Quality improvement project to improve blood culture volumes in haematology patients

Abstract

Sepsis resulting from bloodstream infection is a medical emergency, especially in immunocompromised haematology patients. Identification of causative pathogens through blood cultures is crucial for delivering effective antibiotics promptly.

Adequate blood culture volumes (BCV) are crucial for detecting bloodstream infections. Guidelines recommend 8–10 mL per blood culture bottle, yet achieving these volumes remains challenging. In our haematology wards, baseline mean BCV was suboptimal at 4.5 mL. This quality improvement project aimed to optimise BCVs to recommended levels within 6 months at a Singapore tertiary hospital’s Department of Haematology.

We implemented a multistakeholder quality improvement across three haematology wards from January 2023 to January 2025. Three Plan-Do-Study-Act cycles were executed from July 2024: (1) staff education and stakeholder engagement, (2) equipment and process enhancement and (3) validation of BCV measurements through comparison of an automated BCV measurement system (BD EpiCenter BACTEC Microbiology Data Management System) against manual bottle weighing.

Postintervention data demonstrated clear statistical signals of improvements in BCV through both EpiCenter automated system and manual bottle weighing. Manual validation in selected wards demonstrated achievement of target volumes. Importantly, this validation process revealed potential limitations of automated measurement systems in our specialised clinical setting. Overall, our results demonstrate that a well-coordinated, multidisciplinary approach combining staff education and engagement, improved equipment and BCV measurement process can successfully achieve recommended BCV in our complex haematology settings.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Blood culture volumes of 8–10 mL are recommended by guidelines for optimal detection of bloodstream infections.

• Inadequate blood culture volumes are a widespread issue across clinical settings.

• Multiple factors contribute to suboptimal collection, including equipment limitations, insufficient staff training and patient-related challenges.

WHAT THIS STUDY ADDS

• Demonstrates the effectiveness of integrating staff training, improved equipment and interdepartmental teamwork.

• Highlights the critical role of accurate measurements in specialised patient groups, particularly when using automated systems for patients with anaemia.

• Shows how sustainable quality improvements can be achieved in settings where multiple healthcare teams share blood culture collection duties.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Future studies may investigate whether improved blood culture volumes translate to better clinical outcomes and antimicrobial stewardship.

• Automated blood culture volume measurement methods may need to be optimised and validated, especially in specialised patient populations.

• This work could inform the development of institution-specific guidelines for blood culture collection that highlight optimal volumes.

Introduction

Bloodstream infections (BSI) represent a healthcare challenge, particularly among immunocompromised patients, where mortality rates range from 20% to 30%.¹,³ Blood culture (BC) remains the diagnostic gold standard for BSI, with blood culture volume (BCV) of 8–10 mL being the critical determinant factor of detection sensitivity.⁴ Research demonstrates that each additional millilitre of blood cultured increases BSI detection rates by 3%,⁵ highlighting how optimal BCV collection directly impacts pathogen identification and subsequent antimicrobial choices.

At Singapore General Hospital (SGH), a tertiary healthcare institution, the Department of Haematology serves as a leading centre for blood disorders and stem cell transplantation in Singapore. Through regular monthly mean BCV audits conducted by our Department of Microbiology, we identified consistently suboptimal mean BCV of 4.5 mL in our haematology wards, substantially below the recommended 8–10 mL volume. This finding aligns with published literature showing inadequate BCV collection as a widespread issue across various clinical settings, with multiple contributing factors including equipment limitations, insufficient staff training, time pressures and patient-related challenges.⁶,⁸ Our quality improvement project aims to identify and address these systematic challenges to optimise BCV in our vulnerable haematology patients.

We therefore initiated a quality improvement project to address this critical gap in our haematology inpatient service. Our specific aim was to increase the mean BCV to the guideline-recommended 8–10 mL per bottle through systematic improvements in BC collection practices among our healthcare staff.

Mission statement

To increase the mean BCV from 4 mL to 8–10 mL (representing a minimum 100% improvement) from patients with suspected infection in haematology wards within 6 months (June 2024–January 2025).

