A simulation model to analyse automation scenarios in decontamination centers

ABSTRACT

Decontamination centres provide sterilisation services (sort, disinfect, package, and sterilise) for reusable surgical instruments that have a vital impact on patient safety. The market trend is to increase the level of automation in the decontamination process, to increase productivity, and reduce the risk of human error and musculoskeletal injuries. The goal of this research is to study the use of automated guided vehicles (AGVs) in sterilisation departments, to improve safety and efficiency. A generic simulation model is created based on data gathering of various decontamination centres and is validated for a specific centre to analyse various aspects of applying AGVs to automate the internal transfer. Centre’s potential to increase capacity through AGV application is analysed and a Design of Experiments is conducted to identify the most promising implementation scenarios. Results show reductions in treatment time and work in process, while ,maintaining the accessibility of medical instruments, and ensuring worker safety.

1. Introduction

Decontamination of reusable instruments and other articles that have been used on patients for medical purposes is an essential activity in hospitals. In most modern hospitals, there is a Central Sterilisation Department (CSD) to execute this function. The general process in a CSD consists of pre-disinfection, washing, inspection, packing, sterilisation, and storage (Di Mascolo & Gouin, 2013). Instruments arrive in the decontamination area and the staff separate them then clean them by soaking the instruments in water and hand washing them (pre-disinfection). Groups of instruments are then batched in racks to be loaded into washing machines. After being washed, the racks are unloaded from the machines in a separate room, called a sterilisation area, to avoid the contamination of clean instruments. Packing and sterilisation take place in this area, after which the sterile instruments are stored and transferred for re-use.

Although sterilisation centres have no direct patient contact, their performance directly affects patient care. Besides the quality of sterilisation, the timely availability of instruments for surgeries is important (Keseler, 2019). Therefore, overcoming challenges and improving processes in CSDs leads to an increase in the quality of service to patients. These challenges include capacity shortage, long processing lead times, and ergonomic problems for the staff. CSDs often struggle with a shortage of resources. Meanwhile, the ageing of the population, and improvements in medical services, are driving an increase in the volume and variety of surgical instruments (Keseler, 2019). Each instrument used during medical interventions must be cleaned to remove any form of contamination. These instruments arrive at varying rates from operating rooms, clinics and the emergency room. The size and density of arriving packages, as well as the required treatments, varies for different medical specialities. The CSD staff is called upon to perform specific and sequential tasks, and to handle heavy loads that can cause musculoskeletal problems over time. Due to the occupational hazards with this line of work, CSDs risk losing experienced employees to injury and must invest in job training for a constant stream of new hires.

In this paper, we explore opportunities for improvement in CSDs by applying automation to internal transfer processes. Various solutions exist to move instrument racks in a CSD, such as conveyor systems that can come with or without automated loading/unloading systems for the washers. However, these solutions still require an operator to manipulate the racks with transfer carts. More recently, an automated solution based on an Automated Guided Vehicle (AGV) was developed by a company that provides products and services to decontamination centres, to eliminate the need for rack manipulation. What is perhaps most notable about this new AGV is its connection with other machines in the system, which allow it to recognise instruments’ attributes. These AGVs that use more recent technologies are also called Automated Intelligent Vehicles (AIV) and can function autonomously and communicate with the environment and pass the obstacles on their paths (Cronin et al., 2019). The application of this AGV in a CSD in the province of Quebec, Canada provided an opportunity to analyse the potential benefits of and barriers to using this technology in CSDs.

AGVs have been used in manufacturing areas for decades and their design has been updated with technology improvement (Kumbhar et al., 2018). Typically, the application of AGVs has at least three types of benefit, which include increasing the efficiency of processes and staff, supporting the lack of resources, and ergonomic benefits and comfort for the staff (Hellmann et al., 2019; Kumbhar et al., 2018). Industry group Material Handling Industry of America (MHIA) confirmed the ergonomic benefits of AGV application, citing a reduction in workers’ strain (Benzidia et al., 2019). According to Pedan et al. (2017), the application of AGVs in the hospital setting has high potential, as it may improve operational performance by saving time on no-value-added activities that can instead be transferred to patient care (Chikul et al., 2017). Despite the potential benefits, there is a paucity of research considering the application of AGVs in the healthcare setting (Benzidia et al., 2019; Chikul et al., 2017; Pedan et al., 2017). Moreover, studies of AGVs consider technical aspects of automation, while the system engineering approach is not fully considered. It remains important to investigate the unique characteristics of hospitals (locally and generally) to provide a better understanding of the challenges to and benefits of applying AGVs in hospitals, and to facilitate decision-making for managers and staff.

