Sepsis is one of the leading causes of hospital mortality.1 Patients who survive sepsis have longer lengths of hospital stay and higher rates of hospital readmission,1 as well as significant, persistent new cognitive and functional impairments.2 The Agency for Healthcare Research and Quality has reported that sepsis is the most expensive condition treated in U.S. hospitals, with the cost of treatment nearly $24 billion in 2013.3
For the past 20 years, clinical research and sepsis care guidelines have focused on the most severely ill patients.4 Initially, these patients were identified by systemic inflammatory response syndrome (SIRS) criteria, coupled with such severity of illness measures as hypotension and lactic acidosis.5 (See Systemic Inflammatory Response Syndrome (SIRS) Criteria.4) A similarly narrow group of patients has been identified by elevated Sequential Organ Failure Assessment (SOFA) scores.4, 5 (An online SOFA calculator can be found at www.mdcalc.com/sequential-organ-failure-assessment-sofa-score.)
Some research, however, suggests that interventions targeted to prevent progression to septic shock in patients who have infection but do not meet the criteria for severe illness may have a greater effect on survival than interventions targeted to only the sickest group. In fact, Liu and colleagues found that nearly 60% of 1,817 sepsis-related deaths among patients admitted with sepsis occurred in those with normal blood pressure and lactate levels of less than 4 mmol/L.6 Furthermore, bundle adherence in the ED has been associated with lower rates of progression to septic shock.7
Study aim. In this article, we describe a single-center, multiyear quality improvement (QI) initiative designed to promote early recognition and treatment of sepsis, and we report on the retrospective, interrupted time-series cohort evaluation conducted to gauge its effect on sepsis-related mortality rates, bundle adherence, and the need for rapid response team (RRT) calls.
In 2012, the Virginia Mason Medical Center in Seattle, a multidisciplinary health care network in the Pacific Northwest, made an organizational decision to go beyond the standard guidelines of the Surviving Sepsis Campaign (SSC) in addressing early sepsis recognition and treatment. At that time, the medical center employed approximately 450 physicians and 550 hospital-based nurses, and annually received about 800,000 outpatients in addition to admitting roughly 17,000 inpatients to its single urban hospital (a second hospital in Yakima, Washington, joined the system after completion of this initiative). At the time, processes supporting the recognition and treatment of severe sepsis and septic shock bundle adherence were in place in the ED and critical care unit, and the hospital was served by an RRT. The processes focused on sepsis detection and treatment, education, use of standardized order sets, and bundle adherence follow-up. Our plan to focus on early sepsis intervention and the QI initiative that followed were in addition to this more traditional approach.
We initially set as a target the identification and treatment within one hour of all patients affected by sepsis, regardless of illness severity or care unit. To meet this goal, we developed innovative strategies for early intervention, including nurse-directed sepsis care in both inpatient and ED settings. Our subsequent evaluation of this initiative was to determine its effect on patient outcomes as reflected in sepsis-related mortality rates, bundle adherence, and the need for RRT calls. The evaluation was completed as a part of our QI project and therefore was determined to be exempt from the rules governing research on human subjects by our institution's institutional review board. The institution has an established Lean framework for QI and daily management. (For information on applying Lean principles, see https://library.ul.com/wp-content/uploads/sites/40/2015/02/UL_WP_Final_Applying-Lean-Principles-to-Improve-Healthcare-Quality-and-Safety_v11_HR.pdf.)
The QI initiative was introduced through a multidisciplinary, executive-led sepsis guiding team comprising physicians, nurse leaders, data analysts, and a nurse sepsis coordinator. This group defined the vision, provided oversight, and ensured that the initiative's foundational elements—a focus on early sepsis and on empowering the bedside nurse to assess for sepsis and initiate treatment—remained constant over the five-year intervention and postintervention period. The project was executed through a series of Lean improvement events that targeted early intervention and a nurse-directed approach to sepsis care.
