Original article 33
The predictive validity of RETTS-HEV as an acuity triage tool in the emergency department of a Danish Regional Hospital Noel Péreza, Louise Nissena, Rasmus F. Nielsena, Poul Petersena and Karin Bieringb Introduction The Rapid Emergency Triage and Treatment System – Hospital Unit West (RETTS-HEV) is a triage system used in the emergency department (ED) in Herning, Denmark, since 2010. It categorizes patients according to priority and defines a time limit on how long patients can wait before being seen by a doctor depending on the severity of their condition. The purpose of this study was to determine the predictive validity of RETTS-HEV by measuring the association between triage scores and outcomes such as the admission rate, the length of stay (LOS), and mortality. Materials and methods We performed an observational cohort study by examining the medical records of all patients who attended the ED from 1 September 2012 to 30 November 2012, at the Regional Hospital West Jutland in Herning, Denmark (N = 4680). We defined the following outcomes to make associations with the patients’ triage category: in-hospital mortality, and 30, 60, and 90-day mortalities, the hospital LOS and the admission rate, on the basis of complete information from the Danish National Patient Registry.
Introduction Emergency department (ED) triage provides preliminary clinical assessment of a patient according to how severe their condition is, giving priority to the most ill and ensuring that their waiting time is as short as possible [1]. Triage aims to ensure accurate, consistent, and efficient emergency care when needed [2]. When performed properly, triage is an important component for effective ED care that can improve the outcome for patients. Healthcare environments change rapidly, and as such, healthcare providers have been challenged to find new ways of ensuring accurate triage [2]. The Danish healthcare system is currently undergoing reorganization. It aims to close smaller hospitals and to allocate patients into bigger and more specialized hospitals with larger EDs. This requires each ED to handle an increasing number of visits and makes patient prioritization more important than ever [1]. Triage systems are used in all Danish EDs [3]. A unified triage model has been required to be implemented across hospitals in the entire country [3]. To make this possible, EDs need a triage system that is both reliable and valid [4]. 0969-9546 Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
Results The distribution of age, comorbidity, admission, LOS, and mortality over triage categories differed as expected. After making adjustments for these differences, we found a consistent association between triage categories and in-hospital mortality, and 30, 60, and 90-day mortalities, the hospital LOS, and the admission rate. Conclusion RETTS-HEV was found to be closely related to all examined outcomes, and therefore useful in the risk stratification of ED patients. European Journal of Emergency Medicine 23:33–37 Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved. European Journal of Emergency Medicine 2016, 23:33–37 Keywords: emergency department, predictive validity, Rapid Emergency Triage and Treatment System, triage, validation Departments of aEmergency and bOccupational Medicine, Regional Hospital of Herning, Herning, Denmark Correspondence to Noel Pérez, MD, Department of Emergency, Regional Hospital of Herning, Gl. Landevej 61, Herning 7400, Denmark Tel: + 45 5193 0152; fax: + 45 7843 2061; e-mail: [email protected] Received 12 December 2013 Accepted 25 April 2014
The validity of a triage system refers to the degree in which the measured acuity level reflects the patient’s true acuity at the time of triage [5]. A triage system is considered valid if its results agree with its true value [6]. Validity can be determined either by comparing it with a golden standard (criterion validity) or by measuring how well it can predict the outcome of patients (predictive validity). A golden standard by definition has absolute accuracy [5], and as it is not possible to measure the truth for patient acuity, surrogate outcome markers have been used as criteria to assess validity [5]. As a result, predictive validity is the method that has been used in most triage validity studies [4,7–18]. The most important outcomes measured in these studies have been mortality, disposition, resource utilization, and the hospital length of stay (LOS). However, an important problem may arise if there are confounders for the association between triage and the outcomes, but this challenge can be overcome by performing multinominal logistic regression and thus making adjustments for possible confounders. The Rapid Emergency Triage and Treatment System – Hospital Unit West (RETTS-HEV) is a Swedish DOI: 10.1097/MEJ.0000000000000173
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
34 European Journal of Emergency Medicine 2016, Vol 23 No 1
designed triage system being used at the Regional Hospital West, Jutland, Denmark, since 2010 [19]. The inter-rater agreement of RETTS-HEV was presented on a previous study [20]. As the predictive validity of the RETTS-HEV triage system has never been evaluated in this particular setting, we undertook this study to determine the ability of the system to predict various outcomes such as the admission rate, the LOS, and mortality while controlling for likely confounders.