This paper aims to fill this gap by addressing two central questions about implementing AGVs in CSDs.

• What are the potential improvements in process time, capacity, and worker efficiency by applying AGVs in CSDs?

• Among effective factors such as adding AGVs and staff in each area, and changing the arrival schedule, which ones have the greatest impact on the system? And which combination of factors produces better results?

To achieve this goal, a simulation method was applied to address the complexity of the CSD system. We first created a generic simulation model that could be applied to various centres. We then validated it for a specific centre to analyse how AGV application would affect the capacity of the system to increase the volume of instruments processed. Additionally, a Design of Experiments (DOE) was performed to analyse the interaction of factors such as adding staff or changing the arrival schedule. Comparing the application of AGVs to the current scenario with no automation in place, our results showed promising improvements in total treatment time, our primary service level indicator.

By presenting a simulation model and investigating various aspects of applying AGVs in a CSD, this study intends to make the following contributions. First, our simulation provides an analysis and decision-making tool for CSDs in healthcare settings, using a generic approach. Second, for the CSD under study, the analysis results will orient decisions about applying this technology, its benefits, barriers, and interaction with other factors. Moreover, as the adoption of automation in CSDs is still slow (Moatari-Kazerouni & Bendavid, 2017), comparing automation scenarios and highlighting means of improving the system will serve to promote the application of new technologies in this challenging environment.

This paper is organised as follows. A literature review is presented (Section 2), followed by a description of CSD processes and challenges (Section 3). The methodology and simulation model are introduced (Section 4), then the results of the automation scenario, sensitivity analysis and DOE are presented (Section 5). A discussion (Section 6) and conclusion (Section 7) appear at the end of the paper.

2. Literature review

This paper focuses on simulating CSDs and analysing the effects of AGV application in these systems. The first part of the literature review looks at methods and tools in simulation, optimisation, and automation in healthcare settings and CSDs. The second part of the literature describes applying tools and techniques to analyse the application of AGVs in manufacturing and healthcare settings.

2.1. Applying simulation and optimisation for improvement in healthcare and CSDs

2.1.1. healthcare

Hospitals contain complex and interrelated departments facing a high level of uncertainty. Simulation is an appropriate approach to analyse and predict the elements in this system, because it provides a statistically accurate modelling tool (Centeno et al., 2003). Zhang (2018) provided a systematic literature review regarding the application of Discrete Event Simulation (DES) in healthcare systems, which indicated that beginning in 2009 there was a significant increase in the use of DES. The aforementioned works aimed to provide a better understanding of the complex relationships in the healthcare system and to enable managers to make more efficient decisions (we refer readers to Günal and Pidd (2010) and Zhang (2018) for a comprehensive literature on application of simulation in healthcare). Aroua and Abdulnour (2018) applied simulation and experimental design for optimisation in emergency departments. They assessed the sensitivities of patient length of stay to selected parameters such as improvement of admission waiting times and treatment times. Asamoah et al. (2018) presented a simulation model to analyse the impact of applying RFID on patient scheduling in a hospital environment. The study showed that information visibility offered by RFID technology results in decreased wait times and improved resource utilisation. England et al. (2019) developed a DES model and combined it with forecasting methods for hospital beds management and predicting the number of emergency admissions.

