A Third of Positive Blood Cultures in Preemies Are Not Real Infections, 11-Year Study Finds

A positive blood culture in a newborn intensive care unit sets off an immediate chain of decisions: start antibiotics, continue them, intensify them, or stop them. Yet a striking new analysis shows that in routine practice, roughly one in three of those positive cultures is ultimately judged by the treating clinicians to be contamination rather than true bloodstream infection. The finding, drawn from eleven years of data at a German university hospital, offers one of the most detailed real-world pictures yet of how doctors actually make that high-stakes call — and where the diagnostic ground remains shakiest.

The study, published in BMC Infectious Diseases by a team at the Medical Center, University of Freiburg, examined every blood culture-positive episode occurring in neonates hospitalized at the tertiary neonatal intensive care unit between 2014 and 2024. Of 3,246 blood cultures, 232 were positive — a positivity rate of just 7.1 percent. After accounting for follow-up and parallel cultures and other exclusions, the researchers distilled 189 unique culture-positive episodes. Treating teams, in their original documentation from the time of the episode, attributed 125 of these (66.1 percent) to infection and 64 (33.9 percent) to contamination.

The technical heart of the problem is a group of bacteria called coagulase-negative staphylococci, or CoNS. These organisms are common residents of human skin, which makes them frequent hitchhikers on sampling needles and catheter hubs. But CoNS can also cause genuine, sometimes dangerous infections, particularly in extremely premature infants with central venous catheters. Unlike unambiguous pathogens such as Staphylococcus aureus or Enterobacterales, CoNS carry no intrinsic signal of whether they represent real disease or a laboratory artifact. The clinician must infer intent from context.

The Freiburg data make that ambiguity vivid. Among the 189 episodes, CoNS appeared in 114 — 60.3 percent of the total — and their attribution varied more than any other organism. All 13 CoNS-positive episodes in the early-onset period, meaning within the first days of life, were classified as contamination. Yet among late-onset episodes, CoNS-positive cultures were attributed as infection in 64 of 101 cases, or 63.4 percent. The same microbe, in the same unit, flipped from near-certain artifact to majority true infection depending on the timing and clinical setting.

By contrast, other organisms behaved almost predictably. Enterobacterales, Enterococcus faecalis, Staphylococcus aureus, and Candida species were nearly always attributed as infection, reflecting their established pathogenic role. Bacillus species and other environmental organisms were predominantly dismissed as contamination, consistent with their origin in skin flora and hospital dust. In other words, microbiology laboratories do much of the interpretive work implicitly: the identity of the organism sets a prior probability, and the clinical team then adjusts it.

To quantify what drives that adjustment, the researchers ran descriptive analyses and exploratory logistic regression, focusing on late-onset episodes. The strongest signal by far was the maximum C-reactive protein value, an established blood marker of systemic inflammation. Higher peak CRP showed the strongest association with attribution as infection, both in univariable and multivariable analyses, with an adjusted odds ratio of 4.41 (95 percent confidence interval 2.59 to 7.51). In practical terms, a pronounced inflammatory response made clinicians far more likely to believe the culture.

Other candidate markers told a more complicated story. Time to positivity — how quickly the culture broth signals bacterial growth, with faster growth traditionally suggesting a heavier bacterial load and hence true infection — was associated with attribution, but the association weakened and became imprecisely estimated after adjustment for other factors. The same pattern held for CoNS detection itself and for the presence of a central vascular catheter, a known risk factor for catheter-related bloodstream infection. Once CRP and organism identity were accounted for, these features added comparatively little statistical certainty.

The authors are careful about what this study does and does not show. Because attribution was extracted from routine documentation made during the index episodes, the analysis describes real-world clinical attribution, not diagnostic accuracy. There is no gold standard against which the clinicians’ judgments can be graded, and the retrospective single-center design limits generalizability to other units with different case mixes and sampling practices. The logistic regression was explicitly exploratory. What the data capture instead is a decade of collective clinical reasoning under uncertainty — a portrait of practice, not a validated rule.

Even so, the implications for antimicrobial stewardship are substantial. Neonatal sepsis is among the leading causes of newborn death worldwide, and the standard response to a suspicious clinical picture is prompt empiric antibiotics, a strategy that saves lives when infection is real. But CoNS-driven uncertainty pushes in the opposite direction: overtreatment exposes fragile infants to unnecessary drug toxicities, disrupts the developing microbiome, and fuels resistance, while undertreatment of a genuine bloodstream infection can be catastrophic. The finding that a full third of culture-positive episodes were ultimately called contamination illustrates how much antibiotic therapy in neonatal care rides on judgments that current tools cannot reliably settle.

The study’s clearest conclusion is a call for more reproducible criteria. As the researchers note, CoNS-positive late-onset episodes represented the greatest diagnostic uncertainty, and clinical attribution depended on a combined consideration of organism plausibility, inflammatory response, time to positivity, and clinical context rather than on any single decisive test. Future work — whether refined biomarker panels, molecular diagnostics that quantify bacterial load, or standardized attribution algorithms — will need to target exactly this gray zone. Until then, the decision of whether a positive blood culture in a preemie means infection or contamination remains what it has always been: a judgment call, made under pressure, in the smallest and most vulnerable patients on the ward.

Subject of Research: Clinical attribution of blood culture-positive episodes as infection or contamination in neonatal intensive care

Article Title: Infection or contamination – Clinical attribution of blood culture-positive episodes in neonatal intensive care: an 11-year retrospective single-center study

Article References: Infection or contamination – Clinical attribution of blood culture-positive episodes in neonatal intensive care: an 11-year retrospective single-center study. (n.d.). https://doi.org/10.1186/s12879-026-14605-8

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14605-8

Keywords: neonatal sepsis, blood cultures, contamination, coagulase-negative staphylococci, neonatal intensive care, C-reactive protein, antimicrobial stewardship, bloodstream infection, time to positivity, clinical microbiology, epidemiology, Infection