Article

Triage sepsis alert and sepsis protocol lower times to fluids and antibiotics in the ED

a b s t r a c t

Background: Early identification of sepsis in the emergency department (ED), followed by adequate fluid hydration and appropriate antibiotics, improves patient outcomes.

Objectives: We sought to measure the impact of a sepsis workup and treatment protocol (SWAT) that included an electronic health record -based triage sepsis alert, direct communication, mobilization of resources, and standardized order sets.

Methods: We conducted a retrospective, quasiexperimental study of adult ED patients admitted with suspected sepsis, severe sepsis, or septic shock. We defined a preimplementation (pre-SWAT) group and a postimplementation (post-SWAT) group and further broke these down into SWAT A (septic shock) and SWAT B (sepsis with normal systolic blood pressure). We performed extensive data comparisons in the pre-SWAT and post-SWAT groups, including demographics, systemic inflammatory response syndrome criteria, time to intrave- nous fluids bolus, time to antibiotics, length-of-stay times, and mortality rates.

Results: There were 108 patients in the pre-SWAT group and 130 patients in the post-SWAT group. The mean time to bolus was 31 minutes less in the postimplementation group, 51 vs 82 minutes (95% confidence interval, 15-46; P value b .01). The mean time to antibiotics was 59 minutes less in the postimplementation group, 81 vs 139 mi- nutes (95% confidence interval, 44-74; P value b .01). Segmented Regression modeling did not identify secular trends in these outcomes. There was no significant difference in mortality rates.

Conclusions: An EHR-based triage sepsis alert and SWAT protocol led to a significant reduction in the time to intra- venous fluids and time to antibiotics in ED patients admitted with suspected sepsis, severe sepsis, and septic shock.

(C) 2015

  1. Background

Sepsis remains a significant source of morbidity and mortality in the United States. The National Center for Health Statistics estimates that the number of hospitalizations for sepsis increased from 621,000 in the year 2000 to 1,141,000 in 2008 [1]. In 2010, the National Vital Statis- tic Reports described septicemia as the 11th leading cause of death [2]. Although the mortality rate from severe sepsis was noted to decrease from 39% in 2000 to 27% in 2007 in the United States, more patients required discharge to a long-term care facility [3]. Because of its high

? The authors have no commercial associations or sources of support that might pose a conflict of interest. No source of support for this study.

?? Dr. Nietert’s time is funded in part by a grant from the National Center for Advancing

Translational Sciences (award number UL1 TR000062).

* Corresponding author at: Division of Emergency Medicine, Medical University of South Carolina, 169 Ashley Ave, MSC 300, Room 265E, Charleston, SC, 29425.

E-mail address: [email protected] (G.E. Hayden).

morbidity, mortality, cost, and resource utilization, a number of efforts have been directed towards improving outcomes. In 2001, Rivers et al

[4] published a landmark study describing early, aggressive treatment of sepsis in the emergency department (ED), with markedly improved patient outcomes. Earlier treatment, in particular appropriate broad- spectrum antibiotics, decreases mortality [5-8]. Moreover, “early goal- directed” therapy in the form of Sepsis Bundles and standardized order sets has been consistently shown to improve measures such as time to antibiotics, time to fluid resuscitation, Lactate clearance, and mortality [9-12]. Although more recent multicenter studies have failed to prove a reduction in all-cause mortality or patient outcomes with protocol-based care, the role of early identification, fluid resuscitation, and early appropriate antibiotics remains the mainstay of clinical care for patient with severe sepsis and septic shock [13-15].

We sought to evaluate the efficacy of early, rapid identification of sepsis during ED triage followed by a sepsis workup and treatment (SWAT) protocol emphasizing rapid mobilization of resources, standardized order sets, and early broad-spectrum antibiotics and fluid resuscitation. We hypothesized that a “best practice alert” (triage

http://dx.doi.org/10.1016/j.ajem.2015.08.039

0735-6757/(C) 2015

sepsis alert) embedded in our electronic health record (Epic Systems Corporation, Verona, WI) combined with a SWAT protocol would lead to a significant reduction in door-to-antibiotics time, door-to-intravenous- bolUS time, and overall mortality.

  1. Methods
    1. Design, setting, and population

This was a retrospective, quasiexperimental study of adult ED patients (>= 18 years of age) before and after our SWAT protocol implementation. The setting was a single, urban, academic ED with an annual census of 48,000. The study was approved by the institutional review board.

Study protocol

A preintervention group (before implementation of the SWAT pro- tocol [pre-SWAT]) and postintervention group (after implementation of the SWAT protocol [post-SWAT]) were defined. The pre-SWAT and post-SWAT groups were further broken down into SWAT A and SWAT B categories (Fig. 1). We determined that the implementation of the pro- tocol would have a greater likelihood of success if patterned after an al- ready established trauma alert system at our institution. Similar to a “Trauma A” activation which identifies critically injured patients in our trauma alert system, hypotensive patients with findings consistent with sepsis were assigned a SWAT A. Likewise, normotensive patients with 2 or more Systemic Inflammatory Response Syndrome criteria and a concern for an infectious source were assigned to SWAT B. The criteria used to identify pre-and postintervention patients were the same, but the sequence in which they were used was different. The preintervention group was identified by first reviewing the Epic (Epic Systems Corporation, Verona, WI) electronic health records of all pa- tients 18 years and older who were hospitalized from the ED between November 2012 and March 2013 with an infectious process. Second, only those ED encounters meeting criteria for either a SWAT A or a SWAT B were included (Fig. 1). This specific time frame was chosen such that the preintervention group patients presented to the ED after implementation of the Epic electronic health record. A report was creat- ed using the International Classification of Diseases, Ninth Revision, codes for pneumonia-related diagnoses (480-499), skin infections (680-686), pyelonephritis (590), urinary tract infection (595), other bacterial infec- tions (030-041), septic shock (785), and sepsis (995). Medical record reviews of the patients on the report generated were conducted to

confirm an ED diagnosis consistent with suspected sepsis, severe sepsis, or septic shock, and presence of criteria for either a SWAT A or a SWAT B. The postintervention (post-SWAT) group was composed of patients ad- mitted from the ED between April 2013 and October 2014 for whom either a SWAT A or a SWAT B protocol was activated by the treating physician in response to the triage sepsis alert. Specifically, these were ED patients who met the SWAT A or B criteria and in whom the ED provider suspected an underlying infectious process. A sample size of 130 subjects in the postin- tervention group was estimated to be required to achieve a 95% confidence interval (CI) for a time-to-antibiotics reduction of 30 minutes. Medical re- cords of patients on the SWAT A and SWAT B paging list were reviewed, and patients were excluded who did not meet criteria for SWAT A or

SWAT B on ED presentation or were not admitted to the hospital.