Although the initiative was based on the SSC guidelines, in accordance with our goal of achieving early recognition and treatment of sepsis, as well as research concerning the benefit of early antibiotic administration,5, 8-10 we made the following modifications:
- We set as a target the completion within one hour of the SSC three-hour bundle.11
- To reduce complexity, we used a standard fluid challenge of 2 L, instead of 30 mL/kg.11
- We did not limit bundle application to patients with markers of severe illness, but rather applied the bundle to patients with suspected infection and two or more SIRS criteria.
Intervention phase one, ED Code Sepsis, went live in late 2012, after physicians, nurses, and patient care technicians had been trained in early sepsis recognition and treatment. Clinicians who joined the organization over the course of this initiative received training in orientation, from preceptors, in a postresidency seminar, and through a required sepsis education module. Since nurse turnover is a challenge for all QI initiatives, we ensured that sepsis education was included in orientation, on-unit training competency checklists, and required online education. The ED Code Sepsis process began with triage nurses, who screened every patient for SIRS, flagging those who met the criteria for an ED room and immediate evaluation by an assigned ED provider (a physician, NP, or physician assistant). ED point-of-care equipment was used to measure lactate levels promptly. Physicians, NPs, RNs, and physician assistants served as on-the-ground sepsis experts and champions.
Intervention phase two, the inpatient Power Hour, went live in the summer of 2014. Root-cause analysis of our hospital's sepsis process revealed that improvements were needed in the areas of early detection and treatment. As with the triage nurse role in the ED, the guiding team recognized that the nurse at the bedside was uniquely qualified to detect subtle changes in a patient's condition. As other studies have demonstrated, empowering the bedside nurse to assess patients for sepsis and act accordingly can reduce or eliminate delays inherent in prescriber-driven processes.12-14 The inpatient Power Hour was designed to occur at the earliest possible sign of sepsis, before the patient developed septic shock. We anticipated that providing sepsis care before acute decompensation (signaled by hypotension and organ dysfunction) would obviate the need for RRT intervention, while minimizing disruption and stress for the patient and family.
Just as ED Code Sepsis required triage nurses to screen all patients for SIRS, the inpatient Power Hour required bedside nurses to screen all inpatients under their care for SIRS and for the possibility of infection at least once every shift. For any patient who met SIRS criteria, the nurse independently initiated Power Hour through the electronic health record (EHR) order set, which called for point-of-care lactate level measurement, two sets of blood cultures, and delivery of a 500-mL bolus of normal saline. In addition, Power Hour initiation automatically signaled the pharmacy to anticipate new sepsis antibiotic orders. The process could proceed without prescriber input until the lactate result was available for further diagnosis and care planning, at which point the prescriber was responsible for making a final sepsis diagnosis and ordering antibiotics and the remaining 1,500-mL fluid bolus.
Each antibiotic order triggered an expedited pharmacy review, preparation, and delivery. The nurse sepsis coordinator followed up on each Power Hour initiation within one week, sending an e-mail to the nurse and the unit-level nurse leader. The e-mail detailed specifics about the Power Hour (such as patient name, unit, date, time, triggering criteria, possible source of infection, actions taken, and patient outcome) and thanked the nurse, asking her or him to discuss the experience in unit-based huddles. For patients who received a new antibiotic order, the nurse sepsis coordinator sent a similar e-mail to the prescriber and the section head of the prescriber's department. This feedback loop engaged stakeholders at the bedside by celebrating small wins. As in the ED, unit-level champions acted as change agents and peer educators.
Change management. ED Code Sepsis and the inpatient Power Hour required management to modify its approach to implementing change and to helping personnel with the transition to new policies. Though our medical center's leadership was supportive of the initiative and the executive-led sepsis guiding team had gained approval from the hospital's medical staff committee, acceptance by individual prescribers (primarily physicians and physician assistants) varied in the early phases of the project. Some said they were uncomfortable with nurses independently ordering laboratory tests and fluids (expressing concern that the latter could increase patient risk of fluid overload). Education, real-time feedback, and one-on-one follow-up with prescribers during the first year of the initiative was effective in addressing this challenge. Nurse education further promoted staff acceptance. Because the SIRS criteria is an imperfect sepsis screening tool,4 we recognized that not all inpatients for whom nurses initiated Power Hour would have sepsis and that the best outcome for every Power Hour was a meaningful conversation between the prescriber and the nurse about the plan of care, regardless of the diagnosis.
