Article Text

Download PDFPDF

A prospective quality improvement study in the emergency department targeting paediatric sepsis
  1. Elliot Long1,2,3,
  2. Franz E Babl1,2,3,
  3. Eleanor Angley3,
  4. Trevor Duke2,3,4
  1. 1Department of Emergency Medicine, The Royal Children's Hospital, Parkville, Victoria, Australia
  2. 2Murdoch Children's Research Institute, Parkville, Victoria, Australia
  3. 3Department of Pediatrics, Faculty of Medicine, Dentistry, and Health Sciences, University of Melbourne, Melbourne, Victoria, Australia
  4. 4Pediatric Intensive Care Unit, The Royal Children's Hospital, Parkville, Victoria, Australia
  1. Correspondence to Dr Elliot Long, Department of Emergency Medicine, The Royal Children's Hospital, 50 Flemington Road, Parkville, VIC 3052 Australia; Elliot.long{at}rch.org.au

Abstract

Objective Quality improvement sepsis initiatives in the paediatric emergency department have been associated with improved processes, but an unclear effect on patient outcome. We aimed to evaluate and improve emergency department sepsis processes and track subsequent changes in patient outcome.

Study design A prospective observational cohort study in the emergency department of The Royal Children's Hospital, Melbourne. Participants were children aged 0–18 years of age meeting predefined criteria for the diagnosis of sepsis. The following shortcomings in management were identified and targeted in a sepsis intervention: administration of antibiotics and blood sampling for a venous gas at the time of intravenous cannulation, and rapid administration of all fluid resuscitation therapy. The primary outcome measure was hospital length of stay.

Results 102 patients were enrolled pre-intervention, 113 post-intervention. Median time from intravenous cannula insertion to antibiotic administration decreased from 55 min (IQR 27–90 min) pre-intervention to 19 min (IQR 10–32 min) post-intervention (p≤0.01). Venous blood gas at time of first intravenous cannula insertion was performed in 60% of patients pre-intervention vs 79% post-intervention (p≤0.01). Fluids were administered using manual push-pull or pressure-bag methods in 31% of patients pre-intervention and 84% of patients post-intervention (p≤0.01). Median hospital length of stay decreased from 96 h (IQR 64–198 h) pre-intervention to 80 h (IQR 53–167 h) post-intervention (p=0.02). This effect persisted when corrected for unequally distributed confounders between pre-intervention and post-intervention groups (uncorrected HR: 1.36, 95% CI 1.04 to 1.80, p=0.02; corrected HR: 1.34, 95% CI 1.01 to 1.80, p=0.04).

Conclusions Use of quality improvement methodologies to improve the management of paediatric sepsis in the emergency department was associated with a reduction in hospital length of stay.

  • Audit
  • Resuscitation

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic

  • Guidelines for the initial management of paediatric sepsis have not been widely incorporated into standard practice.

  • Sepsis quality improvement initiatives improve adherence to sepsis guidelines.

  • The effect of quality improvement initiatives on patient-centred outcomes remains unclear.

What this study adds

  • Quality improvement initiatives targeting paediatric sepsis are associated with reduced hospital length of stay.

  • Larger studies are required to assess the effect on mortality.

Introduction

Paediatric sepsis resuscitation has been informed by several published guidelines.1 ,2 Despite the lack of high-quality evidence, these guidelines are the framework for Pediatric Advanced Life Support and Advanced Paediatric Life Support courses. Additionally, though no feasibility data exist, their time-based interventions constitute a standard of care upon which practice is benchmarked. Compliance with these guidelines has been poor, with 19% of patients in paediatric emergency departments (EDs)3 and 38% in paediatric intensive care units (ICUs)4 receiving care that complied with the guidelines. As a result, their impact on patient outcomes is uncertain. Nevertheless, quality improvement (QI) strategies aimed at improving compliance with these guidelines have been studied in the paediatric ED and have resulted in improved process and outcome measures.5–7

Over a decade ago, a landmark randomised controlled trial of protocolised sepsis care in adults demonstrated improved survival in patients receiving early goal-directed therapy (EGDT).8 EGDT became incorporated into standard care through the bundled approach advocated by the Surviving Sepsis Campaign (SSC)2 and by standards endorsed by the National Quality Forum. It remained unclear, however, which specific components of EGDT were responsible for any observed survival advantage. Subsequently, in large multicentred randomised trials, EGDT has not been shown to confer any survival advantage compared with current standard care.9–12 The incorporation of early identification, early antibiotic administration and judicious targeted fluid resuscitation into standard care is thought to be responsible for the majority of this lack of observed benefit from further protocolisation in adults.13–15 If these basic interventions are given in a timely manner, adding other interventions in a protocol may make minimal difference.14 It is unclear whether these findings are generalisable to paediatric sepsis management, where EGDT and protocolisation have never been evaluated in a prospective or randomised fashion.