Because both groups (pre- and postintervention) had to meet SWAT A or B criteria, it was assumed that patients with simple (nonsepsis) in- fections were not included. To support this assumption, the infectious processes for pre- and postintervention patients were reviewed. In comparing the rates of infectious processes for the 2 groups, no signifi- cance difference was found (Appendix A).

The triage sepsis alert, or “best practice alert,” was derived from rules written in the background of the Epic electronic health record (Epic Systems Corporation, Verona, WI). If the required fields in the triage process met SWAT A or B criteria (Fig. 1), the triage nurse received a pop-up notification prompting a direct communication with the physician. The decision to proceed with a SWAT activation was left to the discretion of the physician based on a bedside evaluation of the patient. If a SWAT was activated, a standardized sepsis order set was used (computerized physician Order entry), and the specified resources were mobilized (Fig. 2). The SWAT A or B activation page (as gathered through the paging system, including time and patient medical record number) was recorded by the study investigators, and the medical records were reviewed for inclusion.

All data were abstracted retrospectively by 4 reviewers using standard- ized data collection sheets. Abstracted data were then entered into a secure, online database. Ambiguities were settled by consensus between the 3 sec- ondary reviewers. The ambiguities included 3 instances of ED discharge di- agnoses that were initially unclear but agreed upon by the reviewers.

The senior authors on the paper (GH, GH, and RT) reabstracted 10% of included medical records to assess reviewer concordance. Medical re- cords were selected for re-review at random. In total, 768 total data points were reabstracted from 24 medical records, and agreement was found on 751 data points. The percentage of agreement was 97.8% with a 95% CI for a single proportion of 96.4% to 98.6%. There were 100% agreement for ED arrival time, 100% for the time when antibiotics were given, and 100% for the time when intravenous fluids were given.

Fig. 1. Triage sepsis alert (best practice alert), written into the background of the electronic health record. SWAT A and SWAT B are defined above.

Fig. 2. Workflow after the triage sepsis alert for both SWAT A and SWAT B. CBC/diff = complete blood count with differential, CMP = comprehensive metabolic panel, PT/PTT = prothrom- bin time/partial thromboplastin time.

For the 17 data points with discrepancies, the data abstracted by the senior authors were used.

Measurements

Patient demographic data (age, sex, and race) were obtained, and ED vital signs and laboratory data (white blood cell count and lactate level) were collected to determine severity of illness, SIRS scores, and shock indices. We defined triage SIRS as the SIRS criteria available at the time of patient triage (which excluded the white blood cell count), and total SIRS included the white blood cell count. Effectiveness of the protocol was determined by specific treatment intervals and patient outcomes: door to bolus (door to intravenous fluids bolus), door to an- tibiotics, door to admit order, ED length of stay, hospital length of stay, intensive care unit (ICU) length of stay, and overall mortality (including in-hospital mortality and discharge to hospice care).

Statistical analysis

Descriptive statistics included numbers and rates for categorical data as well as means, standard deviation, medians, and ranges for continu- ous data. Comparisons between characteristics of subjects pre- and postintervention were made using ?2 tests (for proportions) and 2- sample t tests (for continuous variables). Group differences were calcu- lated, along with their 95% CIs. To assess whether or not the interven- tion resulted in significant shifts in the time outcomes (ie, door to bolus, door to antibiotics, door to admit order), a segmented regression modeling approach was used [16] to account for any potential secular trends that may have been occurring despite the intervention. In this modeling process, 4 models were considered for each time outcome:

(1) a “full” linear regression model, which allowed for a change in means and a change in slope after the intervention; (2) a model that allowed for a change in means but included identical nonzero slopes pre- and postintervention; (3) a model that allowed for a change in slopes but forced the intercept to be identical pre- and postintervention;

and (4) a model that allowed for a change in means but assumed a zero (flat) slope pre- and postintervention. Akaike information criterion [17] was used to determine which of the 4 models provided the best fit. Once the most appropriate model was selected, patient characteristics identified as being significantly different (pre- vs postintervention) were included in the model as covariates to control for any potential confounding. P values b .05 were considered statistically significant.

  1. Results

A total of 238 patient medical records were abstracted for the study, including 108 medical records in the pre-SWAT group and 130 medical records in the post-SWAT group. All patients were suspected of having an infection by the ED provider team and met SWAT A or SWAT B criteria suggesting sepsis, severe sepsis, or septic shock (independent of the hospital discharge diagnosis).

The pre-SWAT group was composed of 13 patients meeting SWAT A criteria and 95 patients meeting SWAT B criteria. The post-SWAT group was composed of 32 patients meeting SWAT A criteria and 98 patients meeting SWAT B criteria.

