Institute for Machine Learning, Johannes Kepler University, Linz, Austria.
Department of Anesthesiology and Critical Care Medicine, University and University Hospital, Zürich, Switzerland.
Transfusion. 2020 Sep;60(9):1977-1986. doi: 10.1111/trf.15935. Epub 2020 Jun 28.
The ability to predict transfusions arising during hospital admission might enable economized blood supply management and might furthermore increase patient safety by ensuring a sufficient stock of red blood cells (RBCs) for a specific patient. We therefore investigated the precision of four different machine learning-based prediction algorithms to predict transfusion, massive transfusion, and the number of transfusions in patients admitted to a hospital.
This was a retrospective, observational study in three adult tertiary care hospitals in Western Australia between January 2008 and June 2017. Primary outcome measures for the classification tasks were the area under the curve for the receiver operating characteristics curve, the F score, and the average precision of the four machine learning algorithms used: neural networks (NNs), logistic regression (LR), random forests (RFs), and gradient boosting (GB) trees.
Using our four predictive models, transfusion of at least 1 unit of RBCs could be predicted rather accurately (sensitivity for NN, LR, RF, and GB: 0.898, 0.894, 0.584, and 0.872, respectively; specificity: 0.958, 0.966, 0.964, 0.965). Using the four methods for prediction of massive transfusion was less successful (sensitivity for NN, LR, RF, and GB: 0.780, 0.721, 0.002, and 0.797, respectively; specificity: 0.994, 0.995, 0.993, 0.995). As a consequence, prediction of the total number of packed RBCs transfused was also rather inaccurate.
This study demonstrates that the necessity for intrahospital transfusion can be forecasted reliably, however the amount of RBC units transfused during a hospital stay is more difficult to predict.
在住院期间预测输血的能力可能使血液供应管理得以节约，并通过确保为特定患者储备足够数量的红细胞（RBC）来提高患者安全性。因此，我们研究了四种基于机器学习的预测算法在预测患者住院期间输血、大量输血和输血次数方面的精度。
这是一项在澳大利亚西部的三家成人三级保健医院进行的回顾性观察研究，时间为 2008 年 1 月至 2017 年 6 月。分类任务的主要结局指标是接收者操作特征曲线的曲线下面积、F 分数以及使用的四种机器学习算法（神经网络 (NNs)、逻辑回归 (LR)、随机森林 (RFs) 和梯度提升 (GB) 树）的平均精度。
使用我们的四个预测模型，对至少输注 1 单位 RBC 的输血可以进行相当准确的预测（NN、LR、RF 和 GB 的敏感性分别为 0.898、0.894、0.584 和 0.872，特异性分别为 0.958、0.966、0.964 和 0.965）。使用这四种方法预测大量输血的效果较差（NN、LR、RF 和 GB 的敏感性分别为 0.780、0.721、0.002 和 0.797，特异性分别为 0.994、0.995、0.993 和 0.995）。因此，预测输注的总 RBC 单位数量也相当不准确。
本研究表明，可以可靠地预测医院内输血的必要性，但预测住院期间输注的 RBC 单位数量更具挑战性。
