Original Contribution

Implementation of a bundle of actions to improve adherence to the Surviving Sepsis Campaign guidelines at the ED

Jose M. De Miguel-Yanes MD?,1, Javier Munoz-Gonzalez MD ¹, Juan A. Andueza-Lillo MD, Berta Moyano-Villaseca MD,

Victor J. Gonzalez-Ramallo MD, Ana Bustamante-Fermosel MD

Servicio de Urgencias, Hospital General Universitario “Gregorio Maranon“, 28007 Madrid, Spain

Received 9 May 2008; accepted 16 May 2008

Abstract

Purposes: We had previously demonstrated surviving sepsis campaign guidelines had not had enough impact at our Emergency Department.

Basic Procedures: Actions directed to increase the qualification of our staff and residents, to facilitate guidelines divulgation and to improve spatial conditions by creating a high dependency unit were implemented as a bundle. The impact of these actions on the achievement of early objectives of the campaign and on mortality was analyzed.

Main Findings: Following campaign guidelines was more frequent after the implementation of these actions, as shown by less restrictive fluids administration for more severe cases (P = .001), earlier administration of antibiotics (P = .001) and lactate determination rate (46% vs. 12%). In-hospital mortality difference did not reach statistical difference. Physicians were able to identify high-risk patients on clinical grounds.

Principal Conclusions: The bundle of actions has had a moderate beneficial effect on our Emergency Department. High Dependency Units are useful for managing patients not fulfilling criteria for Intensive Care Unit admission.

(C) 2009

Introduction

Sepsis is a leading cause of in-hospital death [1]. In a previous study, we demonstrated that the Surviving Sepsis Campaign had not had enough impact on the medical care

* Corresponding author.

E-mail address: [email protected] (J.M. De Miguel-Yanes).

¹ Jose M De Miguel-Yanes and Javier Munoz-Gonzalez contributed equally to the writing of the manuscript.

delivered to patients with sepsis at our emergency depart- ment (ED) [2]. Whereas some patients were detected early and subsequently admitted to the intensive care unit (ICU), others often spent hours at our ED without receiving standard of care. The main reasons we could identify (“key success factors”) were lack of divulgation of the Surviving Sepsis Campaign among our residents and staff, low sensitivity to detect cases in early stages of the process, and local spatial conditions that hampered supervision and optimization of clinical management [3].

Bearing these peculiarities in mind, several actions were put in practice to improve medical care for patients with

0735-6757/$ - see front matter (C) 2009 doi:10.1016/j.ajem.2008.05.010

sepsis at our ED. We planned this “second-step” study to analyze, as a primary end point, the impact of these actions on the accomplishment of the Surviving Sepsis Campaign “bundle” measures and, secondarily, on mortality [4]. Other parameters were evaluated as well to explore further associations between different variables and the outcome of the patients.

Methods

The aforementioned key success factors prompted us to design actions specifically directed to the residents’ work, to achieve overall staff sensitization to sepsis, and to improve physical structures. The actions implemented to improve medical care of these patients were as follows:

For residents: Each year, first-year residents had to spend a 2-week period of training at the ED, with a daily 45-minute meeting for reviewing relevant processes, one of which was severe sepsis and septic shock for each group. Furthermore, they

all received an introductory course at their arrival to the hospital and a proficiency one some months later; in both of them, local experts exposed and analyzed common clinical issues while managing patients with sepsis.

For widespread divulgation: Guidelines for diagnosing and treating different processes including sepsis were elaborated as “visual” and easy-to-read tools (Fig. 1). They were available at every computer of our hospital through our local Web site.

For achieving a better environment for medical care: We thought many of these patients would benefit from being segregated from other less seriously ill patients [5]. With this purpose, a High-Dependency Unit (HDU) was designed for our ED. This dispositive idea was conceived to give support to the medical ICU, which meant that patients could be admitted as they progressively needed a higher level of care than the one given at an ordinary observation ward at the ED. The HDU started its activity in February 2007 with the commitment of giving care to patients with severe sepsis or septic shock not requiring an ICU bed as estimated by attending physicians.

The HDU was designed to contain 12 beds. Ratio for nursery was 4:1 most of the time. From Monday to Friday, 2 ED staff physicians usually worked in the morning shift; and

Fig. 1 Local institutional Web site. Sepsis page.

at 2:30 PM, a different staff physician started work until 8:00 AM the next day. On weekend days, a staff physician did a complete 24-hour shift. Residents were encouraged to rotate in the unit, but workload was calculated so as to be able to work without them. Among the ED staff members, a group of 8 constituted the initial team who worked at the unit. As the HDU continued with its activity, more physicians have progressively joined the former group. Having started with a small group has allowed a better supervision of medical care, an earlier detection of problems, and an easier up-to-date instruction of the implicated professionals. This policy equally applied for nurses.

