Eur J Clin Microbiol Infect Dis DOI 10.1007/s10096-015-2373-2
ARTICLE
Evaluation of the BD BACTEC FX blood volume monitoring system as a continuous quality improvement measure L. Coorevits 1,2 & A.-M. Van den Abeele 1
Received: 11 February 2015 / Accepted: 23 March 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract The yield of blood cultures is proportional to the volume of blood cultured. We evaluated an automatic blood volume monitoring system, recently developed by Becton Dickinson within its BACTEC EpiCenter module, that calculates mean volumes of negative aerobic bottles and generates boxplots and histograms. First, we evaluated the filling degree of 339 aerobic glass blood cultures by calculating the weightbased volume for each bottle. A substantial amount of the bottles (48.3 %) were inadequately filled. Evaluation of the accuracy of the monitoring system showed a mean bias of −1.4 mL (−15.4 %). Additional evaluation, using the amended software on 287 aerobic blood culture bottles, resulted in an acceptable mean deviation of −0.3 mL (−3.3 %). The new software version was also tested on 200 of the recently introduced plastic bottles, which will replace the glass bottles in the near future, showing a mean deviation of +2.8 mL (+26.7 %). In conclusion, the mean calculated volumes can be used for the training of a single phlebotomist. However, filling problems appear to be masked when using them for phlebotomist groups or on wards. Here, visual interpretation of boxplots and histograms can serve as a useful tool to observe the spread of the filling degrees and to develop a continuous improvement program. Re-adjustment of the software has proven to be necessary for use with plastic bottles. Due to our findings, BD has developed further adjustments to the software for validated use with plastic bottles, which will be released soon.
* L. Coorevits [email protected] 1
Department of Clinical Microbiology, Saint Lucas Hospital, Groenebriel 1, 9000 Ghent, Belgium
2
Department of Laboratory Medicine, Ghent University Hospital, De Pintelaan 185, 2P8, 9000 Ghent, Belgium
Introduction The incidence of bloodstream infections (BSI) is continuously increasing [1]. Worldwide, the mortality associated with nosocomial BSI remains high and varies from 30 to 55 % [2, 3]. Blood cultures remain the most important diagnostic tool for the detection of BSI. As such, analytical improvements attempting to enhance the accurate and rapid isolation of the responsible pathogens have been a major subject of research for the past two decades. Despite considerable technical improvements in continuous blood culture monitoring systems, diagnostic blood culture improvement equally depends on several factors in the pre-analytical phase: sample collection, transport to the laboratory, specimen receipt, and distribution in the laboratory. It is well known that the amount of inoculated blood is critical and, perhaps, the most important variable in recovering microorganisms from patients with bacteremia [4, 5]. In 2012, the approved guidelines for the collection of blood cultures were reviewed, newly available evidence was incorporated, controversies were discussed, and a consensus for each pre-analytical variable applicable to blood cultures was proposed [6]. During a random test in four secondary care hospitals, the authors determined the percentage of blood culture bottles with >2 mL above or below the recommended volume for the vial type. Irrespective of the manufacturer, the percentage of overfilled bottles ranged from 7.6 to 12.8 %, whereas in 26.3–36.0 %, the volume was below the recommended level. As a consequence of this shortcoming, Becton Dickinson (BD) recently developed an automatic blood volume monitoring system (BVMS) within its BACTEC EpiCenter module. The volume of blood in negative aerobic bottles is estimated from blood background signal data, based on the metabolic activity of red blood cells. Because the estimation is based on the initial blood background
Eur J Clin Microbiol Infect Dis
rate, bottles that are delayed before entry into the BD BACTEC FX continuous monitoring blood culture instrument (>24 h at room temperature or >12 h at 35 °C) may impact the results. Blood samples with hematocrit below 30 % may also generate imprecise estimates. The BVMS does not provide a measurement for each individual incubated bottle, but calculates a mean volume whenever data of ≥25 bottles are available. The aim of our study was to evaluate the analytical performance of this BVMS and to assess the usefulness as a tool for continuous quality improvement.
Materials and methods PART 1 – In order to validate the accuracy of the BVMS (software version 4.20B), we compared the BVMS volume of 400 aerobic bottles with the volume calculated based on the net weight of each capped aerobic bottle. This weight was defined by weighing each individual bottle before distribution to the phlebotomists and after completion of incubation in the BD BACTEC FX instrument. We could not use a mean tare weight, as weighing 50 empty bottles showed a spreading as large as ±1 mL, accompanied with inter-lot variability up to 2 mL. The bottles were distributed amongst the laboratory phlebotomy team—which performs approximately threequarters of all patient draws—and to three selected nursing departments (intensive care, emergency room, and oncology). Once the bottles were drawn, they were incubated in the BD BACTEC FX instrument for 5 days or until bacterial growth was detected. All negative aerobic bottles were weighed again and the volume of the blood cultured was calculated using the following formula: [volume = (weight of filled bottle − weight of empty bottle)/density of blood (1.055 g/ mL)]. The patient’s hematocrit at the time of blood culture puncture was recorded. First, we evaluated the phlebotomy skills of the laboratory phlebotomy team and the nursing departments. As the BVMS does not provide the volume for each individual bottle, the mean volume for both the phlebotomy team and the three nursing departments was compared with the mean weightbased volume for both groups. Second, the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a key performance indicator (KPI) of 8–12 mL. We also investigated a possible relationship between overfilling and false-positivity. Finally, we evaluated the mean BVMS volume of all negative aerobic bottles between 01/10/2011 and 27/06/2013 for each nursing department that included ≥25 bottles. PART 2 – Because of a repeatedly reported, though unpublished, inaccuracy in the calculation algorithm, BD developed a new software (version 4.30) which was installed in our laboratory on 02/01/2014. This software was validated by repeating our experiments on 350 aerobic bottles, which were again distributed amongst the laboratory phlebotomy team and the
three selected nursing departments. Again, the mean BVMS volume of the negative aerobic bottles was compared with the mean weight-based volume, the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, and the mean BVMS volume of all negative aerobic bottles between 14/02/2014 and 22/05/2014 for each department that included ≥25 bottles was calculated. PART 3 – As BD has recently developed plastic blood culture bottles to replace the current glass bottles, a limited validation experiment on 200 plastic aerobic bottles was performed. All bottles were distributed amongst the laboratory phlebotomy team. Once again, the mean BVMS volume of the negative aerobic bottles was compared with the mean weightbased volume and the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled. As the plastic and glass bottles were used interchangeably, no retrospective investigation on a large number of plastic bottles could be carried out. During the study course, no feedback on the bottle filling degrees was given to the laboratory phlebotomy team nor the nursing departments.
