Impact of 1-hour and 3-hour sepsis time bundles on antibiotic use in emergency departments in Queensland, Australia: a before-and-after cohort study

Abstract

Objective

To determine if the introduction of an emergency department (ED) sepsis screening tool and management bundle affects antibiotic prescribing and use.

Design

Multicentre, cohort, before-and-after study design.

Setting

Three tertiary hospitals in Queensland, Australia (median bed size 543, range 520–742).

Participants

Adult patients, presenting to the ED with symptoms and signs suggestive of sepsis who had blood cultures collected. These participants were further assessed and stratified as having septic shock, sepsis or infection alone, using Sepsis-3 definitions. The study dates were 1 July 2017–31 March 2020.

Intervention

The breakthrough series collaborative ‘Could this be Sepsis?’ Programme, aimed at embedding a sepsis screening tool and treatment bundle with weighted-incidence syndromic combined antibiogram-derived antibiotic guidelines in EDs.

Main outcome measures

The primary outcome was the rate of empirical prescriptions adherent to antibiotic guidelines during the ED encounter. Secondary outcomes included the empirical prescriptions considered appropriate, effective antibiotics administered within 3 hours and assessment of harm measures.

Results

Of 2591 eligible patients, 721 were randomly selected: 241 in the baseline phase and 480 in the post-intervention phase. The rates of guideline adherence were 54.0% and 59.5%, respectively (adjusted OR (aOR) 1.41 (95% CI 1.00, 1.98)). As compared with baseline, there was an increase in the rates of appropriate antibiotic prescription after bundle implementation (69.9% vs 57.1%, aOR 1.92 (95% CI 1.37, 2.68)). There were no differences between the baseline and post-intervention groups with respect to time to effective antibiotics, adverse effects or ED rates of broad-spectrum antibiotic use.

Conclusion and relevance

The use of an ED sepsis screening tool and management bundle was associated with an improvement in the rates of appropriate antibiotic prescription without evidence of adverse effects.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Large, detailed dataset derived from integrated electronic medical records (ieMR) and automated medication dispensing machine data, allowing accurate time stamping and determination of sepsis status using Sepsis-3 definitions.

• Cohorts from multiple, geographically diverse healthcare services, but may not be generalisable to all health services.

• Several definitions of ‘quality’ of antibiotic prescribing, including those focusing on reducing unnecessarily broad-spectrum antimicrobial use (appropriateness), as well as considering microbiological effectiveness, were used to assess the impact of sepsis bundle implementation on antibiotic use.

• Limited as a before-and-after observational study with potential for bias by residual confounding.

• Evaluation of antibiotic usage was confined to a nested cohort predominantly among patients who had blood culture collection in sites with ieMR.

Introduction

Sepsis affects 55 million patients worldwide and causes 11 million deaths annually.¹ Institution of sepsis pathways to streamline the recognition, management and review of sepsis cases in hospital settings reduces sepsis mortality and morbidity.^{2 3}

The Surviving Sepsis Campaign (SSC) guidelines advocate time-based sepsis bundles to facilitate early recognition, timely antimicrobials, source control and supportive treatments. The original SSC guidelines in 2012 recommended a 3-hour (septic shock) and 6-hour bundle.² A 2018 update of the SSC guidelines combined the guidelines into a 1-hour bundle with the explicit intention of beginning resuscitation and management immediately.⁴ However, these bundles have been challenged due to the conflation of sepsis and septic shock management leading to the potential for unnecessary testing and increased demand on emergency department (ED) and intensive care unit (ICU) resources.³ There also are concerns of excess antimicrobial use with overprescribing of broad-spectrum antibiotics⁴ and increased complications of antibiotic use including evolution of multidrug resistance.⁵ The appropriateness of antimicrobial prescription as part of bundle implementation has not been evaluated in a systematic manner.

We recently reported that implementing an evidence-based sepsis bundle, using a Breakthrough Series Collaborative approach developed by the Institute for Healthcare Improvement,⁶ resulted in improved bundle compliance and a reduced need for ICU admission.⁷ We used a 1-hour bundle for septic shock and 3-hour bundle for sepsis and by using two time windows, we avoided conflation of sepsis and septic shock.

