Research study leverages big data to help predict inpatient mortality
Investigators at the University of Florida College of Medicine – Jacksonville are part of a novel approach that utilizes machine learning to help doctors predict which patients admitted to the hospital are at the greatest risk of death, regardless of demographics and care setting. This can help health care teams target high-risk individuals and adjust the care approach and resources accordingly.
The activities are part of a study titled, “A Multi-domain machine learning approach to predicting in-hospital mortality.” The study — also referred to as MONITOR— uses de-identified, historical patient data to determine features that are associated with a high-risk of inpatient mortality, and was recently published in the Journal of Medical Artificial Intelligence. The efforts combine vital signs, lab work, medications, consults, diagnoses, demographics and other data to better predict real-time mortality risk for intensive care and non-intensive care units.
According to the Centers for Disease Control and Prevention, there are more than 700,000 inpatient deaths in the United States each year. With tools like MONITOR, a hospital can better align resources like rapid response teams to patients at greatest risk of needing them. It also continuously evaluates a patient’s risk throughout their hospital stay rather than assigning a one-time assessment at the time of hospital admission. Additionally, unlike previous risk assessment tools that focus on single diseases, MONITOR is unique in that it assesses the complete health status of patients.
“Machine learning gives us the capability to broaden our view of potential risks regardless of the specific cause of admission, so we can help save more lives,” said Jennifer Fishe, MD, an assistant professor and director of the Center for Data Solutions at the UF College of Medicine – Jacksonville, and senior author on the publication.
The team of investigators that developed MONITOR utilized over 150,000 patient electronic health records from UF Health Jacksonville between January 2014 and January 2022.
“This new tool continuously evaluates a patient’s risk of dying while in the hospital and thus it can help doctors better identify those patients who need enhanced surveillance and management,” said Alexander Parker, PhD, senior associate dean for research affairs at the UF College of Medicine – Jacksonville. “This is a great example of how our researchers and clinicians are leveraging big data and artificial intelligence to address key issues in health care. True to our vision to be the region’s most valuable health care asset, these efforts focus on issues that have an impact at the patient bedside as well as at the population level.”
Moving forward, the machine learning model needs to undergo further internal validation using prospective data as well as external validation using data from other hospital populations to determine how the model performs with other patient cohorts. Further research is required to help integrate this model into the clinical workflow and become part of the operational setting.
“As an academic health center, we are committed to seeing this all the way to the finish line,” Parker said. “The ultimate goal is not a publication or more research grants, but the integration of a new tool that will help transform how we practice medicine, differentiate the care our doctors provide and drive better outcomes for our patients.”
The UF College of Medicine – Jacksonville has attained national recognition for its role as a leader in transformative, cutting-edge research. Among physicians and scientists, the institution is known for fostering a rich, collaborative environment that fuels discovery and encourages creativity. Research is centered around changing lives and moving medicine forward. Visit med.jax.ufl.edu/research for more information.
For more information, please contact:
UF Health Media Relations
Uf Health Shands Patient Portal