Comprehensive Summary
This study attempts to optimize operation room use by predicting total knee arthroplasty (TKA) and total hip arthroplasty (THA) time using machine learning (ML). 2 ML models were trained using data collected internationally from 302,490 TKA and 196,942 THA surgeries. The resulting model had an accuracy of 78.1% for TKA and 75.4% for THA. Three optimization formulation for the operation room usage were also considered. The rooms could be taken by any surgeon (Any), by only two surgeons (Split), or by a single surgeon (Multiple Subset Sum Problem/MSSP). The results indicate that the Split and MSSP formulation performed better than the Any method. Compared to using the mean for the operation time, the ML prediction method reduced overtime but was not efficient in preventing underutilization of rooms. Nevertheless, the ML model outperformed the mean schedule 97.1% of times during simulations.
Outcomes and Implications
Total knee arthroplasty and total hip arthroplasty are frequently performed procedures that are rising in demand. Due to it demanding more and more resources, it is important to make use of the available operation rooms more efficiently. This study suggests that machine learning based scheduling system would improve operating room efficiency. Even though the method does not lead to a perfect schedule, it outperforms the schedule with means most of the time. As a result, this method would help healthcare institutions that are struggling with financials and waitlists.