We aimed to prospectively evaluate ED sepsis management using published sepsis guidelines as a gold standard. Barriers to adherence with these guidelines were identified, a local guideline was developed and educational interventions introduced using the LEADER approach advocated by the SSC.2 The impact of these interventions on changes in processes, outcome and balancing measures (as defined by the Institute for Healthcare Improvement) was evaluated.16 The study was designed and reported in accordance with the SQUIRE guidelines.17

Methods

Design and setting

Prospective observational cohort study of children presenting to a tertiary ED (The Royal Children's Hospital (RCH), Melbourne, Australia) with sepsis before and after interventions aimed at improving compliance with sepsis guidelines (pre-intervention cohort February–July 2012, post-intervention cohort February–July 2014). RCH is a 340-bed tertiary paediatric hospital with an annual ED census of >85 000.

Planning the intervention

We became aware of quality gaps in ED sepsis management through sentinel events reviewed during monthly departmental morbidity and mortality meetings. Recurring delays in antibiotic administration in patients with sepsis seemed to be associated with adverse patient outcomes, venous blood gas was not routinely ordered in septic patients and intravenous fluid boluses were often administered through an infusion pump. We designed a prospective clinician-driven cohort study to determine whether these quality gaps were isolated events or part of a more systematic problem.

Over a 6-month period, children aged 0–18 years meeting predefined clinical criteria for the diagnosis of sepsis (see definitions) on arrival to the ED were identified by their treating clinician for inclusion. Patients not meeting clinical criteria but whose senior treating clinician judged the patient to have a clinical diagnosis of sepsis were included on an intention-to-treat basis. Laboratory tests (including venous blood lactate), retrospectively applied definitions (such as international consensus definitions)18 and discharge diagnoses were not used for patient identification. Patients whose goals of therapy were palliative and those with structural congenital heart disease with residual defects were excluded. We collected demographic information, microbiological diagnosis, time to initial physician assessment, time to intravenous access, time to fluid resuscitation therapy (FRT), method of delivery of FRT, time to antibiotic administration, total volume of FRT administered, disposition and 28-day mortality for all patients.

Through analysis of this cohort, we identified systematic delays in delivering critical interventions during early sepsis management in the ED. In particular, we found that antibiotic administration was delayed, FRT delivery was often delayed and was administered using an infusion pump, and that initial laboratory tests often did not include a venous blood gas. Other processes were performed well, including early recognition and appropriate triage, early physician assessment, performance of blood cultures on intravenous cannulation and appropriate antibiotic administration.

Using focused discussion groups among senior medical and nursing staff, we targeted early antibiotic administration, timely FRT delivery and venous blood gas as an initial laboratory test for our QI interventions. Key to achieving any sustainable change was thought to be education and empowerment of nurses. Nursing leadership and education staff were therefore involved in development and delivery of QI interventions from the outset. A local guideline was developed through a Clinical Practice Guideline committee, comprising representatives from state-wide critical care and medical teams. Reminder systems for clinicians were developed in the form of printed lanyard cards, smart phone applications and online publication (see online supplementary appendix 1). Education was developed in the form of sepsis workshops, presentations, bedside teaching and simulation-based initiatives (see online supplementary appendices 2 and 3) (figure 1).

Figure 1

Quality improvement interventions undertaken to improve early sepsis management in the paediatric emergency department.

