Background:

Sepsis is among the most ubiquitous problems in health care. Readmission rates after hospitalization, increased morbidity, and high levels of mortality are not uncommon in patients diagnosed with sepsis. Sepsis bundles are treatments and diagnostic tests designed to mitigate the deleterious effects of this pathologic state have been published and recommended for nearly a decade. Despite this, mortality remains high, sepsis bundle requirements controversial, and bundle compliance remains low. This quality initiative reports on the successful implementation of a hospital system wide sepsis initiative.

Methods:

This southcentral Pennsylvania health system made a strategic decision to decrease the mortality rate of sepsis. Their tactics were to review those aspects of census care which corresponded to successful mitigation of the disorder, and to ensure that they did so in a multi-disciplinary environment. This was done by creating a Sepsis Guiding Coalition and adopting a Model Cell mental model to move forward. The Sepsis Clinical Effectiveness Team (CET) created an RN staffed Central Alert Team (CAT) which utilized the strengths of the electronic health record to serve as a systemwide partner to the bedside clinical team caring for the patient. Data were gathered from four acute care hospitals in the system, and compliance data were analyzed. Data on mortality, as reflected by compliance rate, was also obtained and evaluated. Finally, the observed mortality data was compared to predictive data based on other comparable acute care facilities.

Results:

Regression analysis showed significant increases in bundle compliance rates over a 28-month period at each of the four acute care clinical sites (p < 0.05 for each site). The regression reflected the continuous use of the methods described above in increasing compliance rates. Mortality system wide decreased significantly in response to increased bundle compliance (r = 0.80, r2 = 0.64, p < 0.001), with compliance alone accounting for nearly 2/3 of the variance in the linear model. When compared to predicted mortality models, the observed results revealed a median mortality rate of 5.7% (95% CI: 5.1% to 7.3%, n = 23) which was 1.9% lower than predicted when compared to similar institutions. When using only the final twelve months of the SQI, the median mortality drops further to 5.3% (95% CI: 3.9% to 5.6%, n = 12) which was 2.5% less than predicted.

Conclusions:

The Model Cell SQI intervention was successful in increasing bundle compliance, which then decreased mortality. The use of extensive stakeholder input, directed coaching, the presence of a Sepsis CET and the use of CATs contributed greatly to the increased rates of compliance. Of perhaps equal importance was the interaction between the EHR and informatics teams and the CET and CAT groups. The clinical needs were used to drive the electronic outputs, which then augmented human performance. We believe this model can be enhanced as technology improves and is also well positioned for artificial intelligence to help drive further success.

Sepsis continues to be a significant concern for hospitals and health care systems across the world, as well as across the United States. It is the leading cause of death among hospitalized patients1. Globally, there are an estimated 49 million cases of sepsis, resulting in 11 million deaths (22% mortality rate)2. In the United States, one report of 136 hospitals revealed over 95,000 cases of sepsis, whose patients had higher intensive care unit (ICU) admission rates (61% compare to 44% in non-septic patients), longer hospital lengths of stay (median of 19 days vs. 8 days), and a higher in-hospital mortality rate (33% vs 17%)1. In Pennsylvania (the state in which the authors’ study was completed), one group of researchers reported 115,213 sepsis encounters in 152 hospitals, with mortality rates ranging from 12 to 24.5% 3. Against this background, the authors report on the efforts of a health system to mitigate the adverse outcomes related to sepsis in their institutions.

As sepsis is a significant and ubiquitous concern, numerous quality improvement initiatives from varied agencies, governmental and private, have been set in place with varying degrees of success. A recent study from the United Kingdom focused on the administration of a “Septic Six” bundle within 1-hour of notification of sepsis, combined with an intensive staff education program and dedicated multidisciplinary sepsis team4. The authors reported a significant drop in the 90-day mortality rate, from 35.4% in their baseline group to 14.5% in their post QI initiative group4.

In another study, more than 18,000 patients at 21 community-based hospitals were compared over a three-year period before and after the implementation of a quality initiative to standardize a sepsis bundle5. In addition to the standard therapies, the researchers initiated two scoring systems, a measure of comorbidity disease burden and a lab physiology panel based on 15 laboratory and 6 vital sign measurements. These were coupled with education and training for hospital staff5. These researchers found their post-initiative mortality decreased from 8.8% (year 1) and 9.3% (year 2) to 7.9% (year 3). This was associated with increased compliance in meeting their “all bundle goals” (32.2% year 1, 33.8% year 2, and 44.9% year 3).

