Effect of delayed entry of blood culture bottles in BACTEC automated blood culture system in the context of laboratory consolidation

Delayed entry of blood culture bottles is frequent in consolidated laboratories. A retrospective study evaluated time from insertion to detection and total detection time as a function of preincubation time, and we prospectively looked for false negative results. 69,604 blood culture bottles were reviewed for preincubation time, incubation time and total detection time. Positive cultures for specific bacterial subtypes were reviewed to assess the effect of preincubation time on likelihood of detection. 492 negative blood cultures were prospectively tested by 16S RNA PCR and Staphylococcus-specific PCR for the presence of bacterial DNA. Mean preincubation time for samples collected within the city-limits was 3.94 h versus 9.49–18.89 h for other client sites. Higher preincubation times were partially mitigated by a lower incubation time, with an overall increase in total detection time. A lower odds ratio of recovery of Staphylococcus spp was identified, but not confirmed by terminal subcultures and molecular assays. Prolonged preincubation of blood cultures affects total detection time despite a reduction in incubation time. Successful centralization of microbiological services may depend upon optimization of courier routes for inoculated blood culture bottles. Our data supports consideration for an increase in suggested maximum preincubation times.

. Preincubation times for negative blood culture bottles for all study sites and incubation (INC) and total detection times (TDT) for positive blood culture bottles for all study sites. Time values are listed in hours and sites are presented in reverse order of testing volume. INC is defined as time from insertion into the automated blood system to growth detection and TDT is defined as time from collection to growth detection. The lowest and highest values for POS %, Mn INC and Mn TDT are presented in bold for emphasis. Dist: distance, N: number, Hours: hours of operation, Runs: daily number of couriers, Mn: mean, Md: median, TT90%: time to reception of 90% of samples, > 24 h: percentage of samples received after 24 h for each sites, POS %: monomicrobial blood culture positivity rate, CONT %: contamination rate. www.nature.com/scientificreports/ to 18.7 h for cultures collected at one of the most distant sites. For this site, which has one daily courier run, nearly 40% of blood culture bottles had a preincubation time of more than 24 h. All sites within the Ottawa city limits had 8 or more daily courier pick up time, whereas community sites had between 1 and 4 daily couriers (Table 1). Of the culture bottles referred in, 0.19% were delayed > 24 h. Sites in cluster 2 ( Fig. 2), located approximately 100 km away from the testing site, had 3 and 4 daily pick ups with mean preincubation times of 9.2 and 9.6 h, whereas some sites from cluster 4 which were either equidistant or closer to the testing site but had fewer daily pick ups recorded mean preincubation times between 13.8 and 14.3 h. Median preincubation time increased as a function of the distance traveled, which explained approximately 60% of the variation noted between sites (R 2 = 0.593) (Fig. 1). Distribution of preincubation times for each site showed greater variability for smaller community sites (clusters 3 to 5) when compared to inner city sites and larger community centers (clusters 1 and 2) (Fig. 2). The mean preincubation time was statistically significantly different across all cluster sites (P < 0.0001). Distribution of TDT for all sites (Fig. 3) showed less spread than the distribution of preincubation times, and sites from clusters 3 to 5 tended to have lower INC (Table 1) than sites with shorter preincubation times. The shortest TDT was seen for one of the smaller inner-city sites with 20.7 h. TDT were statistically significantly different across all clusters (P < 0.0001), with a difference of 11.9 h between the site with the shortest TDT (20.7 h) and the site with the longest TDT at 32.67 h. Inner-city sites had on average an increase of 1.14 h in TDT as compared to the testing site (24.13 h Positivity rates as a function of preincubation time, and per bacterial species or groups. There was no statistically significant difference in positivity rates across all time clusters, and no clear trend in positivity rates was