Clinical Artificial Intelligence Implementation in Routine Care: Real-World Operational Outcomes from a Provincial...

Clinical artificial intelligence (AI) technologies are increasingly being introduced into hospital practice, yet evidence describing their operational integration and performance after deployment in routine clinical settings remains limited. This study examined the real-world implementation and operational integration of clinical AI within a provincial tertiary health system in China over an 18-month observation period.

This retrospective longitudinal observational study used aggregated institutional data generated during routine platform deployment, including electronic medical record-linked system logs, deployment records, quality-monitoring summaries, and operational reports. The analysis focused on implementation patterns, workflow integration, selected operational indicators, and user acceptance during routine clinical use. The study evaluated implementation and operational integration rather than algorithmic accuracy, diagnostic performance, or patient-level clinical effectiveness.

Three AI-supported clinical pathways were included: an intelligent pre-consultation system, a multidisciplinary tumor decision-support system, and a duloxetine therapeutic drug-monitoring pathway. During the observation period, 127 clinicians across 53 specialties participated in AI-assisted clinical activities involving more than 27,000 patient encounters, 850 multidisciplinary tumor decision-support cases, and 320 therapeutic drug-monitoring episodes. Patient waiting time decreased from 18 to 13 minutes, patient satisfaction increased from 95.40% to 98.92%, consultation efficiency improved by approximately 40%, and documentation completion efficiency improved by approximately 80%. Implementation patterns differed substantially across pathways, reflecting differences in workflow position and clinical accountability rather than deployment effort alone.

In this provincial tertiary health system, clinical AI implementation was associated with sustained operational use, cross-specialty workflow integration, and measurable changes in selected workflow indicators. These findings suggest that technical functionality alone is insufficient for sustained clinical AI use and that workflow compatibility and organizational readiness are central to routine implementation. Because this was a single-site observational study using aggregated operational data, the findings should be interpreted as implementation evidence rather than proof of clinical effectiveness.