Predictive Analytics in 30-Day Hospital Readmissions for Heart Failure Patients
Si-Chi Chin, Rui Liu, Senjuti Basu Roy · 2016
Surgical operations involve a great level of complexity. In order to cope with complexity and increase information visibility, modern healthcare facilities are investing in advanced biomedical sensing and information technology. For example, when patients are moved into an intensive care unit after surgical operations, clinicians and nurses will closely monitor a large number of clinical variables such as heart rate, pulse oximetry, blood pressure, gas exchange, and blood test results (e.g., metabolic panel, complete blood count). Heterogeneous sensing gives rise to data-rich environments in hospitals. However, clinicians are not trained and do not have effective decision-support systems to help estimate clinical status and optimize healthcare management policies considering large and heterogeneous data sets (e.g., multiple attributes describing a patient's health conditions). There is a dire need to go beyond current clinical practice and develop data-driven methods and tools that will enable and assist (i) the extraction of pertinent knowledge about clinical status from heterogeneous healthcare recordings, (ii) the prediction of mortality risks, and (iii) the provision of personalized decision-support systems. This present study focuses on the predictive modeling of mortality rates in intensive care units using patient-specific healthcare recordings. Note that variable heterogeneity, patient heterogeneity, and time asynchronization are common characteristics in the postsurgical monitoring. To tackle these challenges, this chapter presents a postsurgical decision-support system that consists of a suite of analytical tools, including data categorization, data preprocessing, feature extraction, feature selection, and predictive modeling. Experimental results show that the proposed system performs favorably over traditional approaches and yields better results based on the evaluation of real-world data from 4000 subjects. This research shows great potentials for the use of data-driven analytics to improve the quality of postoperative healthcare services.