identified between time clusters (Table S2). Logistical regression analysis using the < 5 h time clusters as a reference showed no statistically significant decrease in the odds ratio of recovering growth in time clusters containing samples with preincubation times > 5 h. We also analyzed the full data set using logical regression and time as a continuous variable, yielding an OR of 0.998 (95% CI 0.994-1.003) for every additional hour of preincubation, not reaching statistical significance (P = 0.4456). The same analysis was repeated after clustering positive cultures by bacterial species or groups, while keeping the same site clusters and preincubation time stratifications (Tables 2, 3). There were no beta-hemolytic Streptococcus, S. aureus, viridans group Streptococcus, S. anginosus group and K. pneumoniae recovered from the site with the highest mean preincubation times. There were significant differences noted in positivity rates between time clusters for S. aureus, coagulase-negative Staphylococcus (CoNS) and E. coli: trends towards lower positivity rates with increasing preincubation times for S. aureus and CoNS, and towards higher positivity rates with increasing preincubation times for E. coli. For S. aureus, logistic regression analysis showed significantly lower odds ratio for the 5-10 h (OR = 0.82) and > 20 h (OR = 0.50) time clusters when compared to the reference < 5 h time cluster, with an overall OR/h of 0.99 when using time as a continuous variable, which also reached statistical significance. For CoNS, logistic regression using time clusters yielded a significantly lower OR for the 10-20 h time cluster only (OR = 0.61), with an overall OR/h of 0.97 also reaching statistical significance. The opposite trend was noted for E. coli, with significantly higher OR at 10-20 h and > 20 h (OR = 1.65 and 2.15 respectively), and OR/h of 1.04. A significant increasing trend was also noted for other Enterobacterales, with a significantly higher OR at 10-20 h (OR = 1.34) and OR/h of 1.02. Of note, the small total number of anaerobic organisms (n = 101) and fastidious organisms such as Haemophilus spp (n = 42) or Neisseria spp (n = 5) recovered across all clusters and sites precluded analysis.
Prospective recovery of undetected bacterial growth. A total of 451 bottles were prospectively assessed by terminal subculture and molecular assays (Table S3), including 207 blood bottles with preincubation times < 12 h, used as controls to establish baseline contamination rates. This was done in order to assess both for undetected growth due to loading of blood culture bottles in the automated blood culture system after bacteria had exited their exponential phase of growth, and for the presence of genetic material derived from non-viable bacteria which may not have replicated after the preincubation period. The control group showed the highest rate of growth on subcultures with 4.83% (10 bottles). Half of these bottles had growth of coagulase-negative Staphylococcus species, as listed in Table S3. Other species isolated included Cutibacterium acnes (2 bottles), Bacillus sp (1 bottle; non-anthracis, non-cereus), Streptococcus oralis (1 bottle) and Candida albicans (1 bottle). Overall, 16 blood culture bottles were positive for growth on subculture, and in all cases only between 1 to 3 www.nature.com/scientificreports/ colonies were seen on the plates, in between streaking lines or adjacent to the side of the plate, and outside of the first streaking quadrant, with three exceptions: • one bottle with moderate growth of C. albicans (control group) • one culture plate growing B. cereus (15)(16)(17)(18) h group) with 20 colonies seen in the path of a drop of condensation with no colonies seen on the inoculation streaks • one culture plate with mixed growth (21 h group), located off the inoculation streaks in the 3rd and 4th quadrant of the chocolate agar plate Amplification of genetic material was noted in one blood culture bottle using an assay targeting Staphylococcus spp. Decision to use this assay was made when we noted that there were no isolates of S. aureus recovered from the cluster 5 site (Table 2). A positive reaction was noted with the Staphylococcus 16S probe, but not with the nuc and mecA/mecC probes, indicating the presence of a methicillin-susceptible coagulase-negative Staphylococcus isolate. Universal 16S RNA amplification from this bottle also showed detectable signal.

