AI Model Forecasts Postoperative Infection Risks to Improve Patient Care

A groundbreaking AI tool is set to revolutionize postoperative patient care by predicting the likelihood of developing infections within seven and 30 days after surgery. Developed collaboratively by researchers including Ph.D. candidate Siri van der Meijden at Leiden University, this model aims to assist healthcare professionals in early identification and intervention for patients at higher risk of infections.

The AI system, named PERISCOPE, utilizes extensive historical patient data—covering nearly a decade—from electronic health records across three hospitals: LUMC, Radboudumc, and Ziekenhuis Oost-Limburg in Belgium. The data, which includes factors such as past infections, diabetes, heart rate, blood pressure, and body weight, enables the machine learning algorithms to assess individual infection risk accurately.

In rigorous testing involving over 250,000 surgical procedures, PERISCOPE demonstrated performance comparable to experienced clinicians. During studies at LUMC, for example, the AI's predictions matched the judgment of seasoned doctors and provided superior insights when clinician confidence was low, highlighting its potential as a decision-support tool.

The application of PERISCOPE is designed to streamline pre- and post-surgical assessments. Healthcare teams, including surgeons, junior doctors, and nurses across various departments, will be able to view a patient's infection risk as a percentage and categorize it as low, medium, or high. This consolidated information aids in quicker decision-making, targeted monitoring, and timely interventions, potentially reducing hospital stays, readmissions, and the need for multiple treatments.

Implementation of the tool into clinical practice is targeted for mid-2026, following integration with the hospital's electronic records system. The development process has taken around five years, ensuring compliance with safety standards and detailed data curation. Future improvements aim to enable PERISCOPE to predict specific infection types and other complications like bleeding or mortality, leveraging continuous data updates.

This AI predictor exemplifies how data-driven technology can augment medical judgment without replacing healthcare providers. Its adoption promises to enhance patient outcomes, optimize resource use, and offer a proactive approach to managing postoperative risks.

For more details, refer to the full report at Leiden University’s event page or visit source.