2.1.2. CSD

A majority of studies that use simulation and optimisation to model healthcare systems considered the patient flow through different hospital departments and operating rooms (Di Mascolo & Gouin, 2013). However, a few studies consider process improvement for the sterilisation of medical devices in CSDs. In a master thesis, Ethier (2003) presented a simulation model in AutoMod software for a CSD, and analysed the effects of increasing the number of arrival trays from specific clinics or operating rooms, so as to compare the average service level in each scenario. Service level, as the principal key performance indicator (KPI), is calculated by dividing the number of immediate responses to requests for instruments by the number of delayed responses. Di Mascolo and Gouin (2013) proposed a generic simulation model to represent sterilisation services in the Rhône-Alpes region in France. They used the generic model to compare the performance of several sterilisation services. To gather data from one special centre and eight hospitals, they used a questionnaire containing 96 questions about the schedule, human resources, and sterilisation processes. Ngameni (2014) provided a Monte Carlo simulation of a CSD and analysed the scenario of using a case cart system. The purpose of using a case cart system was to create a cart with all of the medical devices necessary for a specific operation (Bonnefoy, 2015). In this study, transfers from operating rooms to the CSD and the steps needed to apply the case cart system were considered in order to estimate how many case carts were required and the potential increase in capacity. The article aimed to provide a feasibility study for adding case carts but did not test analytical or improvement scenarios.

Applying optimisation models, Xu and Wang (2018) presented a batch scheduling model for the batch of instruments in the washing machine to minimise washing time. The KPIs considered were process time and work in process. The specific purpose of the study was to improve the washing process, well-known as a bottleneck in CSD operations. The problem was solved by applying a genetic algorithm, which showed reliable results. However, the model was not applied to a real case. Van De, Muls and Schadd (2008) studied the flow optimisation for sterile instruments in hospitals between the CSD and operating rooms. They considered changing the logistics management principles, using technologies such as radio-frequency identification (RFID), and optimising the composition of the batch of sterile materials to reduce the process lead time (the principle KPI). Concerning automation in CSDs, Moatari-Kazerouni and Bendavid (2017) analysed automation scenarios in CSDs using the application of RFID technology. This research showed promising improvements in total process times and in the reduction of defects in processes.

Investigating simulation models for healthcare systems, as well as simulation and optimisation models for CSDs, provides a better understanding of the processes and the KPIs. In the following, studies considering the application of AGVs in manufacturing and healthcare are discussed.

2.2. Application of AGVs

AGVs have been applied in both manufacturing and healthcare settings. Application of AGVs in manufacturing is studied mostly under the subject of material handling which considers the movement, packing, and storing of materials and subsystems. Material handling is important as 80% of the time spent by the materials inside the plant area is spent for moving between places, waiting for processing or finding a place to be stored (Gaur & Pawar, 2016). AGVs were first introduced to the manufacturing industry to reduce product transfer times from one cell to another (Cronin et al., 2019). The highest motivation for applying AGVs for material handling is described as decreasing the production time and letting the workers focus on other activities (Hellmann et al., 2019). AGVs are the most widely automated devices used today in flexible manufacturing systems (FMS) and computer integrated manufacturing (CIM) environments. Kumbhar et al. (2018) stated that using AGVs for material handling increases flexibility and responsiveness in FMS. They mentioned that the two main barriers for applying AGVs in manufacturing systems are the high initial cost and the navigation in the layout such as the height of metal floors that must be crossed. Gaur and Pawar (2016) defined throughput, unit load, flow path design, and fleet size as the major elements influencing the design and operation performance of AGVs in a manufacturing unit.

Hellmann et al. (2019) analysed the decision-making process to choose an appropriate material handling technology. They proposed a framework that integrates failure modes and effects analysis (FMEA) and analytic hierarchy process (AHP) for evaluating possible solutions. The proposed decision-making method is applied for choosing between three choices of hiring additional material handlers, instal an AGV system or an AIV. The important factors to make the decision included the payback period, the potential increase in demand or production, the ease of implementation and control as well as safety and reliability. These factors should be weighted by the managers and the safety should be analysed using the FMEA method. In a more recent literature review of innovative technologies adopted in logistics management, Lagorio et al. (2020) considered articles about RFID, Big Data, Internet of things, AGVs, and automated storage/retrieval systems (AS/RS). Comparing studies that consider AGVS with other investigated technologies, the authors concluded that AGVs suffer from a lack of scientific studies, even though their applications are more widespread and frequent. Their study highlighted the existing gap in the literature about analysing the application of AGVs. Jaghbeer et al. (2020) emphasised that the decision-making process of automation in automated systems requires more academic research, particularly in the links between the different performance and design aspects.