Results PART 1 – We compared the mean weight-based volume of 339 negative aerobic bottles with the volume calculated by the BVMS. For the laboratory phlebotomy team, the mean weight-based volume was 9.9 mL [95 % confidence interval (CI): 9.5–10.4] and the mean BVMS volume was 8.8 mL (95 % CI: 7.6–10.0); this is an absolute deviation of −1.1 mL or −12.2 %. For the nursing departments, the mean weightbased volume was 9.8 mL (95 % CI: 8.9–10.7) and the mean BVMS volume was 8.1 mL (95 % CI: 7.5–8.7), an absolute deviation of −1.7 mL or −18.6 % (Fig. 1 – PART 1). The mean absolute deviation for both groups was −1.4 mL or −15.4 %. The weight-based volume of each individual bottle from the first experiment was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. The laboratory phlebotomy team drew 203 aerobic blood cultures, of which 105 (51.7 %) were correctly filled (8–12 mL), 48 (23.7 %) were filled 12 mL. The selected nursing departments drew 136 aerobic blood cultures, of which 46 (33.8 %) were correctly filled (8– 12 mL), 53 (39.0 %) contained 12 mL. When we excluded the patients with a hematocrit of 30 %, of which 42 (33.3 %) were correctly filled (8– 12 mL), 51 (40.5 %) contained 12 mL (Table 1 – PART 1). Within the 339
Eur J Clin Microbiol Infect Dis Fig. 1 Mean weight-based volume (♦) and mean blood volume monitoring system (BVMS) volume (▲) ±95 % confidence interval (―) of the negatively flagged aerobic bottles in each part of the experiment for drawings by the nursing departments and the laboratory phlebotomy team. a PART 1 – nursing departments; b PART 1 – phlebotomy team; c PART 2 – nursing departments; d PART 2 – phlebotomy team; e PART 3 – phlebotomy team)
included bottles, one was flagged positive, though no bacteria were seen nor grown. This bottle was slightly overfilled (12.1 mL). The anaerobic bottle, however, became positive with Staphylococcus aureus. Between 01/10/2011 and 27/06/2013, 8,991 aerobic blood culture bottles were incubated in and reported negative by the BD BACTEC FX instrument. The BVMS can calculate a mean volume for all those bottles together or split up by nursing department or phlebotomist whenever ≥25 bottles are included in each category. In the current setting, the identification of the individual phlebotomist is not yet registered in the EpiCenter module. This is why we evaluated the filling degree of the bottles for each nursing department, using the KPI of 8– 12 mL. In the selected period, 28 departments drew ≥25 bottles (total n=8,834). The mean BVMS volume was within the KPI for 18/28 (64.3 %) departments. The remaining 10/28 (35.7 %) departments all had a mean BVMS volume below the KPI. Taking into account the systematic deviation of BVMS volume software version 4.20B (−15.4 %), recalculation of these volumes resulted in only 3/28 (10.7 %) departments with a mean BVMS volume below the KPI, whilst the remaining 25/28 (89.3 %) departments all had a mean BVMS volume within the KPI (Table 2 – PART 1). PART 2 – The mean weight-based volume of 287 negative aerobic bottles was compared with the volume calculated by the BVMS. For the laboratory phlebotomy team, the mean weight-based volume was 9.9 mL (95 % CI 8.9–10.8) and the mean BVMS volume was 9.3 mL (95 % CI: 9.7–10.5); this is an absolute deviation of −0.6 mL or −5.9 %. For the nursing departments, the mean weight-based volume was 10.2 mL (95 % CI: 9.7–10.6) and the mean BVMS volume was 10.1 mL (95 % CI: 9.7–10.5), an absolute deviation of −0.1 mL or −0.8 % (Fig. 1 – PART 2). The mean absolute deviation for both groups was −0.3 mL or −3.3 %.
The weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. The laboratory phlebotomy team drew 187 aerobic blood cultures, of which 97 (51.9 %) were correctly filled (8–12 mL), 38 (20.3 %) were filled 12 mL. The selected nursing departments drew 100 aerobic blood cultures, of which 42 (42.0 %) were correctly filled (8–12 mL), 33 (33.0 %) contained 12 mL. When we excluded the patients with a hematocrit of 30 %, of which 28 (41.8 %) were correctly filled (8–12 mL), 26 (38.8 %) contained 12 mL (Table 1 – PART 2). Within the 287 included bottles, none were falsely flagged positive. Between 14/02/2014 and 22/05/2014, 1,557 aerobic blood culture bottles were incubated in and flagged negative by the BD BACTEC FX instrument. Split up by nursing department, 17 departments drew ≥25 bottles (total n=1,350). The mean BVMS volume was within the KPI for 15/17 (88.2 %) departments. The two other departments had a mean BVMS volume below or above the KPI (Table 2 – PART 2). PART 3 – The mean weight-based volume of the 173 negative aerobic plastic bottles was 9.1 mL (95 % CI: 8.7–9.5), whilst the mean BVMS volume was 11.9 mL (95 % CI: 11.5– 12.4). This is an absolute deviation of +2.8 mL or +26.7 %. The weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. Of the 173 aerobic blood cultures, 100 (57.8 %) were correctly filled (8–12 mL), 52 (30.1 %) were
33.3 % 40.5 % 26.2 % 100 % 41.8 % 38.8 % 19.4 % 100 %
%
Number of nursing departments
42 51 33 126 28 26 13 67 40.0 % 20.0 % 40.0 % 100 % 42.4 % 21.2 % 36.4 % 100 % 33.8 % 4 39.0 % 2 27.2 % 4 100 % 10 42.0 % 14 33.0 % 7 25.0 % 12 100 % 33 46 53 37 136 42 33 25 100
n n % n
%
Patients with hematocrit 1.5 mL. Because the estimation of blood volume is based on the initial blood background rate, BD also warns that bottles with a delayed entry into the BACTEC instrument (>24 h at 20–25 °C or >12 h at 35 °C) will also result in an RMSE >1.5 mL.
