ACADEMIC EMERGENCY MEDICINE 2011; 18:483–487 © 2011 by the Society for Academic Emergency Medicine
Background: Implementation of rapid response systems to identify deteriorating patients in the inpatient setting has demonstrated improved patient outcomes. A “trigger” system using vital sign abnormalities to initiate evaluation by physician was recently described as an effective rapid response method.
Objectives: The objective was to evaluate the effect of a triage-based trigger system on the primary outcome of time to physician evaluation and the secondary outcomes of therapeutic intervention, antibiotics, and disposition in emergency department (ED) patients.
Methods: A separate-samples pre- and postintervention study was conducted using retrospective chart review of outcomes in ED patients for three arbitrarily selected 5-day periods in 2007 (pretriggers) and 2008 (posttriggers). There were 2,165 and 2,212 charts in the pre- and posttriggers chart review, with 71 and 79 patients meeting trigger criteria. Trigger criteria used to identify patients at triage were: heart rate of <40 or >130 beats/min, respiratory rate of <8 or >30 breaths/min, systolic blood pressure of <90 mm Hg, and oxygen saturation of <90% on room air. Median times (in minutes) were compared between pre- and posttrigger groups with interquartile ranges (IQRs 25–75), with the Wilcoxon rank sum test used to determine statistical significance.
Results: Overall median times were decreased among the posttriggers group. Median times to physician evaluation (21 minutes [IQR = 13–41 minutes] vs. 11 minutes [IQR = 5–21 minutes]; p < 0.001), first intervention (58 minutes [IQR = 20–139 minutes] vs. 26 minutes [IQR = 11–71 minutes]; p < 0.01), and antibiotics (110 minutes [IQR = 74–171 minutes] vs. 69 minutes [IQR = 23–130 minutes]; p < 0.01) were significant. Median times to disposition (177 minutes [IQR = 121–303 minutes] vs. 162 minutes [IQR = 114–230 minutes]; p = 0.18) were not significant.
Conclusions: Implementation of an ED triggers program allows for more rapid time to physician evaluation, therapeutic intervention, and antibiotics.
Emergency departments (EDs) are busy, crowded, and occasionally chaotic. There are new patients arriving at all times, in various conditions, with different complaints. Accurate sorting and processing of these patients is a priority for EDs. Triage systems date back to the battlefields of the Napoleonic Wars and the American Revolution.1 In more recent times a number of ED triage scores were created, implemented, and validated to address this priority.2–8 Triage scores are varied in their makeup, their application in the United States is not standard, and “vital” signs are not the main decision point in most triage systems.4,8,9
A number of systems or alerts are implemented in the ED to identify and bring resources to patients with traumatic injuries, ST-elevation myocardial infarctions, or acute strokes, but there is currently no system that we are aware of directed toward vital sign abnormalities. Because abnormal vital signs often portend overt deterioration, it is logical to alert emergency physicians to the presence of patients with abnormal vital signs at presentation, so that they may rapidly assess patients to determine if immediate actions are needed. We aimed to utilize the same philosophy of early identification and clinician awareness in place for other time-dependent disease processes on the larger, undifferentiated group of patients who present with grossly abnormal vital signs.
Early physician and precode team involvement in unstable patients is an area of study for in-hospital patients. Rapid response teams (RRT) or medical emergency teams were created in many in-hospital settings as an effective mechanism to get physicians and nurses to the bedside of patients with acute changes in clinical stability.10–13 Clinical triggers were implemented in the Denver Health Medical Center, as a variation of the RRT. They were based on physiological criteria as well as a change in clinical status. The triggers included a respiratory rate of fewer than 8 breaths/min or more than 28 breaths/min, an oxygen saturation of <90% on room air, a systolic blood pressure less than 90 mm Hg, a heart rate more than 120 beats/min or under 50 beats/min, and several other markers that would fall under “marked nursing concern or clinical change.” When these clinical “triggers” were met, it required the physician taking care of the patient to evaluate the patient.11 Our goal was to evaluate the effect of a clinical triggers program at ED triage on timing of patient evaluation and management.