Data collection. To gauge the effects of our QI initiative, we conducted a retrospective, interrupted time-series cohort evaluation of the preintervention (January 2010 to July 2012), intervention (August 2012 to August 2015), and postintervention (September 2015 to December 2016) periods, using the in-hospital sepsis-related mortality rate as the primary outcome.
We also considered as process metrics the initiation of ED Code Sepsis and the inpatient Power Hour; order set use; bundle adherence; and sepsis-related RRT calls, which were viewed as adverse events, as they represented a potentially missed Power Hour and thus a delay in the initiation of sepsis care. In addition to their evaluative purpose, these measures supported management of the program in real time.
We identified activations of ED Code Sepsis from paging data and activations of Power Hour from order set usage reports in the EHR. A nurse sepsis coordinator performed a retrospective chart analysis of all such cases to determine completion times for each bundle element and documented any nonsepsis acute interventions, such as blood product administration or changes in the plan of care, that were put into effect within an hour of Power Hour. Sepsis bundle adherence was extracted from the EHR for severe sepsis cases (those with lactate levels greater than 2 mmol/L or sustained hypotension).15 Every case of severe sepsis was retrospectively audited by the nurse sepsis coordinator to ensure that severe sepsis criteria were met. Both inclusion and exclusion criteria for this group aligned with those of the Centers for Medicare and Medicaid Services sepsis core measure.16 Paging data and RRT documentation in the EHR were cross-referenced with sepsis incidence to provide a list of sepsis RRT calls.
Statistical analysis of the primary outcome of in-hospital sepsis-related mortality rate was performed using interrupted time-series regression, after adjusting for patient age, sex, and type of admission (medical or surgical), variables well documented in medical and nursing literature as affecting sepsis outcomes. Interrupted time-series regression allowed us to understand temporal trends, increased our awareness of changes over time, and enabled us to adjust for autocorrelation in our time-series data. We identified autocorrelation using the Durbin–Watson statistic and the Cochrane–Orcutt procedure and then used the Prais–Winsten estimation for correction. For each time-series model, the final value of interest was the comparison of outcomes between the postintervention and preintervention periods. Preintervention and postintervention patient characteristics and outcomes were compared using the t or χ2 test. Data were also explored graphically using run charts. All analyses were performed using Stata 12 (StataCorp, College Station, TX).
As in other sepsis research, we defined the in-hospital sepsis-related mortality rate as the number of deaths among patients diagnosed with sepsis. Although our approach in evaluating our QI initiative was to focus on early detection of sepsis, one challenge with this approach, as other researchers have pointed out, is the likelihood that early detection may increase the apparent prevalence of sepsis by identifying less severe cases, resulting in deceptive reductions in the sepsis-related mortality rate.13 Accordingly, we also determined the prevalence of sepsis over all three time periods evaluated and, as a secondary outcome, sepsis-related deaths among all hospital patients (those with and without sepsis diagnoses).
Over the course of the seven-year evaluation period, there were 106,220 hospital discharges. Female patients represented 52.1% of hospital discharges during the preintervention period, but less than half of discharges (49.3%) during the postintervention period. Mean patient age decreased from 65.1 to 64.4 years, and length of stay increased from 4.5 to 5.2 days. Sepsis discharges represented 8.4% of preintervention hospital discharges and 9.4% of postintervention hospital discharges (see Table 1).