We evaluated the accuracy of the BVMS by comparing the mean weight-based volume of 339 aerobic bottles with the volume calculated by the BVMS. Data were interpreted per phlebotomy team (laboratory phlebotomy team vs. three included nursing departments). We noted a mean negative deviation of −1.4 mL (−15.4 %) between the mean weight-based volume and the BVMS volume in both groups. Given the high number of bottles included, we considered the mean deviation (−1.4 mL or −15.4 %) as a representative correction factor to be applied on the mean volumes measured by the BVMS. To amend for this deviation, BD developed a new software version, 4.30, which was installed and evaluated using 287 aerobic bottles. The negative deviation disappeared nearly completely, leaving an acceptable mean deviation of −0.3 mL (−3.3 %). A first validation of this amended software version performed in our laboratory on 173 plastic aerobic bottles resulted in an unacceptable deviation of +2.8 mL (+ 26.7 %). In response to this finding, BD has developed further adjustments on the software, which is currently in the final stages of the validation phase. The weight-based volume of the 339 bottles from the first part of the experiments was subjected to an individual analysis of the filling degree towards the KPI of 8–12 mL. Within the laboratory phlebotomy team, no less than 48.3 % of all bottles were inadequately filled, with an almost equal division between the under- and overfilled bottles. The nursing departments scored even worse, with 66.2 % inadequately filled bottles, of which the majority was underfilled. The 287 bottles from the experiments ran on the new software were subjected to the same analysis: 48.1 % of the bottles drawn by the laboratory phlebotomy team was inadequately filled (with a tendency to overfilling), compared to 58.0 % of the bottles drawn by the nursing departments (where mainly underfilling was observed). A correlation between overfilling and false-positivity could not been shown. These poor amounts of adequately filled bottles clearly demonstrates the need to monitor the filling degree actively, (re)educate the phlebotomists, and to observe trends on a routine basis. However, in the daily routine
Eur J Clin Microbiol Infect Dis
microbiology lab, it is not feasible to calculate the weightbased volume of each bottle, leaving the mean BVMS volumes for specific groups to monitor. The disadvantage of those mean values is that they are less sensitive because under- and overfilled bottles compensate each other, masking possible shortcomings. This is shown by calculating the mean BVMS volumes of the 339 and 287 bottles in the first two experiments. Neither for the laboratory phlebotomy team nor for the nursing departments could the BVMS detect a filling problem. The respective mean volumes in the first experiment (n=339) were 8.8 mL (95 % CI: 7.6–10.0) and 8.1 mL (95 % CI: 7.5–8.7), whilst in the second part (n=287), they were 10.1 mL (95 % CI: 9.7–10.5) and 9.3 mL (95 % CI: 8.6– 10.0). To bypass this limitation, histograms (Fig. 2) or boxplots (Fig. 3) can be automatically generated, on which deviations from the KPI can be detected visually. Histograms can be generated when at least 100 data are included in one category, whereas boxplots only need 25 data. In extension, we evaluated the BVMS volumes of all data collected between 01/10/2011 and 27/06/2013 from 28 nursing departments (software version 4.20B) and between 14/02/ 2014 and 22/05/2014 from 17 nursing departments (software version 4.30). Taking into account our own calculated correction factor for the first software version, only 3/28 (10.7 %) departments had a mean BVMS volume outside (below) the KPI. From the 17 nursing departments observed in the second period with the amended software version, only 2 (11.8 %) had a mean BVMS volume outside (below) the KPI. This confirms the observation that the mean BVMS volume, per se, is not adequate to monitor the filling degree of a department as a whole. Nevertheless, these values could be useful to monitor a single phlebotomist, as one would expect a constant pattern, either under- or overfilled. In conclusion, there is a distinct need to actively monitor the filling degree of blood culture bottles. The BVMS (software version 4.20B) calculated the volume of aerobic bottles with a mean deviation of −1.4 mL or −15.4 % as compared with the weight-based volumes. The software update (version 4.30) lead to an acceptable deviation to −0.3 mL or −3.3 % on the glass bottles. The current algorithm proves to be inadequate on plastic bottles, as we noted a deviation of +2.8 mL or +26.7 %. However, BD has developed further adjustments and an improved software update is to be expected soon. The use of mean BVMS volumes to monitor the skills within a group of different phlebotomists can mask a potential filling problem. However, together with visual interpretation of histograms and/or boxplots produced by the BVMS, it is considered a useful tool to guide quality improvement policies. Acknowledgments The authors would like to thank Cindy Germis, Liesbeth Van Acker, Caroline Pieters, and Line Coucke, coworkers in the microbiology lab of the AZ Sint-Lucas Hospital, for their outstanding technical assistance in the experiments.
References 1.
2.
3.
4.
5. 6.
7.
8.
9.
10.
11. 12. 13.
14.
15.
16.
17.
18.
