Lean thinking by integrating with discrete event simulation and design of experiments: an emergency department expansion

Discrete Event Simulation and Lean in healthcare systems

Performing DES in healthcare is not new, dating back to the 1960s (Pitt, 2008). However, since then, there has been considerable growth for its interest. According to Arisha & Rashwan (2017), this progress is strong evidence that the simulation provides better decisions in health services management without compromising patient safety. This advantage has increasingly attracted the attention of hospitals and health authorities (Cheng et al., 2017). However, experts in healthcare simulation claim that its use is more complicated than in other areas (Tako & Robinson, 2015). The main problems found in the healthcare simulation are less evident structure; the system is complex; more significant effort to collect and access data; barriers due to ethical issues; client’s shortage of time; and more difficulty in ensuring implementation.

Despite the difficulties, it is possible to find studies with positive results linked to costs, capacity, wait and stay time, and levels of service and losses. Zhou & Olsen (2018) applied DES for medical supplies management and decreased the costs involved in the process by reducing expired drugs. Hussein et al. (2017) presented effective results reducing overcrowding in a hospital. Similar results were shown by Babashov et al. (2017) and Shim & Kumar (2010) by decreasing patients’ waiting time in an emergency department. Rau et al. (2013) and Uriarte et al. (2017) also used DES to reduce patient waiting time in treatment centers and the radiotherapy sector, respectively. Furthermore, Al-Araidah, Boran & Wahsheh (2012) applied the tool in an ophthalmology laboratory to decrease patients’ waiting and appointment time. Regarding planning and capacity analysis, Pinto et al. (2015) defined the ideal number of beds for a Brazilian hospital. In the balancing of staff work, Reynolds et al. (2011) applied the simulation to reduce the workload of an English pharmacy and (Pongjetanapong et al., 2019) used the tool to evaluate the change effect to staff levels in a cytology department.

Although these studies present positive results, we found some limitations in them. Some papers did not consider variables and statistical distributions for different workgroups (Reynolds et al., 2011) and processes (Rau et al., 2013). Pinto et al. (2015) state that the simulation results have a slight underestimation, although it seems reliable and more effective for decision making than empirical calculations. Although articles use powerful tools jointly with DES, some studies do not test the interaction of resources, locations, and factors using the “best guess” approach (Al-Araidah, Boran & Wahsheh, 2012; Uriarte et al., 2017). In contrast, Hussein et al. (2017) use DoE to observe the interaction of two factors in their scenarios. Nevertheless, they did not observe the interaction in the resource number in each shift (present in this study). Shim & Kumar (2010) state that the simulation did not take into account the varying resource capacities (doctors and nurses) and locations (workstations from the sorting station to the payment station) involved in the emergency care process. Finally, Babashov et al. (2017) use simulation to determine the average waiting time for all patients overall, not for each patient type.

It was also possible to verify improvements using only LH. Even when not implemented systematically and comprehensively in the organization, LH can provide several benefits to health services (D’Andreamatteo et al., 2015). Among these benefits, the authors highlight improvements in productivity, costs, financial results, quality of service delivery, and patient and team safety and satisfaction. However, implementing Lean is long-lasting and challenging work in the health sector (Toussaint & Berry, 2013). Lean transform the organizational culture from the inside out, requiring managers and leaders to become facilitators, mentors, teachers, and enable employees to take the initiative in making improvements.

Studies that show LH implementation can be found for the elimination of processes that do not add value to the patient (Teichgräber & Bucourt, 2012) and reduction in the instrument collection stage (Kimsey, 2010). Laganga (2011) used lean methodology to increase patient care capacity, while Papadopoulos, Radnor & Merali (2011) used LH to reduce delays in receiving samples, prioritize urgent work, standardize processes and anticipate problem identification. Although LH application is useful and essential in waste elimination and increasing patient value-added time, it can be time-consuming, and the staff involved may resist the methodology.

When used jointly, Lean and DES have three goals: to teach, evaluate, and facilitate the process. In order to evaluate, it allows the execution of experiments and the evaluation of their results. It should be employed after holding a team meeting, testing ideas, and creating new solutions (Robinson et al., 2012). The use of DES and LH in hospital settings may bring more quality and efficiency to patients and management (Gaba, 2004). Also, the patient flow may be optimized and served as a motivational factor for employees (Salam & Khan, 2016). Swick et al. (2012) state that hospitals that integrate these tools offer an efficient method of strategic planning and provide employees with a privileged view of how to reduce waste and add value. Moreover, it is possible to reduce patients’ waiting time, decreasing employee workload, and promoting resource reallocation (Haddad et al., 2016; Bhat, Gijo & Jnanesh, 2014). Finally, Doğan & Unutulmaz (2016) and Robinson et al. (2014) used the tools to transform static mappings into dynamic.