Planning the study of the intervention

A before-after study design was used to monitor changes in targeted process measures. A time gap of one and a half years was allowed to enable development and implementation of interventions, to remediate errors and overcome obstacles, and for process measures to be taken up as part of routine care. Data from the post-intervention cohort were recorded over the same calendar months (February–July) as the pre-intervention cohort to allow for seasonal changes in disease frequency and severity. It was not possible to calculate a study sample size because the effect of the study interventions on process and outcome measures was unknown. Inclusion criteria for the post-intervention cohort were the same as for the pre-intervention cohort. We targeted three process measures and monitored the effect of changes in process measures on two outcome measures and one balancing measure. Process measures included time to intravenous antibiotic administration, rapid administration of FRT and measurement of venous blood gas and lactate on initial intravenous cannulation. The outcome measures were hospital length of stay (LOS) and 28-day mortality. The balancing measure was neonatal hospital LOS, in the event that antibiotic administration prior to completing a full septic screen (including lumbar puncture and sterile urine sample collection) resulted in diagnostic uncertainty and delayed discharge.19

Definitions

Sepsis was defined as systemic inflammatory response syndrome (SIRS) plus suspected infection requiring FRT. SIRS was defined as heart rate or respiratory rate above threshold values for age.18 ,20 Suspected infection was defined as fever (>38°C) or hypothermia (<36°C). FRT was defined as ≥20 mL/kg of crystalloid or colloid delivered to treat signs of poor perfusion or hypotension. Rapid administration of FRT was defined by method of administration (via pressure bag or push-pull method, not via infusion pump). Time zero (the time from which process measures were referenced) was defined as time of ED presentation. Time of antibiotic administration was defined as the time when nursing staff signed off on the antibiotics having been given. Hospital bed occupancy cost was based on 2013/2014 average daily costs for a medical patient in a general ward bed unadjusted for inflation. Comorbidities were defined by Feudtner codes.21 Triage category was defined according to the Australian Triage Scale.

Analysis

Data were collected prospectively by the treating clinician and entered into a Microsoft Excel 2010 (Microsoft, Washington, USA) database by study investigators. Descriptive statistics were used for key proportions. Relevant categorical data were compared using χ2 testing, continuous data were compared using t tests for parametric and Wilcoxon rank-sum tests for non-parametric data. Cox regression analysis was performed by a statistician. STATA V.14 (StataCorp, 2015, Stata Statistical Software: Release 14, College Station, Texas) was used for statistical analysis.

Results

Demographics

In total, 102 patients were included in the pre-intervention cohort (February–July 2012) and 113 patients in the post-intervention cohort (February–July 2014). The median age was older in the pre-intervention (1.9 years, IQR 0.2–6.8) compared with the post-intervention cohort (0.8 years, IQR 0.1–5) (table 1). The two cohorts were similar in terms of sex and presence of comorbidities. Also, 7 out of 102 (7%) patients pre-intervention and 3 out of 113 (2%) patients post-intervention were diagnosed with sepsis based on clinician concern only and did not meet predefined criteria for sepsis diagnosis. All patients included in the study were admitted to hospital.

Table 1

Demographic information for patients in pre-intervention (2012) and post-intervention (2014) cohorts

Process measures

The time from arrival to initial physician assessment was similar in both cohorts (table 2). There was a significant reduction in median time to intravenous access, median time to intravenous antibiotic administration and median time to intravenous fluid administration in the post-intervention cohort.

Table 2

Process measures for children presenting to the emergency department with sepsis pre-intervention (2012) and post-intervention (2014)

Outcome measures

Hospital LOS was significantly shorter in the post-intervention cohort, though mortality was similar (table 3). Similar total volumes of bolus fluid were administered in both cohorts.

Table 3

Outcome measures pre-intervention (2012) and post-intervention (2014) in children presenting to the emergency department with sepsis

The HR for reduced hospital LOS in the post-intervention group was 1.36 (95% CI 1.04 to 1.80; p=0.02). Using multivariate Cox regression analysis to correct for patient age, number of patients with comorbidity, number of patients presenting with central venous access devices, number of patients with hypotension on arrival, number of patients with an initial venous lactate >4 mmol/dL and the number of patients with positive bacterial culture from a sterile site between pre-intervention and post-intervention groups, the HR for reduced hospital LOS remained unchanged (corrected HR: 1.34, 95% CI 1.02 to 1.8; p=0.04).

Balancing measures

Subgroup analysis of neonates (<28 days of age) revealed that there was no significant change in hospital LOS between pre-intervention and post-intervention cohorts (median 68 h IQR 60–108 h pre-intervention vs median 67 h IQR 51–120 h post-intervention; p=0.16).