Rodríguez-Créixems M, Alcalá L, Muñoz P, Cercenado E, Vicente T, Bouza E (2008) Bloodstream infections: evolution and trends in the microbiology workload, incidence, and etiology, 1985–2006. Medicine (Baltimore) 87:234–249 Engel C, Brunkhorst FM, Bone HG, Brunkhorst R, Gerlach H, Grond S, Gruendling M, Huhle G, Jaschinski U, John S, Mayer K, Oppert M, Olthoff D, Quintel M, Ragaller M, Rossaint R, Stuber F, Weiler N, Welte T, Bogatsch H, Hartog C, Loeffler M, Reinhart K (2007) Epidemiology of sepsis in Germany: results from a national prospective multicenter study. Intensive Care Med 33:606–618 Suetens C (2008) Surveillance of nosocomial bloodstream infections in Belgian hospitals: results from the national surveillance network, 1992–2004 and 2007 (preliminary). Scientific Institute of Public Health, Brussels York MK, Henry M, Gilligan P (2010) Section 3.4.1. Aerobic bacteriology, blood cultures, general detection and interpretation. In: Isenberg HD (ed) Clinical microbiology procedures handbook. American Society of Microbiology (ASM) Press, Washington, DC Riedel S, Carroll KC (2010) Blood cultures: key elements for best practices and future directions. J Infect Chemother 16:301–316 Willems E, Smismans A, Cartuyvels R, Coppens G, Van Vaerenbergh K, Van den Abeele AM, Frans J; Bilulu Study Group (2012) The preanalytical optimization of blood cultures: a review and the clinical importance of benchmarking in 5 Belgian hospitals. Diagn Microbiol Infect Dis 73:1–8 Bouza E, Sousa D, Rodríguez-Créixems M, Lechuz JG, Muñoz P (2007) Is the volume of blood cultured still a significant factor in the diagnosis of bloodstream infections? J Clin Microbiol 45:2765– 2769 Cockerill FR 3rd, Wilson JW, Vetter EA, Goodman KM, Torgerson CA, Harmsen WS, Schleck CD, Ilstrup DM, Washington JA 2nd, Wilson WR (2004) Optimal testing parameters for blood cultures. Clin Infect Dis 38:1724–1730 Clinical and Laboratory Standards Institute (CLSI) (2007) Principles and procedures for blood cultures; Approved guideline. CLSI document M47-A. CLSI, Wayne, PA, pp 1–53 Cohen J, Brun-Buisson C, Torres A, Jorgensen J (2004) Diagnosis of infection in sepsis: an evidence-based review. Crit Care Med 32: S466–S494 Lassmann B, Gustafson DR, Wood CM, Rosenblatt JE (2007) Reemergence of anaerobic bacteremia. Clin Infect Dis 44:895–900 Hecht DW (2006) Anaerobes: antibiotic resistance, clinical significance, and the role of susceptibility testing. Anaerobe 12:115–121 Bannister ER, Woods GL (1995) Evaluation of routine anaerobic blood cultures in the BacT/Alert blood culture system. Am J Clin Pathol 104:279–282 Horvath LL, George BJ, Hospenthal DR (2007) Detection of fifteen species of Candida in an automated blood culture system. J Clin Microbiol 45:3062–3064 Morris AJ, Wilson ML, Mirrett S, Reller LB (1993) Rationale for selective use of anaerobic blood cultures. J Clin Microbiol 31: 2110–2113 Miller JM (1998) A guide to specimen management in clinical microbiology, 2nd edn. American Society of Microbiology (ASM) Press, Washington, DC Thompson RB Jr, Miller JM (2007) Specimen collection, transport and processing: bacteriology. In: Murray PR, Baron EJ (eds) Manual of clinical microbiology, 9th edn. American Society of Microbiology (ASM) Press, Washington, DC, pp 291–310 James PA, Al-Shafi KM (2000) Clinical value of anaerobic blood culture: a retrospective analysis of positive patient episodes. J Clin Pathol 53:231–233
Eur J Clin Microbiol Infect Dis 19.
van Ingen J, Hilt N, Bosboom R (2013) Education of phlebotomy teams improves blood volume in blood culture bottles. J Clin Microbiol 51:1020–1021 20. Wilson ML, Harrell LJ, Mirrett S, Weinstein MP, Stratton CW, Reller LB (1992) Controlled evaluation of BACTEC PLUS 27 and Roche Septi-Chek anaerobic blood culture bottles. J Clin Microbiol 30:63–66 21. Weinstein MP, Mirrett S, Wilson ML, Reimer LG, Reller LB (1994) Controlled evaluation of 5 versus 10 milliliters of blood cultured in aerobic BacT/Alert blood culture bottles. J Clin Microbiol 32: 2103–2106 22. Koontz FP, Flint KK, Reynolds JK, Allen SD (1991) Multicenter comparison of the high volume (10 ml) NR BACTEC PLUS system and the standard (5 ml) NR BACTEC system. Diagn Microbiol Infect Dis 14:111–118
23.
24.
25. 26.
