AI tool predicts surgical sepsis

Sepsis is the body’s dangerous overreaction to an infection, and a Spanish research team says genetic data may help flag surgical patients at higher risk before an operation. (who.int) (ciberisciii.es) The group at Hospital Clínico Universitario de Valladolid, working with CIBER research networks, the University of Valladolid and the University of Leicester, reported the study on January 20, 2026 and published the paper in *Frontiers in Medicine* on December 15, 2025. (ciberisciii.es) (pmc.ncbi.nlm.nih.gov) Their model used genome-wide association study data — a scan across a person’s DNA for common variants — from 750 patients who developed sepsis after surgery and 3,500 population controls. The researchers said the explainable artificial intelligence system predicted risk and ranked which variants contributed most. (ciberisciii.es) (pmc.ncbi.nlm.nih.gov) The study highlighted variants in three genes — PRIM2, RBSN and SYNPR — and linked them to functions such as gene regulation, DNA replication, cell signaling, cell growth and heart dysfunction. Those are clues about biological pathways, not a diagnosis on their own. (ciberisciii.es) (pmc.ncbi.nlm.nih.gov) The proposed use is a preoperative blood test that helps sort patients into higher- and lower-risk groups before surgery. The Valladolid team said that could support earlier monitoring and more tailored care after the procedure. (ciberisciii.es) That matters because surgery and hospital care are established risk settings for sepsis, and the condition remains a major cause of death worldwide. The World Health Organization says sepsis was linked to 11 million deaths globally in 2020, while the United States Centers for Disease Control and Prevention says at least 1.7 million adults in the United States develop sepsis each year. (cdc.gov) (who.int) (cdc.gov) The paper does not present the tool as a replacement for surgeons or anesthesiologists. It presents the model as a clinician aid, and the authors say more validation is needed before routine hospital use. (pmc.ncbi.nlm.nih.gov) (ciberisciii.es) That caution fits the wider field. A separate *Scientific Reports* study published on April 2, 2026 used more than 150 clinical variables from 2,571 emergency surgery patients at 29 medical centers to predict postoperative sepsis, showing that researchers are testing both genetic and bedside-data approaches. (nature.com) For now, the Valladolid result is a research-stage example of how DNA data could be used before the first incision. The next test is whether hospitals can show the model improves decisions and outcomes in real patients, not just on past datasets. (pmc.ncbi.nlm.nih.gov) (nature.com)