In a groundbreaking effort to tackle the persistent issue of postoperative complications leading to patient mortality, a team at the University of Pittsburgh Medical Center (UPMC) has developed an innovative artificial intelligence (AI)-based tool. This tool aims to provide physicians with a real-time risk assessment of patients scheduled for surgery, helping identify those at higher risk and potentially optimizing their health before the procedure. The research, published in JAMA Network Open, sheds light on the potential of AI in revolutionizing surgical outcomes.
UPMC’s AI Tool Outperforms Standard Risk Metrics
The AI-based tool, designed by an interdisciplinary team comprising anesthesiologists, surgeons, and data scientists, has been trained on an extensive dataset of 1.4 million patients who underwent surgery at UPMC between 2012 and 2019. By analyzing more than 3,000 variables from patient history, including social determinants of health like proximity to healthy food options and access to housing, the model calculates each patient’s mortality risk 30 days post-surgery.
Aiming for Better Surgical Outcomes
With over 4.2 million global deaths annually attributed to surgical complications, and an estimated 100,000 deaths occurring in U.S. hospitals each year after surgery, the need for better risk assessment and preoperative optimization is evident. The AI tool equips physicians with valuable information to identify patients who might benefit from additional medical intervention before undergoing surgery. This “prehabilitation” program is aimed at improving patient health status, such as managing chronic conditions or quitting smoking, to enhance the chances of a successful surgery.
Complementing Human Expertise
While AI offers powerful predictive capabilities, the team at UPMC emphasizes that human judgment and decision-making remain integral to the process. The AI tool is designed to work alongside physicians, enabling them to generate their own risk scores and make the final call. This collaborative approach ensures that medical decisions are well-informed and tailored to individual patient needs.
Addressing Potential Challenges
Despite the promise of AI in optimizing surgical outcomes, there are challenges that demand attention. Ensuring that the AI model does not perpetuate bias or contribute to health disparities is crucial. Transparency with patients is essential, as they need to be aware when an AI model is used to assess their risk. Additionally, patient privacy concerns must be addressed, although the researchers acknowledge that the risk of reidentifying patients exists even without AI applications.
Looking to the Future
The study’s success marks a significant step forward in AI-assisted medical care. However, experts highlight the importance of demonstrating the tool’s actionability and its impact on patient outcomes. Conducting randomized controlled trials to evaluate the AI tool’s effectiveness compared to conventional methods will be crucial in guiding its widespread implementation.