Discussion
With regards to laboratory services consolidation, expert opinion recommends localized provision of tests with expected turn-around-times of less than 4 h as well as tests for which delays in generating results would negatively impact patient care 1 . This includes blood cultures, with current recommendations for incubation within 2 to 4 h of collection. Not only does this limit opportunity for transportation, but even sites that have in-house microbiology struggle to fulfill this parameter. In a previous study showing that an increase in transport time is partially mitigated by a decrease in incubation time 3 , an interquartile (IQR) range of 2 to 15 h of transport was noted for cultures originating from the site where the main laboratory was located. An earlier study in a tertiary care hospital also showed mean transportation time from ward to laboratory of 10.4 h 14 , and a subsequent study at the same site yielded a median transport time of 3.52 h, with an IQR of 1.33 to 13.90 h 15 . Our own findings, based on the audit of a very large dataset, generally reflect those observed in previous studies: preincubation times varied greatly between sites, a general trend towards decreasing time to detection was noted with an increase in preincubation time, but this did not fully compensate for the lengthy preincubation and led to an overall increase in total time to detection. Only the testing site and cluster 1 sites fulfilled, on average, the 4 h guideline www.nature.com/scientificreports/  www.nature.com/scientificreports/ www.nature.com/scientificreports/ for preincubation: inoculated blood culture bottles reached the main laboratory promptly, with an IQR of 0.64 to 1.72 h, which compares favorably to previous studies 14,15 . Overall, the preincubation time data we have gathered here has prompted our organization to reassess courier routes, with an attempt being made to optimize the use of regional hubs and transport of cultures on an "as needed" basis for distant sites for which one daily pick up in the setting of low, unpredictable volumes, may not be optimal. While distance may very well be the main factor explaining the differences seen in preincubation times, other variables are at play including the number of daily couriers and operating hours of the peripheral sites' laboratories, which are involved in the reception of the samples and their preparation for shipping. The data from sites which seemed to have worse preincubations times than what would be expected based on distance alone (specifically those sites falling above the trendline on Fig. 1) was reviewed, and the total volume of blood culture bottles collected for each hour of the day was plotted along with the mean preincubation time for blood culture bottles collected at specific times for each site (data not shown). In most instances courier pick up times were not synchronized with periods of higher volumes of blood culture collections. In some instances, these periods of higher volumes occurred after the local laboratory had closed, worsening delays. Modification of opening hours or strategies for handling of blood culture bottles after hours will need to be devised to allow for optimal management of these clinically sensitive samples.
The main concerns surrounding prolonged preincubation times relate to the fragility of some bacterial species which may affect recovery or lead to potential overgrowth during preincubation, and thus may impact growth detection post-incubation. There has been a previous report of false negative results for Pseudomonas aeruginosa, Candida albicans, Acinetobacter baumanii and Stenotrophomonas maltophilia possibly due to overgrowth during preincubation prior to incubation in a BacT/ALERT system 16 , findings which were not reproduced subsequently in a continuous-monitoring system with upgraded culture media 17 . Another study showed a statistically significant increase in false negative results after 24 h of preincubation either at room temperature or 35 ºC for A. baumanii, B. fragilis, E. coli, E. faecalis, P. aeruginosa and S. pneumoniae, and this seemed to occur more frequently with the BACTEC 9240 continuous-monitoring system than with the BacT/ALERT 3D system 18 . This study also showed no significant increase in false negative results for the first 12 h of preincubation at room temperature. Lee & al. also showed increased false negative results for P. aeruginosa but not S. aureus and E. coli in a simulated bacteremia experiment following preincubation at 25 ºC or 37 ºC for 48 h and 24 h respectively, again with the BacT/ALERT system 19 . Finally, detection of fastidious organism following delayed-entry into the BacT/ALERT continuous-monitoring system was also assessed, with a small difference in detection rates noted between direct insertion (92.5%) and preincubation for 24 h at room temperature (90%) at low concentration of inoculum 20 . Only the recovery of Streptococcus sanguinis was affected following preincubation in this condition, with good recovery of S. pneumoniae.
