, helps focus attention on other aspects of cannabis policy.

Broader Impact: Findings can inform clinical guidance and shape public policy, balancing cannabis's medical/recreational benefits with potential psychiatric risks. Provides empirical evidence on psychiatric consequences during ongoing debates about cannabis legalization.

Additional Active Grants

Columbia University Research Stabilization Fund
Amount: $100,000 | Date: June 2025 | Role: Principal Investigator

Completed Research Grants

AACAP Pilot Research Award: Antipsychotic Medications in the Treatment of ADHD
Amount: $26,145 | Period: June 2017 - November 2019 | Role: Principal Investigator
Institution: Columbia University Department of Psychiatry / New York State Psychiatric Institute

Project Summary: This AACAP-funded pilot study examined antipsychotic prescribing patterns among youth with newly diagnosed ADHD. Using the MarketScan Commercial Database (2010-2015), the study analyzed 187,563 youth ages 3-24 with new ADHD diagnoses to determine the percentage prescribed antipsychotics and identify associated clinical and demographic factors.

Key Findings:

• 2.6% (n=4,342) of youth with new ADHD diagnosis were prescribed an antipsychotic in the first year
• Only 52.7% of antipsychotic-treated youth had an FDA-approved or evidence-supported indication for antipsychotic use
• Less than half (47.9%) received stimulant medication before initiating antipsychotic treatment
• Antipsychotic use strongly associated with psychiatric comorbidity:
  • Self-harm/suicidal ideation (aOR 7.5, 95% CI 5.9-9.6)
  • Oppositional defiant disorder (aOR 4.4, 95% CI 3.9-4.9)
  • Substance use disorder (aOR 4.0, 95% CI 3.6-4.5)
  • Inpatient treatment (aOR 7.9, 95% CI 6.7-9.3)
• Preschool-aged children prescribed antipsychotics were notably psychiatrically and neurologically complex

Clinical Implications: The findings highlight concerns about off-label antipsychotic prescribing in youth with ADHD, particularly when evidence-based ADHD treatments (stimulants) are not tried first. Results emphasize the importance of careful diagnostic evaluation and consideration of comorbid psychiatric conditions before initiating antipsychotic treatment.

Published Research: Sultan RS, Wang S, Crystal S, Olfson M. Antipsychotic Treatment Among Youths With Attention-Deficit/Hyperactivity Disorder. JAMA Network Open. 2019;2(7):e197850.

Conference Presentations:

• Sultan RS, Olfson M. Cohort Analysis of Antipsychotic Treatment in ADHD. AACAP Annual Meeting, Chicago, IL, 2019
• Sultan RS, Olfson M. Clinical and Demographic Risk Factors for Antipsychotic Prescribing in Youth With ADHD. AACAP Annual Meeting, Seattle, WA, 2018
• Sultan RS, Olfson M. Identifying Problematic Behaviors among Adolescents with ADHD. AACAP Annual Meeting, Washington DC, 2017

Secondary Research from Grant: The pilot award also supported analysis of adverse behaviors in adolescents with ADHD using the National Comorbidity Survey Adolescent Supplement (NCS-A), examining 6,483 youth ages 14-18. Key findings included significantly elevated risks for suicide attempts (aOR 2.9), physical aggression (aOR 2.3), school expulsion/job termination (aOR 3.3), and alcohol-related problems among adolescents with ADHD compared to those without ADHD.

NIDA K12 Mentored Clinical Scientist Development Award
Amount: $100,000 | Period: July 2021 - Present | Role: Principal Investigator
Focus: Bioinformatics and Substance Use
Mentors: Frances Levin, MD and Tim Wilens, MD
Institution: Columbia Vagelos College of Physicians and Surgeons

NIMH K08 Career Development Award: Natural Language Processing of EHRs for ADHD Pharmacologic Management
Amount: $400,000 (requested) | Period: 2019-2023 (proposed 4 years) | Role: Principal Investigator
Primary Mentors: Mark Olfson, MD, MPH (Columbia) and Carol Friedman, PhD (Columbia Biomedical Informatics)
Advisors: Adler Perotte, MD, MA (Columbia), Thomas McCoy, MD (Harvard/MGH), Frances Levin, MD (Columbia), Jonathan Posner, MD (Columbia)
Institution: Columbia University Department of Psychiatry / New York State Psychiatric Institute

Research Question: Why does pharmacological treatment of youth with ADHD so often depart from evidence-based standards?

Project Overview: This innovative career development award proposes using Natural Language Processing (NLP) of large-scale Electronic Health Records (EHR) to understand why stimulant medications (first-line evidence-based treatment) are underprescribed for youth with ADHD, while less effective medications like antipsychotics are increasingly used in community practice. The project will analyze over 8,000 children and adolescents with ADHD from the NewYork-Presbyterian (Columbia and Cornell) combined EHR database.