Method

The Breakthrough Series Collaborative ‘Could this be Sepsis?’ Programme was a before-and-after study launched by Queensland Health’s Clinical Excellence Queensland (CEQ) involving EDs in 14 major public hospitals. The programme, aimed at embedding a sepsis screening tool and treatment bundle (online supplemental figure 1) with weighted-incidence syndromic combined antibiogram⁸-derived antibiotic guidelines (online supplemental figure 2) in EDs, was conducted between 1 July 2017 and 31 March 2020. The detailed protocol and the results were recently published by our group.⁷

Participating hospitals

Three geographically representative hospitals (online supplemental table 1), using integrated electronic medical records (ieMR), were selected from the 14 hospitals which participated in the collaborative. These sites also employ networked Laboratory Information Services (AUSLAB) and, at two sites, electronic medication dispensing services (Pyxis) linked to a hospital-based corporate information system. Two of the sites were in major cities and the third was in an outer regional zone in the State of Queensland.⁹ These sites were chosen based on availability of electronic patient record systems and geographical diversity providing a broader infectious disease case-mix.

Inclusion and exclusion criteria

The inclusion and exclusion criteria are provided in detail in our earlier report.⁷ While the original collaborative study only enrolled patients with a confirmed bacteraemia, in this study we also included patients who had had a blood culture collected in the ED, irrespective of the results of the culture.

The criteria for inclusion were:

1. Adult patients (≥18 years old).

2. Presenting to the ED with symptoms and signs suggestive of an infection or fever or hypothermia or signs of clinical deterioration.

3. Who had blood cultures collected in the ED.

4. Who required admission to the hospital.

Patients were excluded if they met any of the following criteria: (a) <18 years old; (b) pregnant or post-delivery (up to 6 weeks); or (c) retrieved from other hospitals bypassing the ED and directly admitted to ICU at participating sites.

Cases meeting the above criteria for inclusion in the study were randomly selected and assessed using Sepsis-3 definitions,¹⁰ based on Sequential Organ Failure Assessment (SOFA) score determination and plasma lactate concentrations within 24 hours of presentation, and, in cases with hypotension, response to intravenous fluids. Three subgroups (referred to as sepsis status) were determined:

1. Infection only—SOFA score <2.

2. Sepsis without shock—infection present with SOFA score ≥2 without hypotension.

3. Septic shock—infection present with SOFA score ≥2 and lactate ≥2 mmol/L with refractory hypotension (systolic blood pressure <90 mm Hg or >40 mm Hg below normal, despite fluid resuscitation or requiring vasopressors).

Data collection

Data were collected from patient electronic medical records, Queensland laboratory information service (AUSLAB) records and automated drug dispensing systems (Pyxis), and entered into a secure online database, Research Electronic Data Capture, hosted by CEQ.¹¹ Data included patient demographic and clinical characteristics, comorbidities, colonisation by, and risk factors for, multidrug-resistant organisms, results of blood (and other significant) cultures, characteristics of antibiotic prescriptions (such as antibiotic choice, dose, duration, indication), complications of antibiotic treatment including Clostridioides difficile infection (up to 6 months post-admission) and details of post-antibiotic prescription review of treatment plans. Historical data, such as baseline renal function, were used to assign SOFA scores.

All prescriptions were assessed based on the diagnosed source of infection as documented in the ED medical records and sepsis status (based on clinical observations and laboratory results available) at the time the prescription was ordered (table 1). Antibiotic prescriptions were assessed according to all antibiotics simultaneously prescribed (antibiotic regimen) at the first ED clinical assessment. For example, a prescription for community-acquired pneumonia would require a combination of a macrolide and a beta-lactam to be prescribed together in the assessment. Data were collected by two antibiotic stewardship (AMS) pharmacists and one medical microbiologist (MM)/infectious diseases physician (IDP). Antibiotic ‘appropriateness’, owing to its subjectiveness, was assessed by a minimum of one pharmacist and one IDP. Case types with discordant assessments were presented to a group of five IDPs for review. Those still with discordant assessments after the IDP group review were considered non-assessable.

Table 1.