The in-hospital sepsis-related mortality rate (calculated as sepsis-related deaths per sepsis discharges), the primary outcome, was 12.5% during the preintervention period. This rate dropped by 36% during the postintervention period, with an absolute reduction of 4.5 sepsis-related deaths per 100 sepsis discharges (95% confidence interval [CI], −1.8 to −7.2; P < 0.001) (see Table 2).
The rate of sepsis-related deaths among all hospital patients (calculated as sepsis-related deaths per hospital discharges), a secondary outcome, was 1.05% during the preintervention period. This rate dropped by 31.4% during the postintervention period, with an absolute reduction of 0.33 sepsis-related deaths per 100 hospital discharges (95% CI, −0.07 to −0.59; P = 0.015).
Sepsis prevalence (calculated as sepsis discharges per hospital discharges) was 8.4% during the preintervention period and did not increase significantly in the postintervention period, with an absolute increase of 0.5 sepsis discharges per 100 hospital discharges (95% CI, −0.74 to 1.74; P = 0.43).
Intervention process measures. A total of 145 inpatient Power Hours were initiated in the postintervention period. Of these, 49.7% (72) received a new antibiotic, and an additional 43.4% (63) received treatment for a nonsepsis acute condition. To date, there have been no known adverse effects, critical care unit transfers, injuries, or deaths as a result of the introduction of the in-hospital Power Hour. Sepsis-related RRT calls per hospital discharges decreased from 2.2% to 0.85% between the pre- and postintervention periods, an unadjusted effect size of −1.4 percentage points (95% CI, −1.2 to −1.6; P < 0.001). Bundle adherence at three hours improved from 40.5% (251/620) to 73.7% (140/190, P < 0.001). Bundle adherence at one hour improved from 2.3% (14/620) to 19.5% (37/190, P < 0.001), and antibiotic administration at one hour improved from 6.6% (41/620) to 31.1% (59/190, P < 0.001) throughout the institution. Postintervention ED triage to antibiotic time averaged 80 minutes.
Through this QI initiative we demonstrated that nurse-directed, early sepsis intervention and adherence to a modified bundle approach was successful in reducing our in-hospital sepsis-related mortality rate.
The literature is limited on both early sepsis intervention and nurse-directed sepsis care in the ED. While some studies have shown that a bundled approach and nurse-based sepsis care can significantly reduce sepsis mortality,7, 17 others have struggled to achieve consistent bundle adherence and significant mortality benefits,12 or have achieved significant improvements in bundle adherence with no significant improvement in mortality rates.13 Our work not only supports early nurse-directed sepsis intervention and a bundled approach, but also suggests that bundle adherence combined with aggressive target times may further improve patient outcomes.
Studies focused on improving inpatient sepsis treatment have largely targeted sepsis detection through an automated early warning system (EWS) embedded in the EHR or alerts that call an assessment team or RRT.18-20 The usefulness of such tools, however, has been inconsistent across studies.18 In a multicenter study by Umscheid and colleagues, the EWS had a 16% sensitivity for sepsis.20 Additionally, though the researchers found that early sepsis detection increased with EWS implementation, they did not find a statistically significant reduction in any mortality measure.20 A recent study by Arabi and colleagues reported a significant decrease in hospital mortality after EWS implementation, but the study was conducted in an ED and targeted only patients with severe sepsis and septic shock.19
In 2015, Jones and colleagues reported a significant decrease in the inpatient mortality rate at a tertiary teaching hospital after the introduction of a nurse-directed, nonautomated, electronic SIRS-based screening program.17 Patients were screened by bedside nurses on admission, at 12-hour intervals, and with any change in condition. Those with positive screens were evaluated within an hour by an NP, and treatment was initiated in accordance with SSC guidelines for management of severe sepsis and septic shock. Our inpatient Power Hour was similar in that screening was SIRS based and nurse directed, but instead of bringing in an NP to perform a secondary assessment, we empowered the bedside nurse to initiate treatment and tasked the prescriber with final clinical decision making and bundle completion, as appropriate.
