Beyond accuracy and time saved: implementation outcomes for generative AI in hospital administrative and clinical-administrative workflows—a mini review
Year: 2026
Type: article
Abstract: Although hospitals are introducing generative artificial intelligence (GenAI) into documentation, messaging, and administrative workflows, model accuracy and time saved do not establish safe or sustainable implementation. This Mini Review examines which implementation and downstream outcomes should inform decisions to pilot, scale up, sustain, modify, or discontinue these workflows. A PubMed-led structured narrative search through 25 August 2026, with an update on 27 August, informed a purposive core of 17 empirical studies: seven direct administrative studies and 10 clinical-administrative studies. Supporting reviews and established implementation frameworks informed the interpretation of the findings. Perceived workload, logged efficiency, adoption, safety, and downstream outcomes did not improve consistently together. Favourable average quality could coexist with consequential errors and substantial verification work. Within this selected core, cost, penetration, and sustainability were measured less often than early fit, use, and feasibility. No study was coded as providing an operationalised equity assessment. We present a literature-informed outcome-to-decision crosswalk that links these findings to clinical workflow governance. This distinguishes non-compensable stop rules from minimum progression criteria and comparative value targets, with explicit ownership, denominators and reassessment. Illustrative stop rules address serious uncorrected errors, unauthorised data disclosure and bypassed human review. The crosswalk is an unvalidated normative synthesis, not a universal stage model. Its value requires prospective testing against patient safety, workforce burden and service outcomes.
Source: Frontiers in Medicine
Authors: Xieyi Wang, Hezhuo Wang
Institutions: Ningbo No.6 Hospital, Ningbo University Affiliated Hospital
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W7225087945