Impact of a Coordinated Sepsis Response on Time to Treatment in a Pediatric Emergency Department

Our hospital is an urban, academic, 364-bed pediatric tertiary care center with 56 000 emergency department (ED) visits annually. Clinical care guidelines (CCGs) and associated order sets were developed for our pediatric ED in 2010, which address >20 ED-specific pediatric conditions and are available within the electronic health platform for providers. Approximately 775 patients are managed by using the sepsis CCG annually. In 2016, the ED developed a formal quality infrastructure with nursing and medical leadership and an embedded consultant from the institution’s Center for Quality and Safety. This improvement project was fully supported by both institutional and divisional leadership.

Efforts to standardize sepsis care at our institution began in 2015 with the introduction of an electronic best practice advisory (BPA) in Epic (Epic Systems, Verona WI), modeled after Balamuth et al, as well as the development of a CCG and order set.⁴  Care for patients identified by the advisory prompts a huddle with members of the care team. Identified patients are clinically evaluated and assigned to an appropriate pathway on the basis of the presence or absence of shock (see Supplemental Fig 5 of red and yellow pathway). Our hospital also became part of the Children’s Hospital Association’s Improving Pediatric Sepsis Outcomes (IPSO) Collaborative at this time.¹³ 

An ED Sepsis Quality Improvement (QI) team was formed to ensure that sepsis work was prioritized. This team led foundational work and small-scale improvement projects without significantly affecting targeted process metrics. A multidisciplinary, rapid improvement event, known as a Kaizen, was held in July 2018 to uncover barriers to adherence to the CCG.^13,14  Solutions generated during the Kaizen informed our interventions. A detailed description of the Kaizen event that informed this intervention has been previously published.¹⁵ 

The Kaizen identified 4 key drivers contributing to delays in timely care: (1) the presence of ineffective processes and workflows, (2) ineffective team communication between and among nurses, residents, fellows, paramedics, and attending physicians, (3) a lack of awareness or variability among clinicians related to the early identification and stratification of sepsis patients, and (4) difficult-to-obtain supplies and resources.¹⁵  To address these barriers, we developed a coordinated sepsis response initiated by a huddle to standardize a team-based response. The intervention components and key drivers are presented in Table 1.

Team member roles and responsibilities were delineated with input from frontline clinicians to align with the clinical workflows. After huddle activation, the following actions occur: an overhead announcement is made, and then the charge nurse, support nurse, paramedic, and provider team member(s) arrive at the bedside, and the supply cart and ultrasound machine are brought to the bedside. In the room, a brief description of the patient’s reason for arrival, pertinent history, and vital signs is presented to the team by the bedside nurse, a brief assessment and examination are conducted by providers, and a pathway is chosen for the patient by the provider and communicated to the team.

We developed a communication tool to standardize language and improve team members’ shared understanding of treatment trajectory and goals for immediate care delivery (Supplemental Fig 6). The guidance included how to communicate indications for huddle activation, clinical assessment findings, pathway designation, and time zero. The tool was posted on the sepsis cart for reference. A badge tag with key points from the communication tool and age-based vital sign triggers was printed and attached to hospital-issued identification badges for all roles in the ED (Supplemental Fig 7). A reassessment step was added at 30 minutes from time zero to assess the patient’s physiologic response to the initial fluid delivery and communicate further treatment or pathway escalation.

The addition of an overhead page and an easily visible mobile sepsis supply cart enhanced departmental awareness and directed critical resources to the patient’s bedside. The illumination of pathway-specific lights on the cart ensured team congruency with the treatment plan.

The team identified items critical for sepsis management and stocked the cart with the supplies required for rapid intravenous (IV) or central access and administration of the initial fluid bolus. Supplies included educational reminders, IV and port catheters, central line access kits, IV tubing, stopcocks, fluid bags, and pressure bags. Additional team members included a support nurse for documentation, a paramedic for vascular access, and a charge nurse for personnel resource allocation.

All ED staff were educated in a phased approach over a 3-month period before the November 2018 start date for the ED sepsis response team (SRT). The sepsis QI team sent emails and posted information and a document with sepsis terminology in high-traffic areas within the ED. Presentations on the initiative were given to unit and institutional stakeholders to disseminate results from the rapid-improvement event and discuss the proposed response plan. Individual and small-group sessions with nurses, paramedics, and ED service representatives and clerks were conducted by the nursing QI coordinator to ensure consistency in the messaging of new process information among all team members. Before implementation, ED provider education took place during 2 faculty meetings with the physician QI leader. Chart reviews were conducted between November 2018 and March 2020. An automated list of all patients placed on the yellow and red sepsis pathways was generated weekly and reviewed by members of the sepsis QI team. Cases with time to fluid or antibiotics falling outside of our timeliness goals and cases in which pathway designation did not align with the guideline were reviewed on a biweekly basis. Any opportunities for improvement identified in these reviews were used to inform future plan-do-study-act cycles. We addressed issues with provider adherence, family preference for anesthetic cream, escalation to advanced IV access, and the decision to monitor clinical improvement in patients with suspected sepsis.