Materials and methods We performed an observational cohort study at the ED of the Regional Hospital in Herning, Denmark. The hospital provides healthcare to around 300 000 people and has an annual ED census of ∼ 29 000 patients. We examined the medical records of all patients who attended the ED from 1 September 2012 to 30 November 2012. The ED receives all patients except for cardiology and gynecological–obstetrical patients. Pediatric surgical patients and pediatric trauma patients are seen in the ED; however, pediatric medical patients are seen in the pediatric department. Severe trauma patients and all of those with life-threatening medical conditions are reported prearrival so the ED staff can activate the trauma alarm and they can be ready to assess and treat them upon arrival. RETTS-HEV has been designed to objectively assist the triage nurse in assigning a triage category to the patient. The process of triage is based on an algorithm for vital signs and 45 emergency symptoms and signs (ESS) algorithms for different chief complaints and symptoms. A triage level can be assigned to only one of the following five categories: I, Red (requires immediate attention); II, Orange (patient can wait up to 10 min); III, Yellow (can wait up to 60 min); IV, Green (can wait up to 120 min); V, Blue (can wait 240 min and can be treated by a trained ED nurse). Once a triage category is assigned according to vital signs, nurses select one or more ESS algorithms Table 1
on the basis of the patients’ chief complaint(s), which provide another triage score. In the case of several chief complaints, more than one ESS algorithm can be chosen. When both the triage score based on vital signs and the one determined by the ESS algorithm(s) are assigned, the highest priority score equals the final score. Triage is normally performed by a single nurse who has previously received training using the system. The training course takes less than a day, and is based on learning how to use the computerized system and getting familiarized with the flow charts, which have been designed to be intuitive and easy to understand (Table 1 and Fig. 1). The triage process is estimated to last no longer than 10 min [19]. We derived information on triage categories from the Electronic Patient Journal (EPJ) and from a handwritten list that is used by triaging nurses to supplement the EPJ. The admission rate and the hospital LOS outcomes were defined by data from the Danish National Patient Registry [21]. We recoded LOS according to the method used by Lim and Tongkumchum [22] by defining a prolonged LOS (higher than 7 days) as an outcome. Danish National Patient Registry also provided information about comorbidity, coded with the Charlson index [21]. Information on sex, age, and the date of death was obtained from the Danish Civil Registration System [23]. In-hospital mortality, and 30, 60, and 90-day mortalities were derived from the date of death, the admission date, and the discharge date. All data sources were merged using the patient’s personal identification number, which is the unique number that every person living in Denmark holds. We presented patient baseline information by describing it using either mean with SD, median with percentiles, or proportions. A multinominal logistic regression was performed to analyze the associations between triage and the outcomes: admission, LOS, in-hospital mortality, and 30, 60, and 90-day mortalities. We expected sex, age, and comorbidity to be possible confounders for in-hospital mortality, LOS, and admission rates, whereas LOS and
The triage algorithm for vital signs in the Rapid Emergency Triage and Treatment System – Hospital Unit West Red (Triage I) Time = 0 min Emergency team needed Continuous observation
Orange (Triage II) Time < 10 min Urgent Monitoring
A Airway B Breathing
Airway obstruction Inspiratory stridor POX < 80% without oxygen POX < 90% with oxygen or < 8
Potentially threatened airway
Free airway
Free airway
POX 90–95% without oxygen RR > 25
POX > 95% without oxygen RR 8–25 (normal)
C Circulation D Disability
HR > 140 SBP < 80 mmHg Unconscious GCS < 8 Cramps Body temperature < 32°C
POX < 90% without oxygen POX < 95% with oxygen RR > 30 HR > 120 or < 40 SBT < 90 mmHg Lethargic GCS: 9–13
HR > 110 or < 50
HR 50–100
Confusion GCS: 14
Alert GCS 15
Temperature > 40°C Temperature 32–34°C
Temperature > 38°C Temperature < 35°C
Temperature 35–38°C
E Exposure
Yellow (Triage III) Time < 60 min Less urgent Supervision
Green (Triage IV) Time < 120 min Not urgent Reassessed
GCS, Glasgow Coma Scale; HR, heart rate; POX, pulse oximetry; RR, respiratory rate.