There were no significant differences between the pre-SWAT and post-SWAT groups in regard to age, sex, or race (Table 1). Although there was no difference in “triage” SIRS criteria between the 2 groups (P = .20), patients in the post-SWAT group had a higher number of “total” SIRS criteria compared with the pre-SWAT group (P = .04). The Shock Index was higher in the post-SWAT group with a mean of

1.1 compared with the pre-SWAT shock index of 0.9 (P b .01). Signifi- cant differences were seen between the post-SWAT and pre-SWAT groups in terms of initial temperature and respiratory rate, and systolic blood pressure. The mean systolic blood pressure was, on average,

15.7 mm Hg lower in the post-SWAT group as compared with the pre- SWAT group (P b .01), which accounted for a greater percentage of pa- tients classified as SWAT A in the post-SWAT group (25%) as compared with the pre-SWAT group (12%) (P = .02).

Table 1

Comparison of demographics and severity

Pre-SWAT (108)

s

Post-SWAT (130)

s

Difference

95% CI

P value

% SWAT A

12

NA

25

NA

13

3 to 23

.02

% SWAT B

88

NA

75

NA

13

% Male

53

NA

62

NA

9

-21.6 to 3.6

.21

% Female

47

NA

38

NA

Age (y)

55

+-19

56

+-18

1

-5.7 to 3.7

.68

% White

45

NA

42

NA

3

-9.6 to 15.7

.74

% Black

50

NA

55

NA

5

-17.1 to 7.1

.50

% Other

5

NA

3

NA

2

-2.9 to 6.9

.65

Heart rate (beats/min)

115.2

+-21.5

118.7

+-18.2

3.5

-8.6 to 1.6

.18

Temperature (?C)

37.7

+-1.6

38.3

+-1.5

0.6

0.2 to 1

b.01

Respiratory rate (breath/min)

22.1

+-6.1

23.9

+-7

1.8

0.1 to 3.5

.04

Systolic BP (mm Hg)

130.2

+-33.6

114.5

+-31.6

15.7

7.4 to 24

b.01

# SIRS total

2.8

+-0.7

3

+-0.8

0.2

0.01 to 0.4

.04

WBC (thousands/uL)

14.1

+-9.8

14.4

+-9.3

0.3

-2.7 to 2.1

.81

Lactate (mmol/L)

2.4 (n=75)

+-2.4

2.8 (n=126)

+-2.4

0.4

-1.1 to 0.3

.25

Lactate performed %

69.4

NA

96.9

NA

27.5

18.3 to 36.8

b.01

Shock index (systolic BP/heart rate)

0.9

+-0.3

1.1

+-0.4

0.2

0.11 to 0.29

b.01

BP = blood pressure; WBC = white blood cell.

Of the 108 pre-SWAT patients, a serum lactate test was performed in 75 (69.4%), and of the 130 post-SWAT patients, lactate was performed in 126 (96.9%). This represented a 27.5% increase in lactate tests being performed in the post-SWAT patients (95% CI, 18.3-36.8%; P b .01). A comparison of pre-SWAT A (11/13 = 84.6%) with post-SWAT A (30/32 = 93.8%) patients showed an increase of 9.2% (95% CI, -27.5 to 9.1%; P = .69). A comparison of pre-SWAT B (64/95 = 67.4%) with post-SWAT B (96/98 = 98%) patients showed an increase of 30.6% (95% CI, 20-41.2; P b .01). No difference, however, was observed in the measured mean lactate levels between the pre-SWAT and post- SWAT groups (P = .25).

Results of the segmented regression modeling indicated that there were no significant secular trends in any of the time outcomes (ie, door to bolus, door to antibiotics, door to admit order) and that the best fitting model for each outcome was one that allowed only for a change in means (assuming a zero slope pre- and postintervention). The post-SWAT group demonstrated marked improvements in both the door-to-intravenous fluids time and the door-to-antibiotics time (Table 2). The mean door-to-bolus time improved by 30.5 minutes (P b .01), a finding that was slightly attenuated when covariates were added to the model (adjusted difference = 22.4 minutes; 95% CI, 5.6- 39.2) (Fig. 3). Similarly, the mean door-to-antibiotics time improved by 58.8 minutes (P b .01), a finding that was also slightly attenuated when covariates were added to the model (adjusted difference = 50.6 minutes; 95% CI, 34.7-66.4) (Fig. 4). No significant changes were noted for door-to-pressor given or door-to-admit order times.

Table 3 highlights differences in patient outcomes pre- and postinter- vention. There were no statistically significant increases in the proportion of subjects who were admitted to the ICU (35.2% vs 43.1%, P = .27) or the proportion who died during their hospital stay (9.3% vs 13.8%, P = .38), but there was a notable decline in the proportion discharged to hospice care (7.4% vs 1.5%, P = .05). No significant differences between groups were noted in the average length of ICU stay or in the average length of hospital stay. Further analysis is available in Appendix B.

In comparing pre-SWAT A and post-SWAT A patients, there were no sig- nificant differences in demographic, severity of illness or outcome variables;

mortality rates and lengths of stay in the ED, hospital, and ICU were similar in the two groups (analysis available in Appendix C). However, the post-SWAT A mean door-to-antibiotics time was 67.8 minutes less (P = .02) (Table 4). In comparing pre-SWAT B and post-SWAT B patients, there was no sig- nificant difference between demographic and outcome variables (analysis available in Appendix D). However, there was a significant difference in a number of severity of illness variables. Mean pre-SWAT B “triage” SIRS was 2.2 compared with post-SWAT B of 2.4 (difference, 0.2; 95% CI, 0.07- 0.33; P b .01). Pre-SWAT mean shock index was 0.9 compared with post- SWAT of 1.0 (difference, 0.1; 95% CI, 0.03-0.17; P b .01). Moreover, pre- SWAT B mean systolic blood pressure was 137.5 mm Hg compared with post-SWAT B of 126.7 mm Hg (difference, 10.8; 95% CI, 2.9-18.7; P b .01). There was also a significant difference for the treatment variables. Post-

SWAT B mean door-to-bolus time was 30.7 minutes less (P b .01) and the

mean door-to-antibiotic time was 56.3 minutes less (P b .01) (Table 4).

