Comparison of Early-Onset Sepsis Risk-Stratification Algorithms in Neonates in a Kenyan Nursery

Abstract

BACKGROUND AND OBJECTIVES Risk stratification algorithms (RSAs) can reduce antibiotic duration (AD) and length of stay (LOS) for early-onset sepsis (EOS). Because of higher EOS and antibiotic resistance rates and limited laboratory capacity, RSA implementation may benefit low- and middle-income countries (LMIC). Our objective was to compare the impact of 4 RSAs on AD and LOS in an LMIC nursery. METHODS Neonates <5 days of age admitted for presumed sepsis to a Kenyan referral hospital in 2019 (n = 262) were evaluated by using 4 RSAs, including the current local sepsis protocol (“local RSA”), a simplified local protocol (“simple RSA”), an existing categorical RSA that uses infant clinical examination and maternal risk factors (CE-M RSA) clinical assessment, and the World Health Organization’s Integrated Management of Childhood Illness guideline. For each RSA, a neonate was classified as at high, moderate, or low EOS risk. We used к coefficients to evaluate the agreement between RSAs and McNemar’s test for the direction of disagreement. We used the Wilcoxon rank test for differences in observed and predicted median AD and LOS. RESULTS Local and simple RSAs overestimated EOS risk compared with CE-M RSA and the Integrated Management of Childhood Illness guideline. Compared with the observed value, CE-M RSA shortened AD by 2 days and simple RSA lengthened AD by 2 days. LOS was shortened by 4 days by using CE-M RSA and by 2 days by using the local RSA. CONCLUSIONS The local RSA overestimated EOS risk compared with CE-M RSA. If implemented fully, the local RSA may reduce LOS. Future studies will evaluate the prospective use of RSAs in LMICs with other interventions such as observation off antibiotics, biomarkers, and bundled implementation.