Machine learning at the operating room of the future : a comparison of machine learning techniques applied to operating room scheduling

Abstract

The quality of an operating room schedule is determined by the accuracy of the surgery duration estimation used. State of the art estimation algorithms consider only three surgery variables-procedure type, surgeon identity, and date of surgery-to predict the length of surgeries. We show that if we can take advantage of a richer set of available information, we can significantly improve estimation accuracy. Additional recorded (but unused) variables include patient age, gender, and morbidity, anesthesiologist identity, and surgery location. We implement and compare the accuracy of four standard machine learning algorithms that take advantage of this richer data set: linear regression, nearest neighbors, regression trees, and support vector regression. We conclude that additional variables can improve the accuracy estimate by as much as 20%. Finally, we discuss the implementation challenges and future work necessary to make machine learning techniques available to the data analyst concerned with implementation. Portions of this work are sponsored by the U.S. Dept. of the Army, under DAMD 17-02- 2-0006. The information does not necessarily reflect the position of the government, and no official endorsement should be inferred.