Artificial intelligence for alcohol, opioid, and cannabis use disorders screening and management: a narrative review of barriers and facilitators to clinical implementation.

Frimpong I.; Hlormenu B. · Frontiers in digital health · 2026

**Background & Methods** This is a narrative review synthesizing evidence on artificial intelligence and machine learning applications for screening and management of alcohol, opioid, and cannabis use disorders. The authors integrated the Framework for AI Implementation Research in Healthcare (FAIIR-H) with the Unified Theory of Acceptance and Use of Technology (UTAUT) to analyze barriers and facilitators across four implementation domains: data/model, clinician/workflow, patient, and system/regulatory factors. **Key Findings** - Alcohol use disorder has the largest predictive AI literature but remains methodologically heterogeneous with limited external validation; opioid use disorder demonstrates more methodologically mature evidence with fairness auditing; cannabis use disorder has the most limited evidence base. - Real-world deployments show mixed evidence: a hospital-based opioid AI screener was associated with 47% lower odds of 30-day readmission across 51,000+ hospitalizations (supported by fairness-auditing and implementation-outcome evidence), while an alcohol relapse-management platform showed up to 18% reduction in relapse risk within a 500,000+ patient-day dataset, though with less rigorous implementation and fairness evidence. - No substance-specific AI application has yet demonstrated simultaneous success in both screening and management; implementation is significantly shaped by stigma, data-sharing concerns, digital access barriers, and 42 CFR Part 2 confidentiality requirements specific to substance use disorder treatment. **Dosage & Administration** Not reported. **Safety & Adverse Effects** Not reported. The review identifies implementation challenges including patient trust burden escalating with directness of AI-patient interaction, though adverse effects of AI systems themselves are not quantified. **Evidence Quality** This narrative review demonstrates significant heterogeneity in the underlying literature. Major limitations include: (1) most AI models remain in development/validation phases without sustained clinical implementation; (2) external validation is limited across all three substance classes; (3) fairness auditing is inconsistently performed; (4) implementation-outcome evidence is sparse; (5) no studies demonstrate simultaneous screening and management efficacy. The authors propose a research agenda prioritizing external validation, fairness auditing, implementation science in safety-net settings, and policy framework development to advance clinical translation.

More recent cannabis research