The Use of Scripting at Triage and Its Impact on Elopements
Presented at the American College of Emergency Physicians Scientific Assembly, Boston, MA, October 5–8, 2009.
The authors report no financial disclosures.
Address for correspondence and reprints: Daniel A. Handel, MD, MPH; e-mail: firstname.lastname@example.org.
Objectives: The objective of this study was to measure the effect of scripting language at triage on the likelihood of elopements, controlling for patient volume and other potential confounding variables.
Methods: This was a pre- and postintervention cohort study using the same 5-month period (November 2007–March 2008 and November 2008–March 2009, respectively) that included in the analysis all patients 21 years of age and older, who presented to the triage window in the emergency department (ED) waiting room (not by ambulance). As part of the scripting, triage nurses informed patients of the longest waiting time (at that point in time) for any patients still waiting to be brought back from the waiting room into the ED. Rates of elopement were compared between patients who did and did not receive the scripting, controlling for individual and daily ED variables.
Results: A total of 24,390 ED visits were included in this analysis. The elopement rate was 4.4% among ED patients in the prescripting period, compared to 2.3% in the postscripting period. In a multivariate logistic regression model, the use of scripting was significantly associated with decreased odds of elopement, compared to the nonscripting group (odds ratio [OR] = 0.61, 95% confidence interval [CI] = 0.46 to 0.80).
Conclusions: The use of triage scripting was found to significantly reduce elopement rates in patients placed in the ED waiting room, even after controlling for other confounding variables. Scripting is a simple and underutilized technique that can have a positive effect for patients and the ED.
ACADEMIC EMERGENCY MEDICINE 2010; 17:495–500 © 2010 by the Society for Academic Emergency Medicine
As emergency department (ED) crowding continues to be a concern nationwide, waiting times for patients are increasing.1 While in the waiting room, ambulatory patients are faced with the uncertainty of when they will be seen, leading to decreased patient satisfaction. Patient satisfaction has been an increasing focus in health care, but because of crowding, patients overall have been found to have decreased satisfaction with their ED experience.2 In the classic Harvard Business School case study The Psychology of Waiting Lines, one of the key principles is that “uncertain waits are longer than known, finite waits.”3 Using this technique, customers waiting at restaurants are given an estimation of their wait time to be seated, as are those at amusement parks while they wait for their favorite ride.
In the ED, patient length of stay (LOS) has also been closely tied to patient satisfaction.4 Patients with prolonged boarding times in the ED have been found to have lower levels of satisfaction.5 In the ED setting, one group of authors reported that providing patients with exaggerated wait times led to improved patient satisfaction and perception of their wait times.6,7
Various patient satisfaction surveys are employed nationally to measure patients’ ED experience, each with advantages and limitations. The left-without-being-seen or elopement rate has also been proposed as a proxy for patient satisfaction.8
One technique employed to create environments conducive to consistently satisfying patients is scripting. The use of scripting in the inpatient setting by nurses has been used with success to meet patient needs.9 We developed a triage scripting message to provide ambulatory patients an estimation of the wait time before they could expect to be seen by a provider. The purpose of this study was to measure the effect of this scripting language at triage on the likelihood of elopements, as an important component of patient satisfaction.
This was a pre- and postintervention cohort study to assess the effect of scripting on elopements. The institutional review board at the study hospital approved the study protocol.
Study Setting and Population
The study setting was an ED with approximately 40,000 annual visits. The ED serves as an urban academic teaching hospital and Level 1 trauma center. All ambulatory patients 21 years of age and older who presented to the triage window in the ED waiting area (not transported by ground or air ambulance or by police car) were included in the analysis. Patients under 21 were not included because they are seen in the pediatric ED, which has a distinct process compared to the rest of the ED.
To provide triage scripting, all triage nurses were trained to notify patients of the wait time. Following an ordered triage process, this waiting time was provided to the patient only after the triage process was completed, which included entering the initial chief complaint, vital signs, and medical history into the electronic health record (EHR). Therefore, any patient who eloped immediately after being provided this information was captured and counted as an elopement. The wait time provided was the longest time for a patient already in the waiting room at that moment in the ED, as seen on the EHR patient track board. This was delivered using a standardized script given to the patient that reads as follows:
We will make every effort to have an emergency department physician or nurse practitioner see you as soon as possible. At this time, we are anticipating that you will be seen in approximately [insert time here] minutes [or hours]. If we get a major trauma victim or heart attack patient, it could possibly push your wait somewhat longer, but we do promise to get you seen as soon as possible. Thank you for your patience and understanding.