Weinstein MP, Mirrett S, Reimer LG, Wilson ML, Smith-Elekes S, Chuard CR, Joho KL, Reller LB (1995) Controlled evaluation of BacT/Alert standard aerobic and FAN aerobic blood culture bottles for detection of bacteremia and fungemia. J Clin Microbiol 33:978– 981 Mirrett S, Hanson KE, Reller LB (2007) Controlled clinical comparison of VersaTREK and BacT/ALERT blood culture systems. J Clin Microbiol 45:299–302 Reimer LG, Wilson ML, Weinstein MP (1997) Update on detection of bacteremia and fungemia. Clin Microbiol Rev 10:444–465 Jungkind DL, Thakur M, Dyke J (1989) Evidence for a second mechanism of action of resin in BACTEC NR 16A aerobic blood culture medium, abstr. C-225, p 430. In: Abstracts of the 89th Annual Meeting of the American Society for Microbiology 1989. American Society for Microbiology (ASM), Washington, DC
ARTICLE
Evaluation of the BD BACTEC FX blood volume monitoring system as a continuous quality improvement measure L. Coorevits 1,2 & A.-M. Van den Abeele 1
Received: 11 February 2015 / Accepted: 23 March 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract The yield of blood cultures is proportional to the volume of blood cultured. We evaluated an automatic blood volume monitoring system, recently developed by Becton Dickinson within its BACTEC EpiCenter module, that calculates mean volumes of negative aerobic bottles and generates boxplots and histograms. First, we evaluated the filling degree of 339 aerobic glass blood cultures by calculating the weightbased volume for each bottle. A substantial amount of the bottles (48.3 %) were inadequately filled. Evaluation of the accuracy of the monitoring system showed a mean bias of −1.4 mL (−15.4 %). Additional evaluation, using the amended software on 287 aerobic blood culture bottles, resulted in an acceptable mean deviation of −0.3 mL (−3.3 %). The new software version was also tested on 200 of the recently introduced plastic bottles, which will replace the glass bottles in the near future, showing a mean deviation of +2.8 mL (+26.7 %). In conclusion, the mean calculated volumes can be used for the training of a single phlebotomist. However, filling problems appear to be masked when using them for phlebotomist groups or on wards. Here, visual interpretation of boxplots and histograms can serve as a useful tool to observe the spread of the filling degrees and to develop a continuous improvement program. Re-adjustment of the software has proven to be necessary for use with plastic bottles. Due to our findings, BD has developed further adjustments to the software for validated use with plastic bottles, which will be released soon.
* L. Coorevits [email protected] 1
Department of Clinical Microbiology, Saint Lucas Hospital, Groenebriel 1, 9000 Ghent, Belgium
2
Department of Laboratory Medicine, Ghent University Hospital, De Pintelaan 185, 2P8, 9000 Ghent, Belgium
Introduction The incidence of bloodstream infections (BSI) is continuously increasing [1]. Worldwide, the mortality associated with nosocomial BSI remains high and varies from 30 to 55 % [2, 3]. Blood cultures remain the most important diagnostic tool for the detection of BSI. As such, analytical improvements attempting to enhance the accurate and rapid isolation of the responsible pathogens have been a major subject of research for the past two decades. Despite considerable technical improvements in continuous blood culture monitoring systems, diagnostic blood culture improvement equally depends on several factors in the pre-analytical phase: sample collection, transport to the laboratory, specimen receipt, and distribution in the laboratory. It is well known that the amount of inoculated blood is critical and, perhaps, the most important variable in recovering microorganisms from patients with bacteremia [4, 5]. In 2012, the approved guidelines for the collection of blood cultures were reviewed, newly available evidence was incorporated, controversies were discussed, and a consensus for each pre-analytical variable applicable to blood cultures was proposed [6]. During a random test in four secondary care hospitals, the authors determined the percentage of blood culture bottles with >2 mL above or below the recommended volume for the vial type. Irrespective of the manufacturer, the percentage of overfilled bottles ranged from 7.6 to 12.8 %, whereas in 26.3–36.0 %, the volume was below the recommended level. As a consequence of this shortcoming, Becton Dickinson (BD) recently developed an automatic blood volume monitoring system (BVMS) within its BACTEC EpiCenter module. The volume of blood in negative aerobic bottles is estimated from blood background signal data, based on the metabolic activity of red blood cells. Because the estimation is based on the initial blood background
Eur J Clin Microbiol Infect Dis
rate, bottles that are delayed before entry into the BD BACTEC FX continuous monitoring blood culture instrument (>24 h at room temperature or >12 h at 35 °C) may impact the results. Blood samples with hematocrit below 30 % may also generate imprecise estimates. The BVMS does not provide a measurement for each individual incubated bottle, but calculates a mean volume whenever data of ≥25 bottles are available. The aim of our study was to evaluate the analytical performance of this BVMS and to assess the usefulness as a tool for continuous quality improvement.
Materials and methods PART 1 – In order to validate the accuracy of the BVMS (software version 4.20B), we compared the BVMS volume of 400 aerobic bottles with the volume calculated based on the net weight of each capped aerobic bottle. This weight was defined by weighing each individual bottle before distribution to the phlebotomists and after completion of incubation in the BD BACTEC FX instrument. We could not use a mean tare weight, as weighing 50 empty bottles showed a spreading as large as ±1 mL, accompanied with inter-lot variability up to 2 mL. The bottles were distributed amongst the laboratory phlebotomy team—which performs approximately threequarters of all patient draws—and to three selected nursing departments (intensive care, emergency room, and oncology). Once the bottles were drawn, they were incubated in the BD BACTEC FX instrument for 5 days or until bacterial growth was detected. All negative aerobic bottles were weighed again and the volume of the blood cultured was calculated using the following formula: [volume = (weight of filled bottle − weight of empty bottle)/density of blood (1.055 g/ mL)]. The patient’s hematocrit at the time of blood culture puncture was recorded. First, we evaluated the phlebotomy skills of the laboratory phlebotomy team and the nursing departments. As the BVMS does not provide the volume for each individual bottle, the mean volume for both the phlebotomy team and the three nursing departments was compared with the mean weightbased volume for both groups. Second, the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a key performance indicator (KPI) of 8–12 mL. We also investigated a possible relationship between overfilling and false-positivity. Finally, we evaluated the mean BVMS volume of all negative aerobic bottles between 01/10/2011 and 27/06/2013 for each nursing department that included ≥25 bottles. PART 2 – Because of a repeatedly reported, though unpublished, inaccuracy in the calculation algorithm, BD developed a new software (version 4.30) which was installed in our laboratory on 02/01/2014. This software was validated by repeating our experiments on 350 aerobic bottles, which were again distributed amongst the laboratory phlebotomy team and the
three selected nursing departments. Again, the mean BVMS volume of the negative aerobic bottles was compared with the mean weight-based volume, the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, and the mean BVMS volume of all negative aerobic bottles between 14/02/2014 and 22/05/2014 for each department that included ≥25 bottles was calculated. PART 3 – As BD has recently developed plastic blood culture bottles to replace the current glass bottles, a limited validation experiment on 200 plastic aerobic bottles was performed. All bottles were distributed amongst the laboratory phlebotomy team. Once again, the mean BVMS volume of the negative aerobic bottles was compared with the mean weightbased volume and the weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled. As the plastic and glass bottles were used interchangeably, no retrospective investigation on a large number of plastic bottles could be carried out. During the study course, no feedback on the bottle filling degrees was given to the laboratory phlebotomy team nor the nursing departments.
