From manual to AI-assisted: Improving infusion scheduling in a high-volume cancer center.

e23297 Background: Efficient use of infusion center resources is critical for high-volume cancer centers, yet infusion scheduling is often manual and labor-intensive. More effective scheduling has the potential to enhance patient-centered care through appointment consolidation, improve nursing continuity, and increase operational efficiency. We implemented Epic’s artificial intelligence (AI)-enabled, decision tree–based scheduling workflow as the standard process for infusion appointment booking at a large academic cancer center and evaluated its impact on operational outcomes. Methods: In September 2025, an automated infusion template generator that used historical operational data to optimize scheduling capacity was implemented, replacing manual electronic calendar-based scheduling. The workflow incorporated a decision tree–based tool guiding schedulers to align treatment requests with available resources and was informed by multidisciplinary stakeholder input. Statistical process control (SPC) charts evaluated changes in operational metrics during the three months before and after intervention. Primary outcomes included time spent scheduling appointments, nursing continuity defined by the proportion of appointments booked with a patient’s primary nurse, and chair utilization rate. A balancing measure was patient layover time between same-day appointments. Control limits were set at ±3 standard deviations (99.7%). Special cause variation was identified using Institute for Healthcare Improvement (IHI) criteria: points outside control limits, shifts of ≥7 consecutive points on one side of the center line, and trends of ≥6 consecutively increasing or decreasing points. Results: During the study period, 22,514 infusion appointments were scheduled. After implementation, SPC analysis demonstrated sustained shifts in scheduling time and patient layover time. Mean scheduler time per appointment decreased from 65 to 40 seconds, corresponding to an estimated 20 person-hours saved per month. Mean patient layover time between same-day appointments decreased from 53 minutes to 45 minutes. Following the process change, the proportion of chemotherapy appointments booked with a patient’s primary nurse rose above control limits. In the three months after implementation, the proportion of chemotherapy appointments booked with a patient’s primary nurse rose from 13 to 17%. No special cause variation was observed for chair utilization. Conclusions: AI-enabled decision tree scheduling was feasible in a high-volume oncology infusion center and was associated with improvements in select operational metrics including scheduling time, nursing continuity, and patient layover time, without adversely affecting chair utilization.