Another multi-center study focused on 4,329 adults with severe sepsis or septic shock admitted to one of 18 ICUs in one of 11 facilities6. These researchers established a common sepsis bundle for both the emergency department and the ICU to decrease unit to unit and provider to provider variation in treatment. Mortality due to septic shock decreased from 19.9% to 12.2% during the study period, while compliance with the protocol in patients with septic shock went from 7% to 60%6.

In 2014 WellSpan Health (WSH) established overall mortality reduction as its systemwide clinical priority. Over the next two years significant improvement in rate and the ratio of observed to expected deaths was realized. As WSH moved to focus on reducing preventable readmissions an inability to sustain our improvement in mortality was noted. The previously realized success was largely founded on training specific people and developing their engagement, rather than creating a sustainable process. With the addition of two more acute care hospitals to the system and the transition to a new electronic medical record platform, incrementalism was not going to achieve the WSH goals of significant and sustainable mitigation of adverse sepsis outcomes. Therefore, in FY 2018, WSH began a three-year journey to achieve national top performer status in sepsis care, focusing on the core strategy of achieving clinical reliability to support outcome excellence.

Based on the current state of sepsis research, along with our personal clinical experiences, stakeholder concerns, and system-wide goals, we were prompted to relook at sepsis and how well it was managed in our institutions. Embarking on this work, we learned a few grim facts regarding our patient population. We found sepsis, severe sepsis, and septic shock were the primary reasons patients were admitted, and that our mortality rates were higher than expected at each of the facilities listed above. When starting the process of initiating the Sepsis Quality Initiative (SQI), we made assumptions about existing models, and compared these to the desired outcomes, summarized in Table 1. Next, we used the concept of the Model Cell to develop a new initiative to mitigate the effects of sepsis7. This study adds to current knowledge regarding sepsis treatment leading to decreased mortality by emphasizing the following key concepts used in developing our SQI:

  • 1) Sepsis Clinical Effectiveness Team (CET): The creation of a system-wide, multidisciplinary CET which established clinical criteria that would drive best practice protocols based on the recommendations of the international Surviving Sepsis Campaign (SSC) as well as our clinical experience and stakeholder input,
  • 2) Clinical teams guide technology: Clinical teams engaged with technical and informatics experts to review data and recommendations, resulting in the development of an evidence-based bundle of treatment/care interventions (order sets and protocols) embedded within the electronic health record (EHR) for adoption and spread across our system-hospitals,
  • 3) Automated Alerts and Central Alert Teams (CAT): Creation of customized sepsis identification algorithms generated automated alerts which were built in the EHR and reviewed and validated via the development of a system-wide sustainable, mission-dedicated CAT.

These initiatives, which are more completely described in the methods section, led to a significant and dramatic decline in mortality rates throughout the WSH system at the acute care hospitals. These specific data are addressed in depth in the data analysis and discussion sections.

Methods

This is a retrospective study based on sepsis mortality and sepsis bundle compliance after the initiation of an SQI. Institutional Review Board approval was not sought, based on review of Nerenz, et.al., who state, “…the fundamental goal is improvement in specific processes and systems within specific organizations. These activities are not research…Those that are not research do not require IRB review” (page 159)8. Our current study data come from the WSH system in southcentral Pennsylvania in which patients were hospitalized in one of the following acute care facilities: 1) York Hospital (YH), a large 580 bed rural, community, teaching hospital with Level 1 Trauma, 2) Ephrata Community Hospital (ECH), a 130 bed mid-sized community hospital, 3) Good Samaritan Hospital (GSH), a 151-bed community teaching hospital, and 4) Gettysburg Hospital (GH), a 76-bed community hospital. All these facilities are within a radius of approximately 40 miles. The study authors served as champions for the SQI. The compliance rates of bundle initiation on patients studied were those admitted to the acute care facilities with a diagnosis of sepsis, severe sepsis, or septic shock during the time period of January 1, 2017 through May 1, 2019. Mortality rates system wide, and system wide bundle compliance rates, were tracked from July 1, 2017 through October 1, 2019. The only inclusion criteria for record review was admission to one of the four acute care facilities with one of the three diagnoses, and there were no stated exclusion criteria.

