52
Journal for Healthcare Quality
Evaluation of an Emergency Department Triage Screening Tool for Suspected Severe Sepsis and Septic Shock Catherine Patocka, Joel Turner, Xiaoqing Xue, Eli Segal Abstract: Early identification of septic patients is important to prevent delays in appropriate management. To improve detection of septic patients presenting to the emergency department (ED), we implemented a triage screening tool. Our study sought to determine the effect of this tool on time to antibiotics in patients with suspected severe sepsis or septic shock presenting to the ED. This was a retrospective chart review examining the time interval to antibiotics pre- and postimplementation of the triage tool. Multiple linear regression analyses were conducted to evaluate the effect of the triage tool on time to antibiotics while controlling for the effect of level of triage. We identified 185 patients with severe sepsis or septic shock in the pretriage tool group and 170 patients in the posttriage tool group. The mean time (in minutes) to antibiotics (±SD) in the pre- and postcohorts was 283 (±213) and 207 (±150), respectively. The multivariable analysis showed that the mean time to antibiotics decreased by 21% (95% CI 6–36%, p < .0074) comparing pre- versus posttriage tool implementation. The use of a sepsis triage screening tool significantly decreased the time to antibiotics in patients presenting to the ED with suspected severe sepsis or septic shock.
Keywords acute care hospital emergency research-quantitative
Journal for Healthcare Quality Vol. 36, No. 1, pp. 52–61
C 2013 National Association for
Healthcare Quality
Sepsis is a common cause of morbidity and mortality (Angus, 2001; Angus, Pires Pereira, & Silva, 2006; Shapiro et al., 2006). Early recognition and treatment of patients with sepsis, specifically early goal-directed therapy (EGDT), a time-dependent protocolized approach to hemodynamic optimization, has been shown to significantly reduce mortality (Rivers et al., 2001). A primary difficulty in the management of patients with sepsis or septic shock is ensuring timely recognition and application of basic therapies (Funk, Sebat, & Kumar, 2009). The Surviving Sepsis Campaign (SSC) was created to produce awareness, consensus, and evidencebased guideline for care of septic patients and quality improvement (Dellinger et al., 2004, 2008). In 2008, the SSC developed an EGDT "resuscitation bundle" for patients with severe sepsis. This resuscitation bundle includes lactate screening, blood cultures obtained prior to antibiotic administration, early appropriate antibiotic therapy, fluid resuscitation and initiation of EGDT with central venous pressure (CVP), and central venous oxygen saturation
(ScvO2) monitoring (Dellinger et al., 2008; Osborn, Nguyen, & Rivers, 2005). Efforts to implement the SSC guidelines, including the resuscitation bundle, have focused on activating protocols once patients have been identified; however, little work has been done to improve patient identification (Francis, Rich, Williamson, & Peterson, 2010; Shapiro et al., 2006; Trzeciak et al., 2006). Given the delays related to crowding and the barriers associated with prompt and effective care in congested emergency departments (EDs), there is a need for effective ways to rapidly identify patients meeting the criteria for EGDT (Sills, Fairclough, Ranade, & Kahn, 2011; Weber, McAlpine, & Grimes, 2011). To promote timely recognition of potentially septic patients that would benefit from early treatment, the Jewish General Hospital (JGH) ED implemented a triage tool designed to identify septic patients. The objective of this study was to determine the impact of this triage screening tool on the timeliness of treatment of septic patients. The primary outcome of our study was the time interval from triage to administration of antibiotic therapy in patients presenting with suspected severe sepsis or septic shock before and after implementation of the triage tool.