A discrete data entry location, involving a single click of a button, was created in the EHR to track whenever scripting was provided after implementation of the plan. For those patient visits that did not have the button clicked, it was presumed that scripting had not been provided.
On October 1, 2008, the scripting began. After the implementation of triage scripting, a 1-month period was excluded (October 2008) allowing comparison of elopement between November 2007–March 2008 (prescripting period) and November 2008–March 2009 (postscripting period). We excluded the first month after scripting implementation from the analysis since it was felt that there would be inconsistencies at first, both in the use of the scripting and in the documentation of its use. Similar time periods of the year were compared to control for potential temporal trends and seasonality. We chose a pre/post study design because of concerns that a randomized trial, while controlling for secular trends, would introduce compliance problems when nurses were forced to provide scripting in an inconsistent (randomized) fashion.
We used retrospective data from our EHR with visits as the unit of analysis. All patient identifiers were removed prior to analysis. The primary outcome variable was a dummy variable of whether or not the patient eloped. Elopement was defined as the patient leaving prior to being evaluated by a provider.
The primary independent variable was the intervention period, i.e., whether or not the visit occurred after implementation of the scripting intervention. None of the patients in the prescripting period received the scripted message. There were also some patients who did not receive a scripted message during the postscripting period. Our primary analysis adopted an intent-to-treat approach, and all visits occurring in the postscripting period were included in the postintervention group whether or not they received the scripting. In addition, we performed two sensitivity analyses. For the first sensitivity analysis, we still used the intervention period as the main independent variable, but we excluded those patients who were marked on the triage scripting notification as “direct to room” and those when the difference between the time the patient was placed in a room and the patient was triaged was less than zero (implying that they were brought back immediately to a room and then triaged). Those patients were excluded because elopement was more pertinent to patients who waited in the waiting room, and the scripting did not apply to patients brought directly back to the treatment area. The second sensitivity analysis included the same sample as the first sensitivity analysis, but used three variables, consisting of the before-scripting period, an after-scripting period when no scripting was provided during the visit, and an after-scripting period where scripting was provided during the visit.
Other predictors and potential confounders include both patient-level and ED-level daily variables. Patient level variables included age, sex, triage acuity (Emergency Severity Index [ESI] 1–5), time of arrival, whether the patient was seen on a weekend or weekday, and type of insurance (private, public [Medicare, Medicaid], self-pay, and other). ED-level daily variables represented overall volume and crowding and included total number of patients, average LOS for both admitted and discharged patients, the presence or absence of the expeditor during that date, and the status of ambulance diversion on that date of visit.
The expeditor role was introduced in the middle of the postscripting period. A paramedic worked as the expeditor to help facilitate the flow of patients through the ED and was felt to possibly influence elopement. This person worked during peak times of the day from 1 pm to 1 am.
Patient characteristics were summarized using descriptive statistics. Differences between pre- and postscripting visits were compared using a two-sided t-test for continuous variables and a chi-square test for categorical variables. A general estimating equations (GEE) multiple logistic regression model was used to model the association between elopement and period, after adjusting for potential confounding variables for both the intent-to-treat analysis and the sensitivity analyses. The GEE approach was used to control for clustering within subjects who had more than one ED visit, because ED visits by the same patient were not independent. Associations between elopement and each independent variable were investigated first in a univariate analysis; variables with a p-value of ≤0.20 were then considered in the multivariate model. A p-value of less than 0.05 was considered significant in the final model. The expeditor variable was determined a priori to be a possible confounder and was left in the model regardless of statistical significance. Initial analysis of the data did not find any serial correlation. In addition, linearity between elopement and continuous variables on the logit scale was assessed. If linearity was not satisfied, the continuous variable was categorized and entered into the model as a categorical variable. In particular, the age variable was categorized based on quartiles, and hours on total ambulance diversion was categorized as a three-level variable: no diversion, 0%–25% time on diversion, and >25% of time on diversion during the 24-hour period on the day of visit. We used Stata v10.0 (StataCorp, College Station, TX) for all data analysis.
A total of 16,462 patients, with 24,390 ED visits, were included in this analysis. Overall, the mean (±SD) age was about 44 (±16.8) years, and approximately 48% of the population was male. In the postscripting period 40% of eligible patient visits (4,792/11,944) received the triage script. A summary of patient characteristics and ED daily variables is provided in Table 1. In the intent-to-treat analysis, without excluding any visits, the elopement rate was 3.4% during the prescripting period, compared to 1.6% in the postscripting period (see Figure 1). Among ED visits in the postscripting period, the elopement rate was 2.3% among those who received scripting versus 4.4% among those who did not, a significant decrease based on univariate analysis (p < 0.001, Table 1).
