Blood Culture-Based Classification of Early Bacterial Bloodstream Infection After Autologous Hematopoietic Stem Cell Transplantation: A Retrospective Cohort Study

Estimates of early bacterial bloodstream infection (BSI) after autologous hematopoietic stem cell transplantation (HSCT) depend on the classification of common commensals. We examined the sensitivity of BSI frequency and microbiology to three operational definitions. This retrospective single-center cohort comprised 161 autologous transplant episodes in 158 patients from September 2015 to February 2026. The endpoint covered day 0–30. A permissive rule counted any bacterial growth. The primary definition excluded CoNS-only episodes classified as contamination following retrospective review of blood culture results and clinical records. A restrictive definition additionally excluded episodes with sole viridans-group or unspeciated streptococci or Corynebacterium. Formal adjudication according to the full National Healthcare Safety Network surveillance criteria was not performed. Regression and survival analyses were exploratory. The proportions meeting the permissive, primary, and restrictive endpoints were 56.5% (91/161), 23.6% (38/161; 95% confidence interval 17.3–30.9), and 21.1% (34/161), respectively. The primary definition included 40 isolates, of which 27 (67.5%) were Gram-negative and 21 (52.5%) were Escherichia coli. Gram-positive isolates predominated under the permissive rule (71.0%). Exploratory regression estimates were imprecise. In the day-30 landmark survival analysis, the operational BSI classification was inversely associated with subsequent mortality; however, this finding was based on only four deaths in the BSI-classified group and remained vulnerable to survivor selection, residual confounding, and exposure misclassification. Reported BSI frequency and organism distribution were sensitive to commensal classification. Explicit operational definitions, retrospective clinical assessment of CoNS isolates, and sensitivity analyses aid interpretation; however, the absence of standardized surveillance adjudication limits the estimation of true BSI incidence. The survival findings require cautious, noncausal interpretation.