Developing innovative methods for measuring body temperature in preterm infants to enhance prediction of late-onset sepsis

Abstract

Background and aim: Significant progress has been made in employing machine learning algorithms to predict late-onset sepsis (LOS). Despite the availability of body temperature measurements, it is underutilized due to external influences like incubator temperature. This study aimed to develop new methods to measure body temperature. Methods: In this retrospective cohort study, preterm infants (GA < 32 weeks) from the Wilhelmina Children’s Hospital (WKZ) were included. Patients were divided into LOS or control groups based on blood culture results. Body and incubator temperatures were extracted around the time a positive blood culture collection, and equivalent timestamps for controls. Two methods were evaluated at five time points: at blood culture collection (t=0), two hours after (t=2), and twenty-four (t=-24), four (t=-4), and two (t=-2) hours before. Firstly, the absolute median difference over the past 30 minutes was assessed for each time point. The second method focused on the disparity between body and incubator temperatures. Differences between group were tested using the Wilcoxon singed-rank test. Results: After matching, two groups of 362 patients were analysed. The MAD showed significant differences at t=0 and t=2. The body-incubator temperature difference showed significant results at t=-2, t=0 and t=2. Conclusion: Both methods demonstrated differences in body temperature measures between LOS and control groups at various time points, indicating their potential for integrating body temperature into machine learning algorithms.