Machine Learning Study to Improve Surgical Case Duration Prediction

An accurate estimate of the duration of an intervention is essential to optimize the utilization of the operating room, it plays a fundamental role in reducing the cost of the operating room (OR). The approaches most used by hospitals are based on historical averages based on a specific surgeon or specific type of procedure obtained from the electronic medical record (EMR) scheduling systems. However, the low predictive accuracy of the EMR leads to negative impacts on patients and hospitals, such as rescheduling of surgeries and cancellations which costs a lot of money. Our aim in this study is to improve the prediction of surgeries duration using advanced machine learning algorithms to find a predictive model. We obtained a large data set containing 66,857 surgery cases undergone in Rivoli and Pinerolo from the year 2016 and on. After exploring the data by detecting the outliers and plotting, we proceeded to examine the data and analyze it, then we trained the model. We computed historic averages of each surgery and they were used as a baseline model for comparison with the model we created. After constructing the model, it was implemented on an application to be more user-friendly and be used by anyone by just different variables as input and getting the predicted time of the surgery in addition to a graph showing occupation time of the operation theatre of the chosen type of surgery in the past according to the data we have. The machine-learning algorithm showed higher predictive capability than electronic medical record (EMR), which is a notable advancement towards statistical modeling of case-time duration in all surgical departments, enabling improved operating room efficiency, cost, and scheduling.