Materials and methods

Problem identification and root cause analysis

We used quality improvement tools to systematically analyse the challenges in BC collection. A root cause analysis using cause and effect diagram (figure 1) and Pareto chart (figure 2) identified key contributing factors, including equipment limitations (unavailable collection kits, unclear BC bottle volume markings), procedural issues (inadequate staff training and awareness) and clinical challenges (poor venous access, concern of needle stick injury during transfers).

Figure 1. Cause and effect diagram. BC, blood culture.

Figure 2. Pareto chart of root causes of inadequate blood culture (BC) volumes.

Study setting

This quality improvement initiative used a mixed-methods approach to evaluate the effectiveness of BCV improvement initiatives across three specialised adult haematology wards at SGH, with a preintervention period from January 2023 to June 2024, followed by post implementation monitoring from July 2024 to January 2025. This study included one stem cell transplant unit and two general haematology wards. The study population included all adult patients admitted to the haematology wards who required BCs as part of their clinical care.

Outcome measures

Our primary outcome measure was the monthly mean BCV collected from both peripheral venipuncture and central lines. Volume measurements were obtained using the BD EpiCenter Microbiology Data Management System, an automated system that is part of our hospital’s established protocol employed at our diagnostic microbiology laboratory for detecting microorganisms in BCs.

In essence, the BD EpiCenter specifically uses aerobic, culture-negative bottles to generate virtual BCV estimates. This automated system determines the mean BCV from batches of at least 25 culture-negative aerobic bottles by measuring red blood cell metabolic activity, then generates virtual BCV estimates based on red cell metabolic rates.^{9 10} Given that virtual BCV estimation depends on red blood cell metabolism, our measurement accuracy faced challenges in our haematology patients. Many of these patients have anaemia with reduced haematocrit levels (personal communications), which can lead to significant BCV underestimation. Previous studies have documented this limitation, showing a mean underestimation of 1.4 mL.^{9 10} To overcome this, we further adopted the bottle weighing method as previously published, which involved measuring the weight difference between empty and filled bottles as the gold standard for precise BCV measurement independent of haematocrit levels.⁵

To differentiate our outcome measures, we refer to the BCV estimated by BD EpiCenter BACTEC system as ‘virtual BCV’ and the volume measured by direct weighing as ‘actual BCV’.

Improvement strategy

Following baseline assessment, we undertook three Plan-Do-Study-Act (PDSA) cycles, summarised in figure 3.

Figure 3. Run chart depicting median BCVs (millilitre) from January 2023 to January 2025. The arrows indicate the initiation of interventions. The median BCVs increased after the interventions in July 2024. Postintervention median=5.8 mL. BCV, blood culture volume; PDSA, Plan-Do-Study-Act.

PDSA cycle 1: staff education and stakeholder engagement (July 2024–January 2025)

We implemented our intervention strategy in three distinct phases. In phase 1, we launched a comprehensive staff education programme addressing two distinct BC collection methods—peripheral venipuncture and central line sampling—as these require different protocols. We established stakeholder groups, including:

Clinical stakeholders:

• Nursing staff from haematology wards who are involved in BC collection.

• Medical officers and residents across haematology and multiple departments who are involved in BC collection.

• Department heads from microbiology and haematology.

• Infectious diseases specialists who provided expertise on BC, data analysis and as overall lead of the project.

• Advanced practice nurses (APN) who served as champions and trainers, and monitored compliance to implementing measures.

Non-clinical stakeholders:

• Quality improvement officers who assisted with project design and metrics.

• Hospital administration representatives who approved projects and resources.

• Laboratory technicians who processed and monitored BC specimens, collected data and evaluated effectiveness of interventions.

• Technical personnel who advised on BCV monitoring equipment.

The diverse stakeholder engagement enhanced both the design and implementation phases of our intervention. The educational component began with updating BC collection protocols with clear 8–10 mL volume requirements for both collection methods, with specific emphasis on sterile technique to prevent contamination and line infections. We conducted structured training sessions for nursing staff, with dedicated refresher sessions scheduled for new rotating doctors on starting their rotations into the Department of Haematology, led by APNs and infectious diseases specialists. Trained haematology nurses supervised doctors during central line BC collection to ensure proper sterile technique and sufficient BCV collection. We integrated these sessions into regular clinical teaching programmes and supplemented them with visual aids and posters displayed in wards and BC collection areas. Monthly meetings were established with key stakeholders from microbiology, haematology and infectious diseases teams to monitor compliance with BCV requirements, with monthly BCV data shared across departments to track progress and maintain engagement.

