Intelligent Clinical Decision Support

Area of Science:

• Critical Care Medicine
• Biomedical Informatics
• Artificial Intelligence in Healthcare

Background:

• Early recognition of cardiorespiratory stress and decompensation in critically ill patients is challenging, even with intensive monitoring.
• Patient instability signifies a failure to adequately respond to cardiorespiratory stress.
• Intensive Care Units (ICUs) generate vast amounts of high-frequency numeric, waveform, and Electronic Health Record (EHR) data.

Purpose of the Study:

• To develop Artificial Intelligence (AI) models for detecting and forecasting instability in ICU patients.
• To create AI-driven clinical decision support (CDS) systems for real-time forecasting and mitigation of critical instability.
• To leverage multi-source patient data for unbiased, reliable, and usable AI systems in healthcare.

Main Methods:

• Utilizing large volumes of multi-source patient data, including high-frequency numeric and waveform data from bedside monitors and EHR data.
• Applying machine learning and systems engineering principles to develop AI models.
• Integrating AI models into a real-time CDS for bedside deployment, emphasizing human factors and bias evaluation.

Main Results:

• The study presents an approach for creating an operationally relevant AI-based forecasting CDS system.
• The developed system aims to leverage multi-source data for improved instability detection and prediction.
• Focus on integrating AI into clinical workflows for practical bedside application.

Conclusions:

• AI holds significant potential for improving patient outcomes in ICUs by enabling early detection and forecasting of critical instability.
• The development of unbiased and reliable AI-based CDS systems is a high priority for healthcare.
• Successful implementation requires a multidisciplinary approach, including machine learning, systems engineering, and human expertise.