Machine learning based prediction of perioperative blood loss in orthognathic surgery

Abstract

The aim of this study was to evaluate, if and with what accuracy perioperative blood loss can be calculated by a machine learning algorithm prior to orthognathic surgery. The investigators implemented a random forest algorithm to predict perioperative blood loss. 1472 patients who underwent orthognathic surgery from 01/2006 to 06/2017 at our institution were screened and 950 patients were included and separated 80%/20% in a training set - utilized to generate the prediction model - and a testing set - utilized to estimate the accuracy of the model. The outcome variable was the correlation between actual perioperative blood loss and predicted perioperative blood loss in the testing set. Other study variables were the difference of actual and predicted perioperative blood loss and important factors influencing perioperative blood loss using random forest feature importance. Descriptive and bivariate statistics were computed and the P value was set at 0.05. There was a statistically significant correlation between actual perioperative blood loss and predicted perioperative blood loss (p < 0.001). The mean difference was 7.4 ml with a standard deviation of 172.3 ml. The results of this study suggest that the application of a machine-learning algorithm allows a prediction of perioperative blood loss prior to orthognathic surgery.

Keywords: Blood loss; Machine learning; Orthognathic surgery.