A subanalysis was performed regarding an institutional goal of adminis- tering antibiotics in presumed sepsis within 60 minutes of ED arrival. Com- paring pre-SWAT to post-SWAT patients, antibiotics were administered within 60 minutes in 11.1% (12/108) of patients and 36.2% (47/130) of pa- tients, respectively, which represented a 25.1% increase (95% CI, 14.1-36.1%; P b .01). Comparing pre-SWAT A to post-SWAT A patients, antibiotics were administered within 60 minutes in 15.4% (2/13) of patients and 53.1% (17/32) of patients, respectively, which represented an absolute 37.7% increase (95% CI, 5.9-69.5%; P = .05). Finally, comparing pre-SWAT B to post-SWAT B patients, antibiotics were administered in less than 60 minutes in 10.5% (10/95) of patients and 30.6% (30/98) of patients, respec- tively, which represented a 20.1% increase (95% CI, 8.7-31.5%; P b .01).

More comprehensive data regarding the pre-SWAT, post-SWAT, and SWAT A, and SWAT B subgroups may be found in Appendices B to D.

  1. Discussion

In this retrospective, quasiexperimental study of adult ED patients admitted to the hospital with suspected sepsis, severe sepsis, or septic shock, a SWAT protocol originating from ED triage (automated ED triage sepsis alert) served to lower the time to intravenous fluids and time to

Table 2

treatment times (all in minutes)

Pre-SWAT

s

Post-SWAT

s

Difference

95% CI

P value?

Door to bolus

81.9 (n = 80)

+-66.8

51.4 (n = 120)

+-47.5

30.5

14.6 to 46.3

b.01

Door to antibiotics given

139.4 (n = 108)

+-74.3

80.6 (n = 130)

+-38.8

58.8

44 to 73.6

b.01

Door to pressor given

139.3 (n = 6)

+-218.5

125.9 (n = 15)

+-132

13.4

-155 to 182

.91

Door to admit order

181.5 (n = 108)

+-99.2

168.0 (n = 129)

+-101.9

13.5

-12.3 to 39.3

.61

* P values were obtained from the best-fitting segmented regression model.

Fig. 3. Segmented regression modeling of door to bolus administration in pre- and post-SWAT implementation periods. This analysis allowed for a change in means but assumed a zero (flat) slope pre- and postintervention.

antibiotics. The value of early identification of sepsis is significant, par- ticularly when coupled with evidence-based therapies such as fluid re- suscitation and early, appropriate Empiric antibiotics [7,18]. Ferrer et al

[5] performed a retrospective analysis of more than 28,000 patients with severe sepsis and septic shock and noted a steady increase in hos- pital mortality for each hour of delay in antibiotic administration. Gaieski et al [6] described a similar experience in a study of 261 ED pa- tients, noting that elapsed time to antibiotics served as a primary deter- minant of mortality in patients with severe sepsis and septic shock. However, beyond early identification, rapid empiric antibiotics, and ad- equate fluid resuscitation, strict adherence to particular early goal- directed therapy protocols has not led to improvement in outcomes

[13-15]. This has consequently led to revisions of the Surviving Sepsis Campaign’s (SSC’s) guidelines and performance improvement indica- tors [19].

There is an increasing focus on ways to improve sepsis identification. sepsis screening and alert systems have been studied in both the ED and the ICU, with improvements in early identification of sepsis and the ini- tiation of therapy [20-24]. Computer alerts have also been studied in the context of sepsis, with improved compliance in terms of lactate test- ing [25]. Various Medical Informatics solutions have also been investi- gated, including a Clinical decision support tool that predicted lactic acid levels and mortality based on vital signs and white blood cell levels [26]. A recent study evaluated an electronic medical record screening

Fig. 4. Segmented regression modeling of door to antibiotic (ABX) administration in pre- and post-SWAT implementation periods. This analysis allowed for a change in means but assumed a zero (flat) slope pre- and postintervention.

Table 3

Outcomes

Pre-SWAT

Post-SWAT

Difference

95% CI

P value

% ICU

35.2

43.1

7.9

-20.4 to 4.6

.27

% Mortality

9.3

13.8

4.5

-12.7 to 3.7

.38

% Discharge

7.4

1.5

5.9

0.8 to 11

.05

to hospice

Average LOS in

4.1

6.2

2.1

-4.9 to 0.7

.14

ICU (d)

Average LOS in

8.2

9.0

0.8

-3.3 to 1.7

.53

hospital (d)

LOS = length of stay.

tool that triggered a sepsis protocol based on a defined number of SIRS criteria present during triage [27]. However, this study was not out- comes focused, and the impact of their screening tool was not assessed. Umscheid et al [28] described an early warning and response system for sepsis that was used for adult non-ICU patients who were already admitted to the hospital. This “alert” used a combination of SIRS criteria, systolic blood pressure, and lactate. This study evaluated the test char- acteristics of the early warning and response system. They did note ear- lier sepsis care and a trend towards decreased sepsis mortality (not statistically significant). Nguyen et al [29] also evaluated an electronic health record-based sepsis identification tool that incorporated similar features to the study of Umscheid et al [28]. They evaluated diagnostic accuracy of their sepsis identification tool in an ED setting but did not report on outcome measures such as time to antibiotics or time to fluids. Next, Alsolamy et al [30] studied another electronic alert system in ED patients admitted to the ICU, evaluating the diagnostic accuracy of the sepsis alert in detecting severe sepsis or septic shock. Their screen incor- porated SIRS criteria and evidence of organ dysfunction (systolic blood pressure, oxygen saturation, and lactate). Similar to the aforementioned studies, our triage sepsis alert used SIRS criteria and blood pressures. However, our alert did not incorporate laboratory testing, as this was unavailable at the time of ED triage. The diagnostic accuracy of the triage sepsis alert was also not evaluated. We did report key outcome mea-

sures such as time to antibiotics, time to fluids, and mortality rates.