In hospitals, more than 30% of costs are related to logistics, so using AGVs may be helpful to improve logistics management (Benzidia et al., 2019). Pedan et al. (2017) assumed that an AGV would be used in a hospital for food transportation to the patient rooms, transportation of laundry, and transportation of waste. Results of the simulation model demonstrated that applying AGVs would save 23 percent of total time for medical assistance per day. Other benefits included the transfer of heavy and dangerous items by AGVs. Benzidia et al. (2019) investigated automation and AGVs in healthcare logistics and highlighted the importance of automation in hospital logistics. They presented a qualitative case study by conducting semi-structured interviews and analysed the scenario of adding AGVs in Mercy Hospital in Metz, France. They categorised flows in the hospital as catering and laundry (bed sheets and meals), pharmaceuticals and sterilised products, waste, blood, and patients. They mentioned that patient flows and surgical and laboratory flows are the areas that received some attention, while studies considering other flows in the hospital such as medical equipment were limited. The study provided comprehensive literature about the application of AGVs and revealed several insights about the challenges and benefits of applying them in a hospital setting. Søraa and Fostervold (2021) studied the social interaction of robots through a case study of AGVs at a Norwegian hospital that interacts with patients, staff, and other machinery. The authors claimed that the application of AGVs in social structures such as hospitals should be considered in various dimensions of domestication. Practical domestication concerns the social acceptance of the robot such as using personalised local language or accent for robot voice. Symbolical domestication concerns the novel ways that robots can get the trust of humans such as when the errors on the robot’s function give a perception that they are not perfect and can make mistakes like a human. Cognitive domestication encompasses how human practices change through interaction with technology. Some situations make the staff grateful for using the AGVs, for example, for the transfer of waste. On the other hand, there are situations where AGVs may not be helpful, for example, when there is an emergency but the AGV does not free the path for the patient as it cannot recognise the urgency of the situation.

When comparing the application of AGVs in manufacturing and healthcare settings, it appears that AGVs in healthcare systems are mainly applied for material handling, reducing the processing time, and helping the staff. However, their application is less recognised and studied in healthcare. Søraa and Fostervold (2021) implied that the application of the AGVs can give an industrial-looking to the hospital which is not expected by patients. Another difference is that the layout of manufacturing factories is more adapted to passing internal vehicles while adapting the layout of hospitals, especially old buildings (Granlund & Wiktorsson, 2013). In the next sections, a simulation model will therefore be proposed to explore the dynamics of AGVs in CSDs and evaluate the benefits from their exploitation.

3. CSD processes and problem definition

In order to create a generic simulation model that could be applied for various CSD contexts, three different centres in the province of Quebec, Canada, were first visited and three reports concerning CSDs in the United States were studied. Based on this knowledge, it was possible to identify the type of medical instruments, equipment, processes, staff, average working hours, etc. associated with a typical CSD. All the CSDs studied appeared similar in terms of processes and equipment used. The main differences observed concerned having or not a cart washer and a case cart as well as the size of the CSD. The main objective in a CSD is to sterilise medical instruments and remove any risk of contamination, for their immediate reuse. After being used in operating rooms or clinics, instruments are transferred to the CSD. The schedule of sending instruments to the CSD can vary based on the scheduling and instructions of the hospital. For systems using case carts, instrument packages for each surgery are put in a case cart. For centres that use chariots, instruments from a group of surgeries completed around the same time are transferred together. The CSD staff is responsible for transporting the case carts or chariots and removing the instruments from them. To avoid contamination, soiled and sterilised instruments must be separated, thus a CSD includes three sections for decontamination, sterilisation, and storage. Pre-disinfection by hand, and disinfection in washing machines, are performed in the decontamination section, while packaging and sterilisation are performed in the sterilisation section. The main equipment in a CSD includes washers, drying cabinets, and autoclaves for sterilisation. There is other equipment such as ultrasonic washers and hydrogen peroxide sterilisers, which are used for a smaller number of instruments based on their characteristics (e.g., heat-sensitive instruments). The washers are loaded in the decontamination section and unloaded in the sterilisation section. There is an open space between the sections to return empty racks from sterilisation to decontamination. For centres that employ a case cart system, a cart washing machine is also required. For some instruments, such as flexible endoscopes, there might be special washers in some centres. The level of automation varies in centres. Examples of automation include cart washers, endoscope washers or scanner systems for the traceability of equipment. Figure 1 summarises the principal sterilisation cycle.