Results PART 1 – We compared the mean weight-based volume of 339 negative aerobic bottles with the volume calculated by the BVMS. For the laboratory phlebotomy team, the mean weight-based volume was 9.9 mL [95 % confidence interval (CI): 9.5–10.4] and the mean BVMS volume was 8.8 mL (95 % CI: 7.6–10.0); this is an absolute deviation of −1.1 mL or −12.2 %. For the nursing departments, the mean weightbased volume was 9.8 mL (95 % CI: 8.9–10.7) and the mean BVMS volume was 8.1 mL (95 % CI: 7.5–8.7), an absolute deviation of −1.7 mL or −18.6 % (Fig. 1 – PART 1). The mean absolute deviation for both groups was −1.4 mL or −15.4 %. The weight-based volume of each individual bottle from the first experiment was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. The laboratory phlebotomy team drew 203 aerobic blood cultures, of which 105 (51.7 %) were correctly filled (8–12 mL), 48 (23.7 %) were filled 12 mL. The selected nursing departments drew 136 aerobic blood cultures, of which 46 (33.8 %) were correctly filled (8– 12 mL), 53 (39.0 %) contained 12 mL. When we excluded the patients with a hematocrit of 30 %, of which 42 (33.3 %) were correctly filled (8– 12 mL), 51 (40.5 %) contained 12 mL (Table 1 – PART 1). Within the 339
Eur J Clin Microbiol Infect Dis Fig. 1 Mean weight-based volume (♦) and mean blood volume monitoring system (BVMS) volume (▲) ±95 % confidence interval (―) of the negatively flagged aerobic bottles in each part of the experiment for drawings by the nursing departments and the laboratory phlebotomy team. a PART 1 – nursing departments; b PART 1 – phlebotomy team; c PART 2 – nursing departments; d PART 2 – phlebotomy team; e PART 3 – phlebotomy team)
included bottles, one was flagged positive, though no bacteria were seen nor grown. This bottle was slightly overfilled (12.1 mL). The anaerobic bottle, however, became positive with Staphylococcus aureus. Between 01/10/2011 and 27/06/2013, 8,991 aerobic blood culture bottles were incubated in and reported negative by the BD BACTEC FX instrument. The BVMS can calculate a mean volume for all those bottles together or split up by nursing department or phlebotomist whenever ≥25 bottles are included in each category. In the current setting, the identification of the individual phlebotomist is not yet registered in the EpiCenter module. This is why we evaluated the filling degree of the bottles for each nursing department, using the KPI of 8– 12 mL. In the selected period, 28 departments drew ≥25 bottles (total n=8,834). The mean BVMS volume was within the KPI for 18/28 (64.3 %) departments. The remaining 10/28 (35.7 %) departments all had a mean BVMS volume below the KPI. Taking into account the systematic deviation of BVMS volume software version 4.20B (−15.4 %), recalculation of these volumes resulted in only 3/28 (10.7 %) departments with a mean BVMS volume below the KPI, whilst the remaining 25/28 (89.3 %) departments all had a mean BVMS volume within the KPI (Table 2 – PART 1). PART 2 – The mean weight-based volume of 287 negative aerobic bottles was compared with the volume calculated by the BVMS. For the laboratory phlebotomy team, the mean weight-based volume was 9.9 mL (95 % CI 8.9–10.8) and the mean BVMS volume was 9.3 mL (95 % CI: 9.7–10.5); this is an absolute deviation of −0.6 mL or −5.9 %. For the nursing departments, the mean weight-based volume was 10.2 mL (95 % CI: 9.7–10.6) and the mean BVMS volume was 10.1 mL (95 % CI: 9.7–10.5), an absolute deviation of −0.1 mL or −0.8 % (Fig. 1 – PART 2). The mean absolute deviation for both groups was −0.3 mL or −3.3 %.
The weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. The laboratory phlebotomy team drew 187 aerobic blood cultures, of which 97 (51.9 %) were correctly filled (8–12 mL), 38 (20.3 %) were filled 12 mL. The selected nursing departments drew 100 aerobic blood cultures, of which 42 (42.0 %) were correctly filled (8–12 mL), 33 (33.0 %) contained 12 mL. When we excluded the patients with a hematocrit of 30 %, of which 28 (41.8 %) were correctly filled (8–12 mL), 26 (38.8 %) contained 12 mL (Table 1 – PART 2). Within the 287 included bottles, none were falsely flagged positive. Between 14/02/2014 and 22/05/2014, 1,557 aerobic blood culture bottles were incubated in and flagged negative by the BD BACTEC FX instrument. Split up by nursing department, 17 departments drew ≥25 bottles (total n=1,350). The mean BVMS volume was within the KPI for 15/17 (88.2 %) departments. The two other departments had a mean BVMS volume below or above the KPI (Table 2 – PART 2). PART 3 – The mean weight-based volume of the 173 negative aerobic plastic bottles was 9.1 mL (95 % CI: 8.7–9.5), whilst the mean BVMS volume was 11.9 mL (95 % CI: 11.5– 12.4). This is an absolute deviation of +2.8 mL or +26.7 %. The weight-based volume of each individual bottle was evaluated as correctly filled, underfilled, or overfilled, using a KPI of 8–12 mL. Of the 173 aerobic blood cultures, 100 (57.8 %) were correctly filled (8–12 mL), 52 (30.1 %) were
33.3 % 40.5 % 26.2 % 100 % 41.8 % 38.8 % 19.4 % 100 %
%
Number of nursing departments
42 51 33 126 28 26 13 67 40.0 % 20.0 % 40.0 % 100 % 42.4 % 21.2 % 36.4 % 100 % 33.8 % 4 39.0 % 2 27.2 % 4 100 % 10 42.0 % 14 33.0 % 7 25.0 % 12 100 % 33 46 53 37 136 42 33 25 100
n n % n
%
Patients with hematocrit 1.5 mL. Because the estimation of blood volume is based on the initial blood background rate, BD also warns that bottles with a delayed entry into the BACTEC instrument (>24 h at 20–25 °C or >12 h at 35 °C) will also result in an RMSE >1.5 mL.