Specifically, in this study, we evaluated a SWAT protocol which in- corporates an electronic health record-based triage sepsis alert (best practice alert), direct communication between ED triage nurse and the ED physician, and standardized orders and a mobilization of resources. Implementation of the SWAT protocol led to an approximately 31- minute decrease in door-to-bolus time and a 59-minute decrease in door-to-antibiotics time in the setting of a more ill post-SWAT group who had more SIRS criteria, a higher mean shock index, and lower mean systolic blood pressures. Segmented regression modeling did not identify secular trends in any of these time outcomes, which con- firms that these significant differences are more likely a result of the in- tervention. The lack of significance difference in mortality rates could be attributed to the post-SWAT group being sicker.

Although there was a significant difference in the performance of lac- tate assessment between the pre-SWAT and post-SWAT groups, there was no difference in the lactate levels between the 2 groups. Of note, no standardized diagnostic workup was in place for patients with

suspected sepsis in the pre-SWAT group, whereas lactate assessment was included on the SWAT protocol in the post-SWAT group. This is the likely explanation for a higher percentage of lactate assessments performed in the post-SWAT group, although a difference in illness se- verity between the 2 groups is also possible.

When pre-SWAT A and post-SWAT A groups were compared, the improvement in door-to-antibiotics time was maintained, although the door-to-bolus times were no different. Because hypotension upon presentation is a core criterion for SWAT A, it is not surprising that a rapid fluid bolus was also observed in both groups. For the pre-SWAT B and post-SWAT B groups, significant improvement in both the door- to-bolus time and door-to-antibiotics time was still observed.

In all postintervention groups, the percentage of patients receiving antibiotics within 60 minutes of ED arrival was significantly higher. This institutional goal is based on the SSC’s guidelines from 2012 [7] which recommend “administration of effective intravenous antimicro- bials within the first hour of recognition of septic shock (grade 1B) and severe sepsis without septic shock (grade 1C) as the goal of thera- py.” As of 2010, only 68% of patients in the SSC registry received antibi- otics within 3 hours. Although no definitive causal link between early antibiotics and patient outcomes has been established, there is an asso- ciation in observational data between early antibiotics and survival out- comes in severe sepsis and septic shock. The absolute 25.1% increase in postintervention patients in this study who received antibiotics within 60 minutes is significant.

  1. Limitations

A number of limitations exist to this study, in addition to its retro- spective and single-site design. Different methods were used to identify pre- and postintervention groups, and the groups were not matched for illness severity. The post-SWAT group had significantly greater severity of illness, and this along with implementing the SWAT protocol likely contributed to improved mean times to bolus and antibiotics. Thus, se- verity of illness was a confounding factor that may have limited the in- ternal validity of the results.

In regard to the post-SWAT group, the study did not measure adher- ence to individual elements of the SWAT protocol, which is a limitation of the retrospective design. Furthermore, the study did not track perfor- mance of the triage sepsis alert, so false-positive and false-negative rates for SWAT activations are not known. If the number of false posi- tives was significant, then the effectiveness of the SWAT protocol treat- ment would be difficult to measure because these patients were not septic, and good outcome rates would be falsely elevated. If the number of false negatives was significant, then all potential patients were not in- cluded in the study, resulting in selection bias.

  1. Conclusion

In conclusion, this study demonstrates a significant reduction in the time to intravenous fluids and time to antibiotics in ED patients admit- ted with suspected sepsis, severe sepsis, or septic shock, following im- plementation of an EHR-based triage sepsis alert and SWAT protocol.

Table 4

Comparisons of SWAT A and B

Pre-SWAT

n

SD

Post-SWAT

n

SD

Difference

95% CI

P value

SWAT A door to bolus (min)

55.9

13

+-63.8

34.6

27

+-31

21.3

-8.8 to 51.4

.27

SWAT A door to antibiotics given (min)

124.4

13

+-77

67.8

32

+-37.6

56.6

10 to 103

.02

SWAT B door to bolus (min)

87

67

+-67

56.3

93

+-50.4

30.7

12.4 to 49

.00

SWAT B door to antibiotics given (min)

141.5

95

+-80

85.2

98

+-38.8

56.3

38.6 to 74

b.01

SWAT A mortality or hospice (%)

30.8

13

NA

21.9

32

NA

8.9

-19 to 37

.70

SWAT B mortality or hospice (%)

14.8

95

NA

13.2

98

NA

1.6

-8.2 to 11.4

.83

Appendix A. Comparison of pre-SWAT and post-SWAT infection sources and positive cultures

Pre-SWAT (108 patients)

Post-SWAT (130 patients)

Difference (%)

95% CI

P value

Blood culture + (%)

20.4 (20)

28.5 (37)

8.1

-19.1 to 2.9

.20

Urine culture +

25.0 (27)

26.9 (35)

1.9

-13.1 to 9.3

.85

Any culture +

50.0 (54)

59.2 (77)

9.2

-21.9 to 35.0

.20

Source: lungs

46.3 (50)

36.2 (47)

10.1

-2.4 to 22.6

.15

Source: urine

28.7 (31)

34.6 (45)

5.9

-17.8 to 6.0

.41

Source: SOFT TISSUe

15.7 (17)

14.6 (19)

1.1

-8.0 to 10.2

.46

Source: gastrointestinal

8.3 (9)

7.7 (10)

0.6

-6.3 to 7.5

.94

Source: venous line

3.7 (4)

3.1 (4)

0.6

-4.0 to 5.2

.92

Source: other/unknown

2.8 (3)

6.9 (9)

4.1

-9.7 to 1.5

.25

Appendix B. Comparison of pre-SWAT and post-SWAT data

Pre-SWAT

Pre-SWAT (+-SD)

Post-SWAT

Post-SWAT (+-SD)

Difference

95% CI

P value

Total patients

108

NA

130

NA

NA

NA

NA

% Male

53%

NA

62%

NA

9%

-21.6 to 3.6

.21

% Female

47%

NA

38%

NA

NA

NA

NA

Age (y)

