System building in emergency medicine has been identified as one of the most important steps to ensure high-quality emergency care. A triage system is a crucial process in the emergency department (ED) which can help identification of serious patients and has a huge bearing on patient survival, morbidity, and resource utilization. Effective triage systems reduce death rates and provide efficiency in emergency care. In low- and middle-income countries (LMICs) most EDs have an informal screening process for emergency patients manned by untrained junior nurses and/or junior doctors. As the disease severity, spectrum of diseases, lack of trained emergency personnel, and resources are very different in these countries, there is an increased need for a specific scientific triage system for the emergency patients presenting to the ED in LMICs. The five-level triage system is not an option in the developing world mainly due to complexity related to different numbers, colors and severity which makes it confusing and also challenging for untrained and majority uneducated hospital support staff. Hence, a simplified three-tier triage system, the All India Institute of Medical Sciences (AIIMS) Triage Protocol or ATP for adult patients was developed by faculty and residents of our department using Delphi method [Supplementary Table 1]. The ATP is based on both physiological and clinical parameters to decide on the acuity of the patient's condition. The comparison of our three-tiered ATP with internationally available five-tier triage systems has been elaborated in our previous publication. On such comparison, the Emergency Severity Index category 1 and 2 are equal to red category on ATP. Category 3 and 4 are the same as the yellow category of patients on ATP and category 5 is green triaged patients as per ATP. Similarly, other triage scales such as Canadian Triage Scale, Manchester Triage Scale have also been compared to our three-tier system.
Validity refers to the accuracy with which the triage scale determines the true medical/surgical acuity of the patient. Broadly, there are two kinds of validity. Criterion validity refers to testing of the scale against external criterion like objective criteria developed by expert consensus or existing standardized triage system etc., Construct validity refers to testing of the scale against severity-related variables such as mortality, intensive care unit (ICU)/ward admissions, and length of stay in ED. Construct validity has been used more often for the purpose of triage scale assessment.
Although an important element of emergency care, triage research has not been a priority in LMICs and hence very few triage systems have been validated. In this effort, we attempted to prospectively follow-up emergency patients triaged either red or yellow and validate (construct validity) our triage system (ATP) in terms of mortality and ICU/hospital admission at 24 h. This is the first large prospective validation of an emergency triage system in India.
Ours is a tertiary care hospital located in the capital city of India, New Delhi. The academic program of its Emergency Medicine Department started in 2012. Our ED caters to more than 200,000 patients every year. The triage area is manned by one or two resident doctors, two nurses, four health assistants, and four security guards, and the final decision to triage lies with the resident doctor. All patients 14 years and above who presented to the triage area of the ED and were triaged red or yellow were included in this prospective observational study. The ATP was used to triage patients into red (very serious) or yellow (moderately serious) and patients were included in the study after obtaining consent. Green (least serious) patients were excluded from the study. Patients matching the red criteria as per the ATP were labeled red and moderately serious patients were labeled yellow. The average waiting time for yellow patients before assessment in the treatment area is approximately 150 min. The investigators (SKS, BG, GT) who were residents on shift duties working in ED, covered approximately 24 shifts (equally divided morning, afternoon, and night) each month. At the beginning of each shift, we recruited the first thirty patients (triaged red or yellow) into the study from the triage area of the ED over a 22-month (January 2019-October 2020) period. The investigators were not part of the triage team and recruited patients after they were triaged by the triage team. A total of 15,505 patients who consented were prospectively recruited in this study. Standardized care was provided to both red and yellow triaged patients as per the institutional protocols.
All included patients (triaged red or yellow) were followed up at 24 h and their outcome documented on a standardized data collection form. The outcome was noted by physical verification or collected from the institutional electronic database. Mortality and ICU or ward admission was noted at 24 h. In case the patient was discharged or shifted to another hospital, telephonic follow-up was done.
All the information was collected and collated in Microsoft Excel spreadsheet (MS Office-365). Counts and percentages were used to summarize categorical data. Categorical information between two or more groups were compared by using Chi-square test. Mean and standard deviation were used to summarize normally distributed data, whereas median, range, and interquartile range (IQR) were used to summarize nonnormal continuous data. Diagnostic statistics of ATP for outcome measures like disposition (mortality, overall admissions, and ICU admissions) at 24-h were calculated in terms of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and positive and negative likelihood ratios. (PLR and NLR) Logistic regression was utilized for examining the predictive ability of ATP for the above-mentioned outcome measures. All the above analyses were performed, and graphs were prepared with IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp. All tests of significance used a two-sided P ≤ 0.05.
