HUMAN-CENTERED PREDICTION MODEL TO STREAMLINE DECISION-MAKING IN SEPSIS MANAGEMENT PATHWAYS

Sepsis presents a significant challenge in healthcare due to its rapid progression and high mortality rates. In healthcare settings, the development of a human-centered prediction model for sepsis holds the potential to enhance decision-making processes. Such a model comprehensively integrates systems-related factors, including clinician insights, workflow dynamics, and human-centered design principles. The contribution of this study includes leveraging expert knowledge during the model development. It offers valuable perspectives on the significance of specific features in learning complex relationships and generating accurate predictions. Then, expert knowledge can be combined with data analytics to develop a machine learning-based prediction model, which can aid in the early recognition and management of sepsis. The result shows that the gradient boost outperformed the other algorithms with an accuracy of 0.75, precision of 0.76, recall of 0.75, and F1 score of 0.75. The result of ML-based algorithms in sepsis prediction models will enhance the decision-making process in the healthcare setting, especially for risk assessment during triage. The current alert system is based on a Modified Early Warning Score (MEWS), which is unreliable as it often generates false positive alarms. Our study suggests that healthcare professionals can be involved in developing and implementing the proposed model to ensure it aligns with their needs and practices.