Study design and settings

We conducted a single-center retrospective observational study to examine the weekend effect on BC contamination at Jikei University Hospital, a 1075-bed tertiary medical center in Tokyo, Japan. The hospital operates from Monday to Saturday to provide regular medical services. Off-hours are from 5:00 p.m. to 9:00 a.m. on these days, as well as the entire day on Sundays, public holidays, and university anniversaries. During off-hours, at least one designated physician is assigned to each department to address patient needs as required. The emergency department is open 24 hours a day and 365 days a year. Patients admitted to the emergency department during the off-hours are attended by the emergency department staff or on-duty physicians of the respective department.

This study was approved by the Institutional Review Board of Jikei University School of Medicine (Approval number: 35-177 [11806]). The need for informed consent was waived because the study design was retrospective, and clinical information about the patients was not used.

Blood sampling and BC procedure

Aligning with our hospital’s usual practice, all blood samples for BC were collected by physicians, not phlebotomists or nurses. In principle, two or more sets of BCs should be performed either simultaneously or within a 24-hour period.^{Reference Weinstein15,Reference Lee, Mirrett, Reller and Weinstein16} A set of BCs consists of two bottles, one for aerobic bacteria and one for anaerobic bacteria. Samples for BC are collected primarily by peripheral venipuncture; however, in some cases, they may be collected via a hemodialysis (HD) circuit or long-term indwelling catheter, such as a central venous catheter. While our facility does not have a licensing system for blood sampling for BC, newly employed physicians must take video or on-site lectures on the proper blood sampling technique shortly after they start work. Physicians collecting blood samples are required to implement appropriate procedures to reduce BC contamination, which include hand washing before blood sampling, thorough skin asepsis, and disinfection of culture bottles before use.

The BCs were processed in the hospital bacteriology laboratory. Since April 2017, blood samples collected during the day shift on holidays or weekdays were processed in the laboratory on the same day, while samples collected during the night shift were processed the following day. Before March 2017, an on-duty bacteriologist was not for all holidays. They were assigned to holiday shifts every other day only when there were more than two consecutive holidays. Therefore, until March 2017, the processing of blood samples collected on holidays could have been delayed by up to 24 hours compared with samples collected on weekdays. BC tests were performed using a BD BACTEC FX system with BD BACTEC Plus Aerobe/Anaerobe bottles (Becton, Dickinson and Company, USA). Bacterial identification and susceptibility testing were performed using the MicroScan WalkAway system (Beckman Coulter, USA).

Study subjects and data collection

The BC results, until 31 December 2019, were screened. The coronavirus disease 2019 (COVID-19) pandemic in 2020 led to changes in our practice structure, including restrictions on regular medical services, priority acceptance of patients with COVID-19, and the associated reorganization of staff and beds. Several studies have reported an increase in the frequency of BC contamination during the COVID-19 pandemic.^{Reference Esquer Garrigos, Wingler, Svoronos, Vijayvargiya, Goodman-Meza and O’Horo17,Reference Ohki, Fukui, Morishita and Iwata18} Given this background, this study did not include BC results obtained after 1 January 2020.

The BC data consisted of information on the date of blood sampling, specimen identification number, department ordering of the test, test site, site of phlebotomy, route of blood sampling, and organism species isolated. BCs performed on pediatric patients were excluded from the study, considering the volume of BC samples and the number of bottles collected may differ between adult and pediatric patients.^{Reference Doern, Carroll, Diekema, Garey, Rupp and Weinstein19} Because the BC results did not include patient-level information such as age, the ordering department, the location of the test (eg, pediatric ward), and other comments indicating that the test was performed on a pediatric patient were used to exclude pediatric BC results. In some cases, pleural fluid and ascites were inoculated into BC bottles and cultured; these were also excluded.

Definitions of weekend and contamination

Based on our hospital settings, BC using blood collected on closed days (Sundays, public holidays, or university anniversaries) were considered as tests performed on weekends (weekend group); all other BCs were considered as tests performed on weekdays (weekday group).

Two or more sets of BCs should be collected within 24 hours to accurately detect the causative organisms of bloodstream infections.^{Reference Lee, Mirrett, Reller and Weinstein16} In this study, the two-set sampling proportion was calculated by dividing the number of BCs obtained with at least two sets within 24 hours by the total number of BCs during the same period. BC contamination was defined as the identification of the following organisms in only one set from BC obtained with a two-set sampling within seven consecutive days: coagulase-negative staphylococci (CoNS) other than Staphylococcus lugdunensis, Propionibacterium spp. including Cutibacterium (Propionibacterium) acnes, Micrococcus spp., viridans group streptococci (VGS), Corynebacterium spp., Clostridium perfringens, or Bacillus spp. other than Bacillus anthracis.^{Reference Doern, Carroll, Diekema, Garey, Rupp and Weinstein19–Reference Dargère, Cormier and Verdon21} The proportion of BC contamination was calculated by dividing the number of contaminated BC sets by the number of BC sets obtained with a two-set sampling during the same period.^{Reference Palavecino, Campodónico and She22}

Study period and outcome

This study defined the proportion of BC contamination according to the criteria, which were not based on clinical judgment. If the two-set sampling proportion was low, the calculated BC contamination proportion in this study could potentially deviate from the actual contamination proportion. Therefore, we calculated the two-set sampling proportion for each year and used the data of the years in which the proportion exceeded 90% for the data analysis.

The primary outcome of the present study was the incidence of BC contamination, which was compared between the weekend and weekday groups.

Statistical analyses

Descriptive statistics are expressed as frequencies and percentages for categorical variables. We evaluated the association between weekend sampling and BC contamination using multivariable logistic regression analysis. Our analysis was based on several assumptions. BCs should be performed immediately after a patient presents with symptoms suggestive of bacteremia, such as fever, hypotension, or loss of consciousness. Considering that these events would occur randomly on weekends and weekdays, the decision to perform BCs on weekends or weekdays is not influenced by patient or physician discretion. Therefore, it is reasonable to assume that there was no bias in the background information of the patients whose BCs were performed on weekends or weekdays. However, considering most patients visiting hospitals on weekends are treated in the emergency department, more BCs may be ordered in the emergency department on weekends, and blood sampling settings are expected to differ between weekends and weekdays. Furthermore, differences may exist depending on how samples were collected for BCs on weekends and weekdays. For example, the site and route of blood sample collection may differ between weekends and weekdays because the choice of sampling method is at the discretion of the physician collecting the blood samples. Based on these assumptions, we analyzed the association between weekend blood sampling and BC contamination using a multivariable model adjusted with variables related to the settings in which blood sampling was conducted (general ward, intensive care unit, general outpatient setting, emergency department),^{Reference Liaquat, Baccaglini, Haynatzki, Medcalf and Rupp23–Reference Aiesh, Daraghmeh, Abu-Shamleh, Joudallah, Sabateen and Al Ramahi25} route of blood sampling (via intravascular catheter, HD circuit, or others),^{Reference Boyce, Nadeau, Dumigan, Miller, Dubowsky and Reilly9,Reference Norberg, Christopher, Ramundo, Bower and Berman26} and site of phlebotomy (femoral vessel or other),^{Reference Ota, Oba, Fukui, Ito, Hamada and Mori27} rather than patient-level background information. Multicollinearity was assessed using the variance inflation factor (VIF). A VIF > 5–10 was considered indicative of multicollinearity.^{Reference Kim28} Statistical analyses were performed using R (version 4.1.3) and R Studio (version 2022.02.1). Statistical significance was defined as a two-sided P-value <0.05.