Artificial Intelligence-Enabled Opportunities to Reduce Administrative Bloat in Gastroenterology Practices: Scoping...

Health care administration accounts for approximately 25% of health care expenditures. We performed a scoping review of artificial intelligence-enabled administrative products in gastroenterology and hepatology, including widely deployed commercial platforms, examining administrative functions, validation practices, and alignment with high-burden workflows.

A 2-component search strategy combining systematic literature search and a targeted environmental scan of commercial platforms identified 34 discrete artificial intelligence products and 2 narrative reviews. Each product was assessed using the 4-domain Framework for the Appropriate Implementation and Review of Artificial Intelligence rubric. Validation was categorized as unvalidated, internally validated, vendor-sponsored, or externally validated.

Products spanned 8 administrative function domains, including billing and coding, clinical documentation, patient engagement, and prior authorization. Twenty-six of 34 were general-platform products, and only 8 were gastroenterology-specific, of which 7 remained precommercialized. Four of 34 products reported no validation data, 2 relied on vendor-sponsored evidence, and 4 were externally validated. Framework for the Appropriate Implementation and Review of Artificial Intelligence scores were lowest for Fairness (1.59/5; standard deviation, 0.66), followed by Reliability (2.57/5; standard deviation, 0.97), Interpretability (2.72/5; standard deviation, 1.01), and Accountability (2.78/5; standard deviation, 0.64).

Commercialized products were broadly designed, least validated, and scored lowest on Framework for the Appropriate Implementation and Review of Artificial Intelligence metrics, particularly Fairness, whereas gastroenterology-specific products with greater methodological rigor remained largely precommercialized. Commercial general-platform products lacked gastroenterology-specific validation, and high-burden specialty workflows including hepatology care coordination and transplant documentation remained comparatively underserved. Realizing the potential of administrative artificial intelligence in gastroenterology requires moving toward empirically validated, specialty-specific, and equitable solutions, where success is measured by reduced administrative burden and time returned to patient care, not by revenue generation.