We investigated the relationship between illness severity and accuracy of neonatal sepsis screen. Consecutive neonates with clinically suspected early onset sepsis (EOS) were enrolled and blood culture and sepsis screen [C-reactive protein, absolute neutrophil count, immature to total ratio (ITR) and microerythrocyte sedimentation rate] were performed. Exclusion criteria were prior antibiotic exposure, nonavailable reports, and contaminated cultures. Score for Neonatal Acute Physiology Perinatal Extension (SNAPPE-II) was used to categorize neonates into "mild to moderate" (score <or=40) and "severe" (score >40) illness. Sepsis was defined as positive blood culture, and positive screen as >or=2 parameters positive. Of 125 subjects, 86 had mild to moderate and 39, severe illness. Twenty-eight (22%) subjects had sepsis. Sensitivity, specificity, negative predictive value, positive predictive value, likelihood ratio of positive test, and likelihood ratio of negative test of the sepsis screen and screen parameters were similar between the two groups. The sensitivity of the screen was 37.5% and 25% for mild to moderate illness and severe illness, respectively. Only ITR values correlated with SNAPPE-II scores in patients with "sepsis" (rho 0.4; P = 0.036). The severity of underlying illness does not alter the performance of the sepsis screen in diagnosing culture-positive EOS.