Predictive Analytics of the Risk of In-Hospital Mortality of Heart Failure Patients in Intensive Care Unit
Establishing and assessing appropriate predictive analytics model to forecast the likelihood of in-hospital death among heart failure patients admitted to ICU is the primary objective of this study. The research conducted a retrospective examination of health records obtained from MIMIC-III, a heterogeneous group of intensive care unit-admitted patients with heart failure. The research used machine learning algorithms, including Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boost (XGB), and Light Gradient Boost (LGB), to construct a prediction model using demographic data, vital signs, laboratory findings, and clinical history. To improve the accuracy and performance of the model, feature selection methods and cross-validation are employed. Predictive factors that contribute to the likelihood of in-hospital mortality among heart failure patients in a controlled environment of the ICU are anticipated to be identified in this research. Precision, Metrics such as Recall, F1-score, Area Under the Receiver Operating Characteristic Curve (AUROC), and K-fold Cross-validation were used to assess the performance of the model. This research aims to improve patient management outcomes, and makes valuable contribution to the expanding domain of predictive analytics in healthcare by enabling individualized and focused treatments, which eventually enhance the standard of care delivered to ICU patients suffering from heart failure. The potential of predictive analytics to improve clinical decision-making for patients with heart failure in intensive care units is highlighted in this work.
Download Paper