Physicians discretionally managed their patients at the general ED observation ward or consulted with ED physicians at the HDU or ICU staff, depending on the previous functional state or the severity of the current process. Endotracheal intubation, use of Recombinant human activated protein C, and invasive monitoring of Central venous oxygen saturation were not standard procedures to be performed at the unit. Whenever an ED physician thought ICU admission, HDU admission, or even staying longer at the general observation ward was the appropriate option for a patient, the decision was respected.

For identification of cases, methods were similar to those previously described for our historical 2005 cohort [2]. In brief, this was an observational study carried out in one large tertiary care university hospital in Madrid whose adults’ ED attended to 200504 patients in 2007, after excluding children and gynecology patients, who were seen at another building. The area covered by our hospital includes 720000 people from southeastern Madrid City and province. Our medical ICU had 22 beds at the time of the study. Our institution has coronary, postanesthesia, and posttransplantation units, too. Patients were included whenever physicians from the General Internal Medicine Section of the ED had made the diagnosis of sepsis, severe sepsis, or septic shock for them at any point of their stay at our department. Furthermore, they had to have spent a minimum of 6 hours at our ED; thus, patients with an “earlier than 6 hour” admission to the ICU or those who had died shortly after their arrival were excluded

to make results comparable with our historical cohort.

Physicians on duty at the Observation Units were contacted at least 3 times a day for the recognition of cases. In addition, the informatics admission program of our hospital was consulted to identify all those patients admitted with diagnoses related to sepsis who might have not been detected by other means. The ICU admissions coming from the ED during that period were also checked. The competent hospital authority approved collection and statistical analysis of the data.

Epidemiologic and clinical data were collected. Five items as proposed by the Surviving Sepsis Campaign were surveyed [4]: (1) measurement of blood lactate; (2) obtaining blood samples for culture before starting with antibiotics; (3) early use of antibiotics (b3 hours); (4) Fluid replacement with at least 20 mL/kg of crystalloids in the first hour-unless contraindicated-if hypotensive or if

blood lactate N4 mmol/L; and (5) use of vasoactive drugs— other than dopaminergic doses of dopamine-if hypoten- sive or in shock. Central lines were available for adminis- tration of inotropic drugs, but measurement of central venous oxygen saturation was not routinely performed in this setting because equipment was not available out of the ICUs. Therefore, this issue of the campaign was not recorded as a primary aim.

As secondary end points, additional data were collected to have a deeper glance at the clinical management of these patients and to identify those variables associated with mortality. These data included age; sex; Predisposing factors to infection; vital signs; laboratory results; Acute physiology and chronic health evaluation II score, calculated with the worst values of the first 6 hours since presentation at the ED [6]; number of failing organs, as defined in previously published guidelines; stage of severity according to the attending physician; and actual stage of severity following universal definitions of sepsis stages [7,8]. Suspected source of infection by attending physicians and rates of obtaining blood, urine, or other biological samples for culture were recorded. Total amount of administered fluids per patient was calculated as reflected in the nurses’ charts. Use of vasoactive drugs, steroids, deep venous thrombosis prophylaxis, relatively tight control of glycemia with insulin to keep it less than 150 mg/dL, and antibiotics

use patterns were surveyed.

The number of cases a consultation to the intensive care specialist was done or the reasons not to do it or not to be finally admitted to the ICU were registered. Other collected data were microbiological isolates and rate of positivity of blood samples, and also if antibiotics were considered to be appropriate a posteriori regarding bacterial resistances. The final diagnosis and length of stay until death or discharge, as well as the outcome of every patient, were inquired.

All tests for significance and resulting P values were 2-sided, with a level of significance of .05. Fisher exact test and ?2 test were used where appropriate; for continuous variables, Kolmogorov-Smirnov test served to identify those variables with a normal distribution, for which the t test was used. Other nonparametric tests, such as Mann-Whitney and Kruskal-Wallis tests, were used for the remaining variables as needed. Those variables significantly associated to mortality in the univariate analysis were included in a backward logistic regression multivariate model. The SPSS version 15.0 (LEAD Technologies, Inc, Charlotte, NC) was used as statistical support.