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
Blue (Triage V) Time < 240 min
Predictive validity of RETTS-HEV Pérez et al. 35
Fig. 1
ESS Algorithm Back pain (code DM 54.9) Dorsal pain (code DM 54.6) Red Sudden and continuous pain on a vitally affected patient. Sudden dorsal pain with vegetative symptoms (sweat, nausea) or syncope.
ECG, bladder scan, fast urine test Orange Sudden neurological event Urine retention Fever > 38.5 and chills Sudden dorsal pain without vegetative symptoms.
ECG, bladder scan, fast urine test
Yellow Macroscopic hematuria Sudden debut < 12 h Prior history with pain
Fast urine test Green None of the above mentioned
Urine test Example of an ESS algorithm showing how back pain, as a chief complaint, is used to divide patients into different levels of triage. ESS, emergency symptoms and signs.
admission were also considered as possible confounders of 30, 60, and 90-day mortalities. To analyze LOS, we excluded discharged patients and those who died within their hospital stay. As the study did not alter or interfere with standard patient care and as it was considered a quality assurance project, the local ethics committee found the study to be
exempt from a formal ethics review. The data collection was approved by the Danish Data Protection Agency, and The Danish Board of Health approved the use of the registries.
Results A total of 5385 patients visited the ED during the study period. Seven hundred and five patients were excluded because they had no triage score registered, leaving 4680 patients included in the study. Table 2 describes the characteristics of the patients distributed over triage categories. The mean age of the patients was 47.1 years (SD 25.3). Ages ranged from 5 days to 103 years, and the majority of patients were 18 years or older (84.4%). The triage category with the highest average age corresponded to I, whereas those in category V had the youngest average age. 50.7% of patients were female. As expected, mortality occurred more frequently in triage categories I and II, where also more patients had a comorbidity score larger than zero. The total admission rate was 52% and the proportion of admitted patients increased according to the triage category (Table 2). LOS was also associated with the triage category. Triage categories were strongly associated with the outcomes admission, LOS, and in-hospital mortality with a consistent exposure–response relationship. Four patients died in the ED before being admitted and belonged to either triage category I (n = 3) or II (n = 1). When making adjustments for the possible confounders, sex, age, and the comorbidity score, the associations attenuated, primarily, the association with admission (Table 3). There was a clear association of the triage category and the 30, 60, and 90-day mortalities, as well as a clear exposure–response relationship. Adjustment for sex, age, the comorbidity score, LOS, and the admission rate strongly decreased the associations, as expected (Table 4).
Discussion This study showed that triage categories assigned by RETTS-HEV were closely related to a range of important outcomes such as admission, prolonged hospital LOS and in-hospital, 30, 60, and 90-day mortality rates, even after adjusting for potential confounders. Our results support the fact that RETTS-HEV has a high predictive validity in assigning a triage score to patients attending the ED. Others studies have demonstrated the ability of triage systems to predict patient outcomes. Widgren and Jourak [17] previously evaluated RETTS in Sweden and found an increased in-hospital mortality and hospital LOS that scaled according to priority. The scaling of triage assignment according to priority was comparable to ours, although in our study, mortality rates and LOS were lower. They only included adult in-hospital patients, which explains the higher mean age and mortality rates in
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
36 European Journal of Emergency Medicine 2016, Vol 23 No 1
Table 2
Characteristics of patients within different triage categories I (n = 202)
Triage value
