Case Study: UMC Utrecht reduces unnecessary antibiotic use and achieves 90% sepsis-prediction accuracy with SAS

UMC Utrecht uses data analytics to proactively treat or even prevent infections in premature babies

UMC Utrecht’s Neonatology Department collects vast amounts of patient data and faced a pressing clinical challenge: how to use that data to better detect infections in premature infants and reduce unnecessary antibiotic treatment. As part of its Applied Data Science in Medicine (ADAM) program, the hospital set out to answer—within three months—whether analytics could proactively identify sepsis and improve treatment decisions for vulnerable NICU babies.

A multidisciplinary team used SAS Enterprise Guide, Enterprise Miner and Visual Analytics to extract, anonymize and model ten years of NICU device and patient data. The resulting ML algorithm predicts the bacteria that cause sepsis with about 90% accuracy (versus ~40% for clinician judgment), revealing that 60% of infants had been treated with antibiotics unnecessarily and enabling data-driven decisions that reduce overtreatment and improve neonatal care.

UMC Utrecht

Daniel Vrijbrief

Neonatologist