Stratification of Culture-Proven Early-Onset Sepsis Cases by the Neonatal Early-Onset Sepsis Calculator: An Individual Patient Data Meta-Analysis

Abstract

Objectives To provide a comprehensive assessment of case stratification by the Neonatal Early-Onset Sepsis (EOS) Calculator, a novel tool for reducing unnecessary antibiotic treatment.Study design A systematic review with individual patient data meta-analysis was conducted, extending PROS-PERO record CRD42018116188. Cochrane, PubMed/MEDLINE, EMBASE, Web of Science, Google Scholar, and major conference proceedings were searched from 2011 through May 1, 2020. Original data studies including culture-proven EOS case(s) with EOS Calculator application, independent from EOS Calculator development, and including representative birth cohorts were included. Relevant (individual patient) data were extracted from full-text and data queries. The main outcomes were the proportions of EOS cases assigned to risk categories by the EOS Calculator at initial assessment and within 12 hours. Evidence quality was assessed using Newcastle-Ottawa scale, Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies, and GRADE tools.Results Among 543 unique search results, 18 were included, totaling more than 459 000 newborns. Among 234 EOS cases, EOS Calculator application resulted in initial assignments to (strong consideration of) empiric antibiotic administration for 95 (40.6%; 95% CI, 34.2%-47.2%), more frequent vital signs for 36 (15.4%; 95% CI, 11.0%-20.7%), and routine care for 103 (44.0%; 95% CI, 37.6%-50.6%). By 12 hours of age, these proportions changed to 143 (61.1%; 95% CI, 54.5%-67.4%), 26 (11.1%; 95% CI, 7.4%-15.9%), and 65 (27.8%; 95% CI, 22.1%-34.0%) of 234 EOS cases, respectively.Conclusions EOS Calculator application assigns frequent vital signs or routine care to a substantial proportion of EOS cases. Clinical vigilance remains essential for all newborns.

Neonatal sepsis prediction through clinical decision support algorithms: A systematic review

“…NICU stay. Additionally, in contrast to the paper by Escobar et al 52 reporting on the benefits of using an EOS calculator, a recent systematic review including over 459,000 infants found limited evidence in this regard. Although this may highlight the importance of continuous CDSA development, it also raises concerns about overfitting and the generalisability of CDSA findings to external settings.…”

Neonatal early-onset infections: Comparing the sensitivity of the neonatal early-onset sepsis calculator to the Dutch and the updated NICE guidelines in an observational cohort of culture-positive cases

“…25 Our results also correspond to those of a recent meta-analysis on 234 EOD patients, which reported that the EOSC recommends antibiotics in 41% of EOD patients shortly after birth. 15 The EOSC relies more heavily on signs of clinical illness, resulting in a more specific approach and a substantial reduction in unnecessary EOD treatments. This comes at the expense of sensitivity, as can be expected.…”

“… 25 Our results also correspond to those of a recent meta-analysis on 234 EOD patients, which reported that the EOSC recommends antibiotics in 41% of EOD patients shortly after birth. 15 …”

“…14 A recent meta-analysis on 234 EOD cases showed that the EOSC did not recommend immediate initiation of antibiotic treatment, but close clinical observation or routine care, in 59% of newborns with culture-confirmed EOD. 15 Although it was not a primary outcome for these metaanalyses, there were no short-term adverse consequences when using the EOSC. However, little is known about the sensitivity of the EOSC to detect early-onset sepsis.…”

Machine Learning Used to Compare the Diagnostic Accuracy of Risk Factors, Clinical Signs and Biomarkers and to Develop a New Prediction Model for Neonatal Early-onset Sepsis

“…As the sepsis calculator is highly dependent on clinical observation, it is similar to a population preselected by healthcare workers using serial examinations. 33 Studies of approaches using serial examinations are increasingly conducted, with promising results. 24,25,27 This approach may be combined with our new model as well.…”