Methods The JGH ED is an urban tertiary teaching hospital in Montreal, Quebec, Canada, with a volume of approximately 65,000 adult patients annually. As part of a quality improvement initiative to promote early identification of potentially septic patients, the JGH ED implemented a sepsis triage tool (Figure 1). The sepsis triage tool was developed by one author’s (J. T.) review of the literature and consists of parameters, which both predict illness severity in sepsis (e.g., Systemic Inflammatory Response Syndrome criteria) and are easily measured at triage. This tool was used at triage in addition to the Canadian Triage and Acuity Scale (CTAS) already in place in the JGH ED. CTAS is a triage system that helps define patients’ need
Vol. 36 No. 1 January/February 2014
for timely care and allows EDs to evaluate their acuity level, resource needs, and performance against specific operating objectives (Beveridge et al., 1999). The primary operational objective of the CTAS triage scale is related to time to see the physician, which is graded on a level of I through V as described in Box 1. The CTAS score sorts patients according to the type and severity of their presenting signs and symptoms without consideration of resource utilization as compared to the Emergency Severity Index version 3, used in many centers across the United States, which rates patients based on acuity and predicted resource intensity (Tanabe, Gimbel, Yarnold, Kyriacou, & Adams, 2004).
In the fall of 2004, the ED nursing educator conducted a multimodal educational intervention including e-mail notices and one-onone training sessions (5–10 min) to explain the triage protocol and its components to the 50– 60 ED triage nurses. Also, all nurses working in the resuscitation area (where patients with severe sepsis and septic shock are placed) received approximately 2 hr of training about sepsis. The protocol and project were also introduced to the medical staff during their departmental meeting, along with formal group education about EGDT. The tool was initially piloted in the fall of 2004. Minor modifications were made to the form following the pilot period (i.e., inclusion of the date on the triage form), but no major changes were made to the protocol.
pected severe sepsis or septic shock. Patients in the precohort were identified between January 1, 2005 and December 31, 2005 by review of admission/discharge diagnosis and diagnosis on ED death certificate presenting to the JGH ED. Diagnoses that were used included the following: acute endocarditis, bacteremia, cellulitis/abscess, cholangitis, encephalitis, endometritis, epiglottitis, fasciitis, fever, malaria, mediastinitis, meningitis, peritonitis, pneumonia, pyelonephritis, septic arthritis, septicemia, urinary tract infection/cystitis. Patients in the posttriage tool cohort were identified between January 1, 2008 and December 31, 2008 via two separate methods: (1) the triage tool and (2) a review of admission/discharge diagnosis and diagnosis on ED death certificate (using the same diagnoses as described above). Two methods were used to identify our postcohort because we were interested in evaluating the implementation rate of the sepsis-screening triage tool. The number of patients correctly identified by the triage tool was compared to the number of patients identified by the admission/discharge/death certificate diagnosis. This is noted in the Results section as a percentage of implementation. A 2-year delay was used between the pre- and postcohorts because of a necessary training and implementation interval for the tool (∼6 months). Subsequently, the hospital converted all medical records to scanned electronic copies and the first entire calendar year available in that format was 2008. Identified patients’ charts were reviewed and included if they met the following criteria: a core temperature 38◦ C (measured orally, unless the patient met criteria to have a rectal temperature performed—see Figure 1 for details) and at least 3 illness severity criteria at triage—altered mental status (AMS), heart rate >90 beats/min, oxygen saturation 7.7 mmol/l in the absence of diabetes), respiratory rate >20 breaths/min and/or mottled skin. Only patients who received antibiotics within the first 24 hr after arrival to hospital (regardless of whether they were still in the ED or admitted to inpatient wards) were included.
Study Design, Setting, and Population
Data Collection
After obtaining approval from our Institutional Review Board, we conducted a retrospective pre–post chart review of ED patients with sus-
The sepsis-screening triage tool was prospectively completed by triage nurses in the ED for all patients with suspected sepsis. All other data
Box 1 Level I Level II Level III Level IV Level V
Resuscitation patients who should be seen immediately by a physician. Emergent patients who should be seen by a physician in ≤15 min. Urgent patients who should be seen by a physician in ≤30 min. Less urgent patients who should be seen by a physician in ≤1 hr. Nonurgent patients who should be seen by a physician in ≤2 hr.