Study flow and patient characteristics
During the study, a total of 15,505 patients were recruited at triage. Of these, 13,754 patients (8335 males, 5419 females) were followed up at 24-h, after excluding patients who left against medical advice, absconded, and were lost to follow-up because they were either not found on system or did not respond to the telephonic follow-up [Figure 1]. Among 13,754 patients, 6303 (45.83%) were triaged red and 7451 (54.17%) were triaged yellow. The median age of patients triaged red was 44 years (IQR: 30–60) and that of patients triaged yellow was 40 years (IQR: 27–55).
Disposition at 24-h
At 24-h of triage, a total of 1289 (9.37%) patients got admitted to ICU, 6314 (45.91%) were admitted in the wards, 5475 (38.81%) got discharged and 676 (4.91%) died. Disposition by triage categories is tabulated in Table 1. Mortality at 24 h was 10.31% (650) in red triaged patients and 0.35% (26) in yellow triaged patients. Among yellow patients, mortality and ICU admission at 24 h was seen in 26 (0.35%) and 19 (0.26%) respectively. The 24-h mortality of red triaged patients (10.31%) was significantly higher (P < 0.001) than that of yellow triaged patients 0.35%.
Diagnostic statistics of ATP red criteria for disposition at 24-h
The presence of one or more ATP “Red” criteria was 96.2% (95% CI: 94.42%–97.47%) sensitive and 56.8% (95% CI: 55.92%–57.63%) specific in predicting 24-h mortality. The PPV, NPV, PLR and NLR for 24-h mortality prediction were 10.3%, 99.7%, 2.22, and 0.07, respectively. Similarly, the sensitivity and specificity of ATP 'Red' criteria for 24-h ICU admission were 98.5% (95%CI: 97.7%–99.1%) and 59.6% (95%CI: 58.8%–60.5%), respectively. Detailed data are presented in Table 2.
Predictive ability of ATP red criteria for disposition at 24-h
The presence of one or more ATP Red criteria was able to predict the 24-h mortality significantly, with an odds ratio of 32.84 (95% CI: 22.15–48.67). ATP Red criteria predicted the 24-h ICU admissions with an odds ratio of 98.7 (95% CI: 62.66–155.48) and 24-h overall admission with an odds ratio of 5.03 (95% CI: 4.66–5.43) detailed information is presented in Table 2.
According to results of our study amongst adult nontrauma patients presenting to the ED, the 24-h mortality for red triaged patients was significantly higher than the yellow patients, 10.31% and 0.35% respectively, and was statistically significant. Similarly, 24-h ICU admission was significantly higher among red patients. The sensitivity and NPV of ATP for both 24-h mortality and 24-h ICU admission was more than 95%. The ATP triage system could predict 24-h mortality and ICU admission significantly with an odds ratio of 32.84 and 98.7, respectively. This high sensitivity of ATP in predicting outcome in red triaged patients suggests that ATP is highly accurate in identifying critically ill patients in our setting. The admission in ward for red patients at 24 h was only 60.14%. This was because of lack of immediate availability of beds for emergency patients and hence most patients had to wait long before they could get admitted. Hence, ward admission was not a great reference point for testing the triage scale in our setting.
In published literature, there are wide variations in assessing triage system validation. Both construct validity and criterion validity are used in validation studies to assess the triage tool. There are very few validated triage systems from LMICs. One such scale, the South African Triage Scale was tested and validated using a criterion validity with standardized triage algorithm developed by experts in emergency medicine. Another triage scale, one-two-triage developed at Stanford University in the USA for utility in Cambodia, also validated their triage system where physician decision to triage was used as the gold standard. Highlighting the lack of standardized objective criteria (existing triage systems) and presence of variability in care, Twomey et al. suggested Delphi approach (a variant of construct validity) for triage testing. Although criterion validity remains the preferred validation model in validating triage systems, construct validity is most commonly used in the validation of triage scales around the globe. We did not have a standard reference of urgency or an existing triage scale in the region and hence criterion validity was not feasible in our setting. Construct validity may be limited by bias, as causes of mortality may be many and possible variability in care. At our academic center, however, the initial care for all emergency patients is standardized hence minimizing bias due to variability in care. We proceeded to test the validity of our screening tool with the reference standard as 24-h mortality and ICU/ward admission as this was most feasible in our setting.
The study validated the performance of ATP in our high-volume public hospital setting. More than fifteen thousand patients were prospectively recruited over 22 months which included patients from all shifts (morning, evening, and night) presenting to the ED throughout the year including all seasons. The patients were followed up 24 h after they were recruited to record their outcome. The authors believe that ATP will have a similar performance in high-volume public hospitals in the region. However, further external validation will be required for its performance in different clinical settings such as private hospitals, low-volume EDs, and rural settings where the clientele and clinical presentations in the ED might be different.
