Abstract Background Bloodstream infections (BSI) remain a major cause of morbidity and mortality in preterm infants. Although BSI pathogens vary geographically, the role of local microbial colonization patterns and clinical practices in driving these differences is not well understood. Methods We conducted a prospective cohort study on 127 preterm infants from two geographically distinct NICUs: University of Cincinnati Medical Center (UCMC, USA) and Children's Hospital, Zhejiang University School of Medicine (ZCH, China). Six hundred and sixty-nine longitudinal stool and skin samples collected during the first three weeks of life underwent metagenomic sequencing. Associations between microbiome composition, clinical factors, and BSI epidemiology were evaluated using Generalized Linear Mixed Models and Random Forest. Results Distinct gut and skin microbiome profiles were observed between NICUs and corresponded closely with local BSI patterns. Staphylococcus aureus predominated at UCMC, while Klebsiella pneumoniae and Enterococcus species were more common at ZCH. Skin microbiota showed strong association with BSI isolates, implicating the skin as an underrecognized potential reservoir for pathogen translocation. Linear mixed models and Random Forest machine learning approaches revealed that clinical practices, including intravenous catheter placement and antibiotic exposure had greater influence on microbiome composition than geographic location alone. Conclusions Our findings demonstrate that modifiable clinical care practices shape the developing microbiome of preterm infants and contribute to geographic differences in BSI epidemiology. The skin microbiome represents a potentially significant risk factor for invasive infection. Further work to clarify how specific clinical practices influence pathogen colonization may inform strategies to reduce BSI incidence in preterm infants.
Xu et al. (Thu,) studied this question.