Sepsis Clinical Effectiveness Team (CET)

In creating this sepsis quality initiative, the Model Cell structure was used7. In this QI process we mobilized the existing system-wide Sepsis Clinical Effectiveness Team (CET). Our mobilization task was to focus its work and add the necessary subgroups. A first step was establishing a Sepsis Guiding Coalition which had representation from each stakeholder group. The purpose of guiding coalition meetings was defined. Rather than a forum for book reports about our tireless efforts and groundbreaking work, the gatherings were opportunities with cross-functional representation to share ideas, seek help, and garner feedback about items that required resolution to support that which was necessary for us to succeed. A function of the meetings was to create a forum, with the expectation that other conversations would occur post-meeting, and, consequently, add value to the SQI. We ensured that those attending had appropriate advance materials and that we managed the agenda, especially with respect and adherence to start and end times. Ultimately, the forum was the primary opportunity whereby leadership could actively manage, maintain focus, and channel energy. The coalition monitored key performance indicators (e.g. specific measures using dashboards and delineated accountabilities). This in turn fostered efficiency and effectiveness of the structures, processes and resources to achieve our the SQI goals. The result was the collaboration of clinicians from across the health system, who in turn established consensus to rely upon the evidence and recommendations of the international Surviving Sepsis Campaign (SSC) as the foundation for clinical criteria that would drive best practice protocols9-12. SSC recommended a bundle of sepsis treatment interventions, some of which were to be completed within 3 hours of sepsis identification, and others within 6 hours. Sepsis CET subgroups were established at each entity to ensure engagement, adoption, and clinical reliability.

These CET subgroups found that the key drivers of the newly revised sepsis mortality reduction efforts should include early recognition, appropriate stratification, bundle adoption, coding and documentation, and proactive care management. The WSH approach focused upon tactics and interventions founded in clinical relevance. There was an active decision making not to focus on Centers for Medicare & Medicaid Services Sepsis Management Bundle (CMS Sep-1) and other administrative measures with related myopic limitations13. Instead, the evidence-based processes and interventions selected demonstrated improved clinical outcomes and were not merely “administrative checklists.”

Clinical Teams Guide Technology

Other healthcare systems have embedded treatment protocols and/or automated sepsis alerts, only to continue to struggle with reliably of delivering best care. Historically, an early intervention to augment use of these tools implemented at one WSH hospital was utilization of a conventional RN staffed rapid response team, or “boots on the ground,” to respond to electronic sepsis alerts This care model was initially successful in improving clinical reliability and outcomes, but was limited by dependency on one person’s ability to review, prioritize, and respond to sepsis alerts among a growing list of competing priorities. Historical analysis of FY 16 (performance baseline) revealed an average time of greater than 67 minutes for the alert to be screened. More concerning, 22% of sepsis alerts never received rapid response team response, due to competing urgencies. Additionally, “time zero,” or the time at which sepsis was first identified and treatment should commence, was often unknown and monitoring for patient treatment bundle compliance poor.

In the creation of this section of the initiative, our mental model was that we would integrate digital science (technology) and biological science (best practice medicine) to augment the human individual strengths of our clinical teams. We focused on how the existing technology could ease workload, provide rapid alerts to the presence of septic patients, and speed the implementation of the sepsis bundle. Clinical teams engaged with technical and informatics experts to review data and recommendations, resulting in the development of an evidence-based bundle of treatment/care interventions (order sets and protocols) embedded within the electronic health record (EHR) for adoption and spread across our hospitals.

Sepsis CET subgroups were established at each clinical entity to ensure engagement, adoption, and clinical reliability. The subgroups also listened to, vetted and researched physician, nurse, and other clinician elevated concerns and recommendations regarding the developed algorithm, order sets and protocols. Because of these multidisciplinary clinical discussions modifications, revisions or clarification were made to the alert system. The related responsiveness of the Sepsis CET subgroups encouraged continued positive engagement.

A vital component of this innovative model was timely reporting of unit, team, and individual performance. Before the initiative, abstracted bundle compliance would be generally 12 weeks old by the time it reached the hands of the clinical team. After three months, when clinicians were approached about variances and outliers, they could no longer recall the details of the case, making performance improvement challenging. In the new care model design, the EHR system allowed sepsis cases to be identified more rapidly. The current system identifies sepsis cases one week after discharge. Further, each Friday a detailed weekly report including bundle performance outlier details, and mortalities, is released to entity clinical teams. These reports are transparent, and available to physicians, nurses and hospital units. Knowing where improvement was needed and being able to make those improvements in a targeted and efficient manner rather than “one-size fits all” model is a crucial aspect of improving compliance and quality of sepsis care.

