MONDAY, March 18, 2019 (HealthDay News) -- Machine-learning models that use electronic health record (EHR) data can identify infants with sepsis in the neonatal intensive care unit hours before clinical recognition, according to a study published online Feb. 22 in PLOS ONE.
In an effort to develop a model capable of recognizing infant sepsis at least four hours before clinical recognition, Aaron J. Masino, Ph.D., from the University of Pennsylvania in Philadelphia, and colleagues used EHR data from infants hospitalized ≥48 hours in the neonatal intensive care unit at the Children's Hospital of Philadelphia (between September 2014 and November 2017) who received at least one sepsis evaluation before 12 months of age.
The researchers considered both culture-positive outcomes (positive blood culture for a known pathogen; 110 evaluations) and clinically positive outcomes (negative cultures but antibiotics administered for ≥120 hours; 265 evaluations) as cases. Data were extracted from the 44-hour window ending four hours before evaluation. Control data included a random sample of 1,100 44-hour windows from all times ≥10 days removed from any evaluation. Thirty-six EHR data features were incorporated into eight machine-learning models. Six models achieved a mean area under the receiver operating characteristic curve (AUC) between 0.8 and 0.82 for discriminating culture-positive cases from controls, with no significant differences between them. When both culture-positive and clinically positive cases were included, the six models achieved an AUC of 0.85 to 0.87, again with no significant differences between them.
"Because early detection and rapid intervention is essential in cases of sepsis, machine learning tools like this offer the potential to improve clinical outcomes in these infants," Masino said in a statement.