Antibiotic prescription assessment and definitions

Empirical antibiotics	The initial antibiotic regimen started in the ED for suspected infection/sepsis/septic shock before results of microbiological and other diagnostic tests available.
Antibiotic guideline adherence	Antibiotic guideline adherence for choice of antibiotic was based on the Queensland State-wide Antibiotic Prescribing Guidelines (online supplemental figure 2) integrated into the pathway document or Therapeutics Guidelines: Antibiotics.23
Antibiotic appropriateness	National Antimicrobial Prescribing Survey15 16 criteria were used for analysis of ‘appropriate’ and ‘optimal’ prescribing. Prescriptions were classified as ‘appropriate’ if they were consistent with local or national therapeutic guidelines or were reasonable alternative choices to the guidelines. Optimal prescriptions were a subset of appropriate prescriptions with optimal adherence to guidelines for choice, dose, route, frequency and allergy status, or, if there were no endorsed guidelines, the prescribed antibiotic was likely to cover the causative pathogens and there was not a narrower spectrum antibiotic regimen available.
Antibiotic effectiveness	Prescriptions were ‘effective’ if the susceptibility profile of organisms isolated from sterile sites (determined by VITEK2 (bioMérieux, Marcy l'Étoile, France)) and using EUCAST breakpoints24 matched the choice of empirical antibiotics delivered. ‘Time to effective antibiotics’ was determined as the time to the first antibiotic prescribed that was deemed effective against the subsequently isolated organism, in contrast to ‘time to antibiotics’ which only measured time to the first antibiotic that may or may not have been effective.
Time zero	Time zero for sepsis (and infection without sepsis) was defined as the time stamp of triage in the ED. Time zero for septic shock was the time that the first SBP of <90 mm Hg (or >40 mm Hg decline from normal) was recorded in the observation charts.
Antibiotic delivery time for septic shock	The time from prescription order to initiation of administration of the first antibiotic (delivery time), as documented in the ieMR Medication Administration Record and confirmed with Pyxis dispensing time stamp.
Antibiotic post-prescription review	Medical records and laboratory information system notation were accessed for evidence of inpatient team or specialty post-ED prescription review by AMS pharmacists, IDPs or MMs. Advice provided by specialty consultation was scrutinised for any contemporaneous change or ‘de-escalation’ made in antibiotic prescription, such as conversion to oral or narrower spectrum antibiotics, indicating that the recommendations were followed.
ED broad-spectrum antibiotic use	Broad-spectrum antibiotic use adjusted for ED patient presentation rates. Number of patients prescribed broad-spectrum antibiotics per 1000 ED patient presentations. Intravenous ceftriaxone, amoxicillin–clavulanate, piperacillin–tazobactam, cefepime and meropenem were considered broad spectrum.

AMS, antibiotic stewardship; ED, emergency department; EUCAST, European Committee on Antimicrobial Susceptibility Testing; IDPs, infectious diseases physicians; ieMR, integrated electronic medical records; MMs, medical microbiologists; SBP, systolic blood pressure.

Clinical deterioration during the ED encounter was based on sepsis status at the time of prescription compared with that at the end of the ED episode.

Outcomes

The primary outcome was empirical prescriptions adherent to antibiotic guidelines (online supplemental figure 2) during the ED encounter. Secondary outcomes included the administration of effective antibiotics, empirical prescriptions considered appropriate or optimal (table 1), effective antibiotics administered within 3 hours, the antibiotic delivery time for septic shock, post-prescription review, clinical deterioration and change of antibiotic prescription.

Assessment of harm measures included dosing errors, adverse effects of antibiotics, allergy and intolerances, C. difficile infection and whole-of-department rates of broad-spectrum antibiotic use. Definitions of these outcome measures are provided in table 1.

Statistical analysis

The primary unit of analysis was the admission and the primary comparison of interest was baseline versus post-intervention. Multivariable models (logistic regression for binary outcomes; Cox regression for length of stay) were performed to quantify the relationship between the exposure variable (baseline vs post-intervention) or sepsis status (septic shock and sepsis, vs infection), controlling for demographics, admission features and comorbidities. The following covariates were included in the adjusted models based on clinical and statistical significance: sepsis without shock (when comparing baseline vs post-intervention), Charlson comorbidity score, methicillin-resistant Staphylococcus aureus infection risk, febrile neutropenia and palliation status.