We evaluated the accuracy of the BVMS by comparing the mean weight-based volume of 339 aerobic bottles with the volume calculated by the BVMS. Data were interpreted per phlebotomy team (laboratory phlebotomy team vs. three included nursing departments). We noted a mean negative deviation of −1.4 mL (−15.4 %) between the mean weight-based volume and the BVMS volume in both groups. Given the high number of bottles included, we considered the mean deviation (−1.4 mL or −15.4 %) as a representative correction factor to be applied on the mean volumes measured by the BVMS. To amend for this deviation, BD developed a new software version, 4.30, which was installed and evaluated using 287 aerobic bottles. The negative deviation disappeared nearly completely, leaving an acceptable mean deviation of −0.3 mL (−3.3 %). A first validation of this amended software version performed in our laboratory on 173 plastic aerobic bottles resulted in an unacceptable deviation of +2.8 mL (+ 26.7 %). In response to this finding, BD has developed further adjustments on the software, which is currently in the final stages of the validation phase. The weight-based volume of the 339 bottles from the first part of the experiments was subjected to an individual analysis of the filling degree towards the KPI of 8–12 mL. Within the laboratory phlebotomy team, no less than 48.3 % of all bottles were inadequately filled, with an almost equal division between the under- and overfilled bottles. The nursing departments scored even worse, with 66.2 % inadequately filled bottles, of which the majority was underfilled. The 287 bottles from the experiments ran on the new software were subjected to the same analysis: 48.1 % of the bottles drawn by the laboratory phlebotomy team was inadequately filled (with a tendency to overfilling), compared to 58.0 % of the bottles drawn by the nursing departments (where mainly underfilling was observed). A correlation between overfilling and false-positivity could not been shown. These poor amounts of adequately filled bottles clearly demonstrates the need to monitor the filling degree actively, (re)educate the phlebotomists, and to observe trends on a routine basis. However, in the daily routine
Eur J Clin Microbiol Infect Dis
microbiology lab, it is not feasible to calculate the weightbased volume of each bottle, leaving the mean BVMS volumes for specific groups to monitor. The disadvantage of those mean values is that they are less sensitive because under- and overfilled bottles compensate each other, masking possible shortcomings. This is shown by calculating the mean BVMS volumes of the 339 and 287 bottles in the first two experiments. Neither for the laboratory phlebotomy team nor for the nursing departments could the BVMS detect a filling problem. The respective mean volumes in the first experiment (n=339) were 8.8 mL (95 % CI: 7.6–10.0) and 8.1 mL (95 % CI: 7.5–8.7), whilst in the second part (n=287), they were 10.1 mL (95 % CI: 9.7–10.5) and 9.3 mL (95 % CI: 8.6– 10.0). To bypass this limitation, histograms (Fig. 2) or boxplots (Fig. 3) can be automatically generated, on which deviations from the KPI can be detected visually. Histograms can be generated when at least 100 data are included in one category, whereas boxplots only need 25 data. In extension, we evaluated the BVMS volumes of all data collected between 01/10/2011 and 27/06/2013 from 28 nursing departments (software version 4.20B) and between 14/02/ 2014 and 22/05/2014 from 17 nursing departments (software version 4.30). Taking into account our own calculated correction factor for the first software version, only 3/28 (10.7 %) departments had a mean BVMS volume outside (below) the KPI. From the 17 nursing departments observed in the second period with the amended software version, only 2 (11.8 %) had a mean BVMS volume outside (below) the KPI. This confirms the observation that the mean BVMS volume, per se, is not adequate to monitor the filling degree of a department as a whole. Nevertheless, these values could be useful to monitor a single phlebotomist, as one would expect a constant pattern, either under- or overfilled. In conclusion, there is a distinct need to actively monitor the filling degree of blood culture bottles. The BVMS (software version 4.20B) calculated the volume of aerobic bottles with a mean deviation of −1.4 mL or −15.4 % as compared with the weight-based volumes. The software update (version 4.30) lead to an acceptable deviation to −0.3 mL or −3.3 % on the glass bottles. The current algorithm proves to be inadequate on plastic bottles, as we noted a deviation of +2.8 mL or +26.7 %. However, BD has developed further adjustments and an improved software update is to be expected soon. The use of mean BVMS volumes to monitor the skills within a group of different phlebotomists can mask a potential filling problem. However, together with visual interpretation of histograms and/or boxplots produced by the BVMS, it is considered a useful tool to guide quality improvement policies. Acknowledgments The authors would like to thank Cindy Germis, Liesbeth Van Acker, Caroline Pieters, and Line Coucke, coworkers in the microbiology lab of the AZ Sint-Lucas Hospital, for their outstanding technical assistance in the experiments.
References 1.
2.
3.
4.
5. 6.
7.
8.
9.
10.
11. 12. 13.
14.
15.
16.
17.
18.
