INDIANAPOLIS — A new study co-led by two Indiana University School of Medicine scientists used machine learning to create vital sign benchmarks in intensive care unit patients that, when reached, coincided with better survival rates.
The report, recently published in npj Digital Medicine, focuses on Dynamic Cohort Ensemble Learning, or DynaCEL, a generic framework supporting a wide range of artificial intelligence, machine learning and statistical models. The published model was trained on ICU data from IU Health and other sources.
The framework scanned through medical records to create unique heart rate and systolic blood pressure goals for each patient. Conventional medical wisdom standardizes these benchmarks at 80 beats per minute and 120 mmHg, but the DynaCEL system generated more accurate and personalized estimates for ideal heart and blood pressure rates.
"The ultimate goal is to reduce deaths in the ICU," said LZ Meng, MD, vice chair for clinical and outcomes research in the Department of Anesthesia and a co-leader of the study.
"Right now, doctors often use general thresholds for heart rate and blood pressure that may not be ideal for every patient," Meng said. "Our work shows that personalized targets, estimated using artificial intelligence, are associated with lower risk of dying within 24 hours compared with fixed, one-size-fits-all targets."
Although the results are promising, Meng said, they are based on past medical records. Clinical trials that assess the benchmarks in real time are needed to determine if meeting the AI-generated benchmarks definitively led to better survival outcomes.
AI has its limits, though. It relies on medical history forms, which can sometimes be inaccurate or incomplete.
Meng hopes to continue studying possible applications of machine learning, AI and statistical models in precision medicine.
"By precision medicine, we mean identifying optimal physiological targets for individual patients, not only heart rate and blood pressure, but also any physiological variables that clinicians monitor and manage," Meng said. "We are also interested in precision interventions, including personalized drug choices, ventilator settings and other treatment strategies, with the ultimate goal of improving outcomes through more individualized care."
Jing Su, PhD, who co-led the study with Meng, is an associate professor of biostatistics and health data science at the IU School of Medicine and an expert in medical informatics and artificial intelligence.
"This work is one of many examples of the successful strategic collaboration of two IU School of Medicine departments — anesthesia and biostatistics and health data science — established since 2023," Su said. "As an engine of research, innovation and training, this strategic collaboration attracted researchers from IU Bloomington and Purdue University, provided training opportunities to MD, PhD and master’s degree graduate students, as well as high school interns, and demonstrated regional influence in clinical research."
Other IU School of Medicine authors include Jiangqiong Li, Xiang Liu, Yanhua Sun, Zuotian Li, Jinjin Cai, George Lu, David C. Adams and Ziyue Liu.
The research team also included researchers from the Luddy School of Informatics, Computing and Engineering at IU Bloomington and Purdue’s Polytechnic Institute.
About the Indiana University School of Medicine
The IU School of Medicine is the largest medical school in the U.S. and is annually ranked among the top medical schools in the nation by U.S. News & World Report. The school offers high-quality medical education, access to leading medical research and rich campus life in nine Indiana cities, including rural and urban locations consistently recognized for livability. According to the Blue Ridge Institute for Medical Research, the IU School of Medicine ranks No. 13 in 2024 National Institutes of Health funding among all public medical schools in the country.
Writer: Rory Appleton, rapplet@iu.edu
For more news, visit the IU School of Medicine Newsroom: medicine.iu.edu/news