Discussion

We found that the implementation of a BCx algorithm in the NICU significantly decreased the BCE rate without corresponding increases in adverse effects. Importantly, we observed no rise in antibiotic days of therapy (DOT), central line-associated bloodstream infections (CLABSIs), 30-day readmission rates, or in-hospital mortality following the algorithm’s introduction. To our knowledge, this is the first study to report the use of a BCx algorithm specifically in an NICU setting.

Unnecessary BCxs lead to a cascade of negative outcomes. These include prolonged hospital stays, increased healthcare costs, a heightened risk of false positives, overuse of antibiotics, and a higher likelihood of antibiotic-related side effects and interactions.^{Reference Goyal, Garg and Bithal3,Reference Hocker, Tian, Li, Steckelberg, Mandrekar and Rabinstein4} Furthermore, these unnecessary BCxs cause patient discomfort due to repeated blood draws, which can lead to anemia and subsequent transfusions. The burden extends to nursing staff and laboratory personnel, who spend significant time collecting and processing these specimens.

The NICU patient population presents a distinct challenge when it comes to BCxs, largely due to the prevalence of central fever and or noninfectious (reactive) leukocytosis. Central fever most commonly arises in patients suffering from brain tumors, intraventricular hemorrhage, subarachnoid hemorrhage, cerebral vasospasm, and traumatic brain injury and is thought to result from hypothalamic dysfunction and damage to brain thermoregulatory centers.^{Reference Goyal, Garg and Bithal3,Reference Badjatia5} Reactive leukocytosis is also a common occurrence post-seizure, which can be a complication of a subarachnoid hemorrhage or traumatic brain injury. These two groups are also prone to central fever.^{Reference Hocker, Tian, Li, Steckelberg, Mandrekar and Rabinstein4,Reference Gusdon, Savarraj and Shihabeddin6}

In addition, the NICU population also includes patients who present with altered mental status, stroke, coma, or other neurologic impairments that prevent effective communication. These patients are therefore unable to communicate whether they are having symptoms that are typical indicators of bacteremia. As a result, there is an increased likelihood of ordering BCxs to rule out bacteremia, even in the absence of clear clinical signs.

Our findings align with other studies that have implemented BCx algorithms in various hospital settings, including emergency departments and ICUs. For example, Theophanous et al. demonstrated that the implementation of a BCx algorithm in emergency department patients as a diagnostic stewardship intervention significantly reduced the number of unnecessary BCxs, leading to more targeted use of resources without compromising patient outcomes.^{Reference Theophanous, Ramos and Calland7} Seidelman et al. also implemented a diagnostic stewardship intervention aimed at optimizing BCx utilization in two surgical ICUs, highlighting the effectiveness of algorithm-driven approaches in reducing BCx events and promoting more efficient patient care.^{Reference Seidelman, Moehring and Gettler2} Finally, Fabre et al. reported on the DISTRIBUTE study, which applied a diagnostic stewardship intervention among adult nonneutropenic inpatients, showing significant improvements in BCx use and reductions in false positives.^{Reference Fabre, Klein and Salinas1} These studies collectively support our findings, reinforcing the utility of BCx algorithms in reducing unnecessary tests while maintaining patient safety and enhancing antimicrobial stewardship.

Our study does have a few notable limitations. First, it is a single-unit, interrupted, unblinded time-series analysis and is accompanied by the limitations thereof. Second, we used historical NICU data as a control instead of comparing the results to another hospital unit during the same period. However, the NICU’s unique characteristics make direct comparisons with other units difficult. Additionally, we used a nonstandard definition of “contaminants” to increase the sensitivity of our algorithm to potential cases of bacteremia, acknowledging the complexity of our patient population.

In summary, our study demonstrates that the implementation of a BCx algorithm in the NICU can reduce unnecessary BCE while maintaining patient safety. Future studies should seek to replicate these findings in multicenter designs to improve generalizability. Our study highlights the importance of utilizing targeted interventions like BCx algorithms in complex, high-risk patient populations, offering a pathway toward more efficient, evidence-based care.