Overall Demographics of Nonscripted and Scripted Periods
|Age (yr), mean (SD)||44.0 (±16.9)||44.4 (±16.7)||0.070|
|Sex (% male)||48.1||49.6||0.051|
|Private insurance (%)||36.3||30.5||<0.001*|
|Publicly insured (%)||34.6||38.1|
|Patient arrival time||0.138|
| 7 am-3 pm (%)||40.9||42.1|
| 3 pm-11 pm (%)||43.9||42.7|
| 11 pm-7 am (%)||15.2||15.2|
|Arrived on weekend (%)||27.7||27.6||0.003*|
|Daily LOS admitted (hours), mean (SD)||6.3 (±1.6)||5.5 (±1.0)||<0.001*|
|Daily LOS discharged (hours), mean (SD)||3.9 (±0.7)||3.7 (±0.5)||<0.001*|
|Daily boarding hours for admitted patients (hours), mean (SD)||2.2 (±1.4)||1.7 (±1.0)||<0.001*|
|Daily volume of patients, mean (SD)||113.8 (±14.6)||110.7 (±17.5)||<0.001*|
|Daily ambulance diversion (% of day)||<0.001*|
|Expeditor present that day (%)||0.0||70.9||<0.001*|
Results from the multiple GEE logistic regression model are shown in Table 2. Visits when scripting was provided were significantly associated with a decreased likelihood of elopement compared to nonscripted visits (odds ratio [OR] = 0.61, 95% confidence interval [CI] = 0.46 to 0.80) after adjusting for confounding variables. In contrast, there was no significant association between the use of an expeditor and elopements (p = 0.474).
Multiple GEE Logistic Regression for Likelihood of Patient Elopement
|Scripting provided||0.61 (0.46–0.80)||<0.001*|
| 30–42||1.07 (0.89–1.29)||0.470|
| 42–55||0.88 (0.73–1.08)||0.229|
| >55||0.48 (0.37–0.61)||<0.001*|
| Private insurance||Referent||<0.001*†|
| Publicly insured ||1.59 (1.31–1.94)||<0.001*|
| Self-pay ||2.03 (1.68–2.45)||<0.001*|
| Other||0.73 (0.41–1.32)||0.303|
|Patient arrival time|
| 7 am-3 pm||Referent||<0.001*†|
| 3 pm -11 pm||1.48 (1.26–1.73)||<0.001*|
| 11 pm-7 am||1.18 (0.94–1.48)||0.157|
|ESI triage level|
| 1 or 2||Referent||<0.001*|
| 3||2.24 (1.68–2.99)||<0.001*|
| 4 or 5||1.74 (1.25–2.42)||0.001*|
| Patient seen on weekend||0.70 (0.58–0.85)||<0.001*|
|Daily ambulance diversion (% of day)|
| 0%–25%||1.27 (1.06–1.51)||0.010*|
| >25%||1.51 (1.22–1.86)||<0.001*|
| Total daily ED patients||1.02 (1.01–1.02)||<0.001*|
| LOS for discharged patients||1.27 (1.14–1.42)||<0.001*|
| Expeditor present that day||1.12 (0.83–1.50)||0.474|
Among patient-level variables, age, insurance status, arrival time, and whether the patient was seen during the weekend were significantly associated with elopement (Table 2). Compared to patients less than 30 years old, those greater than 55 years of age were less likely to elope, but this was not true for patients between 20–42 and 42–55 years (p < 0.001). In addition, publicly insured patients (Medicaid and Medicare) and self-pay patients were more likely to elope compared to those privately insured. Patients seen between 3 pm and 11 pm were more likely to elope than those seen between 7 am and 3 pm (Table 2). Patients seen on the weekend were less likely to elope compared to those seen on weekdays (p < 0.001). Those who were not a Level 1 or 2 triage acuity were also more likely to elope.
Among ED-level daily variables, ambulance diversion was found to be associated with increased odds of elopement, especially when occurring for greater than 25% of the day (OR = 1.51, 95% CI = 1.22 to 1.86). An increase in daily ED volume had a small but significant effect on elopement (OR = 1.02, 95% CI = 1.02 to 1.42, p < 0.001). Increasing LOS for discharged patients was also significantly associated with an increased likelihood of elopement for every hour increase of LOS (OR = 1.27, 95% CI = 1.14 to 1.42; Table 2).
