Data-Driven Exploration: Analyzing Triage Data's Role in Predicting Intensive Care Unit Transfers within 24 Hours

In the realm of healthcare analytics, the importance of data quality and comprehensiveness cannot be overstated. This research focuses on conducting a detailed exploratory data analysis (EDA) of triage data to support the development of an automated model with the aim of addressing the high false positive rate associated with existing early warning systems in healthcare. Our investigation centers on the triage data of patients admitted to the Emergency Department who subsequently require ICU transfers within 24 hours. Our analysis reveals data distribution patterns, such as a normal temperature range of 95 to 102 degrees Fahrenheit, heart rates spanning 50 to 125 beats per minute, and a gradual decline in oxygen saturation levels from 100% to around 90%. These findings enable us to effectively manage hidden outliers and appropriately encode and score the features within the dataset. While our implemented feedforward neural network classifier has made progress in reducing the false positive rate, it maintains a moderate level of accuracy. This indicates that while we have improved the system's performance, there is still room for enhancement. Our ongoing commitment is to continue refining the real-world application of our system, with a primary focus on minimizing false positives and maximizing its impact on real-time patient care. This combination of analytical insights and clinical knowledge holds great potential to enhance patient outcomes and streamline patient management. Ultimately, our research aims to pave the way for a future in which healthcare professionals can more effectively direct their attention to where it is most urgently needed, thereby improving the overall quality of care.