Changing culture: An intervention to improve blood culture quality in the emergency department

Blood cultures (BCs) remain a key investigation in ED patients at risk of bacteraemia. The aim of this study was to assess the effect of a multi‐modal, nursing‐led intervention to improve the quality of BCs in the ED, in terms of single culture, underfilling and contamination rates.


Introduction
Sepsis is a major public health burden, with increasing incidence and disproportionate impact on community health, hospital and intensive care unit resources. 1Blood cultures (BCs) remain a key investigation in sepsis, providing opportunities to identify causative micro-organisms and optimise antimicrobial therapy. 2 ED sepsis management guidelines recommend collection of BC prior to broad-spectrum antibiotics, as these significantly reduce the sensitivity of the test. 3,4Because antimicrobial administration is widely regarded a time-critical intervention in sepsis, 5 time and resources allocated to quality BC collection practices may potentially be diminished.
Guidelines recommend taking at least two BC sets from different sites, inoculation of 8-10 mL blood per bottle, and restricting contamination to <3% of all cultures. 6Quality of BC collection is a key determinant of clinical utilityvolume of blood cultured is proportional to yield of pathogens, and contamination has been associated with increased antibiotic use, hospital stays and costs. 7However a recent quality assurance review from Australasian microbiology laboratories revealed compliance with BC quality standards was monitored in Correspondence: Dr Julian Williams, Emergency and Trauma Centre, Royal Brisbane and Women's Hospital, Butterfield Street, Herston, QLD 4006, Australia.Email: ray4jay@hotmail.comJulian Williams, BSc, MBBS, PhD, Emergency Physician; Angela Hills, BN, Clinical Nurse; Mercedes Ray, BN, Clinical Nurse; Jaimi Greenslade, BPsych (Hons), BHlthSc (Paramedic), MSc (Stats), PhD, Principal Research Fellow, Associate Professor.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Accepted 21 September 2023
Key findings  10,11 The aim of the present study was to determine the effect of an intervention to improve the quality of BCs taken in an ED in terms of three quality indicators: single set, underfilling and contamination rates.The study hypothesis was that each of these indicators would be improved in the post-intervention period.

Study design and setting
This before-and-after intervention study was performed in the ED of a large tertiary adult hospital in Brisbane, Australia, with annual census over 82 000 presentations.The local research ethics committee sanctioned the study and deemed participant consent was not required (HREC/18/ QRBW/231).BCs were collected in standard BacT/ALERT ® (bioMerieux, Craponne, France) aerobic and anaerobic culture bottles, and a BC set was defined as an aerobic and anaerobic BC bottle collected from a single blood draw.Two BC sets from separate sites were recommended as per departmental and national guidelines. 12

Intervention
A multimodal intervention to improve quality of BC collection was designed, including small group education, posters, brief educational videos, social media presence, quality feedback, small group/individual mentoring and availability of BC collection kits containing all materials required (further detail is provided in Appendix S1).The intervention was delivered by two experienced emergency nurses (MR and AH) over 7 weeks and focused on reduction of contamination through appropriate skin antisepsis and technique, collection of at least two BC sets from each patient, and filling BC bottles adequately (optimally 10 mL).

Participants
All ED patients from whom BC were collected during routine management over the study period were identified retrospectively from the hospital microbiology database.Data characterising quality of BC collection in the ED (single set and contamination rates) were recorded over discrete study periods: pre-intervention (18

Data sources
Study nurses measured BC volumes using an incremented ruler designed for the purpose between 4 May 2019 and 21 June 2020, excluding the interval between post-intervention and sustainability phases (Fig. 1).Underfilled BC bottles were defined as those containing ≤7 mL; this effectively allowed a margin of up to 1 mL below the published 8-10 mL recommended range as previously described. 6,13sitive BCs were classified as either false positive (contaminant) or true positive (indicating bacteraemia) according to accepted methods 7,14 and specialist infectious disease input as documented in clinical records.For all study patients, data including demographics, presumed source of infection, key investigation results, disposition and mortality were obtained from relevant hospital records.All study data were de-identified and stored in a password-protected database.