Rodríguez-Créixems M, Alcalá L, Muñoz P, Cercenado E, Vicente T, Bouza E (2008) Bloodstream infections: evolution and trends in the microbiology workload, incidence, and etiology, 1985–2006. Medicine (Baltimore) 87:234–249 Engel C, Brunkhorst FM, Bone HG, Brunkhorst R, Gerlach H, Grond S, Gruendling M, Huhle G, Jaschinski U, John S, Mayer K, Oppert M, Olthoff D, Quintel M, Ragaller M, Rossaint R, Stuber F, Weiler N, Welte T, Bogatsch H, Hartog C, Loeffler M, Reinhart K (2007) Epidemiology of sepsis in Germany: results from a national prospective multicenter study. Intensive Care Med 33:606–618 Suetens C (2008) Surveillance of nosocomial bloodstream infections in Belgian hospitals: results from the national surveillance network, 1992–2004 and 2007 (preliminary). Scientific Institute of Public Health, Brussels York MK, Henry M, Gilligan P (2010) Section 3.4.1. Aerobic bacteriology, blood cultures, general detection and interpretation. In: Isenberg HD (ed) Clinical microbiology procedures handbook. American Society of Microbiology (ASM) Press, Washington, DC Riedel S, Carroll KC (2010) Blood cultures: key elements for best practices and future directions. J Infect Chemother 16:301–316 Willems E, Smismans A, Cartuyvels R, Coppens G, Van Vaerenbergh K, Van den Abeele AM, Frans J; Bilulu Study Group (2012) The preanalytical optimization of blood cultures: a review and the clinical importance of benchmarking in 5 Belgian hospitals. Diagn Microbiol Infect Dis 73:1–8 Bouza E, Sousa D, Rodríguez-Créixems M, Lechuz JG, Muñoz P (2007) Is the volume of blood cultured still a significant factor in the diagnosis of bloodstream infections? J Clin Microbiol 45:2765– 2769 Cockerill FR 3rd, Wilson JW, Vetter EA, Goodman KM, Torgerson CA, Harmsen WS, Schleck CD, Ilstrup DM, Washington JA 2nd, Wilson WR (2004) Optimal testing parameters for blood cultures. Clin Infect Dis 38:1724–1730 Clinical and Laboratory Standards Institute (CLSI) (2007) Principles and procedures for blood cultures; Approved guideline. CLSI document M47-A. CLSI, Wayne, PA, pp 1–53 Cohen J, Brun-Buisson C, Torres A, Jorgensen J (2004) Diagnosis of infection in sepsis: an evidence-based review. Crit Care Med 32: S466–S494 Lassmann B, Gustafson DR, Wood CM, Rosenblatt JE (2007) Reemergence of anaerobic bacteremia. Clin Infect Dis 44:895–900 Hecht DW (2006) Anaerobes: antibiotic resistance, clinical significance, and the role of susceptibility testing. Anaerobe 12:115–121 Bannister ER, Woods GL (1995) Evaluation of routine anaerobic blood cultures in the BacT/Alert blood culture system. Am J Clin Pathol 104:279–282 Horvath LL, George BJ, Hospenthal DR (2007) Detection of fifteen species of Candida in an automated blood culture system. J Clin Microbiol 45:3062–3064 Morris AJ, Wilson ML, Mirrett S, Reller LB (1993) Rationale for selective use of anaerobic blood cultures. J Clin Microbiol 31: 2110–2113 Miller JM (1998) A guide to specimen management in clinical microbiology, 2nd edn. American Society of Microbiology (ASM) Press, Washington, DC Thompson RB Jr, Miller JM (2007) Specimen collection, transport and processing: bacteriology. In: Murray PR, Baron EJ (eds) Manual of clinical microbiology, 9th edn. American Society of Microbiology (ASM) Press, Washington, DC, pp 291–310 James PA, Al-Shafi KM (2000) Clinical value of anaerobic blood culture: a retrospective analysis of positive patient episodes. J Clin Pathol 53:231–233
Eur J Clin Microbiol Infect Dis 19.
van Ingen J, Hilt N, Bosboom R (2013) Education of phlebotomy teams improves blood volume in blood culture bottles. J Clin Microbiol 51:1020–1021 20. Wilson ML, Harrell LJ, Mirrett S, Weinstein MP, Stratton CW, Reller LB (1992) Controlled evaluation of BACTEC PLUS 27 and Roche Septi-Chek anaerobic blood culture bottles. J Clin Microbiol 30:63–66 21. Weinstein MP, Mirrett S, Wilson ML, Reimer LG, Reller LB (1994) Controlled evaluation of 5 versus 10 milliliters of blood cultured in aerobic BacT/Alert blood culture bottles. J Clin Microbiol 32: 2103–2106 22. Koontz FP, Flint KK, Reynolds JK, Allen SD (1991) Multicenter comparison of the high volume (10 ml) NR BACTEC PLUS system and the standard (5 ml) NR BACTEC system. Diagn Microbiol Infect Dis 14:111–118
23.
24.
25. 26.
Weinstein MP, Mirrett S, Reimer LG, Wilson ML, Smith-Elekes S, Chuard CR, Joho KL, Reller LB (1995) Controlled evaluation of BacT/Alert standard aerobic and FAN aerobic blood culture bottles for detection of bacteremia and fungemia. J Clin Microbiol 33:978– 981 Mirrett S, Hanson KE, Reller LB (2007) Controlled clinical comparison of VersaTREK and BacT/ALERT blood culture systems. J Clin Microbiol 45:299–302 Reimer LG, Wilson ML, Weinstein MP (1997) Update on detection of bacteremia and fungemia. Clin Microbiol Rev 10:444–465 Jungkind DL, Thakur M, Dyke J (1989) Evidence for a second mechanism of action of resin in BACTEC NR 16A aerobic blood culture medium, abstr. C-225, p 430. In: Abstracts of the 89th Annual Meeting of the American Society for Microbiology 1989. American Society for Microbiology (ASM), Washington, DC