PDSA cycle 2: equipment and process enhancement (October 2024–January 2025)

In phase 2, we focused on equipment and process improvements. We transitioned from the existing 7-inch to an updated 12-inch tubing extension butterfly needle for peripheral BC collection from our existing vendor, incorporating improved flow characteristics. The longer tubing allowed upright bottle positioning for more precise BCV assessment, while the needleless transfer system provided safety features, and the specialised needle design enabled faster blood flow.¹¹ To ensure successful implementation, we conducted hands-on training sessions for all staff to familiarise them with the new device’s features and proper handling techniques and receive user feedback.

PDSA cycle 3: validation of BCV measurements through comparison of EpiCenter automated system against manual bottle weighing (November 2024–January 2025)

Phase 3 focused on validating our BCV measurement protocols for haematology patients. While continuing to use the BD EpiCenter system’s virtual BCV estimation for routine monitoring from aerobic culture-negative bottles, we recognised its potential measurement bias in patients with anaemia.^{9 10} Therefore, we initiated a validation study comparing EpiCenter’s virtual BCV readings with actual BCV obtained by manual weighing of the same types of bottles (aerobic culture negative) to ensure direct comparison. Due to the labour-intensive nature of manual weighing, we conducted this validation over a 1-month period in two representative wards—one stem cell transplant unit and one general haematology ward. This targeted dual measurement approach provided crucial insights into the accuracy of our BCV measurements and validated our improvement methods. Furthermore, we did not have actual BCV from the pre-implementation period for comparison because this was not part of our standard laboratory protocol.

Data analysis

Preintervention and postintervention changes in monthly mean BCVs are compared using basic descriptive statistics (mean, median) using Microsoft Excel. Run charts were used to visualise trends in mean BCVs over time and identify shifts in trends. Virtual BCVs were determined using EpiCenter software system’s automated analysis method, as previously described.⁹

Balancing measures

The following balancing measures were tracked to monitor some potential unintended consequences:

• Number of BCs performed: to ensure increase in BCV would not reduce the number of BCs performed.

• Number of positive BCs: to determine whether an increase in BCV would affect BC positivity.

• Number of contaminated BCs: to determine whether an increase in BCV would increase the number of contaminated cultures.

In our laboratory, BCs with growth of coagulase-negative Staphylococcus aureus, Micrococcus spp, Bacillus spp, Corynebacterium spp and Cutibacterium spp are considered to be contaminated.

Results

Initial measurements using the EpiCenter automated system showed baseline mean BCV of 4.0 mL (median 4.2 mL) during the 18-month preintervention period (figure 3). Following our intervention bundle implementation in July 2024, mean BCV increased to 5.8 mL (median 5.9 mL) over the subsequent 6 months. Run chart analysis demonstrated clear statistical signals of improvement, with consistent upward trends from baseline and reduced BCV variation (figure 3). There were no missing data during the preintervention and postintervention study periods.

While EpiCenter measurements suggested we fell short of our 8–10 mL target, validation through manual weighing revealed higher actual BCV. In the stem cell transplantation ward (4–30 November 2024), manual weight measurements showed mean actual BCV of 9.1 mL by weighing compared with EpiCenter’s virtual BCV readings of 7.2 mL. Similar findings in the general haematology ward (6 December 2024 to 6 January 2025) showed actual BCV of 8.0 mL by weighing versus virtual BCV readings of 7.3 mL by EpiCenter (online supplemental table 1). This systematic difference (mean 1.3 mL) between measurement methods suggested achievement of target BCV when accounting for the known underestimation bias in our anaemic population.

As balancing measures, we determined the number of BCs performed (online supplemental figure S1a), number of positive BC bottles (online supplemental figure S1b) and number of contaminated BCs (online supplemental figure S1c). Overall, the number of BCs performed preintervention and postintervention was similar. The number of contaminated BCs was low (1 bottle in May 2024, preintervention period; 1 bottle in December 2024, postintervention period).