Figure 1.

Cycle of processes in a CSD.

Four main groups of medical instruments are typically used in CSDs: containers, trays, basins, and peel-packed individual instruments. Containers include the container box and a tray with instruments inside. Container boxes and basins go directly to the washers. Trays and instruments require a pre-disinfection process before going to the washer. Items are placed in racks for the washer. The capacity of a rack depends on the type of instrument and the design of the rack. It is usually equal to four basins or five containers and eight trays. A certain number of individual instruments (on average nine to ten) can be considered equivalent to one tray. After unloading the racks from the washer in the sterilisation section, some of the instruments must undergo a drying process. Container boxes and their trays are put together and packed. Other instruments are also packed, and finally a batch of instruments is placed in the autoclave for sterilisation. After the autoclave, the instruments are stored in a storage area. Based on a schedule, or an unscheduled request, CSD staff must transfer the stored instruments to be used in clinics or operating rooms.

There are many challenges in CSD work. The quality of sterilisation is essential as any mistake can affect patients. Some instruments are unique and might need to be used on the same day, so it is important to reduce waiting time. Workers must constantly load, unload, and move full racks of items. This highly physical work can lead to detrimental health effects over time, and is one of the reasons why many workers eventually quit their jobs. When a washing cycle finishes, workers must interrupt their tasks to unload and reload the washer, which leads to lack of concentration and productivity. During the periods of high volume, it is difficult for the staff to constantly stop what they are doing to transfer, load, and unload the racks. Investigating the impact of transfer automation in CSDs, by measuring its effect on KPIs, could yield new insights into how to overcome some of these challenges.

4. Methodology and model

According to Zhang (2018), DES is an appropriate technique to model complex healthcare systems as they usually consist of queuing networks, individual entities, and discrete events. As mentioned in section 3, processes that are common in all CSDs were identified by visits and studies. A generic simulation model was then created, based on all the information gathered, using the Arena simulation software (version 15). An external Excel file was also developed and linked to the simulation model so as to specify the schedule for instruments arriving into a CSD, in terms of type, number, period, and frequency (i.e., the model had to be flexible to be used for any input data). The cases of having or not a cart washer and a case cart were both modelled. This model was validated by questioning the staff of services provider who had the experience of working with many CSDs.

The main reason for making a generic model was the capacity to apply the model to various centres without requiring a time-consuming data collection phase (Di Mascolo & Gouin, 2013). According to Fletcher and Worthington (2009), there are four levels of genericity from the most generic to the most specific models. Level 1 concerns generic principal that covers multiple providers, multiple service chains, and multiple industries. An example of Level 1 is a generalised theoretical queuing model. Level 2 is a generic framework that covers multiple providers and service chains, but a single industry. Level 3, which is a generic model, covers multiple providers but a single service chain and a single industry. The generic model provided for CSDs is at this level. We visited multiple providers of the same service and created a model based on common processes while keeping the flexibility to adapt it to a specific centre. Level 4 is a specific model that is applicable for a single provider, service chain, and industry. The generic model applied to a specific centre, CHUL, is at this level.

