The impact of a sepsis performance improvement program in the emergency department: a before–after intervention study

Sepsis is a major global health problem defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection [1, 2]. In 2017, the World Health Assembly of the World Health Organization declared sepsis a global priority and adopted a resolution to improve its prevention, diagnosis, and management [3]. With a recent estimate of 49 million sepsis cases each year, the global burden of sepsis may be more significant than previously anticipated [4, 5].

In the early 2000s, the Surviving Sepsis Campaign (SSC) was established to provide evidence-based guidelines for managing sepsis and septic shock [6]. The SSCs goal is to reduce sepsis morbidity and mortality worldwide, and its guidelines were most recently updated in 2021 [7, 8]. By bundling the guideline recommendations into core groups of clinical actions that should be performed within a specific timeframe, the SSC aims to facilitate implementation [8]. Several observational studies have shown that compliance with sepsis care bundles is associated with reduced mortality rates [9‐13]. However, bundle adherence still remains a significant challenge, and non-compliance is especially prominent in the microbiological workup and timely administration of antibiotics in emergency departments (EDs) [14‐16]. As the ED often represents a sepsis patient’s first interaction with the healthcare system, it is crucial to promptly initiate the appropriate care processes in this setting [16].

In response to the suboptimal compliance with sepsis guidelines, hospitals and healthcare organizations have initiated sepsis performance improvement programs. These initiatives often include interventions such as educational programs, screening tools, or changes in sepsis care pathways (e.g., activating dedicated sepsis response teams) [17]. Performance improvement programs have been associated with better adherence to SSC or local sepsis guidelines and decreased mortality rates [18]. The latest SSC guideline thus advocates that all hospitals and health systems implement sepsis performance improvement programs [8].

Although the use of sepsis performance improvement programs is now recommended, there is limited evidence on how these programs should optimally be structured [17, 18]. Furthermore, their potential impact on the ED population is relatively unknown, as most studies target the intensive care unit (ICU) population [17, 18]. In this study, we prospectively evaluate the implementation of a multidisciplinary sepsis response team and performance improvement program in our ED and analyze the management and outcomes of sepsis patients before and after.

We conducted a before–after intervention study in the ED of the Amsterdam University Medical Centers—location VUmc, in the Netherlands. The Medical Ethics Review Committee waived the review of this study as it was a quality improvement project within regular care (IRB number: IRB00002991; case:19.449). Study outcomes were measured between November 2019 and February 2020 (preintervention phase) and December 2021–May 2022 (postintervention phase), while the period in between March 2020 and November 2021 was used to implement all aspects of the sepsis performance improvement program appropriately (implementation phase). Patients were sent a letter to opt out of the use of their data for this project. We adhere to “The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies” [19].

A total of 265 patients were included in the primary study, 132 patients preintervention and 133 postintervention. The median age was 68 years (interquartile range [IQR]: 56–77), and 61.5% of the patients were male. The median MEWS score in the ED was 4 (IQR: 4–6), and the Charlson Comorbidity Index was 5 (IQR: 2–7). Baseline characteristics are provided in Table 1. The most common site of the suspected Infection in the ED was the respiratory tract in both the pre- and post-intervention phases. However, respiratory tract infections were relatively more common in the preintervention phase. e-Fig. 1 of the supplementary appendix shows the distribution of the suspected infection sites and the most likely infection type at discharge in the different phases.

In this study, we found that the implementation of a sepsis performance improvement program in the ED including the use of a specialized multidisciplinary sepsis response team resulted in better identification of sepsis, an improved diagnostic process, and a > 50% reduction in time to antibiotic treatment in suspected sepsis patients. However, these improved process-related outcomes did not translate into improvements in length of stay, admission rates, or mortality rates. The only patient-related outcome which improved was the number of immediate versus delayed ICU admissions.