Given the above, the lack of recovery of S. aureus and beta-hemolytic Streptococcus spp, which are common agents of community-acquired skin and soft tissue or joint infections with potential for secondary bacteremia, from the site in cluster 5 was a major cause for concerns. While it is reassuring that four blood cultures with S. pneumoniae, which is considered more fastidious than other bacteria due to its tendency to autolyze, were recovered from that site, these bottles were collected at the same time on the same patient and had preincubation times of 6.49 h only, well below the mean preincubation time for this distant site. Statistical analysis seemed to confirm our concerns, with significant lower odds ratio for recovery of Staphylococcus species, including S. aureus, with each additional hour of preincubation. A statistically significant lower OR for recovery of S. aureus was also noted with 5-10 h of preincubation, but clinical relevance of this signal is unclear given that this was not noted for the 10-20 h preincubation group. It should be noted that a lower number of S. aureus isolates, but not a complete absence, may be expected from smaller community centers: patients are typically less medically complex, with lower absolute number of intravascular device-associated infections which are commonly seen in tertiary centers. This would lead to an increase in the proportion of cases of bacteremia caused by S. aureus in larger care centers and as such it is possible that this might have introduced a bias in our logistic regression analysis: community centers would be expected to have a lower proportion of cases of bacteremia linked to S. aureus or CoNS, and since they are located farther way from the testing site these would preferentially be included in higher preincubation time clusters. This bias might also explain why logistic regression also showed greater odds of recovery of E. coli and other Enterobacterales with increasing preincubation times as bacteremia with these organisms may represent a greater share of positive blood culture from peripheral, community sites in the context of a higher proportion of cases of secondary bacteremia due to urinary tract infections.
Terminal subcultures mainly yielded growth of organisms generally recognized as contaminants except one case of moderate growth of Candida albicans from an aerobic blood culture bottle from the control group. This was surprising given that there is no known limitation for the detection of C. albicans in BACTEC FX even at low inoculums, with blood cultures flagged as positive within 24.5 h of insertion into the instrument on average 21 . Careful clinical and microbiological review done at the time showed a preincubation time of 88 min, did not show evidence of antifungal treatment prior to the collection of the blood culture, and no evidence of symptoms concerning for ongoing, untreated fungemia. The initial blood culture result was nonetheless corrected and, the primary care providers were made aware of this unexpected finding to allow for appropriate clinical correlation. Antifungal treatment was not initiated thereafter indicating that this result was felt to represent contamination. It should be noted that the molecular assays we used to assess for undetected growth of non-viable organisms was targeting bacterial agents only and would not have been adequate to assess for the presence of fungal agents. Our Staphylococcus spp molecular assay led to the identification of one blood culture bottle which may have contained a methicillin-susceptible coagulase-negative Staphylococcus isolate, and which was not identified by terminal subculture. This same bottle showed detectable signal by 16S rRNA amplification, and this was felt to represent the same CoNS isolate. Of note, this blood culture bottle was from the control group, with a preincubation www.nature.com/scientificreports/ time < 12 h. None of our molecular assays identified previously undetected growth of pathogenic bacterial isolates, which also provided reassurance that preincubation did not prevent bacterial recovery in our study. Other limitations than the ones already stated may have impacted our study. First, while standard practice in our tertiary care institution is to collect blood culture bottles prior to initiation of antibiotherapy, we cannot confirm if this practice was enforced in our smaller community hospitals. This could have introduced a bias in the recovery of organisms for sites located further from the testing site and contributed to the lack of recovery of Staphylococcus spp and beta-hemolytic Streptococcus from the cluster 5 site. Second, while our study included pediatric blood culture bottles, the vast majority would have come cluster 1 site E, a regional tertiary care pediatric center located within walking distance of the testing site. Subgroup analysis to assess bacterial recovery as a function of time for pediatric blood culture bottles was therefore not possible, but this subgroup would not be expected to affect our main conclusions. Thirdly, the prevalence of specific pathogens in each cluster sites prior to regionalization could not be assessed as this data, largely based on paper records and laboratory worksheets, was not available to the research team. As prevalence of specific pathogens could vary over time, as an example in the context of outbreaks, comparison with historical data may not have allowed to infer success or lack-thereof microbial recovery post-regionalization. Lastly, we only recovered 101 anaerobic isolates and 89 Neisseria spp or Haemophilus spp isolates, small numbers which precluded analysis across all sites.