When excluding those patients who were marked on the triage scripting notification as “direct to room,” and those who were brought back immediately to a room then triaged, we still observed a significant decrease in elopement between the pre- and postscripting period. In particular, 5.1% of visits (496/9,828) eloped during the prescripting period, compared to 3.4% of visits (207/6,137) during the postscripting period, and the adjusted OR from the GEE logistic regression model was 0.73 (95% CI = 0.54 to 0.996, p = 0.047). When comparing the elopement for three groups, the nonscripted visits during the postscripting period had a similar elopement rate (4.4%, 89/2006) to those in the prescripting period, but the elopement rate decreased significantly among visits provided with scripting (2.9%, 118/4,131). The adjusted ORs were 0.61 (95% CI = 0.43 to 0.86) and 0.66 (95% CI = 0.49 to 0.89), respectively, when comparing visits provided with scripting to prescripting visits and nonscripted visits during the scripting period. Associations between elopement and patient level, and ED-level daily variables in the two sensitivity analyses, were similar to those from the intent-to-treat analysis (Table 2) and are not reported here.
The role of the expeditor was not found to have a significant impact on elopement rates. The overall elopement rates during the 3-month period when an expeditor was in place did not differ significantly from the 3-month period prior to having an expeditor (1.9% vs. 2.0%; p = 0.889).
We found the use of scripting at triage to be associated with significantly lower rates of elopement, both in a univariate analysis (Table 1) and in a multiple logistic regression analysis (Table 2). During the scripting intervention period, it appears that scripting was used more often during busier periods in the ED. Thus, the actual effect of this intervention may be blunted in this study, as previous studies have demonstrated that increased ED volume may be associated with an increased elopement rate.10 Unlike prior studies, which provided exaggerated waiting times,6,7 this study used a simple metric, that being the longest wait time of a patient already in the waiting room at that moment. Other interventions to reduce elopement rates, including physicians at triage,11–13 additional physician staffing,14 ED acute care bed expansion,15 and the creation of an ED observation unit,16 all require a significant amount of resources to have a positive effect. It is encouraging that the use of a simple intervention such as a triage script providing estimated wait times also appears to have a positive effect on elopement rates.
There are a number of ED-level factors that have been shown to be associated with elopement rates including ED volume, the number of boarding hours for admitted patients, the percentage of patients arriving by ambulance, and LOS for discharged patients.17 Our results confirm the importance of ED volume and LOS for discharged patients as predictors of elopement, but because ambulance arrivals and boarding times were excluded from our analysis, these other variables could not be assessed.
Patients were more likely to elope between 3 pm and 11 pm when compared to the 7 am to 3 PM time period. Weekend patients and those seen between the hours of 7 am and 3 pm were less likely to elope. These periods are usually lower volume in our ED. The highest volume of both adult and pediatric patients occurs during afternoon hours.18 Older patients, and those with increased acuity (based on triage score), were also found to be less likely to elope.
In terms of insurance status, patients who were classified as “self-pay” (i.e., uninsured) were more likely to elope when compared to those with private insurance. This is similar to studies in pediatric patients that found self-pay patients were more likely to elope.18 Publicly insured patients were more likely to elope compared to privately insured patients, which may be partly due to the fact that there is no financial disincentive for them to leave prior to completion of their ED evaluation.
There was no proof that filling out the “triage scripting” field in the EHR meant that it was actually given, as no direct observation of the triage encounter was included in the study due to limited resources. Conversely, it is possible that the triage script was given to a patient and the triage nurse forgot to click on the field in the EHR. Also, it is possible that more cooperative patients were more likely to receive the scripting. In addition, scripting was more likely to be given during busier periods in the ED. If anything, the increased use of scripting during the busiest periods in the ED might be expected to blunt the overall effect on elopement rates due to the increasing effect that busier periods may have on elopement. Another important consideration is how the message was delivered. For example, someone who appears to be simply reading a script (as opposed to talking to the patient) may be less likely to positively affect the patient’s behavior and experience. Furthermore, even though the presence of an expeditor was included in the multivariate model, there is a concern that it may have had an impact on elopement rates.
Several other factors that have been found to have an effect on elopement rates in other studies were not available for this analysis. The primary language of the patients was not a variable obtainable in this study. This is an important consideration as patients who are non-English speaking have been found to be less likely to elope, possibly because they have limited outpatient resources.19 We did not look at inpatient occupancy rates, which has been shown to have an effect on elopement rates in the ED.20 No ED occupancy rates were available, and these also have been found to correlate positively with elopements.21 We were unable to tell whether or not the ED was crowded or on diversion at the specific time of any patient visit. Finally, given the retrospective nature of the data, we were only able to measure association and not causality in this study.
Scripting, or notifying a patient at triage of an approximate wait time using the longest actual wait time for a patient already in the waiting room, is a simple and underutilized technique that requires minimal additional resources and can have a positive effect on elopement rates in ED patients.