Results

Although we had planned a 2-month study for our historical cohort (September and October 2007), collection of cases had to be interrupted after 30 days because organizational changes that might have biased some results occurred: residents started

to join their ED shifts after 3:00 PM, which meant that morning shifts began to be supplied with staff only.

Comparison between 2005 and 2007 cohorts

For that 1-month period, 50 cases were identified by attending physicians as having sepsis at any stage of severity; it represents a higher sensitivity for diagnosing sepsis than our historical 2005 cohort (53 cases within 2 months). All patients had an infectious disease at discharge or death, and all but one actually fulfilled systemic inflammatory response syndrome criteria. Mean age was

70.7 years (71.8 years in 2005, P = .44); median, 79.5 years. No significant difference was found for sex (52% male, P =

.3). The APACHE II score tended to be lower than that in 2005 (15.9 +- 6.9 vs 18.3 +- 6.5, P = .07), but significantly more patients had basal conditions that made them more prone to have sepsis (P = .014) (Fig. 2).

Thirty-five patients were considered to have “sepsis”; 7, “severe sepsis”; and 8, “septic shock.” When analyzed according to actual severity, the proportion was 25, 15, and 10, respectively. No difference was found regarding number of failing organs between both cohorts (P = .45) (Fig. 3). Underestimation of actual severity took place in a somewhat lesser degree than in the 2005 cohort, although it did not reach statistical significance (P = .2).

Concerning primary end point, a complete achievement of the guidelines’ objectives took place in 26% (0% in 2005 cohort), representing 4% of the cases at the general observation ward vs 41% at the HDU (P = .05). In detail, lactate was measured in the first blood sample for 46% of the cases (12% in 2005). Blood cultures were obtained for 67% of the patients (85% in 2005, P = .37). When attended at the HDU, the rate increased to 78%. More cases of respiratory sepsis, associated to a poorer outcome in our 2005 cohort, were found, although it did not reach significance (38% vs 18%, P = .24).

Discretional administration of fluids was less restrictive than in 2005 (P = .1) (Fig. 4). More seriously ill patients received more fluids (P b .001). Furthermore, the number of hours a patient spent before receiving his or her first dose of

Fig. 2 Basal conditions.

Fig. 3 Failing organs.

antibiotics was reduced as compared with 2005 (3.8 vs 6.0, P = .001); median time was 2 hours (5 hours in 2005), and 62% of the patients received it within the first 3 hours (Fig. 5). Actual severity was not significantly associated to earlier administration of antibiotics (P = .13).

Rate of blood transfusion was low: only 35% of the patients with a hematocrit less than 30% received it (5 out of 14). Barely 7 patients received vasoactive agents at the ED; all of them were treated at the HDU and met criteria for shock, yet 3 patients fulfilling these criteria did not receive these drugs.

For secondary aims, measurement of hourly urinary output (P = .04), avoidance of dopaminergic doses of dopamine-associated to a poorer prognosis in our historical cohort-(0%, P = .05), and a tighter Glycemic control (P =

.005) showed improvement as compared with 2005. There was a trend to use steroids more frequently (30%, P = .23), to a better prophylaxis of deep venous thrombosis (P = .23), to a more common use of combinations of antibiotics (P = .08), and to prescribe cephalosporins less frequently (P = .4). A priori use of antibiotics was deemed to be correct in 90%. Distribution according to microbiological isolation was virtually identical to the 2005 cohort.

If we excluded those cases attended at the HDU, only time until first dose of antibiotics would then remain significantly better than in 2005 (4.7 vs 6, P = .04), whereas use of

Fig. 4 Volume of fluids.

Fig. 5 Hours until first dose of antibiotics (in light color, the improvement for the 2007 cohort).

vasoactive drugs (P = .02) and blood culture drawing rate (55% vs 84%, P = .006) were significantly worse, although no significant differences could be detected regarding fluid administration (P = .34), measurement of hourly urinary output (P = .79), or use of steroids (P = .26).

Specialists from the ICU were called for 4 cases from this cohort; in the 46 remaining cases, ICU was not called because of advanced age (n = 2), comorbidity/poor basal status (n = 25), or “not severe enough condition” (n = 19). Only 1 patient from the cohort was admitted to the ICU (survived); 3 were rejected: 2 because of “not severe enough condition” (survived) and 1 because of poor basal status (died). During this period, 8 other patients were admitted to the ICU from the ED with the diagnosis of sepsis under different stages of severity, out of the criteria of inclusion of the study (5 admitted to ICU having spent less than 6 hours at the ED and 3 others diagnosed as sepsis only after ICU admission, although not deemed to have sepsis at the ED).

