For a study, researchers sought to develop and calculate a machine learning model that predicted short-term mortality in the intensive care unit utilizing the trends of 4 easy-to-collect vital signs. The primary training cohort involved 1,968 Veterans Health Service Medical Center patients. The external validation study comprised 409 patients at Seoul National University Hospital. Datasets of heart rate, systolic blood pressure, diastolic blood pressure, and peripheral capillary oxygen saturation (SpO2) calculated each one for 10 h were used. The process of mortality prediction models generated by issuing 5 machine learning algorithms, Random Forest (RF), XGboost, perceptron, convolutional neural network, and Long Short-Term Memory, were evaluated and determined utilizing area under the receiver operating characteristic curve (AUROC) values and an external validation dataset. The machine learning model that came using the RF algorithm revealed the best performance. Its AUROC was 0.922, which was much better than the 0.8408 of the Acute Physiology and Chronic Health Evaluation II. The machine learning model developed using SpO2 showed the best performance (AUROC, 0.89). This mortality prediction process could have been helpful for initial testing of probable mortality and accurate medical intervention, especially in sudden deteriorating patients.
Source – sciencedirect.com/science/article/abs/pii/S0883944122001356
