July 7, 2023
Pittsburgh — Researchers and physicians at the University of Pittsburgh and UPMC used machine learning to create and deploy an accurate and flexible model for predicting patients at high risk for complications after surgery, says today, 10 October 2016. The results of a new study published in the month were announced. JAMA network open.
Before the COVID-19 pandemic, complications 30 days after surgery were the third leading cause of death worldwide, killing about 4.2 million people each year. Identifying patients at high risk for complications before undergoing surgery is critical to saving lives and reducing healthcare costs.
“Improving the overall health status of patients before surgery through pre-rehabilitation can go a long way toward improving outcomes for high-risk patients,” said Pitt, dean of anesthesiology and perioperative medicine at the School of Medicine. Aman Mahajan, M.D., Ph.D., MBA said. , Director of UPMC Perioperative and Surgical Services. “However, identifying high-risk patients can be challenging for busy clinicians who need to integrate the wealth of available health data and frequently perform additional tests and clinical evaluations. wanted to build an easy-to-use model that would use existing data in electronic medical records to quickly provide medical teams with automated and accurate risk assessments.”
To create the model, Mahajan, UPMC’s Chief of Medical Data Analytics, Dr. Oscar Marroquin, and his team trained the algorithm to learn from the medical records of more than 1.25 million surgical patients. The model focused on mortality and whether patients had significant brain or cardiac events such as stroke or heart attack after surgery. This model was then validated on an additional 200,000 patients undergoing surgery at his UPMC.
After validation, the model was deployed in 20 UPMC hospitals. Each morning, the program reads the electronic medical records of patients scheduled for surgery and flags those deemed high-risk. This notification will allow the clinical team to better coordinate care, make healthier decisions, and perform some pre-rehabilitation prior to surgery, including referral to the UPMC Perioperative Care Center. and reduce the risk of complications. Clinicians can run models on demand at any time.
To better understand how their model compares to industry standards, Mahajan and his team compared the model to the American College of Surgeons’ National Surgical Quality Improvement Program (ACS NSQIP). . ACS NSQIP is used in hospitals nationwide, but requires clinicians to manually enter patient information and is unpredictable when information is missing. Mahajan and his team found that their model outperformed ACS NSQIP in identifying high-risk patients.
“We designed the model with healthcare workers in mind,” says Marroquin. “Our model is fully automated and can make educated predictions even when some data is missing, so there is little additional burden on the clinician and a high level of reliability. It provides a highly useful tool.”
As the model continues to refine and develop, Mahajan and his team aim to train a program to predict the likelihood of sepsis, respiratory disease, and other complications that often hospitalize patients after surgery. increase.
Additional authors of this study are Stephen Esper, M.B.A., Jaime Artman, M.S.C., Cynthia Klahre, M.D., CRNP, Senthil Sadhasivam, M.D., M.S., University of Pittsburgh. Dr. Thein Htay Oo, Dr. Jeffery McKibben, Dr. Michael Garver, Dr. John Ryan, and Dr. Jennifer Holder-Murray, all from UPMC.
Photo Details: (Click image for high resolution version)
Credit: UPMC
Photograph Caption: Aman Mahajan, M.D.
Credit: UPMC
Photograph Caption: Oscar Marroquin, MD
