Acute Clinical Deterioration Calculator - Leveraging Predictive Modeling to Enhance Patient Deterioration Identification
This research project tackles a challenge in healthcare: the early recognition and intervention for in-hospital patient deterioration. Delayed responses to clinical deterioration contribute to preventable deaths and adverse outcomes, underscoring the need for reliable systems that identify early warning signs. Many existing tools rely on late-stage indicators and are limited by poor sensitivity, specificity, and integration into clinical workflows.
To address these gaps, this project develops and evaluates the Acute Clinical Deterioration Calculator (ACDC), a machine learning-based predictive tool designed to identify early signals of patient decline by analyzing electronic health record (EHR) data. ACDC leverages over 100 variables, including vital signs and laboratory results, to anticipate deterioration before it becomes clinically apparent. These variables are linked to cellular and molecular markers that reflect the underlying physiological changes and genetic influences driving patient outcomes, such as inflammation, hypoxia, and metabolic imbalance.
By working to improve the detection of cellular biomarkers and physiological changes—such as altered oxygenation (hypoxia), systemic inflammatory markers, and shifts in electrolyte balances—the ACDC model connects high-level clinical observations to the underlying molecular and cellular processes. This integration enables proactive intervention, offering opportunities to prevent adverse events and improve patient safety.
This project includes collaborating with clinical teams to refine the design and implementation of the ACDC early warning system, and develop skills in incorporating feedback from healthcare providers to improve the usability and interpretability of alerts.