Adapted from Beveridge et al. (1999).
53
54
Journal for Healthcare Quality
Figure 1. The Jewish General Hospital Emergency Department Triage Tool
were gathered by retrospective chart review in accordance with established standards for chart review in Emergency Medicine (Gilbert, Lowenstein, Koziol-McLain, Barta, & Steiner, 1996; Lowenstein, 1999). Data were recorded using standard database software (Access; Microsoft Corporation, Redmond, WA). Data collection and entry on standardized forms were done by five data collection personnel who were trained before the start of the study. Training took approximately 3 hr and involved one-on-one detailed review of the standardized documents with the lead author (C. P.) followed by observed data collection of ∼15–20 charts. Each reviewer used a priori determined
rules, available in a standardized document, for data entry in each category including triage date and time, CTAS triage category, illness severity criteria, and time to first antibiotics (Appendix A and B). We also examined several secondary outcomes, determined a priori, based on components of the "resuscitation bundle" defined by sepsis guidelines (Dellinger et al., 2008; Osborn et al., 2005). Information regarding data on admission to hospital (defined as the patient being seen or discussed with a consulting service who then agreed to take the patient under their care), admitting diagnosis, and death (in the ED or during the admission for which they presented with suspected severe
Vol. 36 No. 1 January/February 2014
sepsis or septic shock) was also collected. All information was collected using the standardized documents. One author (C. P.) met at least once every month with each data abstractor to review data entry and to resolve conflicts about data coding and timing intervals. Data collection took approximately 300 hr. Interrater reliability on a random sample of 15% (160) of charts comparing one investigator’s rereview of the time intervals revealed an 85% agreement between raters.
Outcomes The primary outcome in this study was the time interval from triage time (as recorded directly on the triage sheet by the nurse at the time of triage) to administration of antibiotics (as recorded by the nurse on the medication administration sheet). Secondary outcomes included the time interval from triage time to first intravenous fluid bolus (>500 cc crystalloid), time placed in a monitored bed, time central line placed, time first CVP measured, time CVP > 8 mm Hg, time of mean arterial pressure >65 mm Hg, lactate values, time of blood culture collection and disposition.
Data Analysis In order to determine approximate sample size, we screened and recorded data on a pilot sample of 150 charts. Given the large uncertainty of data it was not practical to calculate sample size, though given the number of patients enrolled during the pilot phase and taking into account seasonal variability in patients presenting with sepsis, it was felt that a study period of 12 months would be both reasonable and practical. Descriptive statistics including mean ± standard deviation (SD), median with interquartile range (IQR), and proportions were used to describe the data collected pre and postperiod. The outcomes of this study are time intervals, with primary outcome being time from triage to time of antibiotic administration. Since the distributions of time intervals are highly skewed, natural log transformations were used for each interval (with 1 min added to all time intervals in order to prevent invalid arithmetic calculations). To compare the effects of the triage tool on various time intervals, multiple linear regression analyses were conducted, and through which the potential effects of confounding vari-
ables have been adjusted for. Three patientlevel variables, age, gender, and CTAS score (as defined by nurses at triage), were considered as confounding factors. Results are presented as percentage of change and their 95% confidence intervals (95% CI). Data were analyzed using SAS 9.2 (SAS Institute Inc., Cary, NC).
