In the U.S., trauma is the leading cause of death for those 1-45 years old and hemorrhage remains the largest contributing factor to preventable death. Providers must rapidly identify those suffering from hemorrhage to optimize outcome, but internal bleeding remains difficult to diagnose even for experienced clinicians. Little is known on presentation about those suffering from occult hemorrhage and providers must quickly make treatment decisions in these time-pressured, time-sensitive clinical scenarios. This proposal seeks to develop through artificial intelligence, a type of advanced machine learning, prediction algorithms that could be deployed at the bedside of patients to assist clinicians with more timely recognition of hemorrhage. By doing so, we hypothesize that this approach (integrating diverse data sources that have not previously been combined to one another) could identify patterns in our patients that far surpass current capabilities to quickly detect and act on the critical components contributing to outcome. The ability to rapidly pinpoint these patterns and display them to the bedside clinician could allow more timely intervention and precise therapeutic approaches for hemorrhage control. Beyond the challenges in rapidly identifying bleeding, current treatment of hemorrhage is rudimentary with a standard resuscitation approach for all patients. This reflects attempts to optimize outcome based upon the average treatment effect, rather than being adaptable for unique patient phenotypes. Hemorrhage is believed to initiate a complex chain of events involving crosstalk between the coagulation and inflammatory systems that are hypothesized to play a key role in outcome. Trauma has a known time zero of onset, making it an ideal model to study the immediate pathophysiologic changes associated with hemorrhage. This complex, individual patient biology is believed to explain why those suffering similar injury have differing outcomes. However, to date, these individual characteristics are poorly understood and not factored into initial treatment approaches. Through this proposal, I also seek to define novel digital biomarkers representing patient phenotypes that require precision resuscitation approaches to maximize outcome. Fundamental to reducing hemorrhagic deaths is the need to elucidate a deeper understanding of these mechanistic models of patient states. Strategies that help to identify novel patient phenotypes that could benefit from more tailored treatment pathways may provide important advances in decreasing preventable death. The net result of this proposal will be a deeper insight into the mechanistic models contributing to evolving patient states following hemorrhage, and identify the key phenotypes or digital biomarkers associated with mortality, complications, and occult hemorrhage. Finding solutions to advance our resuscitation approaches following hemorrhage has potential to decrease complications, save lives, and reduce health care costs.

Public Health Relevance

In the US, traumatic injury is the number one cause of death for those under 45 years old and these deaths include many patients dying from the consequences of bleeding. Through leveraging the power of artificial intelligence (a scientific analysis approach similarly used in non-medical fields to help answer questions from large amounts of complex information), an integrated approach to measuring outcome will be developed utilizing biologic, clinical, and electronic medical record (EMR) data. The goal of this project is to lay the ground work for developing early warning detection systems that could identify those at risk of complications early and assist care providers in selecting the best treatment to minimize complications and death from hemorrhage.

National Institute of Health (NIH)
National Heart, Lung, and Blood Institute (NHLBI)
Research Project (R01)
Project #
Application #
Study Section
Bioengineering, Technology and Surgical Sciences Study Section (BTSS)
Program Officer
Kindzelski, Andrei L
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of California San Francisco
Schools of Medicine
San Francisco
United States
Zip Code