Statistical methods
The study was powered to detect a reduction in contamination rates from 6% 9 to the internationally accepted baseline of 3%.Assuming power of 80% and alpha of 0.05, 748 cultures would be required in each of the preand post-intervention periods.This sample size was obtained in 18 weeks for the pre-intervention period, and so the post-intervention and sustainability periods were conducted over the same time.
Analyses were performed using Stata version 17 (StataCorp, College Station, TX, USA).Descriptive statistics were reported by study period for baseline characteristics of the study cohort, process outcomes (single sets and volume collected) and clinical outcomes (contaminants, positive BCs, length of stay, and mortality).Single sets, contaminants, positive BCs, length of stay and mortality were reported at the encounter level.Volumes, contaminants and positive cultures were reported at the culture level.Statistical comparisons between pre-and post-intervention were reported and calculated using chi-squared for categorical data or Wilcoxon rank-sum test for continuous data.P < 0.05 was considered significant.
For the outcome data, pre-and postintervention groups were compared.For categorical endpoints, general linear models were computed with a binomial error distribution.For continuous data, quantile regression was used to compare the medians.Clustered robust standard errors were used to account for non-independence due to the nested structure of the data.
Segmented regression analyses were conducted for each of the outcomes to ensure that differences in the outcome variables pre-and post-intervention were not attributable to downward trends across the entire study period.Within such analyses, each outcome variable was regressed on: (i) time (weeks from the beginning of data collection) to assess for trends before the intervention; (ii) time after implementation (weeks), to assess for a change in trend during the intervention; (iii) time after completion of the intervention (weeks), to assess for a change in trend after completion of the intervention; and (iv) an indicator variable for study period (pre-, peri-or post-intervention) to assess whether there was a quantitative change in the outcome on the start of each period.Analyses were conducted using restricted maximum likelihood with direct modelling of autocorrelation.

Results
Across the four study periods totalling 61 weeks (Fig. 1), data from 4908 BC sets collected during 2347 ED episodes of care were included in the study.Table 1 summarises clinical characteristics by episode of care for each study period.BCs were obtained from only 570 ED patients in the 18-week sustainability period compared to 765 and 734 patients in the 18-week pre-intervention and postintervention periods, respectively.There were no significant differences in the clinical characteristics of study patients pre-and post-intervention.There was no difference in the median ED length of stay pre-intervention (6.6 h [4.4-10.6]) to post-intervention (6.9 h [4.7-10.6],P = 0.27).Compared to pre-intervention, ED length of stay was slightly shorter in the sustainability period (5.0 h [3.6-7.1],P < 0.001).Throughout the study, there were no significant differences in the proportions of patients admitted to an inpatient unit (80.4% pre-intervention, 80.2% intervention, 82.4% post-intervention, 83.2% sustainability), or who died in hospital (4.2% pre-intervention, 4.6% intervention, 3.0% post-intervention and 3.0% sustainability).
The proportions of single sets and underfilled BCs were significantly reduced between pre-and postintervention periods (Table 2).Patients from whom a single BC was collected comprised 56.2% of the pre-intervention group, reducing to 22.8% and 18.8% of the postintervention and sustainability groups, respectively (P < 0.01).Rates of aerobic bottle underfilling were reduced from 52.8% to 19.2% between pre-and post-intervention phases, an effect maintained in the sustainability phase (18.8%).Similar improvements were seen in anaerobic bottle volumes.Skin contaminants were grown from 3.7% of BC sets in the pre-intervention period, improving to 1.5% in the postintervention period (P < 0.001).