Results A total of 6,799 charts were screened and 1,052 were reviewed. Of these, 355 met our diagnostic criteria for suspected severe sepsis or septic shock pretriage tool implementation (n = 185) and posttriage tool implementation (n = 170). Patients in the two cohorts were similar in terms of patient demographics, CTAS triage code, and baseline illness severity criteria (Table 1). Although the total number of illness severity criteria fulfilled by patients in both cohorts were similar, patients in the pretriage cohort more often had AMS and mottled skin, whereas patients in the posttriage tool cohort were more often hyperglycemic. The mean time (in minutes) to antibiotics (±SD) in the pre- and postcohorts was 283 (±213) and 207 (±150), respectively. The median time to antibiotics (in minutes, IQR) was 235 (115–395) and 179 (90–270), respectively. Statistical analysis revealed that time to antibiotics decreased by 21% (95% CI 6–33%, p = .0074) posttriage tool implementation compared to pretriage tool period. There is also a significant decrease in time to antibiotics by 41% (95% CI 30–50%, p < .0001) in the subgroup of patients with CTAS levels 1–2 compared to the CTAS 3–5 group (Table 2). Figure 2 demonstrates the number of patients receiving antibiotics at hourly intervals from time of triage. The number of patients receiving antibiotics in the first hour after triage was similar in both cohorts; however, there were more patients in the posttriage tool cohort who received antibiotics 1–2 hr and 3–4 hr after triage. Time intervals to monitoring, first bolus, central line placement were similar, and there was a shorter interval to CVP >8 in the pretriage cohort as compared to the posttriage tool cohort. Very few patients in both cohorts had data relating to central line placement and CVP monitoring, and these data were not included in the multivariable analysis. Data from secondary outcomes are shown in Table 3. Compared to the pretriage tool cohort, patients in the posttriage tool cohort were more
55
56
Journal for Healthcare Quality
Table 1. Characteristics of Patients Who Met Criteria for Severe Sepsis or Septic Shock Before and After Implementation of the Sepsis Triage Tool No. of patients (%) Pretriage Tool N = 185
Posttriage Tool N = 170
75.3 (17.0) 87 (47)
72.4 (18.4) 74 (46)
8 (4.3) 95 (51.4) 77 (41.6) 5 (2.7) 0
6 (3.5) 99 (58.2) 64 (37.6) 0 0
Number of illness severity criteria 3 112 (60.5) 4 49 (26.5) 5 18 (9.7) 6 6 (3.2)
111 (65.6) 43 (25.4) 13 (7.7) 2 (1.2)
Characteristic Age, years (SD) Female gender CTAS triage code 1 2 3 4 5
Illness severity criteria 174 (94) Hyperthermia (>38◦ C) Hypothermia 11 (5.9) (90 beats/min) Tachypnea (RR 143 (77.2) >20 breaths/min) Altered mental 88 (47.6) status Mottled skin 97 (52.4) Hyperglycemia 71 (38.4) (glucose >7.7 mmol/l) 61 (33) Hypoxia (SpO2 4.0 mmol/l Admitted Admitting diagnosis (presumed etiology) Pneumonia Fever Cellulitis/abscess UTI/cystitis/pyelonephritis Septicemia Other Died (either in the ED or during the admission for which they presented with severe sepsis or septic shock)
UTI, urinary tract infection. * The data for lactate were missing for one patient in the postcohort, therefore the n value for this was 169. ** There were no mortality data available for two patients in the postcohort, therefore the n for this value was 168.