Specific Aims:

• Aim 1: Identify patient, caregiver, provider, and clinical care characteristics associated with stimulant prescription initiation in ADHD
  • Hypothesis: Greater hyperactivity, caregiver-initiated referrals, and psychoeducation will be associated with stimulant initiation
• Aim 2: Compare stimulant prescription initiation between psychiatrists and primary care providers
  • Hypothesis: Youth treated by primary care providers will be more likely to receive stimulants; caregiver discomfort with stimulants will partially mediate this difference
  • Preliminary data shows primary care providers prescribe stimulants more often than psychiatrists (replicated across NSCH, MEPS, and MarketScan datasets)
• Aim 3: Identify characteristics associated with antipsychotic prescription as initial treatment for ADHD
  • Hypothesis: Aggressive behavior and child/parent conflict will be associated with antipsychotic initiation
  • Context: Less than half of antipsychotic-treated youth with ADHD receive stimulants first (see JAMA Network Open 2019 publication)

Innovative Methodology:

• Applies NLP using MedLEE (Medical Language Extraction and Encoding) system developed by Dr. Friedman
• Extracts patient characteristics (symptoms, severity, ADHD subtypes), caregiver factors (concerns about side effects, parent-child conflict, psychoeducation received), clinical care interventions from unstructured EHR narrative notes
• Combines NLP-extracted data with structured EHR elements (diagnoses, treatment settings) and provider characteristics (specialty, training)
• Extends traditional pharmaco-epidemiologic methods beyond administrative data limitations

Training Objectives:

1. Expertise in secondary use of EHR data in research (quality, biases, data management, privacy safeguards)
2. Foundational understanding of clinical informatics, machine learning, and NLP principles
3. Advanced knowledge in clinical epidemiology and mental health services research methods
4. Professional development in grantsmanship and publishing

Career Development Plan: Extensive coursework at Columbia Mailman School of Public Health and Department of Biomedical Informatics including:

• Acculturation to Programming and Statistics, Computational Methods in Biomedical Informatics
• Symbolic Methods in Biomedical Informatics, Computer Applications in Health Care
• Translational Bioinformatics, Analysis of Large-Scale Data, Analysis of Longitudinal Data
• Strategic Issues in Health Care Quality and Implementation
• Applied Methods in Health Services and Outcomes Research

Public Health Significance: Identifying determinants of prescribing practices can shape initiatives to improve ADHD pharmacologic care. Preliminary data shows that following new ADHD diagnosis, only 48.1% of youth receive stimulant medication. Administrative data lacks clinical detail to identify key factors limiting evidence-based treatment. NLP-extracted EHR data can reveal patient presentations, caregiver concerns, and psychoeducation details buried in narrative notes that influence treatment decisions.

Background Context:

• ADHD affects 5-10% of children and adolescents, most prevalent major childhood psychiatric disorder
• Stimulant medications have effect sizes approaching 1.0, improve academic performance, reduce motor vehicle accidents, improve social functioning
• Despite evidence, 1/3 to 1/2 of ADHD youth are not prescribed stimulants
• Antipsychotic use in ADHD has increased, yet only half have FDA-approved indication
• Caregiver knowledge variable: most US adults unable to provide detailed ADHD information, perceive psychotropic treatment as adversely affecting development

Planned Publications (6 manuscripts):

1. Review of Long-Term Outcomes in ADHD (Year 1)
2. Using NLP to Develop Categories of Clinical Characteristics (Year 2, Methods)
3. Evaluating Natural Language Processing Accuracy of Narrative Descriptions of ADHD (Year 2, Methods)
4. NLP-Extracted Factors of Stimulant Treatment in ADHD (Year 3, Aim 1)
5. Provider Specialty Effect on Pharmacological Management in ADHD (Year 3, Aim 2)
6. NLP-Extracted Factors Associated with Antipsychotic Treatment in ADHD (Year 4, Aim 3)

NIMH Strategic Alignment: Strengthens public impact of research (4.0), applies new computational modeling and data analytics to EHRs (4.1), advances precision medicine by characterizing provider specialty variation (3.2), aligns with NIH Big Data to Knowledge (BD2K) program objectives.

Future R01 Direction: Expand methods to multi-site EHR databases (Harvard-affiliated hospitals via Dr. McCoy, national databases via Dr. Perotte), combine NLP-extracted EHR data with administrative datasets to examine medication dosing and polypharmacy, extend methods to other child mental health disorders.

Using AI to Improve Mental Health Screening Outcomes
Date: December 2023 | Role: Principal Investigator
Focus: AI/machine learning applications in mental health screening

The Christopher D. Smithers Foundation, Inc.
Role: Principal Investigator | Focus: Addiction research

Explore Related Content

PUBLICATIONS FROM THESE GRANTS - Cannabis & Psychosis publications
CONFERENCE PRESENTATIONS - APSARD 2026 plenary, AACAP cannabis posters
DIGITAL MENTAL HEALTH - JAMA Psychiatry viewpoint on telepsychiatry integration
GRANT TIMELINE - All grants in chronological context
COMPLETE GRANT HISTORY - All grants, funding amounts, dates, roles in complete CV
RESEARCH OVERVIEW - High-level overview of active research projects