Automated Alerts and Central Alert Teams (CATs)

The Sepsis CET, in conjunction with information technologists, developed algorithms based on hemodynamic changes (or thresholds for a single reading) and lab value combinations. These focused not only on those with diagnosed sepsis, but those patients who had increased risk of developing sepsis. These data spawned alerts which were immediately sent to the CAT. Leveraging the EHR, the combination of alerts (tools) and treatment bundles (mechanisms) have supported early recognition and timely adoption of evidence-based care. Other healthcare systems have embedded treatment protocols and/or automated sepsis alerts, only to continue to struggle with reliably of delivering best care. An array of solutions was considered, including Google Glasses, beepers and alarms to support a reliable vehicle by which WSH could provide real time, care team decision support. Applying observations of air traffic control towers and virtual ICU bunkers, WSH ultimately chose to resource a Central Alert Team (CAT) to provide continuous system-wide oversight to review and validate sepsis alerts, offer clinical decision support, and monitor for treatment timing and compilation; supporting timely, evidenced-based treatment and related positive patient response.

The CAT development process focused on the following: 1) the presence of enhanced judgement (algorithms to support evidence-based decisions), 2) the integration of human elements of care (empathy, ethical choices, emotional connections with patient needs and desires), and 3) communication across the continuum of care. Based on these principles, a team of registered nurses (RNs) with intensive care unit and/or emergency department experience was brought together and trained in sepsis, to devise and implement this innovative idea, in the form of a new care team model16. The training included formal instruction in the pathophysiology of sepsis, a review of evidence based accepted practices in its treatment, system-wide reporting methods and IT training. The CAT was originally comprised of 4.8 FTE registered nurses (RNs) working 24 hours a day, 7 days a week, 365 days a year. As sepsis can be seasonal in nature, PRN RNs have been added to the CAT, providing scalability during times of increased sepsis surge17. The CAT screened, validated, and researched EHR best practice advisories (BPAs) for SIRs, sepsis, pediatric and OB sepsis, and ED triage at all hospitals. While all sepsis alerts continued to register in the patient’s chart, the CAT also received these alerts directly. This resulted in a continuously monitored work queue (mailbox) and supported the bedside team’s awareness and timely intervention. Time zero and the time requirements for components of the 3- and 6-hour treatment bundles are documented clearly in the record for ease in monitoring and tracking patient care and intervention progress by the CAT and the patient’s clinical bedside team.

The initiative was begun at different points in time at the four acute care hospitals in the WSH system. These start times and implementation periods reviewed for this study are reflected in Table 2, along with the length of the baseline period used for comparisons in the data analysis section.

Statistical Data Analysis

Data for the study was generated in the four acute care hospitals during the dates indicated in Table 2 and organized by the WSH Center for Data and Analytics. There were three primary questions related to the SQI to be answered; 1) did the SQI change the bundle completion rate in each of the four acute care hospitals, 2) did the SQI completion rates effect mortality system-wide, and 3) did the number of patients diagnosed with sepsis increase after the SQI? Question 4, did the SQI result in lower than predicted mortality, based on mortality rates in like hospitals provided by Premier Quality Advisor18? All statistical analyses were run using Medcalc software and XLSTAT statistical software19 20.

Question 1 was analyzed using both a univariate linear regression and a Cox-Stuart Trend Test21. The results are found in Figures 1 through 4 and Table 3. The dashed lines on the figure represent a 95% confidence interval of the linear regression line. Each of the acute care facilities demonstrated significant increases in compliance levels over time. The regression models were statistically significant, and demonstrated strong correlations ranging from a low of 0.8 to high of 0.91. The variance accounted for by the model (r2) ranged from a low of 0.64 to a high of 0.84 (see Table 3). In the formulae shown in Table 3, y represents the compliance rate, and x represents time in months (i.e. month 1 is the first month data were collected, month 6 the sixth month, and so on). The start dates for each of the acute care facilities are indicated by arrows on the x-axis in each individual scatter plot. The initial goal set by the SQI team was to achieve a compliance rate of 74% in each of the four acute care hospitals. As seen in figures 1 through 4, compliance goals were being met within 6 to 10 months after starting the SQI .

Question 2 also utilized the same statistical analyses but evaluated the system-wide data of the merged results of all four acute care hospitals. The results are found in Figure 5. There was a significant reduction in system-wide sepsis mortality as the compliance with the SQI increased. Compliance with the bundle explains 64% (r2) of the variance in the linear model, thus nearly 2/3 of the cause of decrease in sepsis mortality was directly due to bundle compliance.

Question 3 sought to determine if the presence of the SQI an effect on the number of sepsis diagnoses had made, irrespective of the compliance with the SQI. In this case the time period from July 1, 2017 through May 1, 2019 was split in three parts. Three of the smaller acute care facilities started the SQI by July 1, 2017, and the largest facility came began the SQI October 1, 2017. Therefore, the first of the three time periods was measured from the earliest date through the first six months and can be referred to as the initiation stage. This reflects the fact that while the SQI use was still in its nascent stages. The next stage is referred to as the implementation stage. At this point, usage of the SQI was becoming more commonly used, though still new. The final stage is referred to as the sustainment stage and indicates the SQI is now the normal way of providing care to septic patients.