Figure 2. Percentage of Patients and Time to Antibiotics at Hourly Intervals Based on Time From Triage (as Recorded by the Nurse at the Beginning of Triage) to Administration of Antibiotics (as Recorded on the Medication Administration Sheet) Time to Antibiotics % of patients receiving antibiotics
25.0 20.0 15.0 Pre-Triage Tool
10.0
Post-Triage Tool
5.0 0.0 8hrs
Time (hrs)
Table 3. Percent Change in Time Intervals Postimplementation of Triage Tool Outcome Time to antibiotics Time to monitoring Time to first bolus(>500 cc crystalloid)
Independent Variable
Percent change (95% Confidence Interval)
p-Value
Posttriage tool vs. pretriage CTAS 1–2 vs. CTAS 3–4 Post triage tool vs. pretriage CTAS 1–2 vs. CTAS 3–4 Post triage tool vs. pretriage
−21% (−6% to −33%) −41% (−30% to −50%) −8% (−32% to 25%) −52% (−43% to −70%) −18% (−28% to 111%)
.0074
Journal for Healthcare Quality
Evaluation of an Emergency Department Triage Screening Tool for Suspected Severe Sepsis and Septic Shock Catherine Patocka, Joel Turner, Xiaoqing Xue, Eli Segal Abstract: Early identification of septic patients is important to prevent delays in appropriate management. To improve detection of septic patients presenting to the emergency department (ED), we implemented a triage screening tool. Our study sought to determine the effect of this tool on time to antibiotics in patients with suspected severe sepsis or septic shock presenting to the ED. This was a retrospective chart review examining the time interval to antibiotics pre- and postimplementation of the triage tool. Multiple linear regression analyses were conducted to evaluate the effect of the triage tool on time to antibiotics while controlling for the effect of level of triage. We identified 185 patients with severe sepsis or septic shock in the pretriage tool group and 170 patients in the posttriage tool group. The mean time (in minutes) to antibiotics (±SD) in the pre- and postcohorts was 283 (±213) and 207 (±150), respectively. The multivariable analysis showed that the mean time to antibiotics decreased by 21% (95% CI 6–36%, p < .0074) comparing pre- versus posttriage tool implementation. The use of a sepsis triage screening tool significantly decreased the time to antibiotics in patients presenting to the ED with suspected severe sepsis or septic shock.
Keywords acute care hospital emergency research-quantitative
Journal for Healthcare Quality Vol. 36, No. 1, pp. 52–61
C 2013 National Association for
Healthcare Quality
Sepsis is a common cause of morbidity and mortality (Angus, 2001; Angus, Pires Pereira, & Silva, 2006; Shapiro et al., 2006). Early recognition and treatment of patients with sepsis, specifically early goal-directed therapy (EGDT), a time-dependent protocolized approach to hemodynamic optimization, has been shown to significantly reduce mortality (Rivers et al., 2001). A primary difficulty in the management of patients with sepsis or septic shock is ensuring timely recognition and application of basic therapies (Funk, Sebat, & Kumar, 2009). The Surviving Sepsis Campaign (SSC) was created to produce awareness, consensus, and evidencebased guideline for care of septic patients and quality improvement (Dellinger et al., 2004, 2008). In 2008, the SSC developed an EGDT "resuscitation bundle" for patients with severe sepsis. This resuscitation bundle includes lactate screening, blood cultures obtained prior to antibiotic administration, early appropriate antibiotic therapy, fluid resuscitation and initiation of EGDT with central venous pressure (CVP), and central venous oxygen saturation
(ScvO2) monitoring (Dellinger et al., 2008; Osborn, Nguyen, & Rivers, 2005). Efforts to implement the SSC guidelines, including the resuscitation bundle, have focused on activating protocols once patients have been identified; however, little work has been done to improve patient identification (Francis, Rich, Williamson, & Peterson, 2010; Shapiro et al., 2006; Trzeciak et al., 2006). Given the delays related to crowding and the barriers associated with prompt and effective care in congested emergency departments (EDs), there is a need for effective ways to rapidly identify patients meeting the criteria for EGDT (Sills, Fairclough, Ranade, & Kahn, 2011; Weber, McAlpine, & Grimes, 2011). To promote timely recognition of potentially septic patients that would benefit from early treatment, the Jewish General Hospital (JGH) ED implemented a triage tool designed to identify septic patients. The objective of this study was to determine the impact of this triage screening tool on the timeliness of treatment of septic patients. The primary outcome of our study was the time interval from triage to administration of antibiotic therapy in patients presenting with suspected severe sepsis or septic shock before and after implementation of the triage tool.