Implementing our sepsis performance improvement program was associated with various improved process-related outcomes. The MEWS score, which already was the preferred screening tool according to hospital policy in the preimplementation phase, was recorded completely in significantly more cases. Interestingly, the MEWS recordings were also more complete when comparing postimplementation patients for whom the sepsis team was or was not activated (post–post comparison). This indicates that the triage nurses indeed linked the use of the MEWS as a screening tool to the sepsis performance improvement program but did not use it when they did not consider the patient as a potential sepsis case. In addition, the number of lactate measurements also significantly increased after implementation and remained significant in the post–post comparison. The SSC guideline recommends using both sepsis screening tools (including MEWS) and lactate measurements, and we thus show increased SSC guideline adherence in these instances [8]. Regarding the workup with blood cultures, we found mixed results. Though there seems to be an increase in the number of blood cultures performed when comparing preimplementation and postimplementation patients, the post–post comparison does not reinforce this effect. ED nurses seem to have been more inclined to draw blood cultures postintervention, irrespective of whether or not the patient was seen by the sepsis team. This could be due to the attention given to blood cultures in the educational meetings as part of the sepsis improvement program, but it could also be a reflection of the higher rate of bacterial infections in the post-phase, as seen in Fig. 1B of the supplementary appendix. In line, sepsis team utilization was associated with a decrease in time to blood culture draws when compared to both control groups.

The implementation of the sepsis team was associated with a considerable reduction in the time to the first administration of antibiotics. Although the benefits of early antibiotics for all sepsis patients remain debatable, there are specific subgroups of patients who may experience benefits [25‐29]. The sepsis team also started antibiotic treatment in the ED in significantly more cases than in the preimplementation phase or in the postimplementation patients when no sepsis team was involved. This indicates that the improved recognition of sepsis may have led to an increased use of antibiotics. Notably, not all patients seen by the sepsis team were treated with antibiotics in the ED. In cases with a relatively low probability of sepsis or shock, the latest SSC guidelines suggest conducting a time-limited investigation first and only initiating antimicrobial therapy when the concern for infection persists [8].

Despite many improved process-related outcomes, we found only a single patient-related outcome that improved after implementing the sepsis team in the ED, which contrasts with previous literature [16, 30, 31]. Viale et al. found that their infectious diseases team improved SSC guideline adherence in a general ED in Italy [16]. Their pre–post comparison including 382 (195 vs. 187) severe sepsis and septic shock patients with a high median age of 82 years (IQR 70–88) showed that the infectious diseases team implementation was associated with higher rates of lactate measurements (90 vs. 76%; p < 0.001) and blood cultures before antibiotics (58 vs. 42%; p < 0.001). The time to first antibiotic treatment did not significantly decrease (154 vs. 169 min; p = 0.42). Interestingly, the all-cause 14 days mortality was significantly lower in univariate and multivariate analyses (29 vs 39%; p = 0.02), but the 30 days all-cause mortality was not (37 vs. 45%; p = 0.102). Arabi et al. implemented a multifaceted intervention similar to ours, including a sepsis response team, in their ED in Saudi Arabia [31]. In that postintervention cohort of 699 patients, most process-related and patient-related outcomes improved significantly. For example, the percentage of patients receiving antibiotics within 3 h improved (89.4 vs. 67.7%; p < 0.001), and the hospital mortality rate was lower (16.9% vs. 47.7; p = 0.003). A recent publication by Simon et al. also shows improvements associated with a sepsis team implementation in the ED of a tertiary hospital in the United States. The pre–post analysis among 863 patients (393 vs. 470) showed that the time to antibiotics was reduced (81 vs. 107 min; p < 0.001), just as the in-hospital mortality (15.1 vs. 28.2%; p < 0.001). A notable difference with all of these cohorts is that their preintervention mortality rates of 45%, 48%, and 28% were much higher than in our cohort (9.1%). This finding is not completely unexpected since other Dutch studies and various international sepsis studies in the ED setting have also reported relatively low sepsis mortality rates [26, 27]. Furthermore, the aim of this sepsis performance improvement program was to screen for sepsis and detect and treat it early. Mortality rates in such a screening cohort will be lower than in cohorts looking only at definite and severe sepsis cases. We may argue that sepsis performance improvement programs are more likely to improve mortality at those higher mortality rates or that establishing a significant mortality benefit is at least easier in such a population. Still, we would have expected to find other improvements in patient-related outcomes through our intervention, especially a shorter length of ED stay. Unfortunately, overcrowding of the ED and exit blocks toward the hospital wards due to staff shortages in our postintervention phase made it challenging to transfer patients to the wards [32, 33]. During the extraction of data from the EHR system, the study team had the impression that patients were ready for hospital admission earlier when they were seen by the sepsis team, but this could not be reflected in shorter ED stays due to the logistical constraints. This hypothesis is further supported by the fact that we were able to show a significant improvement in the number of direct versus delayed ICU admissions. Delayed ICU admission (patients who will eventually need an ICU admission but are first admitted to the regular ward) is an independent risk factor for sepsis mortality, but none of the patients seen by the sepsis team were being admitted to the ICU with a delay [34, 35]. This suggests that our intervention helped bring together the experts needed to make the most appropriate and timely decision about where the patient needed to go next.

