Emergency centre triage category allocations and their associated patient flow timeframes in a private healthcare group in the Middle East

Abstract Aims and Objectives To identify and describe triage category allocations and their associated patient pathway timeframes in four emergency centres of a large private healthcare group in the United Arab Emirates. Background The classification of patients in accordance with their acuity level is a complex task that requires quick and accurate allocation. Triage system categories have predetermined timeframes in which patients should be seen by a physician or treatment initiated for the best possible outcome. Design and Methods An observational, cross‐sectional study was conducted through the prospective capture and evaluation of medical records from patients triaged in each of the four emergency centres (two hospitals and two clinics) over a period of a month. The STROBE statement was used as a reporting framework. Descriptive statistics were used to determine the timeframes associated with the patient pathway through each EC and contrasted against their allocated triage category. Results A total of 4,432 patient records were eligible for analysis from the four emergency centres. Triage category 4 (54.7%) was allocated the most with only a single category 1 patient seen between the four emergency centres. The median time from registration to triage was <10 min and triage to physician consult was <25 min. The overall length of stay of high‐acuity cases was between 1 hr 13 min–2 hr 44 min, compared with low‐acuity cases being 32–49 min. Overall time to physician was substantially lower than the targets set by the triage systems itself.

through the prospective capture and evaluation of medical records from patients triaged in each of the four emergency centres (two hospitals and two clinics) over a period of a month. The STROBE statement was used as a reporting framework.
Descriptive statistics were used to determine the timeframes associated with the patient pathway through each EC and contrasted against their allocated triage category.
Results: A total of 4,432 patient records were eligible for analysis from the four emergency centres. Triage category 4 (54.7%) was allocated the most with only a single category 1 patient seen between the four emergency centres. The median time from registration to triage was <10 min and triage to physician consult was <25 min. The overall length of stay of high-acuity cases was between 1 hr 13 min-2 hr 44 min, compared with low-acuity cases being 32-49 min. Overall time to physician was substantially lower than the targets set by the triage systems itself.

K E Y W O R D S
emergency department, Middle East, nurses, time, triage 3 | ME THODS An observational, cross-sectional study was conducted through the prospective capture and evaluation of patient medical records from the four ECs (two hospitals and two clinics) of the private hospital group in the Emirate of Dubai. The STROBE statement checklist (See File S1) for reporting observational studies was used as a framework (von Elm et al., 2008 Medical records from patients triaged in each of the four ECs over a period of 1 month were evaluated and considered for inclusion. Electronic and manual platforms were used to collect the required data. The initial electronic data were sourced through the hospital groups' medical records department at the end of the 1-month period. EC staff were instructed to specifically include electronic data fields that was set out for this study, in addition to the usual EC data they capture. The electronically captured data from the four ECs during the month were collated and provided in a single Microsoft Excel (2016) spreadsheet. This included triage category allocations and patient flow timeframes (i.e., registration -> triage -> physician -> discharge). It was necessary to capture manual data that were not contained in the hospitals electronic information system. The manual data were captured by the triage nurses completing a one-page form during their triage assessment of patients presenting to their ECs. Entries included the triage category allocation, time of triage, time of physician consult and time the patient leaves the EC. Internal training by the EC unit managers was conducted to familiarize the staff with the content of the data collection form. Medical record stickers with patient identifiers were attached to the form so that the data could later be merged with the electronic data. Clerical staff from the four ECs captured the manual data daily from the forms onto a custom spreadsheet. The researcher collected the spreadsheets from the four ECs and used the patient identifiers to merge the electronic and manual data sets through the merge data function. Patient identifiers (e.g., names, surnames and medical record numbers) were included in the data set shared with the researcher to provide an identifier for merging electronic data with manual data.
Following this, all identifiers were stripped from the sample prior to analysis. The manual data capture forms were collected from the four ECs and handed back over to the hospital groups' medical records department at the end of the study. Only records with all the relevant data points were included.
Records with missing data points were identified, filtered and removed from the database prior to analysis. Removing records from the data set may have introduced exclusion bias that could have resulted in the removal of potential outliers such as high-acuity cases.
Removing incomplete records before analysis ensured that a complete data set was available with all the data points present. There were no obvious reasons for data points to be missing other than random omission from the staff to make entries and thus obtaining these missing data points would not be possible.
The timeframes as patients moved through the ECs were captured at specific points in their journey from entering to leaving the EC by either being discharged or admitted to hospital. An observa-  Table 1. Unfortunately, the ECs were not using a single triage system exclusively at the time of this study, which made a direct comparison unrealistic. The proponent of this study was to identify current timeframes and match them with existing triage systems in the aim to create a standardized locally appropriate triage system, with realistic timeframe expectations.