After the implementation of the coordinated sepsis response, adherence to the intervention was monitored in a sample of patients via direct observations between the hours of 7 am and 7 pm, and chart reviews were performed on patients falling outside the timeliness goal parameters. The primary method used to monitor progress was statistical process control charts (SPC). This study was submitted and deemed exempt by our hospital’s institutional review board.

The QI nursing coordinator conducted direct observations of sepsis responses to evaluate the implementation process between November 2018 and February 2019 from 7 am to 7 pm. Patients presenting outside of this time frame were not formally observed. An observation checklist was used to collect data on process adherence, which were entered into a REDCap (Data tool, Northwestern University) database^16,17  (Supplemental Fig 8). Real-time feedback was provided to the team during the observations. The data were aggregated to inform future plan-do-study-act cycles.

Outcome measures included the average time (in minutes) to fluid and antibiotic administration from the identification of physiologic sepsis (time zero) and the proportion of patients meeting timeliness goals for fluid resuscitation and antibiotic administration. Time zero was defined as the earliest time among 3 discrete events that may indicate sepsis within the ED (inclusive of triage): the firing of an automated sepsis screen, the first sepsis huddle, or the order time of a sepsis order set. Timely fluid and antibiotic administration were defined as administration of the first fluid bolus ≤20 minutes and administration of the first antibiotic within ≤60 minutes of time zero.¹⁸  The proportion of patients meeting timeliness goals for fluid and antibiotics monthly was also monitored with a goal of 80% compliance.

The study cohort included all visits among patients 1 month to 18 years of age on ED arrival who were placed on an ED sepsis pathway (from nursing flowsheet documentation or sepsis order set utilization) and met IPSO sepsis treatment criteria^17,18  between March 2017 and June 2020. Patient visits were excluded from the analysis if the patient was transferred from an outside hospital, the first treatment bolus was received after ED departure, or if time zero occurred at the same time as the initial fluid administration.¹⁹  Patients receiving interventions before time zero were excluded from metric-specific analysis (ie, if antibiotics were administered before time zero, the encounter was excluded from SPC analyses on antibiotic delivery). We used SPC charts as the primary method for evaluating the implementation of the SRT intervention.

SPC charts were generated to monitor progress toward and changes in achieving timeliness goals over the study period. X-bar S charts were used to analyze the continuous data for time to fluid and antibiotic administration and p-charts were used to analyze the proportion of each process meeting timeliness goals. We used established rules to determine special cause variation, including 1 point outside of the control limits, a 6-point trend up or down, and 8 points on 1 side of the center line.²⁰  SPC charts were generated by using QI Charts (Process Improvement Products, San Antonio, TX, Version 2.0.23) and Microsoft Excel 2021 MSO (Version 2310 Build 16.0.16924.20054).

Between March 2017 and June 2020, 1584 patient visits met the inclusion criteria for the study cohort. The characteristics of the pre- and postintervention patient visit populations are shown in Table 2. The only notable difference was that the mean number of visits per patient was significantly higher in the preintervention period, indicating that there were more patients with repeat visits before the intervention period.

We evaluated the timeliness of fluid and antibiotic administration using SPC charts. Before the coordinated sepsis response intervention, the average time to the initial fluid administration was 33 minutes (Fig 1). After a July 2018 Kaizen event, this time decreased to 26 minutes and was sustained after the implementation of the SRT intervention in November 2018. Special cause variation was noted in September 2019. Of the 16 cases identified, 3 cases had delays in fluid administration exceeding 100 minutes. Causes of delays included 1 patient with difficult access, requiring ultrasound assistance, and 2 patients whose symptoms progressed to concern for sepsis. A shift in the proportion of visits receiving fluid bolus administration within 20 minutes of time zero was observed, indicating sustained progress toward meeting timeliness goals (Fig 2). We did not observe similar trends for average time to antibiotic administration and the proportion of visits receiving antibiotic administration within 60 minutes (Figs 3 and 4).

A total of 41 observations of the sepsis response were performed after intervention implementation. We timed the responses to determine the length of the new process without a specific target in mind. In all observations, overhead pages were announced at the start of each huddle, and supply carts arrived at the bedside in <3 minutes. Among all observations, at least 1 support nurse responded to 73% of all huddle observations, and 1 ED paramedic responded to 64% of huddles. In 90% of observations, nurses communicated the reason that the sepsis BPA had triggered the sepsis huddle to the provider. In 95% of cases, the provider at the bedside communicated the pathway choice for the patient (68% yellow, 2% red, and 30% no pathway indicated). A 30-minute reassessment was conducted in 86% of eligible patients who had been put on a pathway.