Therefore, in the next step, a CSD (i.e., the CHUL, located in Quebec, Canada) was studied in greater detail, in order to gather specific information, adapt the generic model, and apply automation scenarios. Value stream mapping and process flows were created to illustrate resource and material flows. A time-and-motion study was also conducted for the different operations of the CSD unit. The work schedule and some statistics were extracted from the CSD database system. Rate of instrument arrivals, staff schedules (hours, number, specialities), process time, transfer time, the number of outputs (number of instruments leaving the system), the equipment capacity, batch numbers, and many other details of the work were gathered. Part of data was gathered by observation and time study (process times, arrival times, etc.), another part was gathered by asking for documentation (surgery schedules, autoclave capacity, etc.) and some by a combination of observations and interviews with the staff and their supervisor (role of staff, process flows, etc.). Data gathering covered a period of almost 4 months.

The simulation model used input data from the centre under study, illustrating the current (i.e., as-is) situation. Validity of the model was ensured by comparing the results with real data, statistical tests, and face-to-face meetings with the CSD supervisor. Performance factors were then defined to analyse the performance of the system. The main scenario concerning the use of AGVs for instrument transfer was finally tested, and the results compared to the as-is case where instruments are moved manually. Sensitivity analysis was performed to investigate how applying the AGVs could affect the increase in the capacity of the CSD. Next, a two-level, five factors full DOE was conducted with the following factors: an AGV in the decontamination area, an AGV in the packing area, staff in decontamination, staff in packing, and the schedule of instruments arriving from the clinic. The methodology is illustrated in Figure 2.

Figure 2.

Summary of the methodology of the research.

The input data, including arrival and work schedules, as well as process time distributions, are described in the following section.

4.1. Input data

Centre Hospitalier de l’Université Laval (CHUL), located in Quebec City, in the province of Quebec, Canada, was the CSD selected for the purpose of this study. This centre has the capacity to process around 3,000 instruments per week. Instruments from operating rooms arrive every hour during opening hours. There are three additional deliveries every day from clinics and three deliveries from emergency rooms. As a result, the distribution of arriving instruments in the simulation model was modelled as a nonstationary Poisson (Kelton et al., 2015), based on a Poisson distribution with a different mean (λ) for each arrival time, for each instrument. The distribution was also different for each day of the week. For example, the distribution of a container for the first arrival at 8 am on Mondays is POIS (2), while the distribution for the second arrival at 9 am is POIS (5). Table 1 illustrates the arrival distribution for the first day of the week as an example.

Table 1.

Arrival of instruments per hour (timeframe: Monday from 8 am to 9 pm).

Origin of use	Hour	Instrument type
		Container	Tray	Individual	Basin
Clinic	8 am	2	12	30	1
Operating Rooms	9 am	5	5	5	3
	10 am	16	18	18	10
	11 am	8	9	9	5
	12 pm	5	5	5	3
OR +clinic	1 pm	6	27	63	4
Operating Rooms	2 pm	6	7	7	4
	3 pm	10	11	11	6
	4 pm	5	5	5	3
	5 pm	5	5	5	3
OR +clinic	6 pm	2	22	58	1
Emergency Room	7 pm	0	8	15	0
	8 pm	0	8	15	0
	9 pm	0	8	15	0

There are three working shifts for staff: day, evening, and night. One employee in the sterilisation section starts earlier in the morning to test autoclaves before the day shift starts. The number of employees for the first, second, and third work shifts is 3, 5, and 1, respectively, in the decontamination section and 2, 4, 1, respectively, in the sterilisation section. There are two employees responsible for autoclaves (one for the day shift and one for the evening shift). A detailed staff schedule is registered in the model. Other input data concern the times required for all processes. Based on a time study, a distribution of each process time was established, as shown in Table 2.

Table 2.

Distribution of process times.

Process	Distribution (minutes)
Transfer to CSD	Uniform (13,16)
Pre-disinfection	Triangular (8,10,12)
Endoscope cleaning	Uniform (4,5)
Endoscope disinfection	Constant (12)
Packing small container	Normal (µ = 5, δ = 0.5)
Packing large container	Normal (µ = 9, δ = 0.8)
Packing trays	Normal (µ = 3, δ = 1.2)
Packing basin	Uniform (2,3)
Packing individuals	Uniform (1,2)
Sterilisation	Normal (µ = 90, δ = 0.75)
Drying	Triangular (5,8,10)
Washing basin	Constant (29)
Washing container	Constant (32)
Washing tray	Triangular (30,32,36)

After running the model with the input data, verification and validation methods were applied.

