OR Scheduling and Planning Optimization using ML algorithms for surgery duration prediction
Anna Mosenzon, M.Sc. student at the Department of Industrial Engineering
12 May 2022, 14:30 PM, Room 206& via zoom
Abstract:
The use of Operational Research in healthcare has developed considerably over the years. Healthcare has become a major service sector, with many people involved either as employees or as consumers. The rising costs of healthcare due to new technologies and demographic trends, have had an impact on the demand for elective surgeries. The operating room (OR) is a vitally important issue for healthcare policy makers as it is the hospital’s largest cost and revenue center as well as a constrained source. Operating room utilization is typically affected by numerous factors that can have conflicting objectives with respect to productivity, quality of care, and quality of labor. These factors clearly stress the need for efficiency and necessitate the development of adequate planning and scheduling procedures. A further complicating factor of the OR planning is the stochastic nature of the process. One of the major uncertainties, is the stochastic durations surgeries. However, nowadays huge amount of data is available to better estimate those durations. In this study we suggest an efficient elective surgery planning method by developing an automated allocation system based on Mixed Integer Linear Planning (MILP) that relies on predicted surgeries durations, based on machine learning algorithms and that considers the constraints that the hospital is subjected to. We cooperated with Assuta hospital, that provided real data records as well as guidelines for a correct implementation. We showed that using the proposed approach we were able to i) better predict the surgeries durations while taking into account available data; ii) reduce unused capacity in by more than 60% comparing in comparison to the current method used by Assuta, while keeping the overtime almost the same. Also, the difference between the unused capacity using the ML algorithm prediction and the unused capacity using the actual durations (as a lower bound) is was found to be lower in 71% comparing to Assuta’s estimation.
Bio:
Anna Mosenzon, M.Sc. student at the Department of Industrial Engineering in Tel Aviv University, specializing in Data Science. Anna holds an MBA from the Technion and a B.Sc. in Industrial Engineering from Ben Gurion University. Her research focuses on OR Scheduling and Planning Optimization using ML algorithms for surgery duration prediction. This research was supervised by Prof. Irad Ben-Gal and Prof. Yossi Bukchin.