To fully understand the results of this study, it is essential to acknowledge the role of the COVID-19 pandemic. Shortly after our preimplementation measurements, the SARS-CoV-2 virus emerged [36]. The pandemic put unprecedented pressure on healthcare workers and hospitals, which caused a significant delay in the implementation of our sepsis team [37]. Consequently, a near 2 years interval was needed before the postimplementation measurements could be performed. Even then, the healthcare system, and certainly the ED, continued to operate under high pressure, which led to imperfect compliance rates with the sepsis team activation. In the meantime, the national report on infectious diseases in the Netherlands and several international publications showed that the distribution of infectious agents had changed, with, for example, a much lower prevalence of influenza [38‐40]. Our study observed similar changes, where influenza was much less prevalent in the postimplementation phase, while bacterial infection rates were higher. Of note, proven COVID-19 cases at presentation were excluded. Although sepsis guidelines are created for a heterogeneous group of patients with all types of infections, the results of our before-after comparison may have been influenced by the COVID-19 pandemic-induced changes in the causative agents of sepsis in our population. Bacterial infections seemed to have been much more prevalent during the postintervention phase. Fortunately, the imperfect compliance rates with the sepsis team activation led us to have an additional cohort of patients from the postimplementation phase in whom no sepsis team was activated and who could serve as an unexpected but essential second control group.

Besides the potential confounding through COVID-19, several other limitations of this study must be addressed. First and foremost, this study was powered to detect a difference in the length of ED stay after implementing a sepsis team. In hindsight, this was an unattainable result due to the logistical constraints (e.g., exit blocks) discussed above. Second, the compliance rate with the sepsis team activation was only 64%. Consequently, selection bias may have been introduced since the ED nurses may have had an unconscious bias to activate the sepsis team only in more severe cases, in whom the diagnostic workup and start of treatment would already happen more timely. Fortunately, we could negate part of this confounding by comparing the pre–post results to the post-post comparison. Still, the fact that patients in the postintervention phase may have been more severely ill and less likely to survive compared with the preintervention phase should be considered when interpreting these results. Interestingly, it seems that the ED nurses could identify the patients with a higher likelihood of having sepsis, as the rate of progression to sepsis was significantly higher in postimplementation patients who were seen by the sepsis team. This finding supports our approach of implementing a sepsis response workflow based on the SSC recommendations but with relative flexibility to maneuver according to clinical judgment. Lastly, the before–after study design has its inherent limitations, such as time-related changes in populations and standards of care. A large (stepped wedge cluster) randomized trial is needed to fully understand the value of sepsis teams and sepsis performance improvement programs in general. Given the limited evidence for the benefits and the proper structure of a sepsis team, we did not have the support base to conduct such a trial. We hope our current work helps create the urgency for this type of study.

In conclusion, implementing our sepsis performance improvement program in the ED was associated with a number of improvements in process-related outcomes but minimal improvements in patient outcomes. The program stimulated physicians to make collaborative and timely decisions regarding diagnostics and treatment of sepsis. The workflow allowed them to incorporate their clinical judgment while still reinforcing the essential elements of sepsis care.