Some limitations were found in these studies. Salam & Khan (2016) did not use arrival and staff modeling. Swick et al. (2012) limited the work to the daytime hours (more considerable resources demand), and there was no adjustment for the night shift. Bhat, Gijo & Jnanesh (2014) conducted the study in an environment where patient arrival cannot be adequately controlled. Haddad et al. (2016) conducted the study limited to a process that ignores interdependencies with internal and external parts and laboratory tests. Doğan & Unutulmaz (2016) used only Value Stream Mapping and did not carry out planned experiments.

IDEF-SIM Technique

The IDEF-SIM (Integrated Definition Methods - Simulation) process mapping technique was developed by Montevechi et al. (2010). It aims to support modeling work in the implementation and analysis phase, reducing project time. The IDEF-SIM was created based on existing logical elements in other mapping techniques, such as IDEF0, IDEF3 and flowchart (Peixoto et al., 2017; Montevechi et al., 2010). To our best knowledge, there is no history of this technique used in healthcare systems, although we have found in manufacturing (Pereira et al., 2015; Montevechi et al., 2010) and transports (Lopes et al., 2017). Nevertheless, we chose to use this technique because the symbols are directly translated from the conceptual model for the computer model, presenting elements of the simulation, e.g., entities, locations, resources, functions, controls, logical rules, and transport (Pereira et al., 2015). The symbols used in the IDEF-SIM are (Peixoto et al., 2017; Montevechi et al., 2010):

1. Entity (circle): items processed in the system, e.g., raw materials, products, people.

2. Functions (rectangle): places where entities change or have delays, e.g., workstations, conveyors, queues, stocks, beds, waiting lines.

3. Entity flow (arrow): direction that the entities follow in the model.

4. Resources (arrow under the rectangle): elements that move the entities or perform the functions, e.g., people, such as operators, nurses, doctors.

5. Controls (arrow above the rectangle): rules used in the functions, e.g., rules of scheduling, queuing, sequencing.

6. Movement (large arrow): it is used when there is a significant entity displacement.

7. Explanatory information (dashed arrow): it is used to include some explanation.

8. Rules for parallel and/or alternate flows: entities must follow in parallel (AND –“&”); alternatively (OR –“X”); or in parallel and/or alternate paths (AND/OR –“O”).

9. Input flow (initial arrow): determines the input or entities’ creation.

10. End of the system (painted circle): determines the end of a path.

11. Connection with another figure (triangle): divide the model into different figures.

Design and analysis of experiments

Design of Experiment (DoE), is considered one of the most relevant methodologies for researchers performing applicable experiments. Moreover, according to Montgomery (2017), DoE is a collection of statistical techniques capable of generating and analyzing experimental projects in which several factors are varied at the same time. In DoE, the modeler chooses different levels for the independent variable and can vary the factors simultaneously to study the effects in the responses (dependent variables). These effects can be separate from each independent variable or the interaction between them (Baril, Gascon & Vadeboncoeur, 2019). When experimental designs are appropriately designed, trial and error analysis are avoided. Also, it is possible to detect unimportant factors and simplify the simulation model. Among DoE techniques, the 2^k factorial design is used as it requires few executions per factor, and data interpretation can proceed mainly through arithmetic and graphs. In DoE, k value is the number of factors that presented the upper (+) level and lower (−) level. In the literature, many studies use DoE to plan and dimension resources around healthcare systems and EDs. Baril, Gascon & Vadeboncoeur (2019) used the DoE to vary the number of doctors, nurses and beds to decrease LOS in Canada. Bhattacharjee & Ray (2018) used DoE and DES to gain insights into appointment scheduling. Moreover, Gul, Guneri & Gunal (2016) dimensioned an emergency department to support natural disasters analyzing the number of nurses and doctors. Although these articles use DoE, our study is different because we use DoE to dimension beds and the number of resources in each shift.