The adjusted odds ratios (aORs) (HR for length of stay outcomes) and 95% CIs are reported for each comparison group.

An a priori sample size calculation was undertaken, indicating that to detect a 20% absolute improvement in antibiotic guideline adherence (from an assumed baseline proportion of 60% to post-intervention proportion of 80%) with 80% power, assuming a type I error of 0.05, 82 records per group (baseline and post-intervention) were required. However, during the data collection process, it became apparent that the 20% improvement was an overestimation of potential effect, and the sample size was recalibrated to reflect the emerging data, resulting in a minimum of 240 cases per group being collected. Due to the extended post-implementation phase, compared with the pre-implementation phase, a decision was made to further enhance statistical power by sampling double the number of cases in the post-implementation phase.

Python V.3.10 was used to undertake the analyses. Corrections for multiple comparisons were not undertaken; however, multiplicity was taken into account when interpreting the results. There were no missing data points for any of the reported outcomes and less than 10% for covariates of interest; therefore, complete case analysis was undertaken.

Patient and public involvement

None.

Results

During the study period of 1 July 2017–March 2020, 2591 eligible patients presented to the three ieMR sites; 721 of these were randomly selected for the study (online supplemental figure 3). The characteristics of the selected patients in baseline (N=241) and post-intervention (N=480) phases were similar for most parameters (table 2).

Table 2.

Demographic and clinical characteristics

Characteristic	Baseline N=241 n (%)	Post-intervention N=480 n (%)	P value
Age (years), median (IQR)	70.0 (61.0–80.0)	71.0 (57.0–81.2)	0.968
Sex (male)	142 (58.9)	285 (59.4)	0.907
BMI (kg/m2) (N=606), median (IQR)	26.6 (22.4–31.1)	25.4 (21.6–30.9)	0.282
eGFR (mL/min/1.73 m2) (N=721), median (IQR)	52.0 (34.0–73.0)	57.0 (34.8–84.0)	0.051
Sepsis status
Septic shock	55 (22.8)	95 (19.8)	0.121
Sepsis without shock	97 (40.2)	232 (48.3)
Infection without sepsis	89 (36.9)	153 (31.9)
Comorbidities
Charlson score, median (IQR)	1 (0–3)	1 (0–2)	0.033
Chronic kidney disease	50 (20.7)	82 (17.1)	0.231
Malignancy—solid organ	50 (20.7)	79 (16.5)	0.157
Malignancy—haematological	3 (1.2)	13 (2.7)	0.220
Palliation status
Palliation within 48 hours	18 (7.5)	56 (11.7)	0.082
Multiresistant organism colonisation
ESBL colonisation	5 (2.1)	13 (2.7)	0.852
MRSA colonisation	15 (6.2)	32 (6.7)	0.820
Infection risk
Intravascular device	18 (7.5)	30 (6.2)	0.536
Risk of Burkholderia pseudomallei*	5 (2.1)	1 (0.2)	0.035
MRSA infection risk factors	79 (32.8)	207 (43.1)	0.008
Febrile neutropenia	19 (7.9)	35 (7.3)	0.776
Source of infection
Respiratory	62 (25.7)	107 (22.3)	0.305
Urinary	83 (34.4)	155 (32.3)	0.563
Gastrointestinal	41 (17.0)	90 (18.8)	0.568
Skin/soft tissue	20 (8.3)	63 (13.1)	0.058
Unknown	38 (15.8)	81 (16.9)	0.706
Other	29 (12.0)	60 (12.5)	0.857
Microorganisms isolated from sterile sites
Staphylococcus aureus	16 (6.6)	71 (14.8)	0.002
MRSA	3 (1.2)	10 (2.1)	0.430
Beta-haemolytic Streptococcus spp	24 (10.0)	38 (7.9)	0.357
Group A Streptococcus	10 (4.1)	15 (3.1)	0.480
Enterobacterales	119 (49.4)	245 (51.0)	0.673
ESBL	6 (2.5)	15 (3.1)	0.633
ampC	12 (5.0)	30 (6.2)	0.493
Pseudomonas aeruginosa	8 (3.3)	17 (3.5)	0.878
Other organisms	40 (16.6)	68 (14.2)	0.389
No organism	33 (13.7)	38 (7.9)	0.015
Clinical outcome
Mortality (inpatient)	24 (10.0)	61 (12.7)	0.281
Mortality (30-day)	11 (4.6)	19 (4.0)	0.701
Hospital length of stay (days), median (IQR)	6.2 (3.6–12.3)	6.1 (3.6–13.6)	0.409

MRSA infection risk factors: chronic disease (eg, renal failure, diabetes), immunosuppression, chronic wounds, living in communities with high MRSA prevalence and colonised with MRSA.

