Evidence · Research paper
Development and validation of a diagnostic model for early differentiation of sepsis and non-infectious SIRS in critically ill children - a data-driven approach using machine-learning algorithms.
- Care and rehabilitation
- Infection and sepsis
- Line and device care
- Medical
- Nursing
- Sepsis
This paper is from 2018, more than five years ago. Newer evidence may have been published since.
Summary
Secondary analysis of a randomised trial at a German tertiary PICU developing a machine-learning (Random Forest) diagnostic model to distinguish sepsis from non-infectious SIRS at disease onset. The model combined four clinical variables (PICU length of stay to onset, central line, core temperature, prior episodes) and four laboratory markers (interleukin-6, platelets, procalcitonin, CRP), and was temporally validated. Discrimination exceeded that of individual biomarkers, with the authors suggesting potential to reduce unnecessary antimicrobial use, though external validation is needed. Relevant to PICU teams interested in early sepsis recognition and antimicrobial stewardship.
Summary written for PICUlearn. Always read the original.
- Licence:
- Open (Creative Commons)
The full text of this openly licensed paper is available to PICUlearn members, and Peeky can answer questions from it.
Guidelines are shown for education. Follow your unit's own guidance and senior advice.