Case study

Cape Breton Regional Hospital (CBRH) is one of four facilities that make up the Cape Breton Health Care Complex, a Canadian public health network of four hospitals, serving approximately 100,000 inhabitants. This paper aims to plan the expansion of a Canadian ED to meet the demand that comes from the small closed care centers. The exponential increase in demand is due to the closure of these small service units in locations near the CBRH.

Regarding the patients’ flow, they arrive at the ED by themselves or by ambulance. Both are triaged and ranked according to severity. This classification is given according to the Canadian Triage Acuity Scale (CTAS). CTAS level I corresponds to the most severe level, resuscitation (blue, to be seen immediately); level II is emergent (red, to be seen <15 min); level III is urgent (yellow, to be seen <30 min); Level IV is less urgent (green, to be seen <60 min) and level V is non-urgent patient (white, to be seen <120 min). Patients arriving by ambulance, after triage, go straight to care. Patients who arrive on their own go to the registry and expect the transfer to their appointment. CTAS level I, II, and III patients go to a bed area (45 beds and 3 for mental health), while CTAS IV and V follow to the vertical area (10 seats).

When the patient arrives at the bed area (BA) or the vertical area (VA), he undergoes two evaluations made by the nurse and the physician. Then the ward clerk receives and records the patient’s prescription. These prescriptions may include laboratory tests (blood, urine), medical procedures, and diagnostic imagining (DI). Such procedures are assigned according to the patient’s category, shown in Table 1. If DI is required, a nurse and a porter must escort CTAS level I patients. For other patients, only the porter is required.

After waiting for the exam results, the patient may go through an appointment with a specialist or move on to the next stage. If the patient goes through the appointment, he must wait until the specialist arrives, since they are not in the ED. In the next step, the patient may be discharged or go through monitoring. After monitoring, the patient goes to the SSU (10 available beds), where it can remain from one to three days.

The ED Nurses are divided into two teams. The first team is responsible for patients CTAS level I and II (EDN1), and the second is responsible for CTAS level III, IV, and V (EDN2). They work in four shifts: 07:00 a.m. to 07:00 p.m.; 09:00 a.m. to 09:00 p.m.; 11:00 a.m. to 11:00 p.m. and 07:00 p.m. to 07:00: a.m. Physicians and Triage Nurse (TN) start working at the same shifts. In addition, three porters work in the following shifts: 07:00 a.m. to 07:00 p.m.; 10:00 a.m. to 06:00 p.m. and 07:00 p.m. to 07:00 a.m.

The study aims to analyze whether the ED can absorb the full demand from the closures. Other issues to be solved are ideal numbers of resources (TN, EDN, physicians, and porters); the ideal number of beds in BA, chairs in VA and beds in SSU; reduction in the time of care after going through the triage, in the time to transfer from BA/VA to SSU and LOS. In this sense, the decision-makers from the hospital proposed an approach to be tested by the simulation, called Initial Planning (IP). We used DES because patients are individual entities with particular characteristics, and the procedural flow depends on the CTAS of each one. Moreover, we chose DES over System Dynamics because it has the most relevant characteristics of the work responses. We can say that DES is better in decision-making, prediction, optimization, and comparison; it behaves well for high rates of change and presents the high importance of tracking individuals. In addition, it is suitable for systems with few entities and relatively small-time scales (Braislford & Hilton, 2001).

Conceptual modeling

We performed the conceptual modeling of the system through the IDEF-SIM language. Figure 1 shows the patient flow. We do not present percentages for each process because there are many combinations based on the disease category and CTAS level.

As mentioned in “IDEF-SIM Technique”, the first circle represents the entities in the model, i.e., patients. Rectangles mean delays in patient flow or procedure waiting (queue, exam results). Large arrows represent transport for the patient, e.g., an exam performed in another room. Processes that need a nurse or doctor are indicated with an arrow at the bottom of the rectangle, while the upper arrows are controls. Symbols with an “x” inside them indicate an option (OR connection), while symbols with an “&” represent an obligation (AND condition). The diamonds represent a stopping point. The entity only follows the flow if the given conditions have been met. In the model, the patient only continues the flow if the test results are ready, and there is an evaluation by a doctor and a nurse. We ensure that the conceptual model presents high reliability since it was validated by face-to-face validation, where experts verify if the model matches the real system (Kleijnen, 1995; Sargent, 2013).