55

+-19

56

+-18

1

-5.7 to 3.7

.68

% White

45%

NA

42%

NA

3%

-9.6 to 15.7

.74

% Black

50%

NA

55%

NA

5%

-17.1 to 7.1

.50

% Other

5%

NA

3%

NA

2%

-2.9 to 6.9

.65

% SWAT A

12%

NA

25%

NA

13%

3 to 23

.02

% SWAT B

88%

NA

75%

NA

NA

NA

NA

# SIRS on ED arrival

2.2

+-0.6

2.3

+-0.6

0.10

-0.3 to 0.1

.20

# SIRS total

2.8

+-0.7

3.0

+-0.8

0.20

0.01 to 0.4

.04

Temperature (?C)

37.7

+-1.6

38.3

+-1.5

0.6

0.2 to 1

.00

Respiratory rate (breath/min)

22.1

+-6.1

23.9

+-7.0

1.8

0.1 to 3.5

.04

Systolic BP (mm Hg)

130.2

+-33.6

114.5

+-31.6

15.7

7.4 to 24

b.01

Heart rate (beats/min)

115.2

+-21.5

118.7

+-18.2

3.5

-8.6 to 1.6

.18

WBC (thousands/uL)

14.1

+-9.8

14.4

+-9.3

0.3

-2.7 to 2.1

.81

% Central line

2.8%

NA

10.8%

NA

8%

1.4 to 14.6

.03

% Pressors given

4.6%

NA

11.5%

NA

6.9%

-14 to 0.2

.09

% ICU

35.2%

NA

43.1%

NA

7.9%

-20.4 to 4.6

.27

% Mortality

9.3%

NA

13.8%

NA

4.5%

-12.7 to 3.7

.38

% Discharge to hospice

7.4%

NA

1.5%

NA

5.9%

0.8 to 11

.05

% Mortality or hospice

16.7%

NA

15.3%

NA

1.4%

-7.9 to 11

.91

Door to bolus (min)

81.9 (80)

+-66.8

51.4 (120)

+-47.5

30.5%

14.6 to 46.4

b.01*

Door to antibiotic order (min)

103.1

+-70.9

46.0

+-32.3

57.1

43.4 to 70.8

b.01

Door to antibiotic given (min)

139.4

+-74.3

80.6

+-38.8

58.8

44 to 73.6

b.01*

Order to antibiotic given (min)

36.4

+-21.3

34.3

+-26.1

2.1

-4.1 to 8.3

.50

Door to pressor order (min)

139.3 (5)

+-218.5

125.9 (15)

+-132.0

13.4

-155 to 182

.87

Door to pressor given (min)

139.3 (5)

+-218.5

125.9 (15)

+-132.0

13.4

-155 to 182

.87*

Door to admit order (min)

181.5

+-99.2

168.0

+-101.9

13.5

-12.3 to 39.3

.30*

ED admit order to OTF (min)

279.1

+-381.0

214.3

+-281.4

64.8

-19.9 to 149.5

.13

ED LOS (min)

460.7

+-387.0

382.1

+-305.2

78.6

-9.8 to 167

.08

LOS ICU (d)

4.1 (38)

+-2.9

6.2 (56)

+-8.4

2.1

-4.9 to 0.7

.14

LOS hospital (d)

8.2

+-9.2

9.0

+-10.1

0.8

-3.3 to 1.7

.53

Shock index (systolic BP/heart rate)

0.9

+-0.3

1.1

+-0.4

0.2

0.11 to 0.29

b.01

BP = blood pressure; LOS = length of stay; OTF = off-the-floor; WBC = white blood cell.

*Corresponding P values were obtained from the best-fitting segmented regression model.

Appendix C. Comparison of pre-SWAT A and post-SWAT A data

Pre-SWAT

Pre-SWAT (+-SD)

Post-SWAT

Post-SWAT (+-SD)

Difference

95% CI

P value

Total patients

13

NA

32

NA

NA

NA

NA

% Male

62%

NA

72%

NA

10%

-40 to 20

.76

% Female

38%

NA

28%

NA

NA

NA

NA

Age (y)

59

+-16

52

+-20

7

-5.6 to 19.6

.27

% White

46%

NA

38%

NA

8%

-24 to 40

.87

% Black

54%

NA

59%

NA

5%

-37 to 27

.98

% Other

0%

NA

3%

NA

3%

-12 to 6

.61

# SIRS on ED arrival

1.8

+-1.0

1.8

+-0.7

0

-0.5 to 0.5

1.00

# SIRS total

2.3

+-1.1

2.6

+-0.8

0.3

-0.9 to 0.3

.31

Temperature (?C)

36.6

+-3.0

37.3

+-1.9

0.7

-2.2 to 0.8

.35

Respiratory rate (breath/min)

21.5

+-10.7

22.1

+-6.7

0.6

-5.9 to 4.7

.82

Systolic BP (mm Hg)

77.5

+-7.4

77.2

+-8.9

0.3

-5.3 to 5.9

.92

Heart rate (beats/min)

105.5

+-29.0

116.7

+-22.8

11.2

-27.6 to 5.2

.18

WBC (thousands/uL)

11.2

+-6.1

15.0

+-10.5

3.8

-10.1 to 2.5

.23

% Central line

7.7%

NA

28.1%

NA

20.3%

-47 to 6

.28

% Pressors given

15.4%

NA

18.8%

NA

3.4%

-28 to 21

.87

% ICU

61.5%

NA

65.6%

NA

4.1%

-35 to 27

.93

% Mortality

7.7%

NA

21.9%

NA

14.2%

-39 to 10

.49

% Discharge to hospice

23.1%

NA

0.0%

NA

23.1%

7 to 39

.03

% Mortality or hospice

30.8%

NA

21.9%

NA

8.9%

-19 to 37

.81

Door to bolus (min)