Triage needs for LMICs are different from the triage needs of high-income countries largely due to a wider spectrum of clinical presentations encountered in ED, lack of trained personnel, high volume of sick patients, and lack of adequate resources. Although poorly specific, the three-tier ATP triage system has a high sensitivity to identify sick patients which makes it an optimal triage in settings of low resource and limited trained staff. This scientific evidence will assist the policymakers to advocate for a common triage system throughout the country. The language of red, yellow, and green will need to be quickly and widely disseminated among all cadres of healthcare personnel including personnel of emergency medical services. The same strategy can help policymakers to stratify hospitals based on red, yellow, and green capacity and capability.
Further validation of triage with expert consensus panel or comparison with another triage system as surrogate for urgency may be possible in the future studies. Another rigorous validation methodology for triage system is identifying under and over triaging. This could be done by an expert panel who assigns the urgency level right after triage by the triage team or through simulated clinical scenarios. Further validity testing using criterion validity and inter-rater agreement will also be needed to test the triage scale. It is likely that in times to come, as systems in emergency care improve and the expertise in emergency medicine grows, AIIMS Triage Protocol will evolve.
Triage is a continuous and dynamic process. We could not gather data on re-triage due to limitations in data collection systems and challenges in data collection. The reference point for testing validity would be better if decision of admission to ICU was taken into account as the availability of ICU bed is a bottleneck in our system. We had to resort to this to simplify the process of data collection in a very high volume ED.
ATP when applied to adult non-trauma patients had a high accuracy in recognizing sick patients presenting to the ED. A time-tested and validated triage system like ATP may be a good starting point for public hospital EDs across LMICs. External validation studies will help create more scientific evidence for ATP. Testing the ATP or a modified ATP system among pediatric and trauma patients should be next steps.
Research quality and ethics statement
This study was approved by the Institutional Review Board / Ethics Committee ( IRB IECPG_528/14.11.2018, RT_17/19.12.2018. The authors followed applicable EQUATOR Network (https://www.equator-network.org/) guidelines during the conduct of this research project.
Financial support and sponsorship
Conflicts of interest
We certify that Authors– Praveen Aggarwal, Sanjeev Bhoi and L R Murmu are members of the Editorial Board of the Journal of Emergencies, Trauma, and Shock.
1. Moresky RT, Razzak J, Reynolds T, Wallis LA, Wachira BW, Nyirenda M, et al Advancing research on emergency care systems in low-income and middle-income countries: Ensuring high-quality care delivery systems BMJ Glob Health. 2019;4:e001265
2. Emergency Care Systems for Universal Health Coverage: Ensuring Timely Care for Acutely Ill and Injured: Report by the WHO Director General to the 72nd World Health Assembly. 2019
3. Dekker-Boersema J, Hector J, Jefferys LF, Binamo C, Camilo D, Muganga G, et al Triage conducted by lay-staff and emergency training reduces paediatric mortality in the emergency department of a rural hospital in Northern Mozambique Afr J Emerg Med. 2019;9:172–6
4. Sahu AK, Bhoi S, Galwankar S, Aggarwal P, Murmu L, Nayer J, et al All India Institute of Medical Sciences Triage Protocol (ATP): ATP of a busy emergency department J Emerg Trauma Shock. 2020;13:107–9
5. Streiner DL, Norman GR, Cairney J. Health Measurement Scales: A Practical Guide to their Development and Use Oxford University Press. 2015 Print ISBN-13: 9780199685219
6. Fan J, Darrab AA, Eva K, Fernandes CM. Triage scales in the emergency department: A systematic review Ann Emerg Med. 2005;46:41
7. Razzak JA, Kellermann AL. Emergency medical care in developing countries: Is it worthwhile? Bull World Health Organ. 2002;80:900–5
8. Twomey M, Wallis LA, Thompson ML, Myers JE. The South African Triage Scale (adult version) provides reliable acuity ratings Int Emerg Nurs. 2012;20:142–50
9. Kuriyama A, Urushidani S, Nakayama T. Five-level emergency triage systems: Variation in assessment of validity Emerg Med J. 2017;34:703–10
10. Twomey M, Wallis LA, Myers JE. Limitations in validating emergency department triage scales Emerg Med J. 2007;24:477–9
11. Khan A, Mahadevan SV, Dreyfuss A, Quinn J, Woods J, Somontha K, et al One-two-triage: Validation
and reliability of a novel triage system for low-resource settings Emerg Med J. 2016;33:709–15