*Risk of B. pseudomallei: travel to, or residing, north of 20^o S latitude and any of: diabetes, hazardous alcohol consumption, chronic kidney disease, chronic lung disease and immunosuppressive therapy.

ampC, ampC-producing Enterobacterales; BMI, body mass index; eGFR, estimated glomerular filtration rate; ESBL, extended-spectrum beta-lactamase-producing Enterobacterales; MRSA, methicillin-resistant Staphylococcus aureus.

Primary outcome: adherence to guidelines

As compared with baseline, there were no significant differences in the post-intervention phase with respect to adherence to guidelines (54.0% vs 59.5%, aOR 1.41 (95% CI 1.00, 1.98)) (table 3).

Table 3.

Antibiotic prescribing outcomes

Parameter	Baseline N=241 n (%)	Post-intervention N=480 n (%)	Adjusted effect size (95% CI)*
Primary outcome
Adherent to antibiotic guidelines†	121 (54.0)	261 (59.5)	1.41 (1.00, 1.98)
Secondary outcomes
Effective antibiotics prescribed in ED‡	110 (78.0)	227 (73.5)	0.75 (0.46, 1.22)
Appropriate prescription§	136 (57.1)	334 (69.9)	1.92 (1.37, 2.68)
Optimal prescription§	77 (32.4)	217 (45.4)	1.90 (1.35, 2.65)
Timing of antibiotics
Effective antibiotics <3 hours (sepsis and septic shock), N=450, median (IQR)†	78 (55.3)	194 (62.8)	1.36 (0.89, 2.06)
Time to first antibiotic (min), median (IQR)	95.0 (51.0–162.0)	92.0 (55.0–177.0)	0.76 (0.62, 0.94)
Antibiotic delivery time (septic shock only) (min),  N=87, median (IQR)¶	15.0 (5.0–24.0)	11.0 (5.0–21.8)	1.02 (0.61, 1.72)

*Adjusted for sepsis without shock, Charlson score, MRSA infection risk, febrile neutropenia and palliation status.

†Guideline adherence excludes cases with no guidelines for given indications or directed therapy.

‡Effectiveness of antibiotics only includes cases with confirmed organism susceptibility available (baseline N=141, post-intervention N=309).

§Prescription appropriateness excludes non-assessable cases (online supplemental figure S4) (baseline N=238, post-intervention N=478).

¶Delivery time (prescription to administration time) for audited shock cases only (baseline N=41, post-intervention N=46).

ED, emergency department; MRSA, methicillin-resistant Staphylococcus aureus.

Secondary outcomes

There were no differences in the post-intervention phase with respect to effective antibiotic prescriptions (78.0% vs 73.5%, aOR 0.75 (95% CI 0.46, 1.22)).

Appropriateness of prescription was greater in the post-intervention phase (69.9% vs 57.1%, aOR 1.92 (95% CI 1.37, 2.68)). Optimal prescriptions, with respect to choice, dose, route, frequency and allergy status, were more likely after introduction of the pathway than before (45% vs 32%, aOR 1.90 (95% CI 1.35, 2.65)).

There was no difference in the proportion of patients receiving effective antibiotics within 3 hours between the two phases. There was no difference in the median antibiotic delivery times for septic shock (15 min vs 11 min adjusted HR 1.02 (95% CI 0.61, 1.72)) before and after pathway implementation (table 3).

