825 Can Applying Machine Learning on Perioperative Team Communication Data Change the Way we Predict Case-time Duration?

Abstract

INTRODUCTION: Accurate case-time prediction is a critical component in improving operating room efficiency. Applying machine learning on patient and booking parameters has been shown to improve accuracy. However, the details of the actual surgical plan as determined by the surgeon pre-operatively have not been included in these predictions. In this study, we applied machine learning methods on data from perioperative team communication platform to predict case time duration. METHODS: The objective data collected from 535 neurosurgical cases by a perioperative team communication platform were divided to training and testing datasets (431 and 104 cases respectively). We developed a machine learning algorithm based on historical case-duration data (last 10 similar cases by the same surgeon) and compared to planning data collected from a perioperative team communication platform. We used the real case-time duration of the cases in the training dataset to train the algorithm. We then tested our model on the testing dataset and compared the accuracy of case-time prediction to the historical data alone prediction model, which is the common method used in our institution. RESULTS: The machine learning model combining historical and team communication data provided better accuracy in case-time prediction than historical data alone. The historical data method averaged 90.6 minutes difference in case-time from real time, and the machine learning model averaged 80.04 minutes (p=0.029). For cases longer than the case-time 50th percentile (183 minutes), the historical data method averaged 126.02 minutes difference from real time, and the machine learning model averaged 104.52 minutes (p=0.014). CONCLUSIONS: The actual case plan as discussed amongst the surgical team in the perioperative period provides information that is not often available otherwise. Applying a machine learning algorithm on the data captured through a team communication platform improved case time prediction compared to common methods.

Ähnliche Arbeiten