Data set

Data collection and modeling were based on historical data (patients’ arrival) and from some databases that are used in the hospital from April 1st, 2017 to March 31st, 2018. Through the historical data, we observed that patients arrive at different rates hourly and daily. These rates are repeated weekly. As soon as a patient is screened in the emergency room their information is entered into the system. After that, Medietch^® is used for registration, diagnostic codes, CTAS level, treatment time, diagnostic test ordering, and blood exams. Figure 2 shows the patient’s arrival rate based on weekday and the two-hour scale.

We notice that the curves show similarity in the shape where the number of patients starts to increase from 08:00 a.m. Moreover, the curves present a peak between 10:00 a.m and 12:00 p.m. The days with the highest arrival rates are Tuesday, Wednesday and Thursday. Regarding the type of patient, the curves do not differ significantly, showing the same behavior throughout the day and week. Patients of CTAS level III and IV are predominant in the ED. As shown in Fig. 1, the patient may be admitted for SSU placement or be discharged. Each of these stages involves different resources, such as doctors, nurses, porters, and the patient suffers a delay (process time). These times were estimates from the ED manager and nurses through Delphi method (Table 2).

Table 2:

Processing time and required staff.

Task	Process time (min)	Staff required
		Triage nurse	Registration	ED nurse	Physician	Admin	Porter
Triage	T(4,7,10)	x
Escort to Bed	T(0.5,1,2)			x
Register Patient	3		x
Nurse Assessment	CTASIV/V: 5; CTASIII: 10; CTASII: 15			x
Physician Assessment	T(7,12,15)				x
Write Patient Orders	T(2,4,5)				x
Initiate DI/Lab	T(2,2.5,3)					x
Prepare Patient for DI	3			x
Receive/Add Tests to Chart	1			x		x
Review Tests (Nurse)	1			x
Review Tests (Physician)	2				x
Take Specimen	T(4,6,15)			x
Medical Procedurea	T(15,20,60)			x	x
Initiate Consult	2					x
Arrange Consultant	2					x
Discuss with Specialist	3				x
Make Disposition	20% - 30; 80% - 3				x
Admission Form	2			x
Determine Care Plan	T(7,10,12)				x
Contact Specialist	5				x
Transport Patient to SSU (I and II)*	T(15,20,25)			x			x
Care Plan and Final Orders	T(2,4,10)				x
Give Instruction	T(2,5,12)			x
Prepare Patient	T(1,2,4)			x
Shift Change Written/Verbal	5				x
EHS Call	1	x		x
EHS Triage	T(4,7,10)	x		x
Prepare Room for Next Patient	1			x
Huddlea	10	x	x	x	x	x

DOI: 10.7717/peerjcs.284/table-2

Notes:

aDenotes a task which requires all staff indicated (x).

As previously stated, ED patients may require a sample, consultation and/or lab work. The patients may carry out a sample, e.g., blood work (BW) and specimen (SP). In addition, patients may have DI exams, such as X-ray (XR), electrocardiogram (EKG), echocardiogram (ECHO), computerized tomography (CT), and ultrasonography (US) according to the CTAS level and patient’s category (Table 1). Table 3 summarizes the probability if the patient needs an exam.

Table 3:

Exam probability for each CTAS level.

		Treatment category
Exam	CTAS	GMI	RE	GA	OR	CA	DE	GE	ENT	MH	NE	OP	GY	SM
CT	I and II	1%	x	✓	50%	x	x	✓	✓	x	✓	50%	x	x
	III	1%	x	50%	10%	x	x	50%	x	x	x	x	x	x
	IV and V	1%	x	x	x	x	x	x	x	x	x	x	x	x
ECHO	I and II	x	x	x	x	10%	x	x	x	x	x	x	x	x
	III	x	x	x	x	x	x	x	x	x	x	x	x	x
	IV and V	x	x	x	x	x	x	x	x	x	x	x	x	x
EKG	I and II	x	✓	x	75%	✓	25%	x	x	x	x	x	x	x
	III	x	x	x	75%	✓	x	x	x	x	x	x	x	x
	IV and V	x	x	x	x	✓	x	x	x	x	x	x	x	x
US	I and II	1%	x	✓	x	x	x	✓	x	x	x	x	✓	x
	III	1%	x	50%	x	x	x	50%	x	x	x	x	✓	x
	IV and V	1%	x	x	x	x	x	x	x	x	x	x	x	x
XR	I and II	50%	✓	✓	✓	✓	25%	✓	✓	x	✓	25%	x	x
	III	50%	✓	50%	✓	75%	x	50%	25%	x	x	x	x	x
	IV and V	50%	25%	x	50%	75%	x	x	x	x	25%	x	x	x
BW	I and II	50%	✓	✓	✓	✓	75%	✓	✓	✓	✓	✓	✓	✓
	III	50%	✓	✓	✓	✓	25%	✓	25%	✓	✓	x	✓	✓
	IV and V	50%	x	x	x	25%	x	x	x	50%	10%	x	50%	75%
SP	I and II	x	✓	50%	25%	x	25%	✓	25%	✓	✓	x	✓	✓
	III	x	75%	50%	10%	x	25%	✓	50%	✓	50%	x	✓	✓
	IV and V	x	25%	x	x	x	x	✓	x	50%	x	x	50%	75%