Age (years): mean (SD) 53.2 Sex [n (%)] Female 94 Male 108 Comorbidity score [n (%)] 0 124 1 40 2+ 38 Mortality [n (%)] In-hospital 29 30-day 29 60-day 36 90-day 39 Hospital length of stay (days) Median (p25; p75) 1 Admission and discharge rate [n (%)] Admitted 173 Discharged 26 Died before admission 3
II (n = 671)
III (n = 1816)
IV (n = 1284)
V (n = 707)
Total (n = 4680)
(25.1)
52.1 (24.6)
51.6 (24.6)
47.3 (25.3)
29.1 (18.7)
47.14 (25.3)
(46.5) (53.5)
334 (49.8) 337 (50.2)
899 (49.5) 917 (50.5)
691 (53.8) 593 (46.2)
351 (49.7) 356 (50.4)
2369 (50.6) 2311 (49.4)
(61.4) (19.8) (18.8)
415 (61.9) 104 (15.5) 152 (22.6)
1262 (69.5) 265 (14.6) 289 (15.9)
972 (75.7) 142 (11.1) 170 (13.2)
643 (91) 45 (6.4) 19 (2.6)
3416 (73) 596 (12.7) 668 (14.3)
(14.4) (14.4) (17.8) (19.3)
20 35 44 50
(0;6)
(3.0) (5.2) (6.6) (7.5)
23 43 71 93
1 (0;4)
(85.6) (12.9) (1.5)
(1.3) (2.3) (3.9) (5.1)
9 22 34 39
1 (0;2)
502 (74.8) 168 (25) 1 (0.2)
(0.7) (1.7) (2.7) (3)
0 1 1 1
0 (0;1)
1131 (62.3) 685 (37.7) 0 (0.0)
(0) (0.1) (0.1) (0.1)
81 130 186 222
0 (0;0)
608 (47.4) 676 (52.7) 0 (0.0)
21 (3.0) 686 (97.0) 0 (0.0)
(1.73) (2.8) (4) (4.7)
1 (0;2) 2435 (52) 2241 (47.9) 4 (0.1)
Association between the triage category and the admission rate, a prolonged length of stay (more than 7 days) and in-hospital mortality
Table 3
LOS > 7 days [OR (95% CI)]
Admission [OR (95% CI)]) Triage category I II III IV V Test for trend (P)
Adjustedb
Crude 217.4 97.61 53.9 29.4
(119.4; 395.5) (61.1; 155.9) (34.5; 84.1) (18.8; 45.9) 1 < 0.001
148.6 63.3 33.9 19.6
Adjustedb
Crude
(80.7; 273.7) (39.4;101.8) (21.6; 53.1) (12.5; 30.9) 1 < 0.001
6.0 3.0 2.4 2.4
In-hospital mortality [OR (95% CI)]a
(0.8; 46.5) (0.4; 23.1) (0.3; 17.8) (0.3; 17.8) 1 < 0.001
5.5 2.3 1.8 1.7
(0.7; 45.3) (0.3; 18.4) (0.2; 14.0) (0.2; 13.8) 1 < 0.001
Crude
Adjustedb
23.8 (11.1; 51.0) 4.4 (2.0; 9.6) 1.9 (0.8; 3.9) 1 Not measured < 0.001
23.8 (10.6; 53.5) 3.4 (1.5; 7.7) 1.6 (0.7; 3.4) 1 Not measured < 0.001
a
In-hospital mortality was associated with triage category 4 instead of 5 because there were no patients who died in the hospital having been triaged as 5. Adjusted for sex, age, and the comorbidity score. CI, confidence interval; LOS; length of stay; OR, odds ratio. b
Table 4
Association between the triage category and 30, 60, and 90-day mortalities 30-day mortality [OR (95% CI)]
Triage category I II III IV V Test for trend (P)
Crude 118.3 38.9 17.1 12.3
(16.0; 874.8) (5.3; 284.4) (2.4; 124.6) (1.7; 91.5) 1 < 0.001
Adjusted 11.8 3.5 1.8 1.5
60-day mortality [OR (95% CI)] a
(1.5; 96.0) (0.4; 27.6) (0.2; 13.9) (0.2; 12.3) 1 < 0.001
Crude 153.1 49.5 28.7 19.2
Adjusted
(20.8; 1124.8) (6.8; 360.6) (4.0; 207.1) (2.6; 140.6) 1 < 0.001
12.6 3.7 2.6 2.1
90-day mortality [OR (95% CI)] a
(1.6; 100.8) (0.5; 28.8) (0.3; 19.8) (0.3; 16.6) 1 < 0.001
Crude 168.9 56.8 38.1 22.1
(23.0; 1238.5) (7.8; 412.7) (5.3; 273.9) (3.0; 161.3) 1 < 0.001
Adjusteda 15.3 4.5 3.7 2.6
(1.9; 121.8) (0.6; 35.0) (0.5; 28.6) (0.3; 20.4) 1
The predictive validity of RETTS-HEV as an acuity triage tool in the emergency department of a Danish Regional Hospital Noel Péreza, Louise Nissena, Rasmus F. Nielsena, Poul Petersena and Karin Bieringb Introduction The Rapid Emergency Triage and Treatment System – Hospital Unit West (RETTS-HEV) is a triage system used in the emergency department (ED) in Herning, Denmark, since 2010. It categorizes patients according to priority and defines a time limit on how long patients can wait before being seen by a doctor depending on the severity of their condition. The purpose of this study was to determine the predictive validity of RETTS-HEV by measuring the association between triage scores and outcomes such as the admission rate, the length of stay (LOS), and mortality. Materials and methods We performed an observational cohort study by examining the medical records of all patients who attended the ED from 1 September 2012 to 30 November 2012, at the Regional Hospital West Jutland in Herning, Denmark (N = 4680). We defined the following outcomes to make associations with the patients’ triage category: in-hospital mortality, and 30, 60, and 90-day mortalities, the hospital LOS and the admission rate, on the basis of complete information from the Danish National Patient Registry.