Table 1 Features of deceased patients

Comparison between management at the HDU and at the rest of the ED

Only 22 patients out of the 2007 cohort (44%) were treated at the HDU at any point during their first 6 hours of care at the ED. The 2 groups differed in actual severity (more cases of severe sepsis or shock vs sepsis with no organ damage in HDU, P = .045) and source of infection (urinary less frequent and abdominal more frequent at the HDU, P =

.015). There was a trend to a younger age (68 vs 74 years, P = .07), a higher number of failing organs (P = .08), and a higher APACHE II score for cases managed at the HDU (17.4 vs 14.7, P = .2). Nevertheless, no differences were found regarding presence of predisposing basal conditions (P = .27), lactate level (P = .5), sex (P = .7), or appropriate estimation of severity (P = .7).

Concerning management of the patients at the unit, both fluid overload (2.6 vs 1.2 L, P = .001) and inotropic drugs (P = .03) were more frequently used. Hourly urinary output was more often quantified (P = .001). Furthermore, a trend to higher rates of steroids use (P = .06), blood culture extraction (P = .07), microbiological isolation in cultures (P = .09), and earlier administration of Antimicrobial drugs (2.8 vs 4.7 hours, P = .3) was patent.

No significant differences were detected for ICU consultation (P = .3), prediction of source of infection (P =

.6), deep venous thrombosis prophylaxis (P = .7), glycemic control (P = .7), use of combinations of antibiotics (P = 1), or appropriate use of antibiotics (P = 1).

Mortality for the 2007 cohort

Mortality during hospital stay was 18%, not reaching statistical difference as compared with our historical mortality (22.6%, P = .62). Eight out of those 9 patients who died had been attended at the HDU at any point during their first 6 hours of care, and only 1 was not (P = .006).

In the univariate regression analysis, variables signifi- cantly associated to mortality were APACHE II score (22.4

Case Age (y) Gender Organs Stage APACHE II Source Consultation Reason Isolation

81 Female 3 Shock 35 Respiratory No Comorbidity None 60 Male 1 Severe 9 Respiratory No Not severe G+ 86 Female 5 Shock 29 Urinary Yes Comorbidity G- 94 Female 2 Severe 16 Abdominal No Age G- 85 Male 3 Shock 27 Abdominal No Comorbidity G+ 66 Male 1 Severe 14 Respiratory No Comorbidity None 94 Female 3 Shock 29 Soft tissues No Age G- 81 Male 1 Severe 15 Urinary No Not severe None 53 Male 5 Shock 28 Respiratory No Comorbidity Mixed

Organs: number of failing organs. Severity: sepsis, severe sepsis, shock. Consultation: it reflects whether ICU was called or not. Reason (just in case ICU was not called): advanced age, comorbidity, “poor basal performance status,” “not severe enough condition,” or “moribund.” Isolation (microbiological): none, gram+, gram-, or mixed.

vs 14.3, P = .001), fluid overload (P = .002), actual severity (P = .002), receiving care at the HDU (36% vs 3.7%, P =

.007), and having one or more failing organs (P = .009). When including these variables in the multivariate model, only actual severity was significantly associated to mortality (odds ratio = 6.6 [1.9-22]) after analyzing potential interactions between variables.

A detailed analysis of the 9 patients who did not survive to hospital discharge is showed in Table 1. Only 2 of them (patients 2 and 8) were deemed to have been candidates for ICU admission by ED staff in case of worsening conditions. Patient 2 died at the ICU after having received medical care at the general ward of the hospital; death was due to malignant Pericardial tamponade secondary to lung adenocarcinoma (both of them necropsy findings) after a later admission to ICU. Patient 8 died at another hospital after a 9-day admission, having finally been considered as nonsubsidiary of ICU by his responsible physicians.

Discussion

The actions implemented to improve medical care of patients (in brief, resident training, divulgation computer tools, and creation of an HDU) have had a moderate positive effect on sepsis at the ED. Both primary and secondary objectives of the Surviving Sepsis Campaign guidelines, such as an increased rate of lactate determina- tion, higher amounts of administered fluids, a lesser time before receiving initial dose of antibiotics, an appropriate use of vasoactive drugs, and a rational use of steroids or glycemic control, were more readily achieved. Sensitivity for detection of sepsis without losing specificity appears to be higher, too.

Notwithstanding these facts, we have the impression of having improved only partially and being still far from good while caring for septic patients at the ED. Different reasons might explain the incomplete implementation of the campaign guidelines, such as low levels of evidence for some of its postulates, recent contradictory evidence against previous recommendations, occasional feeling of “over- treating” less seriously ill cases, the necessity to perform time- and resource-consuming tasks, or even other uni- dentified factors.