Discussion
Following a 7-week multi-modal intervention delivered by two ED nurses, quality of BC taken in our busy tertiary adult ED significantly improved in terms of appropriate bottle filling, obtaining two or more culture sets, and reduced contamination rates.These improvements were sustained up to 9 months postintervention.Collection of at least two BC sets rather than one was associated with significantly increased true positivity rate and adequately filled BC bottles returned a higher true positive rate than inadequately filled bottles.When all study episodes of care were analysed, total volume of blood cultured was significantly associated with a diagnosis of bacteraemia.On the basis of expected values from the regression equation, improving mean volume collected per patient from the pre-intervention figure of 24 mL to an optimal 40 mL could result in up to 46 (95% CI 7-84) additional patients per year diagnosed with bacteraemia in our ED.This figure is based on an estimated 2000 patients each year undergoing BCs.
BCs were taken from only 570 patients during the 18-week sustainability period, compared to more than 700 in each of the 18-week preand post-intervention periods.This was in the context of reduced overall ED presentations during lockdowns for COVID-19 in early 2020.Point estimates for patient disposition locations and true positive rate during sustainability may suggest a slightly sicker cohort underwent BC testing in that period, however corresponding measures of disease severity and morbidity (e.g.MEDs and Charlson scores) were not significantly different.ED length of stay was significantly shorter in the sustainability period, most likely related to reduced ED access block associated with a moratorium on elective surgical admissions during COVID-19 lockdowns.
The aspect of BC quality that has received the most attention in previous literature is contaminationpotentially because this has been associated with patient harm. 7A broad range of interventions including diversion devices, education initiatives, phlebotomy teams, standardised BC collection kits and clinician feedback have been implemented across a variety of settings with associated reduction in BC contamination. 15Both the number of BC sets collected and volume in each bottle contribute to total blood volume inoculated, which is the critical determinant of BC sensitivity.][21] The incidence of solitary BC sets in the ED has been shown to range from 16.9% to >90% 9,22 and from a wider hospital perspective 48% to 94%. 23,24Despite this, published interventions to reduce solitary BC are few.Choi et al. intervened with education and a computerised forcing function to reduce solitary BC rates in the ED from 41.1% to 11.6%. 25nother education and feedback initiative saw solitary BC reduced from 33% to 26% across several medical wards. 26In the only prior study identified to assess and improve multiple aspects of BC quality in the ED, Shaji et al. delivered a multimodal intervention over 6 months, including introducing a standard protocol and education sessions. 27Compared to pre-intervention, the 6-month intervention was associated with reduced contamination (13.7% to 4.2%), less underfilled BC and single sets (63.5% to 20.5%), and increased percentage of true positive BC (from 11.9% to 17.9%).

Limitations
Our before and after study is inherently limited by inability to establish causal relationship between the intervention and observed improvements.Each study period occurred during different intervals in the calendar year, so results were potentially impacted by temporal changes in patient characteristics, department activity or staffing.However no significant differences were found between measured clinical characteristics of patients enrolled in each study period (Table 1), and preintervention rates of single set, contamination and true positive cultures were similar to those measured over the 2017 calendar year. 28Segmented regression analyses were consistent with our intervention changing practice.The observed relationship between total volume cultured and true positive rate must also be considered an association, as clinicians may have been less likely to collect optimal volumes or two sets of BC from patients believed to be a low risk for bacteraemia, and conversely for patients at higher risk.However, a causal relationship has been well established previously. 14The study sample size was calculated to detect a reduction in contamination rate from 6% to 3%.However significant reductions in single set, underfilling and contaminated BC rates were observed postintervention.As a single-site study, findings may not be generalisable to other settings.The intervention was led by and aimed predominantly at nurses because in our department nurses collect the majority of BC; it is appreciated that in other settings different clinicians (such as phlebotomists or physician assistants) may perform this role.Strengths of our study include delivery of a reproducible, multi-modal intervention by ED senior nurses with minimal external resources or funding.Key study measurements including numbers of ED patients cultured, single set and contamination rates, and bottle volume determination were robust.

Conclusions
Significant improvements in BC quality are possible through nursingled interventions in the ED.Given the importance of this test, BC quality metrics should be measured and reported along with other indicators of quality care for patients with sepsis.Further work should explore whether improvements are sustained over longer time periods and refine selection of patients who undergo BCs.

Figure 1 .
Figure 1.Study dates and intervals.The upper bar represents periods over which blood culture single set and contamination rate data were collected, the lower bar blood culture volumes.Numbers are weeks within each interval.

Figure 2 .
Figure 2. Regression of blood culture positivity on total volume collected in ED.Increased volume was associated with higher positivity (RR 1.14, 95% CI 1.02-1.26per 10 mL).Data from all 2347 study episodes of care.Total volume figures are in mL.CI, confidence interval; RR, risk ratio.

TABLE 1 .
Clinical characteristics by study period (episodes of care) Data are presented as median (IQR) for continuous measures, and n (%) for categorical measures.CCI, Charlson comorbidity index; MEDS, mortality in emergency department sepsis; SBP, systolic blood pressure.There were missing data for age (n = 1), lactate (n = 1440) and MEDS score (n = 3).Comparisons are between pre-and post-intervention periods.

TABLE 2 .
Process data, true positive and contamination rates by study period

TABLE 3 .
Association between process outcomes and true positive blood cultures