Overall, the results from our study indicate that an increase in preincubation time does impact total detection time, and that bacterial recovery may be impacted for some species with a prolonged preincubation although the presence of a possible bias in the logistic regression analysis should be considered, and results from terminal subcultures and molecular assays did not support this finding. Based on the prolongation in total detection time alone and the associated clinical impact of a delayed report, centralized laboratories should aim at reducing preincubation times as much as possible through optimization of courier services and processing of collected samples at peripheral sites. However, given the new reality of a centralization of microbiological services in the context of budget constraints, we feel that our data does support considering a modest increase in the currently recommended maximum preincubation times to 8 to 12 h.

Methods
Study sites and samples. EORLA RMRL is located within the General campus of The Ottawa Hospital, a tertiary care academic center, and provides diagnostic microbiology services 24 h a day, 7 days a week, in eastern Ontario. Seven hospitals are located 0.5 to 12.3 km from RMRL within the city of Ottawa, including a pediatric tertiary care centre, a specialized centre for cardiac interventions, a large tertiary care centre with a trauma centre, a large tertiary care centre with an hematology and stem cell transplant ward, an outpatient hospital housing a renal transplant outpatient clinics and one-day elective surgery, and two community hospitals offering secondary care services. Twelve smaller community hospitals are located within 49.6 to 197 km of the RMRL. In order to facilitate interpretation of the results, sites were grouped into clusters (Table 1) based on preliminary data showing similarities in mean and median preincubation times and compared to the testing site; these sites often share common courier routes for sample delivery. Cluster 1 is composed of hospitals located within the city of Ottawa, other than the testing site. Cluster 2 includes two large community hospitals located east of Ottawa. Cluster 3 includes three community hospitals located west of Ottawa and one community hospital located to the south. Cluster 4 is composed of four community hospitals located in suburban areas west and south of Ottawa, and one smaller site located farther west. Cluster 5 is composed of a single site serving a distant, small community, for which mean and median preincubation times were outliers when compared to other sites.
Between July 1st 2017 and June 30th 2018, a total of 70,458 inoculated blood culture bottles were received at the RMRL (Table 1). Included were aerobic, anaerobic and pediatric blood culture bottles (BACTEC™ Plus Aerobic/F, BACTEC™ Lytic/10 Anaerobic/F and BACTEC™ Peds Plus/F culture vials respectively, BD-Canada, Mississauga, ON). For the purpose of this study, the following exclusion criteria were established: • bottles for which no loading time, defined as the time point at which the bottle was inserted into the BACTEC instrument, was recorded • bottles for which time from collection to loading was inferior to the minimum expected transport time from the collecting ward to the testing site, established using Google Maps, under normal traffic conditions. This was established in order to exclude bottles for which the collection time, which is a mandatory information in our laboratory information system, was not recorded at the collection site and instead recorded as the reception time at the testing site. • bottles originating from autopsies • bottles originating from proficiency testing panels This left a total of 69,604 blood cultures bottles (59,077 negatives, 10,527 positives), all of which were incubated in the BACTEC FX continuous-monitoring system for 5 days as per our standard blood culture protocol. Pediatric blood bottles were included in the analysis. Pediatric blood bottles were included in the analysis, and mainly came from Cluster 1 Site E (pediatric hospital, 540 negative and 60 positive bottles). Positivity rates and contamination rates, defined as the recovery of skin organisms (coagulase-negative Staphylococcus) or other common environmental organisms (Paenibacillus spp, Bacillus spp non-anthracis) in single bottles from a set of blood cultures, were also documented for each site.

Preincubation time, incubation time (INC) and total detection time (TDT). Preincubation time,
i.e. the time from drawing of the blood sample on the patient to the insertion of the inoculated bottle into the continuous monitoring system, was calculated for each bottle. We included data from negative bottles in the www.nature.com/scientificreports/ calculation of median and mean preincubation times for all individual sites. Only monomicrobial positive blood cultures were used to establish INC (defined as time from insertion into the automated blood system to growth detection) and TDT (defined as time from collection of samples to growth detection) in order to allow analysis by species clusters. INC was calculated as the time and date at which the bottle was flagged positive (detection time) minus the time and date of insertion into the BACTEC FX. TDT was calculated as the detection time minus the time and date of drawing. Whenever the detection time was not available, this was inferred from the time at which the gram stain result was recorded in the laboratory information system minus 30 min, in order to account for the time required to prepare, stain and read the slides.