DOI: 10.7717/peerjcs.284/table-3

There is a probability of the patient having an appointment with one or two specialists. Moreover, when the patient needs an appointment with an expert, some response delay can happen according to the specialist and the CTAS patient level. This delay occurs because the specialist is not staffed in the ED unit all the time. They release the patient to the attending physician while consulting the patient unless they are CTAS level I. After a specialist arrives in the ED to consult with a patient, he attends all patients that need him, from lowest to highest CTAS level. The specialist will start a new consult any time before the end of their shift. Any patients still in need of a consult at the end of the day will be seen by Internal Medicine or wait until the beginning of a specialist’s shift the following day to receive an appointment. Internal Medicine covers cardiology, nephrology and gastroenterology. Table 4 describes these data.

Table 4:

Specialist probability appointment for each CTAS level and data set.

Treatment category	Consult 1	CTAS I and II		CTAS III		CTAS IV and V
		C1	Time (min)	C1	Time (min)	C1	Time (min)
RE	Respiratory Therapist	75%	60	10%	15	x	x
GA	Gastroenterologist	✓	T(15,22,30)	25%	T(15,22,30)	x	T(15,22,30)
OR	Orthopedist	✓	10	✓	5	50%	3
CA	Cardiologist	✓	T(30,45,60)	50%	T(30,45,60)	10%	T(30,45,60)
DE	Dermatologist	25%	T(30,45,60)	10%	T(30,45,60)	x	x
GE	Urologist	✓	T(15,22,30)	✓	T(15,22,30)	x	x
ENT	Otolaryngologist	✓	60	50%	20	x	x
MH	Crisis Response	✓	T(60,75,90)	✓	T(60,75,90)	✓	T(60,75,90)
NE	Neurologist	✓	60	✓	60	x	x
OP	Ophthalmologist	10%	T(30,45,60)	5%	T(30,45,60)	x	x
GY	Gynecologist	✓	T(15,22,30)	✓	T(15,22,30)	x	x
SM	Crisis	✓	T(60,75,90)	✓	T(60,75,90)	✓	T(60,75,90)

Treatment category	Consult 2	CTAS I and II		CTAS III		CTAS IV and V
		C2	Time (m)	C2	Time (m)	C2	Time (m)
GE	Nephrologist	50%	T(10,15,20)	25%	T(30,45,60)	x	x
MH	Psychologist	✓	T(60,75,90)	75%	T(60,75,90)	25%	T(60,75,90)
SM	Psychologist	✓	T(60,75,90)	50%	T(60,75,90)	25%	T(60,75,90)

DOI: 10.7717/peerjcs.284/table-4

Computer Modeling

We designed the computer model in FlexSim Healthcare^® software, programming three different processes. First, we programmed patient flow according to CTAS levels, because each level may indicate different paths and require different resources. The second process involves the procedures exam, from sample collection to the end result. It directly impacts the patient waiting time, since it can only follow through the flow after the results, characterizing a necessary time, but that does not add value for them. Finally, we construct global processes, such as staff meetings that affect patients’ waiting time and must be reduced to a sufficient level. Figure 3 shows the computer model designed for the IP scenario.