Quality culture within our ED has been developing over the past decade by using improvement methodology to solve complex clinical problems. Sepsis is an institutional priority because of the significant morbidity and mortality associated with delays in care. The addition of an SRT and a mobile sepsis-specific supply cart, as part of a coordinated sepsis response process, was associated with improvement in meeting our timeliness goals. Our model was built on the Advanced Trauma Life Support trauma response process, which shares the need for rapid, effective, and team-based care.²¹ 

After the Kaizen, nurses who participated in the event began optimizing workflows to improve time to fluid administration. We attribute this initial improvement to the socialization of ideas and solutions presented during the event. Assessing local barriers and the creation of a coordinated, intradepartmental sepsis response aligned resources and allowed our team to improve fluid administration time. Patients on both the red and yellow pathways were treated comparably regarding fluid administration goals, which was the easier metric to improve. Nonetheless, our goal of fluid administration within 20 minutes of identification was challenging to consistently achieve for all patients. Multiple decisions and tasks must occur in rapid succession, and delays may occur (difficulty with access, language barriers, simultaneous emergency situations, etc). Although no true gains were made in the time to the initiation of antibiotics, this is attributed to our organizational commitment to antibiotic stewardship for patients who improve with fluid resuscitation without developing signs of shock, along with the broad definition of the IPSO cohort. At the time of this initiative, evidence for the identification of pediatric sepsis had limitations, and the goals defined in this guideline were applied to a spectrum of patients of varying severity, many of whom never went on to develop severe sepsis or septic shock. To achieve 80% compliance in treatment timeliness, a narrow definition of patients at the highest risk for severe sepsis or shock would be necessary and require separate data analysis. Since the initial creation of the SRT, the SSC has published new guidelines specifically addressing these issues, with expanded time goals for patients of lower severity.²²  Subsequent pathway-specific timeliness metrics are now used at our institution.

The strengths of this project included a multidisciplinary, team-based response utilizing standard communication and delineated roles and responsibilities. The use of an overhead page and the addition of visual cues ensured departmental awareness of the presence of a sepsis patient and prompted team members to engage in early recognition and treatment.

One unintended behavior noted among providers postimplementation was the increased use of the sepsis response and order set to rapidly initiate treatment on patients whose presenting complaint was not sepsis-related. This practice, once recognized, was discouraged.

For this project, budgeted costs included the purchase of 2 carts for supplies and pathway designation lights and printing costs for sepsis badge tags. All supplies used in the treatment of sepsis were within normal ED stock. The reallocation of responsibilities within the nursing and paramedic staff allowed this intervention to occur within the existing staffing model. With the estimated cost of a pediatric sepsis hospitalization in 2016 ranging from $10 000 to $168 000, any intervention to avoid sepsis-related morbidity was considered as a significant cost-saving opportunity for this population.²³ 

The benefits of dedicated SRTs have been debated in the literature over the past decade. Much of the debate stems from work within the adult hospitalized population to improve the early recognition of hospital-onset sepsis. Improvement in outcome metrics, specifically mortality rates and the need for mechanical ventilation, was reported after the implementation of an early warning electronic alert coupled with a nurse–physician SRT in the ICU by Arabi et al.²⁴  Similar results were noted hospital-wide after trained nursing teams were deployed to ensure the implementation of sepsis bundle orders for patients meeting early warning criteria identified by electronic triggers.^25–29  Department-specific improvements utilizing SRTs are underreported, and it is currently difficult to determine their effect on outcomes such as mortality, organ damage, and ventilator days. To our knowledge, the utilization of an SRT to improve the timeliness of care was a novel approach in a pediatric ED.

We have been limited in sustaining improvement during periods in which our system is under stress. During the patient surge in the fall of 2019, resources were reallocated to provide clinical care, and ongoing oversight of our processes was constrained. We chose to end our data collection in June 2020 because of the low volume of patients presenting during the initial phase of the coronavirus disease 2019 pandemic. Although we were able to sustain our trend toward our goals, we never fully achieved our initial smart aims.

The implementation of a dedicated ED SRT with designated roles and responsibilities, easily accessible supplies, and directed communication has led to improvements in the timeliness of fluid administration in our pediatric population. At our institution, this was a low-cost initiative that used existing staff and standardized workflows to improve efficiency in care. We believe that this model can be implemented in other EDs to improve the time to treatment in the pediatric sepsis population.

The authors gratefully acknowledge the Children’s Hospital Association IPSO Collaborative, a multihospital, multiyear collaborative that provides centralized, nonmonetary resources to support pediatric sepsis QI. We also acknowledge the time and support of ED leadership, frontline clinicians, data analytics and reporting staff, and data analytics and reporting team member Shan Sun.