| RE SULTS
There were a total of 7,311 electronic and 6,754 manual patient records captured from the four ECs. When the data were combined in a single spreadsheet, there were some records captured electronically but not manually and vice versa, thus resulting in a smaller, It was found that the overall median time from registration -> triage was <10 min (IQR 0-6 min) and registration -> physician consult was <20 min (IQR 0-19 min) ( Table 3). The median triage -> consult times support the notion that patients were seen by a physician within 25 min (IQR 0-22 min) from the time they are triaged. EC1 was the only EC that saw a category 1 case; a physician saw them immediately. Category 2 cases were also seen immediately by physicians in all except EC2. They reported a median of 16 min (IQR 12-19 min).
Timeframe data from EC2 showed a marked increase compared with the other ECs in the time it took for patients to be seen by a physician. In most cases, the median time was three to four times higher than the other ECs. The overall lengths of stay in the ECs were much longer for the mid-to high-acuity cases (i.e., categories 1, 2 and 3) (IQR 1 hr 13 min-2 hr 44 min) with the lengths of stay of the low-acuity cases (i.e., categories 4 and 5) (IQR 32 min-49 min) being markedly less. This decrease in lengths of stay of low-acuity cases as compared with the mid-to high-acuity cases is further evidenced by the decreased times from physician consult -> patients leaving the EC being 15-31 min.

| D ISCUSS I ON
One of the most important validation criteria of any triage system is the time-to-physician variable (Beveridge et al., 1998 They were, however, designed with the setting in mind that the triage systems would be used. Although there were observable differences in the overall timeframes of patients as they moved through the four ECs, especially from the two hospitals and two clinics, it was evident that the median time for patients from entering an EC to be seen by a physician was relatively short when compared with the timeframe targets of the existing triage systems (Beveridge et al., 1998;Gilboy et al., 2011;Manchester Triage Group, 2006;South African Triage Group, 2012 Health Authority Abu Dhabi, 2019). This decreased load allows for patients to be seen at a relatively fast pace, throughout all triage categories. It is noted, however, that EC2 had a markable increased time from triage to physician consult, even with half the patient load as compared with EC1. This could be due to the triage system, or a combination of systems they employ at that EC, or purely be an organizational issue in that EC that requires further investigation. EC1 was able to move higher acuity patients out of their EC quicker than the others, which may open available bed space as patient throughput is faster.
This is a key element when evaluating the time-to-physician times in an EC. Being able to transfer patients out of an EC more readily allows for resources to be freed sooner, resulting in more patients that can be seen in a shorter period of time (Gravel et al., 2013;van der Wulp, 2010). This is especially true when considering this hospital group's largest patient cohort is of low acuity, thus requiring fewer binding resources per patient.
The use of electronic and manual platforms led to exclusions; gaining a full sample was reliant on the data points matching up between the platforms. Data points that did not exist in both electronic and manual data sets, as well as duplicate records or records with missing data points, were removed. However, early reports suggested that for cases where manual records were present but were not reflect electronically, these patients were streamed to outpatient departments and not seen in the EC. For cases where electronic records were present and not reflected manually, operator omission was considered, or the EC operations required a bypass of triage for unknown reasons. It is unlikely that these missing data points would have affected the results of this study. It was found and acknowledged that a large portion of children's records was excluded due to missing blood pressure entries.
Evaluating timeframes were complex with most of the time stamps requiring manual entry and those that were self-generated by the hospital information system still were at the mercy of staff entry time. Incorrect times could have resulted from unsynchronized clocks, forgetting to accurately determine and record the times, or making late entries on the hospital information system. It was accepted that some variation of time records existed as this was dependant on human input. Using the medians as a measure of central tendency helped mitigate any absolute outliers that could have influenced the findings.

| CON CLUS ION
in use. By shortening the time patients wait to see a physician and get appropriate treatment, it will not only improve morbidity and mortality, but also improve the patient experience. These benchmarks would greatly assist this and other private hospital groups in the region to set targets for their own triage system.

| RELE VAN CE TO CLINI C AL PR AC TI CE
Timely care in any healthcare setting is crucial for the effective management of a patients' illness or injury. Mortality and morbidity have shown to decrease when treatment is given sooner rather than later. There is an increasing patient population that presents to ECs, increasing the strain on available resources. To ensure the most critically ill or injured patients are attended to first, it is crucial for a triage system to distinguish acuity accurately and consistently.
Once a patient is triaged, it is vital for them to be attended to within the relative timeframes associated with their acuity, to ensure timely emergency care is provided. Evaluating the performance of an EC to meet these targets (and to make changes where necessary) strengthens the clinical ability of the unit to manage patients effectively.

ACK N OWLED G EM ENTS
The author wishes to acknowledge the contributions of Stevan Bruijns from the University of Cape Town (South Africa) and Albert Oliver