Making the operating room schedule more efficient with artificial intelligence models

The Erasmus MC aims to deliver the highest standard of patient care. Unfortunately, delivering the highest standard of care comes with numerous challenges. Staff shortages and increasing demands due to aging populations result in long-term capacity shortage for the expected demand [1]. Simultaneously, healthcare expenses continue to rise, driven by costs of treatments and personnel, the growing administrative burdens, and inefficiencies in resource allocation. These challenges affect both hospitals and patients, highlighting the urgent need for cost-effective and innovative solutions. Efficient scheduling is one of the practical solutions to tackle these issues. By leveraging tools and strategies to better allocate staff, equipment, and facilities, healthcare organizations can maximize the use of their existing resources. This ensures that patients receive timely and high-quality care and also helps balancing workloads. One key element of scheduling is the coordination of operating rooms (ORs). The OR is one of the most expensive departments of a hospital, therefore improperly coordinated OR schedules can lead to substantial financial losses, a disruption in patient treatment and dissatisfied personnel due to working overtime [2], [3]. Therefore, it is important that the OR schedule is robust and resources are utilised as efficient as possible. The Erasmus MC encounters several challenges within their OR planning that lead to a number of core problems: rescheduling surgeries, growing waiting lists, personnel working overtime, and costs without production. This study will focus on the problems related to scheduling at the operational level. Artificial intelligence and operations research techniques represent innovative solutions to tackle complex scheduling challenges in healthcare. These techniques leverage advanced algorithms, predictive analytics, and models to create more robust schedules. Unlike traditional approaches, artificial intelligence can process vast amounts of data in real time, identify patterns, and adapt to changing circumstances. Operations research techniques provide robust frameworks for optimizing workflows and minimizing inefficiencies. Together, these tools enable healthcare systems to streamline operations, reduce costs, and improve both staff satisfaction and patient care outcomes [4], [5]. Therefore, this research will look into the application of these techniques to create a more robust OR schedule.

The objective of this study is to create a more robust OR schedule for the operational level using artificial intelligence and operations research techniques. The objective is split into four sub-projects. Each sub project will look at an aspect of the OR schedule. 1. Predicting duration of elective operations; 2. Predicting the probability of overtime for an operating room and the OR-complex in general ; 3. Bed allocation at the PACU; 4. Creating insight in emergency patients and their surgery duration.

Single Center retrospective cohort analysis with prospective implementation of the model For this research, medical records of patients are used (numerical, categorical and textual data)). 

Study population: Patients that underwent surgery in the Erasmus MC in the last 10 years, both elective and non-elective surgeries.

Methods The method will consist of the following components: 

• Collection of data: data is collected via the HDP. 
• Processing of data: cleaning and processing of the categorical, numerical and textual data. NON WMO Research protocol V 1.0, 03-12-2024 5 OR schedule and AI 
• Artificial intelligence: selecting, comparing and fine-tuning of relevant machine learning models for objectives that cover a form of prediction. 
• Operations research techniques: selecting appropriate mathematical models for objectives that cover mathematical problems. 
• Verification, validation and implementation: verification and validation of models before implementation. Creating most suitable implementation pathway in consultation with experts and end-users.

none

There are no burden and risks, because there is no involvement from participants needed. Only historical data (i.e. data previously gathered by the hospital) will be used for this research. The data will be processed according the relevant laws and regulations.