55.9

+-63.8

34.6 (27)

+-31.0

21.3

-8.8 to 51

.16

Door to antibiotic order (min)

95.9

+-72.9

40.4

+-30.7

55.5

25 to 86

.00

Door to antibiotic given (min)

124.4

+-77.0

67.8

+-37.6

56.6

10 to 103

.02

Order to antibiotic given (min)

28.5

+-21.6

44.7

+-20.0

16.2

30 to 3

.02

Door to pressor order (min)

53.5 (2)

+-58.7

111.5 (24)

+-147.8

58

-281 to 165

.60

Door to pressor given (min)

53.5 (2)

+-58.7

111.5 (24)

+-147.8

58

-281 to 165

.60

Door to admit order (min)

162.8

+-124.0

148.7

+-115.3

14.1

-64 to 92

.72

ED admit order to OTF (min)

118.4

+-118.0

149.6

+-173.7

31.2

-137 to 75

.56

ED LOS (min)

281.2

+-232.6

298.3

+-219.3

17.1

-165 to 131

.82

LOS ICU (d)

4.3 (8)

+-3.2

5.9 (21)

+-10.3

1.6

-9.3 to 6.1

.67

LOS hospital (d)

10.6

+-8.4

8.8

+-9.3

1.8

-4.2 to 7.8

.55

Shock index (systolic BP/heart rate)

1.4

+-0.4

1.5

+-0.4

0.1

-0.4 to 0.2

.45

Appendix D. Comparison of pre-SWAT B and post-SWAT B data

Pre-SWAT

Pre-SWAT (+-SD)

Post-SWAT

Post-SWAT (+-SD)

Difference

95% CI

P value

Total patients

95

NA

98

NA

NA

NA

NA

% Male

52%

NA

59%

NA

7%

-21 to 7

.41

% Female

48%

NA

41%

NA

NA

NA

NA

Age (y)

54

+-19

57

+-18

3

-8.3 to 2.3

.26

% White

45%

NA

44%

NA

1%

-13 to 15

1.00

% Black

49%

NA

53%

NA

4%

-18 to 10

.68

% Other

6%

NA

2%

NA

4%

-1.5 to 9.5

.29

# SIRS on ED arrival

2.2

+-0.4

2.4

+-0.5

0.2

0.07 to 0.33

.00

# SIRS total

2.9

+-0.7

3.2

+-0.7

0.3

0.10 to 0.5

.00

Temperature (?C)

37.8

+-1.3

38.6

+-1.2

0.8

0.44 to1.2

.00

Respiratory rate (breath/min)

22.2

+-5.3

24.5

+-7.0

2.3

0.53 to 4.1

.01

Systolic BP (mm Hg)

137.5

+-29.0

126.7

+-26.3

10.8

2.9 to 18.7

.01

Heart rate (beats/min)

116.6

+-20.1

119.3

+-16.5

2.7

-7.9 to 2.5

.31

WBC (thousands/uL)

14.5

+-10.1

14.2

+-9.0

0.3

-2.4 to 3.0

.83

% Central line

2.1%

NA

5.1%

NA

3%

-8.3 to 2.3

.47

% Pressors given

4.2%

NA

7.1%

NA

2.9%

-9.4 to 36

.58

% ICU

31.6%

NA

35.7%

NA

4.1%

-17 to 9

.65

% Mortality

9.5%

NA

11.2%

NA

1.7%

-10 to 7

.88

% Discharge to hospice

5.3%

NA

2.0%

NA

3.3%

-2 to 8.6

.40

% Mortality or hospice

14.8%

NA

13.2%

NA

1.6%

-8.2 to 11.4

.91

Door to bolus (min)

87.0 (67)

+-66.7

56.3 (93)

+-50.4

30.7

12.4 to 49

.00

Door to antibiotic order (min)

104.1

+-74.9

48.2

+-33.0

55.9

40 to 72

.00

Door to antibiotic given (min)

141.5

+-79.8

85.2

+-38.8

56.3

39 to 74

.00

Order to antibiotic given (min)

37.4

+-22.3

36.1

+-28.1

1.3

-5.1 to 8.5

.72

Door to pressor order (min)

182.2 (4)

+-266.5

142.4 (7)

+-120.7

39.8

-219 to 299

.74

Door to pressor given (min)

182.2 (4)

+-266.5

142.4 (7)

+-120.7

39.8

-219 to 299

.74

Door to admit order (min)

195.9

+-100.8

176.6

+-97.9

19.3

-9 to 48

.18

ED admit order to OTF (min)

274.2

+-379.8

231.6

+-302.5

42.6

-55 to 140

.39

ED LOS (min)

470.0

+-390.7

407.7

+-322.4

62.3

-39 to 164

.23

LOS ICU (d)

3.9 (30)

+-2.5

6.4 (35)

+-7.2

2.5

-5.3 to 0.26

.08

LOS hospital (d)

7.6

+-5.7

9.1

+-10.7

1.5

-3.9 to 1

.23

Shock index (systolic BP/heart rate)