Introduction Emergency department (ED) triage provides preliminary clinical assessment of a patient according to how severe their condition is, giving priority to the most ill and ensuring that their waiting time is as short as possible [1]. Triage aims to ensure accurate, consistent, and efficient emergency care when needed [2]. When performed properly, triage is an important component for effective ED care that can improve the outcome for patients. Healthcare environments change rapidly, and as such, healthcare providers have been challenged to find new ways of ensuring accurate triage [2]. The Danish healthcare system is currently undergoing reorganization. It aims to close smaller hospitals and to allocate patients into bigger and more specialized hospitals with larger EDs. This requires each ED to handle an increasing number of visits and makes patient prioritization more important than ever [1]. Triage systems are used in all Danish EDs [3]. A unified triage model has been required to be implemented across hospitals in the entire country [3]. To make this possible, EDs need a triage system that is both reliable and valid [4]. 0969-9546 Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
Results The distribution of age, comorbidity, admission, LOS, and mortality over triage categories differed as expected. After making adjustments for these differences, we found a consistent association between triage categories and in-hospital mortality, and 30, 60, and 90-day mortalities, the hospital LOS, and the admission rate. Conclusion RETTS-HEV was found to be closely related to all examined outcomes, and therefore useful in the risk stratification of ED patients. European Journal of Emergency Medicine 23:33–37 Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved. European Journal of Emergency Medicine 2016, 23:33–37 Keywords: emergency department, predictive validity, Rapid Emergency Triage and Treatment System, triage, validation Departments of aEmergency and bOccupational Medicine, Regional Hospital of Herning, Herning, Denmark Correspondence to Noel Pérez, MD, Department of Emergency, Regional Hospital of Herning, Gl. Landevej 61, Herning 7400, Denmark Tel: + 45 5193 0152; fax: + 45 7843 2061; e-mail: [email protected] Received 12 December 2013 Accepted 25 April 2014
The validity of a triage system refers to the degree in which the measured acuity level reflects the patient’s true acuity at the time of triage [5]. A triage system is considered valid if its results agree with its true value [6]. Validity can be determined either by comparing it with a golden standard (criterion validity) or by measuring how well it can predict the outcome of patients (predictive validity). A golden standard by definition has absolute accuracy [5], and as it is not possible to measure the truth for patient acuity, surrogate outcome markers have been used as criteria to assess validity [5]. As a result, predictive validity is the method that has been used in most triage validity studies [4,7–18]. The most important outcomes measured in these studies have been mortality, disposition, resource utilization, and the hospital length of stay (LOS). However, an important problem may arise if there are confounders for the association between triage and the outcomes, but this challenge can be overcome by performing multinominal logistic regression and thus making adjustments for possible confounders. The Rapid Emergency Triage and Treatment System – Hospital Unit West (RETTS-HEV) is a Swedish DOI: 10.1097/MEJ.0000000000000173
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
34 European Journal of Emergency Medicine 2016, Vol 23 No 1
designed triage system being used at the Regional Hospital West, Jutland, Denmark, since 2010 [19]. The inter-rater agreement of RETTS-HEV was presented on a previous study [20]. As the predictive validity of the RETTS-HEV triage system has never been evaluated in this particular setting, we undertook this study to determine the ability of the system to predict various outcomes such as the admission rate, the LOS, and mortality while controlling for likely confounders.
