Perinatal Antibiotic Prophylaxis Exposure and Neonatal Sepsis in Indonesia: A Prospective Cohort Study

Background Perinatal antibiotic prophylaxis is widely used to prevent early neonatal infection, but extensive exposure may alter neonatal-sepsis epidemiology. We assessed the association between perinatal antibiotic-prophylaxis exposure and neonatal sepsis in two Indonesian referral hospitals. Methods We conducted a prospective cohort study of mother-newborn dyads enrolled at delivery and followed through 28 days of life. Maternal and neonatal prophylactic antibiotic exposure was ascertained from medical records. Early-onset sepsis (EOS) occurred within 72 hours and late-onset sepsis (LOS) thereafter. Because conventional inverse probability weighting showed poor covariate overlap and extreme weights, propensity-score overlap weights estimated separately within each hospital were used to estimate adjusted risk differences in the overlap population. Results Among 904 analyzed dyads, 651 newborns (72.0%) were exposed to maternal and/or neonatal prophylactic antibiotics. Ninety-four neonates developed sepsis (46 EOS and 48 LOS); 25 cases were culture proven. In the primary overlap-weighted analysis, the estimated risks of overall sepsis were 3.4% without prophylaxis and 9.4% with any prophylaxis (adjusted risk difference [RD], 0.060; 95% confidence interval [CI], 0.012 to 0.109). The adjusted RD was 0.015 (95% CI, -0.022 to 0.052) for EOS and 0.045 (95% CI, 0.013 to 0.078) for LOS. In exploratory pairwise analyses, neonatal prophylaxis alone was associated with higher risks of overall sepsis (RD, 0.149; 95% CI, 0.044 to 0.255) and LOS (RD, 0.097; 95% CI, 0.023 to 0.171). Conclusions Among mother-newborn dyads with overlapping measured probabilities of receiving prophylaxis, perinatal antibiotic-prophylaxis exposure was associated with a higher risk of neonatal sepsis, particularly LOS. Limited positivity, confounding by indication, sparse events among unexposed neonates, and limited culture confirmation preclude causal interpretation and limit generalizability to the full cohort.