The rates of guideline adherence, effective antimicrobial prescription and prescription appropriateness, and time to effective antimicrobial prescription were stratified by infection or sepsis status in the post-implementation phase. These results are summarised in online supplemental table 2. When stratified for sepsis status, 73.7% of patients with septic shock and 67.7% of patients with sepsis alone received effective antibiotics as part of their empirical regimen, which was significantly greater than those with infection without sepsis (aOR septic shock 3.81 (95% CI 2.11, 6.90), sepsis without shock aOR 2.6 (95% CI 1.69, 4.01)). A greater proportion of patients with septic shock and sepsis received effective antibiotics within 3 hours compared with infection alone (62.0%, 57.5% and 45.9%, respectively, aOR septic shock 2.14 (95% CI 1.21, 3.80), sepsis 1.69 (95% CI 1.09, 2.64)).

Post-prescription review

While in the ED, approximately 30% of cases in both phases of the study had evidence of a senior medical officer supervising the antibiotic plan (29.5% vs 31.0%, aOR 1.11 (95% CI 0.79, 1.57)).

Almost all cases were reviewed by the admitting team consultant within 24 hours of triage (98.6% vs 98.4%, aOR 0.92 (95% CI 0.23, 3.70)) and two-thirds had evidence of consultation with infectious diseases (ID), AMS and/or medical microbiology specialties within 48 hours (combined baseline and post-intervention 384 of 721; 65.8%) (online supplemental table 3). The advice provided by specialty services resulted in antibiotic prescription change or de-escalation in 76% (292 of 384) of cases that required modification. There was no difference in the median total duration of antibiotics between the baseline (11.7 days (IQR 2.2–7.7)) and post-intervention phases (12.0 (IQR 7.1–17.6)).

Rates of clinical deterioration and change of antibiotic prescription

Progression to either sepsis or septic shock occurred in 26.4% of patients initially presenting to the ED with infection alone (N=239). Almost one in five patients (18.6%, N=360) presenting with sepsis deteriorated to septic shock after prescription of the first antibiotic while in the ED. Around one-third of patients in the before and after pathway implementation cohorts had changes made to their antibiotic prescriptions (ie, addition of antibiotics) during the ED stage of their sepsis management (34.0% vs 25%, aOR 1.03 (95% CI 0.68, 1.58)).

Assessment of harm from pathway use

Dosing errors were significantly reduced in the post-implementation phase (aOR 0.41 (95% CI 0.28, 0.63)) (table 4). There was no significant increase in adverse effects of antibiotic use with introduction of the pathway (aOR 0.59 (95% CI 0.28, 1.22)).

Table 4.

Adverse effects

Parameter	Baseline N=241 n (%)	Post-intervention N=480 n (%)	Adjusted effect size (95% CI)*
Dosing errors	75 (31.1)	76 (15.8)	0.41 (0.28, 0.63)
Patients with any complications	15 (6.2)	17 (3.5)	0.59 (0.28, 1.22)
Acute kidney injury	5 (2.1)	1 (0.2)	–
New-onset allergy	3 (1.2)	6 (1.2)	–
Clostridioides difficile infection (hospital-associated)	0 (0.0)	1 (0.2)	–
Neutropenia	1 (0.4)	2 (0.4)	–
Allergy mismatch	1 (0.4)	3 (0.6)	–
Intolerance to medication	4 (1.7)	6 (1.2)	–

*Adjusted for sepsis without shock, Charlson score, MRSA infection risk, febrile neutropenia and palliation status.

MRSA, methicillin-resistant Staphylococcus aureus.

Broad-spectrum antibiotic use, adjusted for ED activity, calculated as number of patients prescribed broad-spectrum antibiotics per 1000 ED presentations, did not increase after pathway introduction (figure 1).

Figure 1.

Control chart: number of ED patients prescribed broad-spectrum^a antibiotics per 1000 ED presentations. ^aCeftriaxone, amoxicillin–clavulanate, piperacillin–tazobactam, cefepime and meropenem. Site 1 and Site 2: study sites using electronic medication dispensing services (Pyxis) linked to a hospital-based corporate information system to allow determination of number of patients prescribed broad-spectrum antibiotics in the ED. Vertical line corresponds to date of pathway implementation at each study site. The upper control limit and lower control limit are CL±3√CL. CL, control limit; ED, emergency department.