4.2. Verification and validation of the model

To verify the model, the details were added step by step and the model was verified at each step. Validation methods included creating output files to read the number of arrivals and items processed, a pause on the entities to follow their status when the model runs, as well as changing the elements to check the logic of changes on the results. The simulation model illustrating the current situation of the CSD studied was first run for one week (5 working days) with a 4-hour warm-up time and 10 replications. The number of outputs for each replication was recorded in detail for each instrument for each day of the week. These values were used to calculate the run number required to achieve a reliability level of ± 2.26 when building a 95% confidence interval. The formulation used was

$$N\left(m\right)={\left[\frac{s\left(m\right)t_{m-1,1-\alpha /2}}{\bar{X}\left(m\right)\epsilon }\right]}^2$$ (1)

where N(m) = the required number of simulation runs; $\bar{X}\left(m\right)$ = the sample mean from m replications; α = the level of significance (considered 95%); ε = an allowance for error (considered 0.05); and m = 10, t_9,0.025 = 2.262. The test showed that N ≥ 16. To ensure good results, 40 simulation runs were executed.

The average result of the simulation runs was compared to the average result gathered from the real system with a statistical t-test. The null hypothesis is H₀: μ₁ – µ₂ = 0, which implies that both samples represent the same systems and the alternative hypothesis is H₁: μ₁ – µ₂ ≠ 0 which implies that there is a significant difference between the real system and the simulation. The level of confidence is 95% which means that the parameter α = 0.05. If the P-value ≤ α then the difference between the means is statistically significant (Reject H₀). If the P-value > α then the difference between the means is not statistically significant (accept H₀). As Table 3 shows, according to the P-value of the t-test, the null hypothesis is acceptable. Thus, we can conclude that the simulation model is valid and represents the real system of this CSD.

Table 3.

Comparing the results of the real system, simulation, and t-test P-value (unit = number of instruments).

Instrument	State	Mon	Tue	Wed	Thu	Fri
Container	Real	65	66	66	82	68
	Simulation	68	69	59	82	68
	T-test p-value	0.68	0.53	0.63	0.54	0.51
Basin	Real	43	50	40	46	42
	Simulation	44	50	39	46	41
	T-test p-value	0.7	0.56	0.54	0.59	0.61
Individual	Real	241	345	337	250	341
	Simulation	219	332	328	243	328
	T-test p-value	0.55	0.6	0.57	0.58	0.56
Tray	Real	170	186	204	179	208
	Simulation	169	180	200	175	205
	T-test p-value	0.6	0.58	0.64	0.52	0.57

4.3. Introducing an AGV

As mentioned, the goal of this research is to investigate the application of AGVs to transfer, load, and unload racks into and from the washer as a solution for improving productivity in a CSD. Automated transfers seem to be beneficial not only for saving time but also for ensuring staff safety. An additional advantage is that the AGV can unload and transfer racks as soon as they are out of the washer, while staff must wait for them to cool down. The AGV is designed to communicate with the staff (receiving commands) as well as with the equipment. Its ability to load and unload as well as to receive signals from other machines and instruments, provides an opportunity to be expanded upon for future implementation of a smart system in CSDs. Figure 3 demonstrates the interactions of the AGV with the system.

Figure 3.

Interactions of the AGV with other system components.

The CSD studied planned to add two AGVs to their system, one in the decontamination section and one in the sterilisation section. In the decontamination section, the AGV would fetch the empty racks, transfer the full racks, and load the washer. The AGV would also be used for the automated return of clean, empty racks from the preparation and packaging area to the decontamination room. Once the rack is on the conveyor in the decontamination area, a signal would automatically be sent to the AGV. The rack identification information would then be sent to the AGV so it could pick up and deliver the rack to the appropriate workstation. In the sterilisation section, it would unload the racks from washers, transfer the racks to the packing area, and return the empty racks to the return point. The objective of testing this scenario is to assess the impact of adding new technology based on three performance indicators:

1. Turnaround: indicates the time interval from the instrument’s arrival in the decontamination section until it is ready to be delivered (Lin et al., 2008).