Discussion

This QI initiative found that sepsis management in the paediatric ED was amenable to improvements in process. Reduction in time to antibiotic administration, rapid administration of FRT and measurement of venous blood gas on cannulation were associated with a reduction in hospital LOS. These improvements were achieved after implementation of a local guideline and education programme. No harms from process improvements were identified.

Baseline mortality in both pre-intervention and post-intervention cohorts was low (<2%). As an outcome measure, therefore, the study was not sufficiently powered to demonstrate a survival advantage through QI interventions. Our finding of reduced hospital LOS following QI interventions was an intermediate measure of benefit only. Based on an average daily ward bed occupancy cost of $1350, this inexpensive intervention may result in substantial annual hospital cost savings.

This study compares favourably to published ED QI projects for the initial management of sepsis in children. Three retrospective studies have reported improved process measures through avoidance of the use of infusion pumps for FRT, standardised order sets and the use of locally modified sepsis protocols.5–7 Use of a pressure bag, rapid infuser or manual push has been associated with improved sepsis guideline adherence.3 Standardised order sets have been shown to reduce variability in practice and hasten the delivery of time-sensitive sepsis interventions.5 ,7 Paediatric sepsis protocols have been associated with improved process measures such as time to intravenous antibiotic administration and time to intravenous fluid administration,5–7 ,22 ,23 and have been associated with improved outcome measures such as an increase in the number of cases of septic shock between each death.7 This stands in contrast to the lack of benefit from protocolised sepsis management in large, randomised trials in adults.9–11 This discrepancy may result from poor baseline compliance with sepsis guidelines in paediatric ED's and ICU's,3 ,4 uncertainty regarding the clinical benefit of individual components of existing paediatric sepsis guidelines and a lack of feasibility data for time-based interventions advocated in existing paediatric sepsis guidelines.1

Neonates included in this QI initiative did not demonstrate a reduction in hospital LOS. Complexities in neonatal sepsis management outside of process measures addressed may have resulted in this lack of observed effect. These might include differences in disease aetiology, host response to infection and inpatient management.24–27 The neonatal cohorts both pre-intervention and post-intervention were small, and our lack of observed effect may also have resulted from type II error.

Interpretation

For individual departments as well as health systems, the implications of this and previous QI studies are several fold. First, defining the patient population is important. We used a pragmatic clinical definition for sepsis based on clinician intention to treat. Other studies have used retrospective diagnoses or included laboratory tests in their definition,5–7 making comparison between studies and direct translation to clinical practice difficult,28 and precluding prospective evaluation.29 ,30 For sepsis QI interventions to be taken up at a more national or global level, the target population needs to be more clearly defined. Second, the definition of time zero in our study was the time of patient arrival to the ED (as advocated by the SSC2 and paediatric sepsis QI reviews28). This allowed us to include pre-triage (clerking) time and triage time in our QI process. Previous studies have used a bedside timeout to declare time of sepsis onset.7 This makes direct comparison between studies difficult and benchmarking of QI processes inconsistent. A widely accepted definition of time zero would be helpful. Third, all published QI interventions for paediatric sepsis have included a locally modified sepsis guideline.5–7 This allowed individual departments to address local barriers to guideline adherence and may partly explain why departmental QI initiatives seem to have been associated with better adherence than sepsis guidelines introduced at a national level.3 ,4 Across studies, however, several QI processes have been universally targeted, including non-use of infusion pumps to deliver FRT and standardised order sets. Lastly, though sepsis QI interventions may be associated with improved outcome, the evidence underlying each intervention is lacking. As became clear when the individual components of EGDT were studied,9–11 QI interventions may target therapies that are not associated with patient benefit.

