211 - Improving Productivity in 39-OR Surgical Center via AI-Powered Auto Scheduling App

Description: Study Question and Background Information
Against the backdrop of China's healthcare transformation from "accessible care" to "high-quality care," the growing public demand for high-quality surgical services increasingly conflicts with limited medical resources.

Particularly in large surgical centers with over 35 ORs, maximizing OR utilization and increasing surgical capacity within regular working hours (8:30 -17:30) has become critical to alleviating resource shortages. Traditional surgical center scheduling, reliant on manual planning by the head nurse, suffers from inefficiencies, uneven resource allocation, and underutilization of ORs, failing to meet the refined operational needs of large surgical centers. Thus, there is an urgent need to optimize scheduling through IT approaches.

Consistent with the AORN Guidelines for Perioperative Practice, which emphasize leveraging health information technology (HIT) to enhance workflow efficiency, this study aligns with recommendations such as "Health care organizations should use electronic patient health care records in the perioperative setting" (1.1) and "The health care organization should use an electronic documentation platform that has an alternate data entry and backup process" (1.5). These guidelines underscore the value of standardized, technology-driven processes to support perioperative efficiency, providing a framework for integrating IT tools into surgical center scheduling.

Design Description
To address these challenges, this study developed an AI-based surgical center scheduling optimization scheme with three core components:
1. AI-driven accurate surgical duration prediction: Minimizing scheduling disruptions caused by duration fluctuations to ensure scheduling accuracy.
2. AI-driven auto-scheduling application: Utilizing AI algorithms to generate the next day's surgical center schedule automatically.
3. Multidimensional evaluation system: Defining operational optimization goals to quantify "optimization status."

Methodology
1. Surgical Duration Prediction: An AI model was trained using historical surgical data. Preliminary tests showed that this model, with graph convolutional neural network (CNN) algorithms, reduced prediction errors to within 20 minutes, outperforming manual estimates (average error: 59.3 minutes) and traditional deep neural networks (DNN) (49.5 minutes) referenced from a paper study at Memorial Sloan-Kettering Cancer Center (MSK).

2. Automated Scheduling Application: A software tool helps the head nurse reduce manual scheduling time from 3 hours to 0.5 hours, significantly improving efficiency.

3. Evaluation Indicators: Three core metrics were established to comprehensively assess efficiency: monthly OR utilization rate, OR load balance, and total OR usage duration, moving beyond single-index evaluations.

Preliminary results demonstrated an 83% reduction in manual operation time for surgical center scheduling. To further validate efficacy and productivity in surgical centers, a 3-month controlled study will compare three groups: manual scheduling, AI automated scheduling, and actual operational outcomes. By analyzing these groups against the defined metrics, the study aims to verify AI's role in enhancing utilization, balancing loads, and extending effective usage time—ultimately identifying optimal productivity directions for surgical centers.