Methods The JGH ED is an urban tertiary teaching hospital in Montreal, Quebec, Canada, with a volume of approximately 65,000 adult patients annually. As part of a quality improvement initiative to promote early identification of potentially septic patients, the JGH ED implemented a sepsis triage tool (Figure 1). The sepsis triage tool was developed by one author’s (J. T.) review of the literature and consists of parameters, which both predict illness severity in sepsis (e.g., Systemic Inflammatory Response Syndrome criteria) and are easily measured at triage. This tool was used at triage in addition to the Canadian Triage and Acuity Scale (CTAS) already in place in the JGH ED. CTAS is a triage system that helps define patients’ need
Vol. 36 No. 1 January/February 2014
for timely care and allows EDs to evaluate their acuity level, resource needs, and performance against specific operating objectives (Beveridge et al., 1999). The primary operational objective of the CTAS triage scale is related to time to see the physician, which is graded on a level of I through V as described in Box 1. The CTAS score sorts patients according to the type and severity of their presenting signs and symptoms without consideration of resource utilization as compared to the Emergency Severity Index version 3, used in many centers across the United States, which rates patients based on acuity and predicted resource intensity (Tanabe, Gimbel, Yarnold, Kyriacou, & Adams, 2004).
In the fall of 2004, the ED nursing educator conducted a multimodal educational intervention including e-mail notices and one-onone training sessions (5–10 min) to explain the triage protocol and its components to the 50– 60 ED triage nurses. Also, all nurses working in the resuscitation area (where patients with severe sepsis and septic shock are placed) received approximately 2 hr of training about sepsis. The protocol and project were also introduced to the medical staff during their departmental meeting, along with formal group education about EGDT. The tool was initially piloted in the fall of 2004. Minor modifications were made to the form following the pilot period (i.e., inclusion of the date on the triage form), but no major changes were made to the protocol.
pected severe sepsis or septic shock. Patients in the precohort were identified between January 1, 2005 and December 31, 2005 by review of admission/discharge diagnosis and diagnosis on ED death certificate presenting to the JGH ED. Diagnoses that were used included the following: acute endocarditis, bacteremia, cellulitis/abscess, cholangitis, encephalitis, endometritis, epiglottitis, fasciitis, fever, malaria, mediastinitis, meningitis, peritonitis, pneumonia, pyelonephritis, septic arthritis, septicemia, urinary tract infection/cystitis. Patients in the posttriage tool cohort were identified between January 1, 2008 and December 31, 2008 via two separate methods: (1) the triage tool and (2) a review of admission/discharge diagnosis and diagnosis on ED death certificate (using the same diagnoses as described above). Two methods were used to identify our postcohort because we were interested in evaluating the implementation rate of the sepsis-screening triage tool. The number of patients correctly identified by the triage tool was compared to the number of patients identified by the admission/discharge/death certificate diagnosis. This is noted in the Results section as a percentage of implementation. A 2-year delay was used between the pre- and postcohorts because of a necessary training and implementation interval for the tool (∼6 months). Subsequently, the hospital converted all medical records to scanned electronic copies and the first entire calendar year available in that format was 2008. Identified patients’ charts were reviewed and included if they met the following criteria: a core temperature 38◦ C (measured orally, unless the patient met criteria to have a rectal temperature performed—see Figure 1 for details) and at least 3 illness severity criteria at triage—altered mental status (AMS), heart rate >90 beats/min, oxygen saturation 7.7 mmol/l in the absence of diabetes), respiratory rate >20 breaths/min and/or mottled skin. Only patients who received antibiotics within the first 24 hr after arrival to hospital (regardless of whether they were still in the ED or admitted to inpatient wards) were included.
Study Design, Setting, and Population
Data Collection
After obtaining approval from our Institutional Review Board, we conducted a retrospective pre–post chart review of ED patients with sus-
The sepsis-screening triage tool was prospectively completed by triage nurses in the ED for all patients with suspected sepsis. All other data
Box 1 Level I Level II Level III Level IV Level V
Resuscitation patients who should be seen immediately by a physician. Emergent patients who should be seen by a physician in ≤15 min. Urgent patients who should be seen by a physician in ≤30 min. Less urgent patients who should be seen by a physician in ≤1 hr. Nonurgent patients who should be seen by a physician in ≤2 hr.