0.9

+-0.3

1.0

+-0.2

0.1

0.17 to 0.03

.01

References

  1. Hall MJ, Williams SN, DeFrances CJ, Golosinskiy A. Inpatient care for septicemia or sepsis: a challenge for patients and hospitals. Natl Cent Health Stat Data Brief 2011;62:1-8.
  2. Murphy SL, Xu J, Kochanek KD. Deaths: final data for 2010. Natl Vital Stat Rep 2013; 61(4):1-118.
  3. Kumar G, Kumar N, Taneja A, Kaleekal T, Tarima S, McGinley E, et al. Nationwide trends of severe sepsis in the 21st century (2000-2007). Chest 2011;140(5):1223-9.
  4. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, et al. Early goal- directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001;345(19):1368-77.
  5. Ferrer R, Martin-Loeches I, Phillips G, Osborn TM, Townsend S, Dellinger RP, et al. Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement pro- gram. Crit Care Med 2014;8.
  6. Gaieski DF, Mikkelsen ME, Band RA, Pines JM, Massone R, Furia FF, et al. Impact of time to antibiotics on survival in patients with severe sepsis or septic shock in whom early goal-directed therapy was initiated in the emergency department*. Crit Care Med 2010;38(4):1045-53.
  7. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, et al. Surviving Sepsis Campaign: international guidelines for management of severe sepsis and sep- tic shock: 2012; 2013 580-637.
  8. Kumar A, Zarychanski R, Light B, Parrillo J, Maki D, Simon D, et al. Early combination antibiotic therapy yields improved survival compared with monotherapy in septic shock: a propensity-matched analysis. Crit Care Med 2010;38(9):1773-85.
  9. Seoane L, Winterbottom F, Nash T, Behrhorst J, Chacko E, Shum L, et al. Using quality improvement principles to improve the care of patients with severe sepsis and sep- tic shock. Ochsner J 2013;13(3):359-66.
  10. Castellanos-Ortega A, Suberviola B, Garcia-Astudillo LA, Holanda MS, Ortiz F, Llorca J, et al. Impact of the Surviving Sepsis Campaign protocols on hospital length of stay and mortality in septic shock patients: results of a three-year follow-up quasi- experimental study. Crit Care Med 2010;38(4):1036-43.
  11. Thiel SW, Asghar MF, Micek ST, Reichley RM, Doherty JA, Kollef MH, et al. Hospital- wide impact of a standardized order set for the management of bacteremic severe sepsis. Crit Care Med 2009;37(3):819-24.
  12. Micek ST, Roubinian N, Heuring T, Bode M, Williams J, Harrison C, et al. Before-after study of a standardized hospital order set for the management of septic shock*. Crit Care Med 2006;34(11):2707-13.
  13. ARISE Investigators and the ANZICS Clinical Trials Group et al. Goal-directed resusci- tation for patients with early septic shock. N Engl J Med 2014;371(16):1496-506.
  14. ProCESS Investigators, et al. A randomized trial of protocol-based care for early sep- tic shock. N Engl J Med 2014;370(18):1683-93.
  15. Mouncey PR, Osborn TM, Power GS, Harrison DA, Sadique MZ, Grieve RD, et al. Trial of early, goal-directed resuscitation for septic shock. N Engl J Med 2015;372(14):1301-11.
  16. Penfold RB, Zhang F. Use of interrupted Time series analysis in evaluating health care quality improvements. Acad Pediatr 2013;13(6 Suppl.):S38-44.
  17. Akaike H. A new look at the statistical model identification. IEEE Trans Autom Con- trol 1974;19(6):716-23.
  18. Paul M, Shani V, Muchtar E, Kariv G, Robenshtok E, Leibovici L. Systematic review and meta-analysis of the efficacy of appropriate empiric antibiotic therapy for sepsis. Antimicrob Agents Chemother 2010;54(11):4851-63.
  19. Surviving Sepsis Campaign “updated bundles in response to new evidence”. Accessed on April 13, 2015 at http://www.survivingsepsis.org/SiteCollectionDocuments/SSC_ Bundle.pdf.
  20. LaRosa JA, Ahmad N, Feinberg M, Shah M, Dibrienza R, Studer S, et al. The use of an early alert system to improve compliance with sepsis bundles and to assess impact on mortality. Crit Care Res Pract 2012;2012(1):1-8.
  21. Sawyer AM, Deal EN, Labelle AJ, Witt C, Thiel SW, Heard K, et al. Implementation of a real-time computerized sepsis alert in nonintensive care unit patients. Crit Care Med 2011;39(3):469-73.
  22. Moore LJ, Jones SL, Kreiner LA, McKinley B, Sucher JF, Todd SR, et al. Validation of a screening tool for the early identification of sepsis. J Trauma 2009;66(6):1539-47.
  23. Girardis M, Rinaldi L, Donno L, Marietta M, Codeluppi M, Marchegiano P, et al. Effects on management and outcome of severe sepsis and septic shock patients admitted to the intensive care unit after implementation of a sepsis program: a pilot study. Crit Care 2009;13(5):1-8.
  24. Campbell J. The effect of nurse champions on compliance with Keystone Intensive Care Unit Sepsis-screening protocol. Crit Care Nurs Q 2008;31(3):251-69.
  25. Berger T, Birnbaum A, Bijur P, Kuperman G, Gennis P, et al. A computerized alert screening for severe sepsis in emergency department patients increases lactate test- ing but does not improve inpatient mortality. Appl Clin Inform 2010;1(4):394-407.
  26. Gultepe E, Green JP, Nguyen H, Adams J, Albertson T, Tagkopoulos I, et al. From vital signs to clinical outcomes for patients with sepsis: a machine learning basis for a clinical decision support system. J Am Med Inform Assoc 2014;21(2):315-25.
  27. Dumont L, Harding AD. Development and implementation of a sepsis program. J Emerg Nurs 2013;39(6):625-30.
  28. Umscheid CA, Betesh J, VanZandbergen C, Hanish A, Tait G, Mikkelsen ME, et al. De- velopment, implementation, and impact of an automated early warning and re- sponse system for sepsis. J Hosp Med 2015;10(1):26-31.
  29. Nguyen SQ, Mwakalindile E, Booth JS, Hogan V, Morgan J, Prickett CT, et al. Automat- ed electronic medical record sepsis detection in the emergency department. Peer J 2014;2:e343.
  30. Alsolamy S, Salamah MA, Thagafi MA, Al-Dorzi HM, Marini AM, Aljerian N, et al. Di- agnostic accuracy of a screening electronic alert tool for severe sepsis and septic shock in the emergency department. BMC Med Inform Decis Mak 2014;14(1):1-6.

Leave a Reply

Your email address will not be published. Required fields are marked *