Study design
We performed a retrospective study of patients who underwent blood culture testing at the National Center for Global Health and Medicine between April 2012 and March 2021. The need for informed consent was waived due to the retrospective nature of the study design. The study information was presented on the Web for the possibility of opting out of consent. This was substituted for the participants’ consent. The protocol of this study including the opt-out consent method was approved by the Certificate Review Board of National Center for Global Health and Medicine (NCGM-G-004168-00) and conformed to the amended Declaration of Helsinki. The data were compiled from the registry of blood culture surveillance, including data on contamination, and the microbiology laboratory.
The registry data of blood culture surveillance
For every case, two or more infectious disease physicians of the National Center for Global Health and Medicine determined whether the case was contaminated from a clinical point of view by reviewing clinical records and laboratory data in accordance with the CUMITECH criteria [3]. Undetermined cases and those with pending determination were excluded from the study.
Identification of bacterial species
All blood culture samples were collected into standard aerobic and anaerobic culture bottles (92F or 94F and 93F, 23F or 20F and 24F Becton Dickinson Microbiology Systems, Sparks, MD, USA) and processed using the BACTEC 9240, 9120, and FX systems (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). These samples were routinely monitored for at least 144 h. The bottles that tested positive were removed and subjected to Gram staining. The specimens were then inoculated into 5% sheep blood agar and BTB agar media (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan), and incubated at 35°C (Depending on the situation, other media may be added or the environment may be changed, such as anaerobic incubation). Conventional bacterial identification and susceptibilities to the predefined antimicrobials were determined in accordance with the Clinical and Laboratory Standard Institutions criteria (M100) [4, 5] by matrix assisted laser desorption/ionization-time of flight mass spectrometry system (MALDI Biotyper system; Bruker, Billerica, MA, USA) and automated broth micro dilution system (MicroScan WalkAway 96 SI system; Beckman Coulter, Brea, CA, USA). All Staphylococci species, except Staphylococcus aureus and S. lugdunensis, were treated as CoNS.
Indicators
We calculated the monthly ConR [(total number of contaminated cases per month) / (total number of blood culture sets collected per month) × 100] [3]. The values of the four indicators were aggregated as follows: the number of CoNS-positive sets (Indicator A), CoNS-positive cases (Indicator B), cases with only one CoNS-positive blood culture set (Indicator C), and cases of CoNS contamination (Indicator D).
Statistical analysis
Correlation coefficients were calculated using Pearson’s correlation test. Receiver operating characteristic (ROC) curves were prepared for all indicators with a ConR of ≥ 2.5, as the objective variable, and cut-off values were calculated using Youden's index. The area under the curves (AUCs) were compared using the Delong method with the Holm correlation. Statistical analysis was performed using EZR for Windows version 1.54 [6]. Figures were created using IBM SPSS Statistics software for Windows (version 26.0; IBM Corp., Armonk, NY, USA). The probability of significance was calculated to be 5%.