Discussion

The implementation of a sepsis pathway in the ED was not associated with a significant difference in antibiotic guideline adherence but resulted in improvement in the rates of appropriate and optimal antibiotic prescription. There were no differences between the two phases with respect to effective antibiotic administration and time to effective antibiotics. There was a significant reduction in dosing errors with institution of the pathway and no evidence of increased adverse effects or increased broad-spectrum antibiotic use.

To our knowledge, this is the first multicentre study to report in detail the impact of implementation of an ED sepsis pathway on antibiotic prescribing for sepsis/septic shock and infection cases, to allow exploration of the impact of the pathway on patients who did not have sepsis. Data were collected systematically as part of a collaborative. Sepsis-3 definitions based on SOFA score allowed assessment of antibiotic use without conflation of sepsis and septic shock. We used multiple definitions of antibiotic prescription quality and ‘appropriateness’, based on the sepsis status and diagnosis of source of infection at the time of prescription, as well as the subsequent organism susceptibility. The availability of data from ieMR and digitised pharmacy dispensing modules facilitated accurate time stamps and precise calculation of antibiotic prescription and delivery times.

Implications of study findings

The guideline adherence rates were comparable with Australian national inpatient prescribing data¹² and other ED studies,¹³ and the rates of effective antimicrobial prescription observed in our study were consistent with previously published data.¹⁴ The finding of higher effective prescription rates as compared with guideline adherence may reflect antibiotic use outside of guidelines or ‘change of mind’ prescribing, as occurred in around one-third of cases, thus increasing the chances of an antibiotic being effective for the organism causing sepsis. Further research into the way guidelines are used for sepsis may help to answer this question.

Appropriate prescribing, as a composite of optimal and adequate prescribing,^{15 16} significantly improved by 8% with introduction of the pathway to around 70%, a figure higher than previously reported.¹⁷

We observed short median duration of time from prescription-to-antibiotic delivery for patients with septic shock. The Infectious Diseases Society of America has proposed reporting this time interval for patients with septic shock to drive improvements in drug delivery systems.¹⁸ Of note, we observed that up to 30% of patients presenting with infection deteriorated to sepsis or septic shock within the ED making ‘time zero’ a difficult target. Over one-third of patients underwent changes in empirical regimen within the ED. These findings have implications for researchers or quality improvement analysts evaluating time-to-antibiotic administration.

The finding of almost all cases having had a treating team review antimicrobial prescription within 24 hours and additional review by ID, AMS pharmacists or MMs within 2 days of presentation accords with the post-prescription review recommendations.^{19 20} Not all patient prescriptions required changes, but in those that did, 76% had alterations made to their management as a result of interactions with these specialist groups.

The antibiotic-associated adverse events rate observed in our study is consistent with previously published data.²¹ We also observed that the institution of the pathway was not associated with higher rates of broad-spectrum antibiotic use, although published data suggest that in specific patient populations, such as in cancer-associated sepsis, sepsis pathways resulted in higher rates of broad-spectrum antibiotic use.²²

Our study had limitations. This was a before-and-after observational study and therefore there was potential for bias by residual confounding. Evaluation of antibiotic usage was confined to a nested cohort predominantly among patients who had blood culture collection in sites with ieMR. These data were collected in hospitals in one state in Australia and therefore limits the generalisability of our findings. As source control was not documented, its impact on the duration of antibiotics could not be assessed.

In conclusion, the use of a sepsis screening and treatment pathway for sepsis management in EDs was associated with an improvement in the rates of appropriate and optimal antibiotic prescription without evidence of adverse effects.

Data availability statement

Data are available upon reasonable request. Deidentified patient data will be made available (participant data with identifiers, data dictionary), beginning 6 months after publication of the study and ending at 2 years. All requests for data sharing must be accompanied by a formal request, a study proposal with clear statement of aims and hypotheses, and a statistical analysis plan. All applications will be assessed by the Sepsis Collaborative Steering Committee. Applications from investigators with suitable academic capability to conduct the proposed work will be given consideration. Any proposal will require approval from the ethics committee which approved the conduct of this trial prior to sharing of any patient data. If a proposal is approved, a signed data transfer agreement will be required before data sharing.

Ethics statements

Patient consent for publication

Not required.

Ethics approval

This study was approved with a waiver of informed consent by the Metro North Human Research Ethics Committee (LNR/2019/QPCH/5089), Brisbane, Queensland.