The clinical setting is likely to affect a sepsis program's success. The Lean framework for QI,21 which incorporates multidisciplinary collaboration and effective plan–do–study–act cycles, was already in place at our institution before our QI initiative was introduced. Strong support for nursing empowerment among executive and physician leaders fostered change management, as did the institution's dedication of such resources as a multidisciplinary clinical guiding team, data analysts, and a nurse sepsis coordinator, all of which facilitated follow-up and accountability.
Limitations. Our evaluation was limited in the following respects:
- It was performed retrospectively, using historical controls.
- Since implementation of multiple program components occurred simultaneously, we were unable to measure the specific contribution of each.
- The intervention was applied and evaluated at a single urban teaching hospital, which may limit generalizability to other settings.
- Our teams used International Classification of Diseases codes to identify patients for sepsis bundle adherence and mortality metrics, which may have caused us to exclude some patients with sepsis from the evaluation.22
- Although we calculated our in-hospital sepsis-related mortality rates, we did not follow patients after discharge.
One challenge in sepsis research is that efforts to improve detection may include less critically ill patients, thereby potentially increasing the overall prevalence of sepsis and biasing mortality measures.13 As in previous sepsis QI reports and clinical trials,9, 13, 20, 23 we report the in-hospital sepsis-related mortality rate (sepsis-related deaths per sepsis discharges) as our primary outcome. However, we also report sepsis-related deaths among all hospital patients and the prevalence of sepsis. Through these two additional measures, we demonstrated that, though we saw a nonsignificant upward trend in sepsis identification, that alone did not provide a plausible explanation for the reduction we observed in both mortality measures. The rise we observed in sepsis diagnoses occurred in the early stages of the intervention and, generally, did not correspond temporally with the decline in either of the two mortality measures. In the later stages of the intervention, sepsis prevalence remained relatively flat, while both the in-hospital sepsis-related mortality rate and the number of sepsis-related deaths among all hospital patients declined. In addition, the magnitude of the reductions we report in both mortality measures is greater than that of the upturn in sepsis identification. The reduction in sepsis-related deaths among all hospital patients—those with and without sepsis diagnoses—suggests that increased diagnoses did not substantially bias our results (that is, since our intervention was effective in decreasing both mortality measures, these decreases were not merely a result of the rise in sepsis diagnoses).
A challenge specific to our evaluation was determining the relative contribution of each component of the intervention. The SSC bundles for patients in septic shock,11 which form one component, have been validated previously.9 We added early intervention protocols in the ED and on hospital units to improve the timeliness of intervention and to direct intervention toward patients with less severe sepsis. The reduction we observed in the rate of sepsis RRT calls underscores the value of detecting and treating sepsis before it progresses to severe sepsis or septic shock, as has been reported in previous studies.7, 24 Because hospital-acquired sepsis is less common than community-acquired sepsis, which presents in the ED, we were unable to distinguish the effects of the in-hospital versus the ED components of our intervention.
The nonspecific nature of the SIRS criteria as diagnostic criteria for sepsis has been discussed by the Third International Consensus Definitions for Sepsis and Septic Shock—or Sepsis-3—task force.4 We too found these criteria nonspecific, as only 40% of patients for whom Power Hour was initiated were determined to have been newly affected with sepsis. Nevertheless, an additional 35% of these patients required some acute intervention unrelated to sepsis or a change in their plan of care, suggesting that the benefit of our nurse-directed process extended beyond the population of patients with sepsis diagnoses.
Implementation of a sepsis program emphasizing nurse-directed identification and treatment of early sepsis before the development of septic shock, and in addition to traditional bundle adherence, reduced our in-hospital sepsis-related mortality rate. Though our program was also associated with a nonsignificant rise in sepsis diagnoses, this increase was insufficient to explain the reductions we observed in both the in-hospital sepsis-related mortality rate and sepsis-related deaths among all hospital patients after introducing this initiative. Nurse-directed care was a critical component of our initiative's success. By leveraging the skills and expertise of nurses, we were able to remove such barriers to sepsis care as ineffective communication among care team members and location of care delivery within the hospital. Our results support the inclusion of nurses in sepsis care both at the bedside and on organizational improvement teams.