After the model was built, we performed a verification process. The process was conducted through a face-to-face technique for each CTAS level. The specialists and the modeling team analyzed the simulated model and agreed that it was as close to the real patient flow as possible. We also performed the validation. If the modeled system is a non-observable system, it is recommended using a subjective approach (Sargent, 2013; Sargent, Goldsman & Yaacoub, 2016). In this sense, we carried out a sensitivity analysis to see the model variability. We have tested patient arrival rates at five levels: 100%, 75%, 50%, 25% and 10%. We got the mean LOS and the percentage of treated patients to perform the validation. For LOS, in minutes, the simulations show: 2,213.7 (100%), 428.2 (75%), 352.9 (50%), 270.9 (25%), 243.4 (10%). We expected to increase the number of patients, LOS would increase in the same way. This situation occurs because the patients have to wait more time for available resources. Although, if we are decreasing the arriving rate, the percentage of treated patients starts to increase because there are more staff available for each one. Figure 4 shows the sensitivity analysis for the validation.

Model execution and initial analysis

After the computer validation, the model was simulated for one week, and data were collected from Tuesday to Tuesday. A warm-up of one day (Monday) was necessary to get a stable process. Consequently, according to the computer simulation, the ED cannot meet the expected demand for IP. Moreover, the obtained results were:

• On average, 1,504 patients arrived in the model;

• On average, 258 (17.2%) patients were completed treated in the simulated period;

• Around 1,157 patients (76.9%) did not even make it triage;

• On average, LOS was 2,213.7 min;

• Patients wait about 404.3 min to be seen after triage;

• Patients who need to go to SSU wait, on average, 367.4 min to be transferred.

Regarding the results, the number of patients that do not go through the risk classification is alarming. In addition, after triage, the average waiting time is around 404.3 min, which corresponds to approximately 3.5 times what the patient with CTAS level V should wait at most. Moreover, patients expect to be attended after the risk classification, on average 25.9, 164.6, 332.2, 649.7 and 419.5 min for the CTAS levels I, II, III, IV, and V, respectively, which does not meet the specifications.

The result is unsatisfactory and below the current quality standard offered by the hospital. We conclude the initial ED planning for the possible scenario after closing the surrounding care centers still has deficiencies. This reinforces the importance of using simulation to ensure that ED, in this scenario, continues to offer a high quality of care. Especially in disruptive situations where planning becomes more difficult.

Moreover, for the initial analysis, most of the resources are idle, e.g., triage nurses and physicians. Most nurses and porters are waiting for empty locations (beds and seats). Consequently, the patient’s flow is stuck. It other words, beds in the BA, MBA, SSU, and VA are occupied most of the time.

In this sense, based on LT, some improvements have been proposed. These improvements aim to reduce patient waiting time, thereby increasing the safety of their care. Therefore, we investigated the interaction between places in the ED and the resources in the patient flow.

Design and analysis of experiments

Confirmation runs

In order to confirm the proposed changes, the model was simulated 30 times and a warm-up from Monday (00:00) to Tuesday (00:00). Data were collected from Tuesday to Tuesday. Table 8 presents the Final Planning (FP) configuration for locations and resources.

According to simulation results, the study recommends two additional beds in the MBA and five additional chairs in the VA. Regarding the resources, it is necessary to hire one TN, one EDN2, four porters, and reduce two physicians. Table 7 lists the shift for each group. It was necessary to dimension the staff, avoiding, when possible, hire them. The number of beds in SSU directly affects patients’ LOS. Then, the ideal number is 30 beds, corresponding, on average, 146.4 min of waiting for transfer to SSU. About the specialist, a psychiatrist, a neurologist, and a crisis response must stay directly in the unit. With the proposed measures, the patient’s LOS decreases by 1,752.6 min. Table 9 presents the results of the IP and FP scenario.

Sargent, Goldsman & Yaacoub (2016) suggest that, for a non-observable system, an objective approach is to compare the results from the developed model with another similar system to perform the validation. In this sense, we decide to compare the model with another ED that works in the same way to verify the results from the future state. In our model, the mean LOS is 461.2 min and a standard deviation of 20.7 min. In the other ED, the LOS is 472.1 min. Then, we carried out a 1-Sample Equivalence test that evaluates if the mean of a population is close enough to a target value, specifying a range to be equivalent. In the test, we can say that the difference between the model’s results and benchmarking is not significant, less than 5% (p-value < 0.001). In other words, the models are equivalents, and we can affirm that the FP scenario and the results are validated.

After the proposed changes, the number of treated patients increased considerably, making their flow continuous during the process (LT principles). In addition, patients are seen as expected after triage, according to the CTAS classification level.