A separate samples pre- and postintervention study was conducted using retrospective chart review of outcomes in ED patients for three randomly selected 5-day periods (15 days total) in both 2007 (pretriggers) and 2008 (posttriggers). Institutional review board approval for this study was granted.
Study Setting and Population
The setting was a 50,000 + -visit ED in a tertiary care center. Eligible patients were those meeting the following inclusion criteria: 1) adult patients age 18 or older, 2) presented during one of the 5-day periods arbitrarily selected for inclusion, and 3) met any one of the predefined triggers vital signs. The trigger vital sign parameters were established based on prior data, in-hospital policies, and ED management consensus. The ED trigger criteria were: heart rate of <40 or >130 beats/min, respiratory rate of <8 or >30 respirations/min, systolic blood pressure of <90 mm Hg, or an oxygen saturation of <90% on room air. Marked nursing concern was also a trigger criteria; however, as there were no explicit criteria for this nor a mechanism to capture these patients in a retrospective manner, and to facilitate identification of comparable groups for comparison, patients who had a sole trigger criteria of marked nursing concern in the posttrigger period were not included in the analysis. Because there was a separate alert mechanism in place, patients were excluded from the study if they were a trauma activation, a code stroke, or having an acute ST-segment elevation mycardial infarction. The pretriggers stage was our standard care in the ED. In 2008 we created the ED trigger program as an offshoot of our quality assurance program. The specific process change for the ED trigger program is an overhead page by the triage nurse “Trigger arriving to room x” for any patient who meets the predefined trigger criteria at triage and the expectation that the ED attending physician, senior and junior ED residents, ED nurse, and technician report to the room immediately. If a patient met trigger criteria, he or she was immediately placed in a room. If a standard room is not available, the patient was placed in a trauma resuscitation room, which is always kept open according to our ED policy for an initial rapid evaluation.
We performed a structured chart review of 2,165 charts for the pretriggers program and 2,212 charts for the posttriggers program. This review yielded 71 patients in the pretrigger study period and 79 patients in the posttrigger study period who meet inclusion criteria. A single data abstractor was used, trained in data abstraction and management. The primary outcome of interest was time from ED presentation (triage) to physician evaluation (electronic time stamp), and the secondary outcomes were time to initial intervention, time to antibiotics if indicated, and time to ED disposition (time of bed requisition, placement in observation status, or discharge from the ED). Initial intervention was defined as a documented, targeted, therapeutic intervention such as the administration of vasoactive medications, medications to control heart rate, intubation, noninvasive ventilation, cardioversion, and antibiotic administration (Table 1). We chose not to include intravenous fluid administration, oxygen administration, intravenous access, electrocardiogram, or laboratory testing as part of this definition in keeping with the Emergency Severity Index (ESI) definition of what is not included in level one life-saving interventions.9 Additionally, these are all interventions that nurses often initiate without specific physician direction. Measurements of time were abstracted from the hard copy of the patients’ ED record and from the time stamps in the electronic ED tracking system.
|First Intervention||2007 (n = 71)||2008 (n = 79)|
|Antibiotics||11 (15)||17 (22)|
|Analgesia||8 (11)||12 (15)|
|Blood||2 (3)||1 (1)|
|Cardiac (medications or electricity)||16 (22)||19 (24)|
|Mood stabilizer||4 (6)||2 (2)|
|Vasopressor||2 (3)||2 (2)|
|Respiratory (medications or ventilatory support)||14 (20)||18 (23)|
|Other||9 (13)||6 (8)|
|None||5 (7)||2 (2)|
Median times were compared between pre- and posttriggers groups (reported in minutes with interquartile range [IQR] 25–75), with the Wilcoxon rank sum test used to determine statistical significance, with p-values reported where appropriate with an alpha set at 0.05 as being significant. Microsoft Excel (Microsoft Corp., Redmond, WA) and Graphpad Statistics (GraphPad Software, Inc., La Jolla, CA) were used for data analysis.