Materials and methods We performed an observational cohort study at the ED of the Regional Hospital in Herning, Denmark. The hospital provides healthcare to around 300 000 people and has an annual ED census of ∼ 29 000 patients. We examined the medical records of all patients who attended the ED from 1 September 2012 to 30 November 2012. The ED receives all patients except for cardiology and gynecological–obstetrical patients. Pediatric surgical patients and pediatric trauma patients are seen in the ED; however, pediatric medical patients are seen in the pediatric department. Severe trauma patients and all of those with life-threatening medical conditions are reported prearrival so the ED staff can activate the trauma alarm and they can be ready to assess and treat them upon arrival. RETTS-HEV has been designed to objectively assist the triage nurse in assigning a triage category to the patient. The process of triage is based on an algorithm for vital signs and 45 emergency symptoms and signs (ESS) algorithms for different chief complaints and symptoms. A triage level can be assigned to only one of the following five categories: I, Red (requires immediate attention); II, Orange (patient can wait up to 10 min); III, Yellow (can wait up to 60 min); IV, Green (can wait up to 120 min); V, Blue (can wait 240 min and can be treated by a trained ED nurse). Once a triage category is assigned according to vital signs, nurses select one or more ESS algorithms Table 1
on the basis of the patients’ chief complaint(s), which provide another triage score. In the case of several chief complaints, more than one ESS algorithm can be chosen. When both the triage score based on vital signs and the one determined by the ESS algorithm(s) are assigned, the highest priority score equals the final score. Triage is normally performed by a single nurse who has previously received training using the system. The training course takes less than a day, and is based on learning how to use the computerized system and getting familiarized with the flow charts, which have been designed to be intuitive and easy to understand (Table 1 and Fig. 1). The triage process is estimated to last no longer than 10 min [19]. We derived information on triage categories from the Electronic Patient Journal (EPJ) and from a handwritten list that is used by triaging nurses to supplement the EPJ. The admission rate and the hospital LOS outcomes were defined by data from the Danish National Patient Registry [21]. We recoded LOS according to the method used by Lim and Tongkumchum [22] by defining a prolonged LOS (higher than 7 days) as an outcome. Danish National Patient Registry also provided information about comorbidity, coded with the Charlson index [21]. Information on sex, age, and the date of death was obtained from the Danish Civil Registration System [23]. In-hospital mortality, and 30, 60, and 90-day mortalities were derived from the date of death, the admission date, and the discharge date. All data sources were merged using the patient’s personal identification number, which is the unique number that every person living in Denmark holds. We presented patient baseline information by describing it using either mean with SD, median with percentiles, or proportions. A multinominal logistic regression was performed to analyze the associations between triage and the outcomes: admission, LOS, in-hospital mortality, and 30, 60, and 90-day mortalities. We expected sex, age, and comorbidity to be possible confounders for in-hospital mortality, LOS, and admission rates, whereas LOS and
The triage algorithm for vital signs in the Rapid Emergency Triage and Treatment System – Hospital Unit West Red (Triage I) Time = 0 min Emergency team needed Continuous observation
Orange (Triage II) Time < 10 min Urgent Monitoring
A Airway B Breathing
Airway obstruction Inspiratory stridor POX < 80% without oxygen POX < 90% with oxygen or < 8
Potentially threatened airway
Free airway
Free airway
POX 90–95% without oxygen RR > 25
POX > 95% without oxygen RR 8–25 (normal)
C Circulation D Disability
HR > 140 SBP < 80 mmHg Unconscious GCS < 8 Cramps Body temperature < 32°C
POX < 90% without oxygen POX < 95% with oxygen RR > 30 HR > 120 or < 40 SBT < 90 mmHg Lethargic GCS: 9–13
HR > 110 or < 50
HR 50–100
Confusion GCS: 14
Alert GCS 15
Temperature > 40°C Temperature 32–34°C
Temperature > 38°C Temperature < 35°C
Temperature 35–38°C
E Exposure
Yellow (Triage III) Time < 60 min Less urgent Supervision
Green (Triage IV) Time < 120 min Not urgent Reassessed
GCS, Glasgow Coma Scale; HR, heart rate; POX, pulse oximetry; RR, respiratory rate.