These three stages were tested using the Kruskal-Wallis test to determine if the number of patients in each stage differed significantly from one another. It was also used to determine if there was a difference in the overall system-wide compliance rate in each stage. The nonparametric test was used as it was not possible to determine if the population underlying our sample was normally distributed. The results can be seen in Figure 6. There was no significant difference in the numbers of patients diagnosed in each of the three periods (p = 0.189). Figure 7, however, shows a significant and dramatic increase in compliance as the SQI moves forward in time through the three stages. The omnibus test of the Kruskal-Wallis is significant, and the post hoc test reveals that all three stages differ from each other, and that compliance goes up with each subsequent stage (initiation stage vs. implementation stage, p = 0.032, implementation stage vs. sustainment stage, p = 0.002, and initiation stage vs. sustainment stage, p < 0.001).

Question 4 compared the predicted rates of mortality due to sepsis in a comparable group of hospitals, provided by Premier Quality Advisor, to the observed sepsis mortality rate during the period of July 1, 2017 through October 1, 2019. The Two-sample paired Wilcoxon non-parametric test was used to compare these data, again due to the lack of confidence that the underlying population data was normally distributed. The outcomes for the original study period, and an additional evaluation of the data for the last 12 months of the study (i.e. the period during which the SQI had matured into the new normal), are presented in Table 4. Note that the median mortality rate of 5.67% is significantly less than the predicted mortality rate of 7.58% in the data sampled during the overall study period. Further, in the final 12-month period, the median mortality rate of 5.29% is significantly less than the predicted mortality rate of 7.75%.

Discussion

The SQI increased bundle compliance at each of the four acute care hospitals, and systemwide. In the last 12 months recorded for this study, the systemwide bundle compliance rate averaged 90.4% (2.8% standard deviation) and the mortality rate averaged 4.5% (1.3%). The increase in bundle compliance resulted in a significant decrease in sepsis mortality.

History and the literature support the reality that simply sharing alert algorithms and workflow diagrams will not directly lead to realization of the WSH outcomes successes. The comprehensive “new care model” as it has been developed and implemented at WSH, is the key to success. The CAT is an illustration of the power of a care platform integrating digital science (technology) and biological science (best practice medicine) to augment humanity (human strengths) through:

  • Enhanced judgement (algorithms to support evidence-based decisions)
  • Leverage humanity: Empathy, Ethical choices, Personal choices, Exceptions, and personalizing these elements via emotional connection
  • Communication across the continuum (See Table 1 and Figure 8).

There were some limitations in this study. The extent to which mortality still occurs in sepsis might have been better had we delineated the type of sepsis diagnosis (sepsis, severe sepsis, or septic shock) was related to the mortality rate. The mortality rate for septic shock, for example, is higher (and the interventions more complex) than for those patients presenting with the diagnosis of sepsis. There are also numerous opportunities for further QI initiatives and research. We noted significant staff turnover during the initiative and are concerned that this might have an impact on the implementation of the SQI.

Nevertheless, our initiative resulted in compliance rates far higher and mortality rates distinctly lower than those noted in other reported studies1-6. A very recent study, conducted during nearly the same time frame as our data were collected, showed compliance rates of 43% to 55%, with the lowest compliance percentage being found in teaching hospitals.

The BPAs and CAT described are our “Version 1.0.” Future versions may soon consider current alert algorithms to leverage machine-based learning and artificial intelligence (AI) in support of more robust predictive analytics23. These may then trigger orders from protocol driven smart order sets (whose foundation is evidence based), combined with local factors. Artificial intelligence will also support identification of sepsis subclassifications and directed therapies. These possibilities will further allow us to leverage humanity, providing the bedside teams more time with the patient. Given the early successes, the possibilities for similar future applications are limitless.

The EHR and this new care model has allowed WSH to learn in new ways, beyond current limited linear and nonlinear learning. WSH has learned the evidence of sepsis and coded algorithms for early identification, including when alert thresholds worked, did not work and best worked, and adjusted, with real time revisions and customizations to learned thresholds/algorithms and triggers. The WSH SQI continues to evolve becoming better prepared to predict patients most likely to recover, those more likely to relapse or deteriorate, and which patients are more vulnerable, or at risk. The EHR has provided the tools to support better, more timely, reliable care. The new platform of care; clinical and technical staff working together and leveraging the EHR, will transform the way in which care is provided.