Adapted from Beveridge et al. (1999).
53
54
Journal for Healthcare Quality
Figure 1. The Jewish General Hospital Emergency Department Triage Tool
were gathered by retrospective chart review in accordance with established standards for chart review in Emergency Medicine (Gilbert, Lowenstein, Koziol-McLain, Barta, & Steiner, 1996; Lowenstein, 1999). Data were recorded using standard database software (Access; Microsoft Corporation, Redmond, WA). Data collection and entry on standardized forms were done by five data collection personnel who were trained before the start of the study. Training took approximately 3 hr and involved one-on-one detailed review of the standardized documents with the lead author (C. P.) followed by observed data collection of ∼15–20 charts. Each reviewer used a priori determined
rules, available in a standardized document, for data entry in each category including triage date and time, CTAS triage category, illness severity criteria, and time to first antibiotics (Appendix A and B). We also examined several secondary outcomes, determined a priori, based on components of the "resuscitation bundle" defined by sepsis guidelines (Dellinger et al., 2008; Osborn et al., 2005). Information regarding data on admission to hospital (defined as the patient being seen or discussed with a consulting service who then agreed to take the patient under their care), admitting diagnosis, and death (in the ED or during the admission for which they presented with suspected severe
Vol. 36 No. 1 January/February 2014
sepsis or septic shock) was also collected. All information was collected using the standardized documents. One author (C. P.) met at least once every month with each data abstractor to review data entry and to resolve conflicts about data coding and timing intervals. Data collection took approximately 300 hr. Interrater reliability on a random sample of 15% (160) of charts comparing one investigator’s rereview of the time intervals revealed an 85% agreement between raters.
Outcomes The primary outcome in this study was the time interval from triage time (as recorded directly on the triage sheet by the nurse at the time of triage) to administration of antibiotics (as recorded by the nurse on the medication administration sheet). Secondary outcomes included the time interval from triage time to first intravenous fluid bolus (>500 cc crystalloid), time placed in a monitored bed, time central line placed, time first CVP measured, time CVP > 8 mm Hg, time of mean arterial pressure >65 mm Hg, lactate values, time of blood culture collection and disposition.
Data Analysis In order to determine approximate sample size, we screened and recorded data on a pilot sample of 150 charts. Given the large uncertainty of data it was not practical to calculate sample size, though given the number of patients enrolled during the pilot phase and taking into account seasonal variability in patients presenting with sepsis, it was felt that a study period of 12 months would be both reasonable and practical. Descriptive statistics including mean ± standard deviation (SD), median with interquartile range (IQR), and proportions were used to describe the data collected pre and postperiod. The outcomes of this study are time intervals, with primary outcome being time from triage to time of antibiotic administration. Since the distributions of time intervals are highly skewed, natural log transformations were used for each interval (with 1 min added to all time intervals in order to prevent invalid arithmetic calculations). To compare the effects of the triage tool on various time intervals, multiple linear regression analyses were conducted, and through which the potential effects of confounding vari-
ables have been adjusted for. Three patientlevel variables, age, gender, and CTAS score (as defined by nurses at triage), were considered as confounding factors. Results are presented as percentage of change and their 95% confidence intervals (95% CI). Data were analyzed using SAS 9.2 (SAS Institute Inc., Cary, NC).
