This study develops a multivariable data analysis technique for subjects who are monitored for four vital signs, namely heart rate, respiration rate, blood pressure and body temperature, and proposes a method to provide early warning of abnormal health condition and visual analytics that identifies the contributing factor that causes the early warning. It proposes the use of the deranged values of the vital signs of a collection of subjects to fix a multivariable PCA model and set the control level based on the Hotelling T2 statistics. The test subject is monitored using the model and the control level. The values are deemed abnormal when the multivariable observation exceeds the control limit by two consecutive observations, and an early warning is issued when there are multiple of such detections. The study shows that there is statistical significance between the distribution of the number of detections in subjects who are well enough to be discharged from ICU and those whose conditions are so bad that they did not survive the stay in the ICU.
Woo, W. L., Koh, B. H. D., Gao, B., Nwoye, E. O., Wei, B., & Dlay, S. S. (2020). Early Warning of Health Condition and Visual Analytics for Multivariable Vital Signs. In ACM International Conference Proceeding Series (pp. 206–211). Association for Computing Machinery. https://doi.org/10.1145/3398329.3398356