2. Iwashyna TJ, et al Long-term cognitive impairment and functional disability among survivors of severe sepsis
JAMA 2010 304 16 1787–94
3. Torio CM, Moore BJ National inpatient hospital costs: the most expensive conditions by payer, 2013
. Rockville, MD: Agency for Healthcare Research and Quality; 2016 May. #204. Statistical Brief; https://www.ncbi.nlm.nih.gov/books/NBK368492
4. Singer M, et al The Third International Consensus Definitions for Sepsis
and Septic Shock (Sepsis
-3) JAMA 2016 315 8 801–10
5. Angus DC van der Poll T. Severe sepsis
and septic shock N Engl J Med 2013 369 9 840–51
6. Liu V, et al Hospital deaths in patients with sepsis
from 2 independent cohorts JAMA 2014 312 1 90–2
7. Miller RR, et al Multicenter implementation of a severe sepsis
and septic shock treatment bundle Am J Respir Crit Care Med 2013 188 1 77–82
8. Kumar A, et al Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock Crit Care Med 2006 34 6 1589–96
9. Seymour CW, et al Time to treatment and mortality during mandated emergency care for sepsis
N Engl J Med 2017 376 23 2235–44
10. Sherwin R, et al Does early and appropriate antibiotic administration improve mortality in emergency department patients with severe sepsis
or septic shock? J Emerg Med 2017 53 4 588–95
11. Dellinger RP, et al Surviving sepsis
campaign: international guidelines for management of severe sepsis
and septic shock: 2012 Crit Care Med 2013 41 2 580–637
12. Bruce HR, et al Impact of nurse-initiated ED sepsis
protocol on compliance with sepsis
bundles, time to initial antibiotic administration, and in-hospital mortality J Emerg Nurs 2015 41 2 130–7
13. Gatewood MO, et al A quality improvement
project to improve early sepsis
care in the emergency department BMJ Qual Saf 2015 24 12 787–95
14. Winterbottom F Nurses’ critical role in identifying sepsis
and implementing early goal-directed therapy J Contin Educ Nurs 2012 43 6 247–8
15. Shankar-Hari M, et al Developing a new definition and assessing new clinical criteria for septic shock: for the Third International Consensus Definitions for Sepsis
and Septic Shock (Sepsis
-3) JAMA 2016 315 8 775–87
16. Faust JS, Weingart SD The past, present, and future of the Centers for Medicare and Medicaid Services quality measure SEP-1: the early management bundle for severe sepsis
/septic shock Emerg Med Clin North Am 2017 35 1 219–31
17. Jones SL, et al Reductions in sepsis
mortality and costs after design and implementation of a nurse-based early recognition and response program Jt Comm J Qual Patient Saf 2015 41 11 483–91
18. Alberto L, et al Screening for sepsis
in general hospitalized patients: a systematic review J Hosp Infect 2017 96 4 305–15
19. Arabi YM, et al The impact of a multifaceted intervention including sepsis
electronic alert system and sepsis
response team on the outcomes of patients with sepsis
and septic shock Ann Intensive Care 2017 7 1 57
20. Umscheid CA, et al Development, implementation, and impact of an automated early warning and response system for sepsis
J Hosp Med 2015 10 1 26–31
21. Blackmore CC, Kaplan GS Lean and the perfect patient experience BMJ Qual Saf 2017 26 2 85–6
22. Rhee C, et al Incidence and trends of sepsis
in US hospitals using clinical vs claims data, 2009-2014 JAMA 2017 318 13 1241–9
23. Seymour CW, et al Delays from first medical contact to antibiotic administration for sepsis
Crit Care Med 2017 45 5 759–65
24. Cross G, et al The epidemiology of sepsis
during rapid response
team reviews in a teaching hospital Anaesth Intensive Care 2015 43 2 193–8