In the structured chart review, there were 71 patients who met the inclusion criteria in the pretriggers group, of whom 60% were female. In the posttriggers group there were 79 patients, of whom 62% were female. The mean age of patients in the pretriggers group was 61.0 (standard deviation [SD] ± 21.0) years of age, and the mean age in the posttriggers group was 60.1 (SD ± 24.5) years. The pre- and posttriggers groups were demographically similar (Table 2). The admission rate for trigger patients was 78% in 2007 and 86% in 2008, and the total admission rates for all patients during the same time periods in 2007 and 2008 were 41 and 43%. Overall median times were significantly decreased among the posttriggers group in time to physician evaluation, therapeutic intervention, and time to antibiotic administration when given (Table 3). The median time to physician evaluation was 21 (IQR = 13–41) minutes in the pretriggers group and 11 (IQR = 5–21) minutes in the posttriggers group (p < 0.001). Comparing the pre-versus posttriggers groups, the median time to therapeutic intervention was 58 (IQR = 20–139) minutes versus 26 (IQR = 11–71) minutes (p < 0.01). The time to antibiotics, when necessary, in the pretriggers group (n = 33) was 110 (IQR = 74–171) minutes and in the posttriggers group (n = 46) was 69 (IQR = 23–130) minutes (p < 0.01). The median time to disposition in the pretriggers group was 177 (IQR = 121–303) minutes and was 162 (IQR = 114–230) minutes in the posttriggers group (p = 0.18), which was not a significant change.
|Demographic||2007 (n = 71)||2008 (n = 79)|
|Mean age, yr||61 (56–66)||60 (55–66)|
|Sex (% Female)||60 (48–72)||62 (50–72)|
|HR < 40||5 (2–13)||1 (0–6)|
|HR > 130||31 (20–43)||36 (26–48)|
|RR > 30 or RR < 8||14 (7–24)||15 (8–25)|
|Oxygen sat < 90%||24 (15–25)||40 (28–51)|
|sBP < 90 mm Hg||31 (21–43)||19 (11–29)|
|1||3 (0.2–7)||8 (2.2–13.1)|
|2||76 (41–70)||81 (49–81)|
|3||21 (8–28)||11 (4–17)|
|Intensive care unit||30 (13–32)||42 (23–46)|
|Admit to ward||48 (26–51)||44 (24–49)|
|ED observation||4 (0.6–9)||1 (0.03–6)|
|Discharge||14 (5–18)||13 (5–18)|
|Intervention||Pretrigger Time, Median (IQR 25–75)||Posttrigger Time, Median (IQR 25–75)||p-value|
|Physician evaluation||21 (13–41)||11 (5–21)||0.0002|
|First intervention||58 (20–139)||26 (11–71)||0.01|
|Antibiotics||110 (74–171)||69 (23–130)||0.01|
|Disposition||177 (121–303)||162 (114–230)||0.18|
We demonstrated that the implementation of a clinical triggers system coupled with an overhead page, and the expectation of emergency physician evaluation, decreased the time to patient evaluation and intervention. It is important to note that the implementation of the ED trigger program was associated with neither an increase in any ED staffing, nor any other significant operational change. During this time period, the annual patient volume increased from 51,087 patients in fiscal year 2007 to 53,313 patients in the fiscal year 2008. Despite rising volume, and without an increase in resources, the implementation of this program was effective in reducing time to physician evaluation and initiation of therapy in these sick patients. The triggers patients accounted for approximately 3% of our total patient volume. Based on our current volume, this amounts to roughly four trigger patients in a 24-hour period.
There are several reported studies on the effectiveness of the RRT and clinical triggers protocol on the inpatient medicine side.11,13 Our novel approach was the implementation of the clinical triggers program in the ED, while focusing its effect on time to evaluation and therapy of this group of patients. Our ED triggers were modeled after the triggers described by Moldenhauser’s group in Denver.11 However, we tailored our triggers to reflect a similar criteria used on our inpatient ward. Furthermore, our triggers were also modified and defined by experienced ED attending physicians with an aim to not overburden the system with alerts.
Alerting physicians of the presence of very sick patients, and the process of getting experienced clinicians to the bedside of these patients quickly, makes intuitive sense. However, we are not aware of any prior ED-based studies on the process and implementation of this alert function. Furthermore, there are state and government guidelines regarding the timing of evaluation of patients based on their triage levels, with several recent studies examining the ability of EDs to meet these state and federal targets.14,15 These studies show that patients with the most emergent complaints and triage levels missed their target time evaluation at a higher rate and that this rate was increasing.