It was difficult to determine which of the study interventions were responsible for the observed reduction in hospital LOS. This was partly because the three targeted processes (early antibiotic administration, timely FRT delivery and venous blood gas as an initial laboratory test) were combined using a guideline and were therefore addressed simultaneously, making their individual contribution unclear. Reduced time to antibiotic administration may be the single most obvious quantifiable difference between the pre-intervention and post-intervention groups that may explain the observed reduction in hospital LOS. Although there was an improvement in timely administration of FRT in the post-intervention group, the total volume of intravenous fluid administered was not significantly different between groups. It was therefore not clear what the impact of timely FRT had in reducing hospital LOS. Lastly, the impact of improved lactate monitoring in the post-intervention group is unclear. Though it may have contributed to the recognition of illness severity in some children and been used to monitor response to treatment in those with an initially elevated lactate, evidence supporting its use for these purposes in children is limited.31–33

The main impact of the study intervention may have resulted from changes in departmental culture rather than from improvements in targeted processes.34 By embarking on the process of QI, there were several noted changes in departmental culture. The first was improved communication between medical and nursing staff. Medical staff were more likely to stay at the patient bedside during FRT because it was occurring in a timely fashion, encouraging frequent re-evaluation of patient condition and allowing nursing concerns to be voiced and addressed rapidly. Communication between nursing and senior medical staff was also improved because deviations from the sepsis guideline were expected to be escalated. This ‘collective mindfulness’ may have resulted in problems being reported when they were still small and easily addressed.35 The study intervention also created a shared mental model for sepsis care delivery between staff members. This meant that the expected course of management was reproducible and could be anticipated. Nursing staff were able to take the initiative in prompting medical interventions, flattening the medical and communication hierarchy and reinforcing the sense of teamwork. The creation of a departmental ‘safety climate’ has resulted in interest in improving care in areas other than sepsis. QI initiatives in airway management and several common paediatric illnesses have subsequently been undertaken.36 ,37

Limitations

As an uncontrolled before-and-after study, the results may be prone to confounding and bias.38 Pre-intervention and post-intervention cohorts were not identical in measured demographic variables (including age, proportion with indwelling central venous access device (CVAD), oncology patients) and may have been dissimilar in multiple unmeasured variables. A reliable indicator of disease severity has not been validated for use in the ED,28 and it was therefore difficult to directly compare cohorts. Nevertheless, correction for differential distribution of known confounders on regression analysis did not suggest that they were significant contributors to the observed reduction in hospital LOS in the post-intervention cohort. We reported multiple process measures and targeted three of these with our intervention. There may have been other unmeasured process measures that changed over the study period that may have contributed to our observed changes in outcome measures. This was thought to be unlikely, however, as our targeted process measures were similar to those identified in previous QI reports5–7 and in keeping with current understanding of important interventions in early sepsis management.13–15 Factors other than the study interventions may have changed over the study period. The ongoing, ward-based care of patients included in both pre-intervention and post-intervention cohorts was likely similar, however, because our QI interventions targeted only initial ED resuscitation. Reporting bias on the part of clinicians completing the research study forms may have resulted in apparent improvements in care over the study period. This was thought to be unlikely, however, as rotating junior medical staff were responsible for filling in the majority of forms and would have been different during the pre-intervention and post-intervention phases. Lastly, the sustainability of this intervention is not certain. The post-intervention cohort was studied 1.5 years after the pre-intervention cohort, to allow time for study interventions to achieve steady state, but ongoing monitoring will be required to ensure sustainability.

Conclusion

Implementation of a low-cost educational and QI initiative to improve sepsis management for children in the ED led to significant improvements in process measures and was associated with shorter hospital LOS.

Acknowledgments

The authors thank the ED clinical staff who participated in the study interventions. In particular, nursing educators were imperative to the success of the study.

References

Footnotes

  • Contributors EL conceptualised and designed the study, developed the study interventions, collected and analysed the data, drafted the original manuscript and approved the final manuscript as submitted. EL had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. FEB contributed to study design, reviewed and revised the original manuscript and approved the final manuscript as submitted. EA designed the data collection instrument, supervised data collection and entry, carried out the initial analyses, critically reviewed the original manuscript and approved the final manuscript as submitted. TD contributed to study design, helped develop the study interventions, reviewed and revised the original manuscript, and approved the final manuscript as submitted.

  • Funding This work was supported in part by a National Health and Medical Research Council Centre of Research Excellence Grant for Pediatric Emergency Medicine, grant number GNT1058560, and by the Victorian Governments Infrastructure Support Program, Melbourne, Australia.

  • Competing interests None declared.

  • Ethics approval Institutional Human Research and Ethics Committee at Royal Children's Hospital, Melbourne, Australia (HREC 31152A).

  • Provenance and peer review Not commissioned; externally peer reviewed.