Results A total of 6,799 charts were screened and 1,052 were reviewed. Of these, 355 met our diagnostic criteria for suspected severe sepsis or septic shock pretriage tool implementation (n = 185) and posttriage tool implementation (n = 170). Patients in the two cohorts were similar in terms of patient demographics, CTAS triage code, and baseline illness severity criteria (Table 1). Although the total number of illness severity criteria fulfilled by patients in both cohorts were similar, patients in the pretriage cohort more often had AMS and mottled skin, whereas patients in the posttriage tool cohort were more often hyperglycemic. The mean time (in minutes) to antibiotics (±SD) in the pre- and postcohorts was 283 (±213) and 207 (±150), respectively. The median time to antibiotics (in minutes, IQR) was 235 (115–395) and 179 (90–270), respectively. Statistical analysis revealed that time to antibiotics decreased by 21% (95% CI 6–33%, p = .0074) posttriage tool implementation compared to pretriage tool period. There is also a significant decrease in time to antibiotics by 41% (95% CI 30–50%, p < .0001) in the subgroup of patients with CTAS levels 1–2 compared to the CTAS 3–5 group (Table 2). Figure 2 demonstrates the number of patients receiving antibiotics at hourly intervals from time of triage. The number of patients receiving antibiotics in the first hour after triage was similar in both cohorts; however, there were more patients in the posttriage tool cohort who received antibiotics 1–2 hr and 3–4 hr after triage. Time intervals to monitoring, first bolus, central line placement were similar, and there was a shorter interval to CVP >8 in the pretriage cohort as compared to the posttriage tool cohort. Very few patients in both cohorts had data relating to central line placement and CVP monitoring, and these data were not included in the multivariable analysis. Data from secondary outcomes are shown in Table 3. Compared to the pretriage tool cohort, patients in the posttriage tool cohort were more
55
56
Journal for Healthcare Quality
Table 1. Characteristics of Patients Who Met Criteria for Severe Sepsis or Septic Shock Before and After Implementation of the Sepsis Triage Tool No. of patients (%) Pretriage Tool N = 185
Posttriage Tool N = 170
75.3 (17.0) 87 (47)
72.4 (18.4) 74 (46)
8 (4.3) 95 (51.4) 77 (41.6) 5 (2.7) 0
6 (3.5) 99 (58.2) 64 (37.6) 0 0
Number of illness severity criteria 3 112 (60.5) 4 49 (26.5) 5 18 (9.7) 6 6 (3.2)
111 (65.6) 43 (25.4) 13 (7.7) 2 (1.2)
Characteristic Age, years (SD) Female gender CTAS triage code 1 2 3 4 5
Illness severity criteria 174 (94) Hyperthermia (>38◦ C) Hypothermia 11 (5.9) (90 beats/min) Tachypnea (RR 143 (77.2) >20 breaths/min) Altered mental 88 (47.6) status Mottled skin 97 (52.4) Hyperglycemia 71 (38.4) (glucose >7.7 mmol/l) 61 (33) Hypoxia (SpO2 4.0 mmol/l Admitted Admitting diagnosis (presumed etiology) Pneumonia Fever Cellulitis/abscess UTI/cystitis/pyelonephritis Septicemia Other Died (either in the ED or during the admission for which they presented with severe sepsis or septic shock)
UTI, urinary tract infection. * The data for lactate were missing for one patient in the postcohort, therefore the n value for this was 169. ** There were no mortality data available for two patients in the postcohort, therefore the n for this value was 168.
Figure 2. Percentage of Patients and Time to Antibiotics at Hourly Intervals Based on Time From Triage (as Recorded by the Nurse at the Beginning of Triage) to Administration of Antibiotics (as Recorded on the Medication Administration Sheet) Time to Antibiotics % of patients receiving antibiotics
25.0 20.0 15.0 Pre-Triage Tool
10.0
Post-Triage Tool
5.0 0.0 8hrs
Time (hrs)
Table 3. Percent Change in Time Intervals Postimplementation of Triage Tool Outcome Time to antibiotics Time to monitoring Time to first bolus(>500 cc crystalloid)
Independent Variable
Percent change (95% Confidence Interval)
p-Value
Posttriage tool vs. pretriage CTAS 1–2 vs. CTAS 3–4 Post triage tool vs. pretriage CTAS 1–2 vs. CTAS 3–4 Post triage tool vs. pretriage
−21% (−6% to −33%) −41% (−30% to −50%) −8% (−32% to 25%) −52% (−43% to −70%) −18% (−28% to 111%)
.0074
