P2.180: Early Recognition of Sepsis in the Pediatric Population

Background: Sepsis is defined as life-threatening organ dysfunction due to a dysregulated response to infection and is a leading cause of death for children worldwide. Early recognition of sepsis in pediatric patients can be improved with a screening tool focused on patient-centered outcomes. Systematic screening improves outcomes in pediatric sepsis by decreasing time to treatment, length of hospital stay, and number of missed diagnoses. Multiple screening algorithms are utilized nationwide with varying sensitivities and positive predictive values with recent shift towards the more recently introduced Phoenix Sepsis Score. The existing system used by our hospital system was reliant on action by nursing staff and alerted with low sensitivity. There was also lack of clarity with high-risk populations such as febrile neutropenia, sickle cell disease, and active cancer. OLOL Children's Hospital developed a Best Practice Advisory (BPA) within the EHR to improve the detection of sepsis in its pediatric emergency department. Alerting re-design was targeted through the implementation of a “Ghost BPA” that is an unseen analytic tool which uses implemented criteria to then actively alert on the EHR of a patient’s sepsis status. The efficacy of the EHR-based screening tool was optimized with data analytics aimed towards retroactive interrogation of novel criteria and utilization metrics.

Methods: Retrospective analysis of the existing sepsis screener showed poor sensitivity and specificity in children and high override rate. The screener was modified to specify patient categorization, improve intra and inter-rater reliability, and incorporate criteria from the Phoenix Sepsis Score that was modified to suit the needs of the emergency department. The new BPA included criteria such as hypotension, existing risk factors, GCS Score, oxygen saturation, temperature, and heart rate. The “Ghost BPA” was then applied to 17,991 patients over a 3-month period in the pediatric emergency department.

Results: Of the 17,991 patients, 88.9% were alerted correctly with the new BPA compared to 54.4% sensitivity with the existing screener. Compared to the standard BPA, the new BPA increased identification of non-septic patients from 50% to 79%. Specificity and accuracy remained equivalent, 91.8% to 91.1% and 91.0% to 91.8% respectively. Positive predictive value was 1.5% with the new BPA compared to 0.98% with the standard BPA.

Conclusion: Early identification of pediatric sepsis remains a challenge considering the low incidence and variable presentation. Our study refined an existing alert by specifying age-specific criteria, using a Ghost alerting system, and extracting from EMR data captured during patient triage. Ghost BPA methodology is a promising tool for not only improving screening performance but reducing false positives while maintaining specificity.