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
Blue (Triage V) Time < 240 min
Predictive validity of RETTS-HEV Pérez et al. 35
Fig. 1
ESS Algorithm Back pain (code DM 54.9) Dorsal pain (code DM 54.6) Red Sudden and continuous pain on a vitally affected patient. Sudden dorsal pain with vegetative symptoms (sweat, nausea) or syncope.
ECG, bladder scan, fast urine test Orange Sudden neurological event Urine retention Fever > 38.5 and chills Sudden dorsal pain without vegetative symptoms.
ECG, bladder scan, fast urine test
Yellow Macroscopic hematuria Sudden debut < 12 h Prior history with pain
Fast urine test Green None of the above mentioned
Urine test Example of an ESS algorithm showing how back pain, as a chief complaint, is used to divide patients into different levels of triage. ESS, emergency symptoms and signs.
admission were also considered as possible confounders of 30, 60, and 90-day mortalities. To analyze LOS, we excluded discharged patients and those who died within their hospital stay. As the study did not alter or interfere with standard patient care and as it was considered a quality assurance project, the local ethics committee found the study to be
exempt from a formal ethics review. The data collection was approved by the Danish Data Protection Agency, and The Danish Board of Health approved the use of the registries.
Results A total of 5385 patients visited the ED during the study period. Seven hundred and five patients were excluded because they had no triage score registered, leaving 4680 patients included in the study. Table 2 describes the characteristics of the patients distributed over triage categories. The mean age of the patients was 47.1 years (SD 25.3). Ages ranged from 5 days to 103 years, and the majority of patients were 18 years or older (84.4%). The triage category with the highest average age corresponded to I, whereas those in category V had the youngest average age. 50.7% of patients were female. As expected, mortality occurred more frequently in triage categories I and II, where also more patients had a comorbidity score larger than zero. The total admission rate was 52% and the proportion of admitted patients increased according to the triage category (Table 2). LOS was also associated with the triage category. Triage categories were strongly associated with the outcomes admission, LOS, and in-hospital mortality with a consistent exposure–response relationship. Four patients died in the ED before being admitted and belonged to either triage category I (n = 3) or II (n = 1). When making adjustments for the possible confounders, sex, age, and the comorbidity score, the associations attenuated, primarily, the association with admission (Table 3). There was a clear association of the triage category and the 30, 60, and 90-day mortalities, as well as a clear exposure–response relationship. Adjustment for sex, age, the comorbidity score, LOS, and the admission rate strongly decreased the associations, as expected (Table 4).
Discussion This study showed that triage categories assigned by RETTS-HEV were closely related to a range of important outcomes such as admission, prolonged hospital LOS and in-hospital, 30, 60, and 90-day mortality rates, even after adjusting for potential confounders. Our results support the fact that RETTS-HEV has a high predictive validity in assigning a triage score to patients attending the ED. Others studies have demonstrated the ability of triage systems to predict patient outcomes. Widgren and Jourak [17] previously evaluated RETTS in Sweden and found an increased in-hospital mortality and hospital LOS that scaled according to priority. The scaling of triage assignment according to priority was comparable to ours, although in our study, mortality rates and LOS were lower. They only included adult in-hospital patients, which explains the higher mean age and mortality rates in
Copyright r 2016 Wolters Kluwer Health, Inc. All rights reserved.
