**Background:** Neonatal sepsis is a leading cause of mortality and morbidity in premature infants. Antibiotics are essential for treatment, but inappropriate use drives antimicrobial resistance and may increase risks of late-onset sepsis (LOS), necrotizing enterocolitis (NEC), severe intraventricular hemorrhage (IVH), and long-term conditions like asthma and obesity. In China, antibiotic abuse in NICUs is particularly severe—a prior survey of 24 NICUs in Hunan Province found that VLBW infants received antibiotics for more than half of their hospital stay. This study aimed to evaluate the impact of an antibiotic stewardship program initiated in 2015 at the NICU of Xiangya Hospital of Central South University on antibiotic usage patterns and short-term clinical outcomes in VLBW infants.
**Methods:** This single-center retrospective cohort study enrolled all eligible VLBW infants (birth weight <1500 g) admitted from January 1, 2014 to December 31, 2016. Infants were classified into three groups: pre-stewardship (2014, n=96), during stewardship (2015, n=77), and post-stewardship (2016, n=76). Exclusion criteria included admission after 24 hours of birth, culture-proven early-onset sepsis, death within the first week, major congenital anomalies, suspected genetic metabolic disorders, and incomplete records. Data from 2021 (n=90 after exclusions) were also analyzed to assess sustainability. The stewardship program included specific diagnostic and treatment guidelines for early-onset and late-onset sepsis, consistent with Chinese national recommendations. Primary outcomes were duration of initial antibiotic course, total antibiotic days, and antibiotic use rate (AUR). Secondary outcomes included a composite adverse outcome (death, NEC ≥ stage 2, LOS, severe IVH ≥ grade III, severe BPD, and ROP ≥ stage 3). Multivariate logistic regression adjusted for confounders including gestational age, birth weight, Apgar scores, and mechanical ventilation.
**Key Results:** Empirical antibiotics were used in over 90% of VLBW infants across all groups. Over the three-year period, the median duration of the initial antibiotic course decreased significantly from 25.0 days in 2014 to 13.5 days in 2015 and 5.0 days in 2016 (p<0.001). The proportion of infants receiving an initial antibiotic course ≤3 days increased from 2.1% to 9.1% to 38.2% (p<0.001), while the proportion receiving >7 days decreased from 95.8% to 79.2% to 39.5% (p<0.001). Total antibiotic usage days during the entire NICU stay fell from a median of 27.0 days in 2014 to 21.0 days in 2015 and 10.0 days in 2016 (p<0.001). The AUR declined from 64% to 39% to 12% (p<0.001). No statistically significant differences were found in mortality, NEC, LOS, severe ROP, or severe BPD across groups. However, severe IVH incidence was lower in 2016 (7.9%) than in 2014 (25.0%, p=0.013), and median days to full enteral feeding decreased from 26.0 days in 2014 to 21.5 days in 2015 and 17.0 days in 2016 (p<0.001). Multivariate logistic regression showed that after adjusting for confounders, the risk of composite adverse outcome increased by approximately 17% for every 10% increase in AUR (aOR=1.175, p<0.01). Infants receiving >7 days of initial antibiotics had a 5.1-fold higher odds of composite adverse outcome compared to those receiving 0–3 days (aOR=5.148, 95% CI: 1.598–16.583, p=0.006). In 2021, the median initial antibiotic course further decreased to 4.0 days (p<0.001 vs 2016), the proportion ≤3 days increased to 56.7% (p=0.022), and total antibiotic days decreased to 7.0 days (p=0.010). No significant differences in adverse outcomes were observed between 2016 and 2021.
**Clinical Implications:** This study provides strong evidence that a structured antibiotic stewardship program can safely and substantially reduce antibiotic exposure in VLBW infants in a Chinese NICU setting. The reduction was associated with improved short-term outcomes, including lower severe IVH rates and faster achievement of full enteral feeds, without increasing mortality or other major complications. The findings challenge the common practice of prolonged empirical antibiotic courses in culture-negative infants and support early discontinuation (≤3 days) when infection is not confirmed. The sustainability of the program through 2021 further demonstrates its feasibility. Limitations include the retrospective design, single-center setting, potential unmeasured confounders, and lack of data on antimicrobial resistance patterns. A multicenter clinical trial is warranted to confirm these findings and support broader implementation across China and other developing countries.