Aims To identify independent predictors of clinical and microbiological treatment failure and develop a predictive model in neonates with bloodstream infections (BSIs).

Methods A total of 1078 episodes of BSI occurred in 793 neonates in a tertiary-level neonatal intensive care unit (NICU) between 2004 and 2012 were enrolled. Patient demographics, underlying chronic comorbidities, clinical features, antimicrobial treatment and microbiological characteristics were evaluated.

Results Presences of underlying congenital anomalies (odds ration [OR] 2.22, 95% confidence interval [CI] 1.15–4.29) and pulmonary hypertension (OR 3.57, 95% CI 1.65–7.70), infections caused by multidrug-resistant gram-negative bacteremia (OR 2.84, 95% CI 1.21–6.66), Group B streptococcus (OR 3.08, 95% CI 1.31–7.26), and fungus (OR 4.06, 95% CI 1.97–8.38), a NTISS score of ≥ 23 (OR 6.61, 95% CI 2.40–26.47), inappropriate antibiotics (OR 2.01, 95% CI 1.31–3.08), and concomitant meningitis (OR 4.35, 95% CI 2.13–8.89) and ventilator-associated pneumonia (OR 2.82, 95% CI 1.26–6.32) were identified as independent risk factors for 28-day treatment failure in neonatal bacteremia. A risk-score model was created by adding points for each independent risk factor, and had a c-statistic of 0.83. Patients with risk scores of 0, 4, 8, 12 and 15 had estimated 28-day treatment failure rates of approximately 3%, 17%, 52%, 85% and 95%, respectively.

Conclusions This predictive model, calculated after documentation of a BSI, reflects spectrum of BSI severity and is associated with subsequent treatment failure through illness severity score and case-mix variables. This simple score could prove useful in clinical and research settings, and practical in estimating the prognosis.