Discussion

BCs are crucial for diagnosing life-threatening BSIs, particularly in immunocompromised patients where early detection significantly impacts outcomes. The Infectious Diseases Society of America and the Royal College of Pathologists UK guidelines both emphasise collecting adequate BCV (8–10 mL per bottle) before initiating antimicrobial therapy,^{12 13} as detection rates increase by 3% by millilitre of blood cultured.⁵ Despite this importance, many healthcare providers remain unaware of optimal collection requirements.^{6 7}

Our quality improvement initiative used a multifaceted approach combining staff education, technical improvements and interdepartmental collaboration to increase BCV in a complex clinical environment. The baseline mean BCV increased from 4.5 mL (median 4.4 mL) to 5.8 mL (median 5.9 mL) postintervention. Manual validation in selected wards revealed higher actual BCV than automated measurements by EpiCenter: the stem cell transplantation ward achieved 9.1 mL (vs 7.2 mL by EpiCenter) and general haematology ward reached 8.0 mL (vs 7.3 mL by EpiCenter). This systematic difference of mean 1.3 mL between measurement methods indicates achievement of target BCV when accounting for the known underestimation bias in our population of patients with anaemia. The discrepancy in BCV between these two methods aligns with a previous study that showed EpiCenter recordings were over 1 mL lower than actual BCV in patients with anaemia.¹⁰

The success of our intervention can be attributed to three main factors. First, our comprehensive approach addressed multiple barriers simultaneously through staff education, equipment upgrades and interdepartmental collaboration. Second, strong engagement from nursing staff, physicians and laboratory personnel ensured consistent implementation across different shifts and departments. Despite the regular rotation of medical officers every 3–4 months, our structured training programme and regular enforcement of protocols effectively maintained quality standards. Third, our dual measurement approach provided crucial validation of improvements, particularly important given our specialised patient population.

Our findings align with previous quality improvement studies in BC collection, though our work extends beyond single-team interventions to coordinate multiple professional groups, particularly given our reliance on both nursing staff and medical officers for BC collection rather than dedicated phlebotomists. This reflects the reality of many hospital settings where BC collection responsibilities are shared across teams, requiring robust coordination and standardisation of practices. The systematic differences we found between virtual and actual BCV measurements highlight the importance of validation methods in specialised populations, an aspect not extensively discussed in previous quality improvement projects.^{6 7}

Several limitations warrant discussion. First, resource constraints prevented actual BCV measurement from project initiation, limiting our ability to quantify baseline-to-endpoint improvements using weight measurements. Second, maintaining BC collection consistency across different staff and shifts posed significant challenges in our hospital setting, as the regular rotation of medical officers every 3–4 months necessitated continuous training programmes and competency assessments. This required considerable resources including dedicated educators for regular teaching sessions during each rotation cycle. Third, while our findings are promising for haematology settings, generalisability to other clinical areas requires careful evaluation due to differences in patient characteristics such as haemoglobin levels, venous access and blood collection practices that may vary across different specialties and institutions. Finally, while we identified systematic underestimation by our automated system in patients with anaemia, implementing continuous actual BCV measurements would strain available resources and ultimately, an improved automated system calibrated for patients with anaemia would be more sustainable long term.

For sustainability, our interventions have been formally incorporated into departmental policy through several key measures. We have updated educational materials specifically for BC collection procedures that emphasise proper volume collection. Additionally, we implemented structured ongoing staff education for both permanent staff and rotating junior doctors, in order to address the variations in adherence particularly among junior medical staff. We have also standardised collection equipment, including introduction of the 12-inch tubing extension butterfly needles for peripheral BCs, and implemented routine BCV monitoring as part of standard laboratory procedures. Collectively, these measures ensure the sustainability of our quality improvements beyond the initial study period.

Our future work will focus on evaluating downstream clinical benefits, including time to pathogen detection, optimisation of antimicrobial therapy and potential reduction in antibiotic courses—important areas for further research. We also aim to integrate these improvements with broader diagnostic stewardship initiatives to maximise their impact on patient outcomes.

Conclusions

This quality improvement project successfully increased BCV in our haematology wards through systematic intervention and strong interdepartmental collaboration. Key sustainable measures include ongoing education, improved collection equipment and regular monitoring with feedback. The project demonstrates the effectiveness of a multifaceted approach in improving clinical practice. Future work will evaluate the impact of these improvements on patient outcomes, including time to pathogen detection and appropriate antimicrobial therapy.

Data availability statement

Data are available upon reasonable request.