Data visualization and statistical analysis. Preincubation time and TDT data was presented visually using the "box and whisker" chart type in Microsoft Excel (Figs. 2, 3). Data points falling more than 1.5 times the interquartile range above the third quartile or below the first quartile were considered outliers. For the statistical analysis of positivity rates by bacterial types and site clusters, all P values are reported as two-sided. Hypothesis testing was carried out with an overall level of significance set using a P value < 0.05. Median and average preincubation times were established for all sites, including the testing site.  Table S3). A total of 207 bottles with incubation times < 12 h were selected as controls. Overall, 500 blood culture bottles were tested. All remote sites were included in the sampling. Bottles with preincubation of > 12 h were flagged upon reception at the testing site, and subsequently incubated in the BACTEC FX as per the manufacturer's instruction. Flagged bottles that exhibited no growth after 5 days of incubation were retrieved. These were subcultured to chocolate agar plates for incubation under aerobic conditions (48 h incubation at 35ºC under 5% CO 2 atmosphere) and to CDC agar plates for incubation under anaerobic conditions (48 h incubation at 35ºC) as per our standard blood culture protocol. Aliquots of blood were simultaneously retrieved for molecular testing (detection of Staphylococcus spp and 16S RNA amplification described below). No gram stain was performed. Aerobic plates were incubated in Kiestra ReadA Compact smart incubators (Becton, Dickinson & Company, Franklin Lake, NJ), and plates were imaged at 24 and 48 h. Anaerobic plates were incubated in anaerobic jars in standard incubators, but also imaged on the ReadA Compact after 48 h. The presence or absence of growth was recorded. Since these were clinical samples, any plate exhibiting growth was rerouted to the technologist performing work up for positive blood cultures, and bacterial identification and susceptibility testing was performed as per our standard clinical protocol, when appropriate. Identification was performed by MALDI-TOF MS (Bruker Ltd, Milton, ON) using Biotyper v.7854.

Molecular detection of Staphylococcus spp and universal 16S RNA amplification.
Aliquots of 1 ml of blood culture broth were retrieved from bottles with prolonged preincubation (> 12 h) and absence of growth in BACTEC FX at 5 days and tested using a laboratory developed assay for the qualitative detection of Staphylococcus spp, S. aureus and MRSA through the simultaneous amplification of genes encoding the Staphylococcus 16S rRNA, the S. aureus endonuclease nuc and the MRSA mecA/mecC. This sample was also tested for the presence of bacterial 16S RNA as previously described 22 . Six microliters of blood culture were added to a MagNA Lyser Green Bead vial (Roche Diagnostics) containing 600 µL of TE buffer. The mix was then lysed in a MagNA Lyzer instrument for 70 s at 5,000 rpm. 2 μL of blood extract was used directly for real-time PCR. Specific primers and probes sequences for mecA/mecC and nuc were previously described 23 , as were the primers and probe sequences for Staphylococcus spp 16S rRNA 24 and the universal 16S RNA primers 22 (Supplementary data, Table S1). For the latter assay, which was developed for molecular detection of micro-organisms directly from clinical samples, published data indicate a limit of detection of 150 CFU per PCR reaction from blood sample 22 , while internal validation showed a limit of detection of 75.5-85 CFU in sterile fluids and tissue samples respectively (data not shown). Real-time PCR was conducted in a total volume of 10 μL, containing 1 × QuantiFast Multiplex PCR Kit (QIA-GEN), MecA, MecAlga251, Nuc, Staph16S primers and corresponding dual-labelled probes (Bioresearch Technologies) as per the final concentrations listed in Table S1 www.nature.com/scientificreports/