36 European Journal of Emergency Medicine 2016, Vol 23 No 1
Table 2
Characteristics of patients within different triage categories I (n = 202)
Triage value
Age (years): mean (SD) 53.2 Sex [n (%)] Female 94 Male 108 Comorbidity score [n (%)] 0 124 1 40 2+ 38 Mortality [n (%)] In-hospital 29 30-day 29 60-day 36 90-day 39 Hospital length of stay (days) Median (p25; p75) 1 Admission and discharge rate [n (%)] Admitted 173 Discharged 26 Died before admission 3
II (n = 671)
III (n = 1816)
IV (n = 1284)
V (n = 707)
Total (n = 4680)
(25.1)
52.1 (24.6)
51.6 (24.6)
47.3 (25.3)
29.1 (18.7)
47.14 (25.3)
(46.5) (53.5)
334 (49.8) 337 (50.2)
899 (49.5) 917 (50.5)
691 (53.8) 593 (46.2)
351 (49.7) 356 (50.4)
2369 (50.6) 2311 (49.4)
(61.4) (19.8) (18.8)
415 (61.9) 104 (15.5) 152 (22.6)
1262 (69.5) 265 (14.6) 289 (15.9)
972 (75.7) 142 (11.1) 170 (13.2)
643 (91) 45 (6.4) 19 (2.6)
3416 (73) 596 (12.7) 668 (14.3)
(14.4) (14.4) (17.8) (19.3)
20 35 44 50
(0;6)
(3.0) (5.2) (6.6) (7.5)
23 43 71 93
1 (0;4)
(85.6) (12.9) (1.5)
(1.3) (2.3) (3.9) (5.1)
9 22 34 39
1 (0;2)
502 (74.8) 168 (25) 1 (0.2)
(0.7) (1.7) (2.7) (3)
0 1 1 1
0 (0;1)
1131 (62.3) 685 (37.7) 0 (0.0)
(0) (0.1) (0.1) (0.1)
81 130 186 222
0 (0;0)
608 (47.4) 676 (52.7) 0 (0.0)
21 (3.0) 686 (97.0) 0 (0.0)
(1.73) (2.8) (4) (4.7)
1 (0;2) 2435 (52) 2241 (47.9) 4 (0.1)
Association between the triage category and the admission rate, a prolonged length of stay (more than 7 days) and in-hospital mortality
Table 3
LOS > 7 days [OR (95% CI)]
Admission [OR (95% CI)]) Triage category I II III IV V Test for trend (P)
Adjustedb
Crude 217.4 97.61 53.9 29.4
(119.4; 395.5) (61.1; 155.9) (34.5; 84.1) (18.8; 45.9) 1 < 0.001
148.6 63.3 33.9 19.6
Adjustedb
Crude
(80.7; 273.7) (39.4;101.8) (21.6; 53.1) (12.5; 30.9) 1 < 0.001
6.0 3.0 2.4 2.4
In-hospital mortality [OR (95% CI)]a
(0.8; 46.5) (0.4; 23.1) (0.3; 17.8) (0.3; 17.8) 1 < 0.001
5.5 2.3 1.8 1.7
(0.7; 45.3) (0.3; 18.4) (0.2; 14.0) (0.2; 13.8) 1 < 0.001
Crude
Adjustedb
23.8 (11.1; 51.0) 4.4 (2.0; 9.6) 1.9 (0.8; 3.9) 1 Not measured < 0.001
23.8 (10.6; 53.5) 3.4 (1.5; 7.7) 1.6 (0.7; 3.4) 1 Not measured < 0.001
a
In-hospital mortality was associated with triage category 4 instead of 5 because there were no patients who died in the hospital having been triaged as 5. Adjusted for sex, age, and the comorbidity score. CI, confidence interval; LOS; length of stay; OR, odds ratio. b
Table 4
Association between the triage category and 30, 60, and 90-day mortalities 30-day mortality [OR (95% CI)]
Triage category I II III IV V Test for trend (P)
Crude 118.3 38.9 17.1 12.3
(16.0; 874.8) (5.3; 284.4) (2.4; 124.6) (1.7; 91.5) 1 < 0.001
Adjusted 11.8 3.5 1.8 1.5
60-day mortality [OR (95% CI)] a
(1.5; 96.0) (0.4; 27.6) (0.2; 13.9) (0.2; 12.3) 1 < 0.001
Crude 153.1 49.5 28.7 19.2
Adjusted
(20.8; 1124.8) (6.8; 360.6) (4.0; 207.1) (2.6; 140.6) 1 < 0.001
12.6 3.7 2.6 2.1
90-day mortality [OR (95% CI)] a
(1.6; 100.8) (0.5; 28.8) (0.3; 19.8) (0.3; 16.6) 1 < 0.001
Crude 168.9 56.8 38.1 22.1
(23.0; 1238.5) (7.8; 412.7) (5.3; 273.9) (3.0; 161.3) 1 < 0.001
Adjusteda 15.3 4.5 3.7 2.6
(1.9; 121.8) (0.6; 35.0) (0.5; 28.6) (0.3; 20.4) 1
