Objective To evaluate the usefulness of a single C-Reactive Protein (CRP) measurement at 18 h of age to identify neonates where antibiotics started for possible early onset sepsis (EOS) could safely be discontinued.

Design/Methods In a prospective cohort of 647 preterm (<35 weeks) and 555 late preterm (35–36 weeks) or term newborns with maternal and/or neonatal risk factors for EOS, CRP levels were measured between 15 and 21 h of age.

Results There were 16, 107 and 1079 neonates with proven EOS, possible EOS and no EOS, respectively. Among the 645 neonates with a CRP<10 mg/L, 1 had proven EOS, 43 had possible EOS and 601 (93.2%) were not infected. All with possible or proven EOS were either less than 35 weeks’ gestation, symptomatic at the time of CRP assessment or remained on antibiotics because of maternal bacteraemia: they would therefore not be considered for discharge. There were 557 neonates with a 18-h CRP≥10 mg/L. Of these, 15 had proven EOS, 64 had possible EOS, and 478 (85.8%) were not infected. Sensitivity and specificity of 18-h CRP for proven or possible EOS were 64% (95% CI 56 to 73) and 56% (95% CI 53 to 59), respectively. The negative predictive value was 93% (95% CI 91 to 95), and the positive predictive value was 14% (95% CI 11 to 17).

Conclusions The duration of antibiotic treatment in neonates born beyond 34 weeks’ gestation and asymptomatic at the time of CRP assessment could be potentially reduced with a diagnostic algorithm that includes a point-of-care 18-h CRP measurement. An elevated 18-h CRP in isolation should not be used as a reason to prolong antibiotics.