Back to all papers

Use of AI-Related CPT Codes by Patient, Geographic, and Payer Characteristics in the U.S., 2021-2026

November 11, 2026medrxiv logopreprint

Authors

Kim, E. J.,Chang, S.,Feng, J.,Bedmutha, P. S.,Poursoltan, L.,Cao, J.,Wellsjo, A.,Floyd, E.,Luo, D.,Zhu, K.,Boussina, A. J.,Holland, J.,Sitapati, A.,Hallaj, S.,Nemati, S.,Gabriel, R.,Khalessi, A.,Longhurst, C.,Singh, K.,Ramsis, M.

Affiliations (1)

  • University of California, San Diego

Abstract

ImportancePayment policy for clinical artificial intelligence (AI) is changing quickly, yet national data on which AI-enabled services are used, and for whom, are limited. ObjectiveTo describe national trends in the use of AI-related Current Procedural Terminology (CPT) codes and how use differs by patient, geographic, and payer characteristics. Design, Setting, and ParticipantsThis serial cross-sectional study used Epic Cosmos, a deidentified electronic health record dataset, to examine 11,605,867,309 encounters among 233,628,292 US patients from January 1, 2021, to April 14, 2026. Exposures: Calendar year, age, sex, race, ethnicity, rurality, area-level social vulnerability, census region, and primary payer. Main Outcomes and MeasuresEncounters with any of 65 CPT codes that require AI, identified by matching the CPT code set to US Food and Drug Administration-listed AI-enabled devices and classified as assistive, augmentative, or autonomous using the CPT Appendix S taxonomy. ResultsOf 11.6 billion encounters, 335,349 (2.9 per 100,000) included an AI-required CPT code. The odds of an encounter with an AI-required code were 9.88 (95% CI, 9.71-10.06) times higher in 2025 than in 2021. In unadjusted comparisons, use was higher among patients aged 65 to 85 years and among those with Medicare (OR, 5.11) or Medicare Advantage (OR, 5.12) vs commercial insurance. Use was lower among Black (OR, 0.72) and Hispanic or Latino (OR, 0.34) patients, Medicaid enrollees (OR, 0.65), residents of the most socially vulnerable areas (OR, 0.53), and residents of the most rural areas (OR, 0.86). Use of automated diabetic retinopathy screening increased after the service was added to a HEDIS quality measure. One newer code (quantitative brain MRI) appeared in rural areas within months of becoming effective, whereas older codes such as retinopathy screening took years. Conclusions and RelevanceIn this cross-sectional study, billed use of AI-enabled services remained rare but grew substantially and was concentrated among older and Medicare-insured patients, with lower use among Medicaid enrollees and residents of socially vulnerable areas. As the Centers for Medicare & Medicaid Services restructures payment for software and launches outcome-based models, tracking who receives AI-enabled care across payment pathways will be important for equitable access. KEY POINTSO_ST_ABSQuestionC_ST_ABSHow has billed use of artificial intelligence (AI)-related CPT codes changed nationally since 2021, and how does use differ by patient, geographic, and payer characteristics? FindingsIn this cross-sectional study of 11.6 billion US encounters, 2.9 per 100,000 included an AI-required CPT code. Odds of use were nearly 10 times higher in 2025 than in 2021, higher among Medicare beneficiaries, and lower among Medicaid enrollees and residents of socially vulnerable areas. MeaningBilled AI use remains rare and uneven across payers and communities, so payment policy for AI may need to be paired with monitoring of who receives AI-enabled care.

Topics

health policy

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.