Localized Semiparametric Prediction: A Precision Medicine Approach in a Trauma Patient Population

• Activity Number: 261
• Type: Contributed
• Date/Time: Monday, August 1, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Sara E. Moore* and Alan E. Hubbard and Mitchell J. Cohen
• Companies: University of California at Berkeley and University of California at Berkeley and University of California at San Francisco
• Keywords: Personalized medicine ; Prediction ; Semi-parametric methods ; Ensemble learning

Abstract

Traumatic injury is the leading cause of death among Americans under 50 years of age, claiming over 136,000 lives in the U.S. in 2014 alone. Physicians providing emergency and trauma care often have limited time and information to make life-saving treatment decisions for severely injured patients. These decisions are typically made univariately or based on rudimentary multivariate scoring systems which are often not targeted for the clinical outcomes of interest. We believe that trauma care would greatly benefit from improved patient-specific or "localized" decision support algorithms. We propose a supervised classification method which performs both dimension and instance reduction data-adaptively to target only the most relevant information for a given patient. The algorithm utilizes an ensemble learner, the Super Learner, which chooses the best combination of individual learners via minimization of cross-validated loss. The predictive performance of the algorithm in real-world patient data is demonstrated to exceed that of existing trauma scoring systems. Future directions include real-time treatment decision support for use by critical care clinicians via a mobile app.