The improvement was not translated into significantly lower in-hospital mortality, although the number of cases was probably not high enough to draw definite conclusions. Populations from both cohorts were slightly different; yet a lower APACHE II score in the 2007 cohort seems to be counterbalanced by a higher number of patients with basal immunosuppressive conditions and more cases or respiratory sepsis, traditionally associated to a worse prognosis.

Patients at the HDU were more seriously ill, with more cases of severe sepsis or shock and higher number of failing organs and APACHE II score, which surely justifies their

mortality rate. Clinical management parameters from patients treated at the HDU might explain most differences in quality improvements when the 2007 cohort is compared with our historical one, as differences between both cohorts disappear when patients treated at the HDU are excluded from statistical analyses.

It is quite surprising that admission to the HDU was not requested for up to 56% of the cases even though the policy of admission to the unit was very permissive. Nevertheless, only 1 out of 28 patients for whom HDU admission was not requested died during hospital stay. This patient was deemed not to be subsidiary of ICU. Thus, it seems that, for this subgroup of patients, admission to any unit with more resources would have added no advantages. Although recent articles have shown that certain score systems allow discrimination between low- and high-risk patients [9-11], our residents and staff were very skillful to detect which patients posed a higher risk just on a clinical basis, as reflected by the 3% vs 36% mortality rates at general ED and HDU, respectively. We do not know whether this ability could be reproducible in other ED or even for other cohorts at our ED.

It is reassuring that only 1 out of 19 patients with a diagnosis of sepsis under different stages with potential indication for ICU admission died during hospital stay because of an unrelated complication of sepsis during a later admission to ICU.

Scientific articles coming from “ideal” American ED settings reveal that they offer, because of their particular organization and the “semi-intensive care physician” qualification of their ED staff, especially good conditions for invasive monitoring or for keeping intubated patients [12]. However, this is probably not the case for most European busy EDs, which must accept that exactly that same type of care cannot be provided at many institutions [13]. Yet, this study confirms that actions can be implemented to help carefully select which patients would benefit from an early admission to ICU and to improve medical care of those patients kept for longer periods at the ED because of age, comorbidity, or being initially deemed not to be so seriously ill.

References

1. Angus DC, Linde-Zwirble WT, Lidicer J, et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and Associated costs of care. Crit Care Med 2001;29:1303-10.
2. De Miguel-Yanes JM, Andueza-Lillo JA, Gonzalez-Ramallo VJ, et al. Failure to implement evidence-based clinical guidelines for sepsis at the ED. Am J Emerg Med 2006;24:553-9.
3. Carlbom DJ, Rubenfeld GD. Barriers to implementing protocol-based sepsis resuscitation in the emergency department-results of a national survey. Crit Care Med 2007;35:2525-32.
4. Dellinger RP, Carlet JM, Masur H, et al. Surviving Sepsis Campaign guidelines for management of severe sepsis and septic shock. Crit Care Med 2004;32:858-73.
5. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001;345:1368-77.
6. Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a Severity of disease classification system. Crit Care Med 1985;13:818-29.
7. Bone RC. Let’s agree on terminology: definitions of sepsis. Crit Care Med 1991;19:973-6.
8. Levy MM, Fink MP, Marshall JC, et al. International sepsis definitions conference. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Intensive Care Med 2003;29:530-8 [electronic publication 2003 Mar 28].
9. Viallon A, Guyomarc’h S, Marjollet O, et al. Can emergency physicians identify a high mortality subgroup of patients with sepsis: role of procalcitonin. Eur J Emerg Med 2008;15:26-33.
10. Sankoff JD, Goyal M, Gaieski DF, et al. Validation of the Mortality in Emergency Department Sepsis (MEDS) score in patients with the Systemic Inflammatory Response Syndrome . Crit Care Med 2008;36:421-6.
11. Shapiro NI, Howell MD, Talmor D, et al. Mortality in Emergency Department Sepsis (MEDS) score predicts 1-year mortality. Crit Care Med 2007;35:192-8.
12. Nguyen HB, Lynch EL, Mou JA, et al. The utility of a quality improvement bundle in bridging the gap between research and Standard care in the management of severe sepsis and septic shock in the emergency department. Acad Emerg Med 2007;14:1079-86.
13. Sivayoham N. Management of severe sepsis and septic shock in the emergency department: a survey of current practice in emergency departments in England. Emerg Med J 2007;24:422.