The standard criteria used in many EDs to identify, sort, and process patients from triage is the ESI, a well-established triage tool that is benchmarked for times by the U.S. Government Accountability Office (GAO). The ESI has five levels, with 1 being the most acute (require emergent life saving intervention) and 5 being the least acute (requiring no nursing resources).8,9 According to the GAO report, more than 73% of Level 1 ESI patients are not seen within the specified time frame of immediately (defined as less than 1 minute), and 50% of Level 2 patients exceed the recommended time frame.16 The GAO data show that a majority of hospitals are not coming close to the time seen recommendations for the sickest patients. Our study did not set out to address the time to physician evaluation based on ESI level, as the ED triggers program is separate from the ESI level; however, our study does show that implementation of the triggers program is successful in reducing the time to evaluation on this set of very sick ED patients. Our intervention shows a benefit with respect to evaluation timing and treatment time; however, it did not affect disposition time.
We did not examine the effect on timing to therapy on other patients in the ED and the possible diversion of resources away from some patients to evaluate the trigger patients. One may postulate that the temporary diversion of resources to this subset of patients may delay therapy on the rest of the patients in the ED. While not within the scope of this project, the triggers program did not appear to lead to an increased length of stay for all of the patients admitted through the ED. Our internal quality data regarding ED length of stay for admitted patients during the time period of this study did not change significantly. Further data need to be collected on daily throughput times based on the number of patients seen and the number of trigger patients seen during that time period to see if the trigger program potentially diverts resources and leads to a delay in other patients’ care.
This study was subject to all the limitations of a retrospective design, including incomplete data and limited ability to control for confounders. We believe we captured all patients who met inclusion criteria during the study period, as every single chart of a patient evaluated during the study period was examined for inclusion. A retrospective study also limits the ability to control for the Hawthorne effect; however, the physicians were not aware that a study was ongoing, and therefore we do not believe that there was a major Hawthorne effect during this study. Another limitation is that we did not include “marked nursing concern” as a trigger for evaluation. Although part of our current trigger criteria, we chose not to include this trigger for analysis as there was no way to back track and identify patients who were triggered by nursing concern. While recognizing that nursing concern is important, it is difficult to reliably and accurately identify these patients retrospectively based on chart review. Going forward, we plan to automate the data collection of ED trigger patients as identified by triage vital signs and will include a method for nurses to identify patients who were triggered for marked nursing concern (i.e., acute mental status changes), and we will then be able to comment on this subset of ED clinical triggers patients.
A further limitation is that the times documented may not accurately represent the actual timing of a given intervention. In our ED, physicians “sign up” for patients on an electronic ED tracking system when they are seeing a patient. While this may not reflect the exact time to physician evaluation, it mirrors our practice, and the method in which one “signs up for a patient” did not change in the pre- or posttriggers time period.
Another possible limitation of the study is confounding regarding the timing of interventions in the pre- and posttrigger era. That is, are the times improved due to nonmeasured operational changes? We did not make any other significant operational changes during these time periods, and the staffing levels of physicians were stable in the pre- and posttrigger periods. While we did not study the throughput time or time to interventions on all patients in the study period, we feel that the lack of other significant operational changes makes this less likely. As there we no other significant operational changes to the ED during this time frame, we believe that the substantial improvement in time to physician evaluation and intervention may be due to the ED triggers program, although further research is needed with adequate control for potential confounding variables.
This study was a process study, and final clinical outcomes were not assessed. There is no definitive documentation that a trigger was paged overhead in the posttrigger phase, so we are unable to say if there were triggers that were missed. Finally, this was a single site, so our findings may not be applicable to other institutions or patient populations.
We believe that the simple process of the introduction of a clinical triggers program is beneficial to patients with respect to time to evaluation. This program can be implemented at minimal cost, as most EDs have either an overhead paging system or a direct paging system. We have shown that the introduction of an ED clinical triggers program was successful at reducing the time to physician evaluation and therapeutic intervention in patients with clearly abnormal vital signs, although not at reducing time to disposition. We were able to institute this program without changing staffing and without adding resources to the ED. This intervention helps to improve the speed of the evaluation and treatment of some of our sickest patients.