2. Queue time: refers to the time that an instrument or a group of instruments wait before a person or machine is available.

3. Work in process (WIP): implies the number of instruments that stay in the system at night, to be treated the following day.

Reducing these factors would be considered an improvement and increase the service level. The results of these indicators are presented in the next section.

5. Results

For the scenario of adding one AGV in decontamination and one in the packing area, three KPIs of turnaround, queue time, and number of WIPs are analysed, as well as the amount of distance and the number of loads that are substituted to AGVs. The results are presented in Section 5.1. Results of a sensitivity analysis about adding the arriving instruments up to 50% are presented in Section 5.2, and DOE for factors of AGVs, staff and arrival schedule is presented in Section 5.3.

5.1. Results of adding AGVs

5.1.1. Turnaround indicator

If we look at the turnaround indicator (Figure 4), when adding AGVs in the CSD, we observe a 26% reduction in turnaround time for containers, 23% for individuals, a 23% reduction for trays, and 24% for basins. These results indicate that applying AGVs improves the most important challenge in a CSD. As a result, special instruments such as containers and trays would be available faster for operating rooms. Furthermore, reducing the workload for staff allows them time to better concentrate on the process.

Figure 4.

Turnaround in manual and AGVs scenario.

5.1.2. Queue time indicator

When looking at the second performance indicator, it can be observed that adding AGVs may have a significant impact on the system (Figure 5). The most interesting change is a 58% decrease in the pre-disinfection queue, as a high number of transfers are done by the AGV, which frees the decontamination staff from this task. On the other hand, the queue time at the washing machines increases. The reason for this is that more instruments are being treated, which makes buffers of racks ready to go to the washer. The queue times of drying and packing show a decrease of 32% and 37%, respectively. The queue for peel packing of individuals does not show any change.

Figure 5.

Queue time in manual and AGVs scenario (unit = minutes).

5.1.3. WIP indicator

Let us recall that the third KPI, WIP, is the number of instruments that remain in the system at night. It is calculated by Equation (2):

$$\begin{array}{rl}& \mathrm{W}\mathrm{I}\mathrm{P}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{a}\mathrm{y}=\mathrm{A}\mathrm{r}\mathrm{r}\mathrm{i}\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{a}\mathrm{y}+\mathrm{W}\mathrm{I}\mathrm{P}\mathrm{o}\mathrm{f}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{i}\mathrm{o}\mathrm{u}\mathrm{s}\mathrm{d}\mathrm{a}\mathrm{y}\\ & -\mathrm{e}\mathrm{x}\mathrm{i}\mathrm{t}\mathrm{s}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{y}\\ & \end{array}$$ (2)

As there are some days where the workload is lighter, the WIP from one day to another may also vary. Figure 6 shows the WIP per day for each type of instrument.

Figure 6.

WIP in manual and AGVs scenario (unit is the number of instruments).

As the figures suggest, the WIP number for trays and individuals is higher compared to basins and containers. The reason is that the later arrivals of instruments are from the clinic and emergency room which do not usually use containers and basins (i.e., they are mainly used in operating rooms). Using the AGVs cause a significant reduction in the number of WIP, which is promising, not only because it is linked to the reduction of turnaround, but also because contaminated instruments should not remain untreated for a long period of time. The results show improvements in all three performance factors, with minimum change in other factors.

5.1.4. Ergonomic aspects

Results of the simulation model provide the effects on selected performance indicators. However, some human factors such as the long-term ergonomic benefits and satisfaction of staff are also important to consider. An interesting element to look at when introducing AGVs in a CSD is the number of wash loads per day. In our case, the average number of wash loads is 329 per week. This number implies that the CSD staff manipulates full racks 658 (329*2) times in a week, including 329 transfers and 329 loading/ unloading operations. There are also transfers for fetching and returning the empty racks. Table 4 shows the number and distance of transfers that can be saved per week by using the AGVs for transfers. Besides time saved, this result confirms the ergonomic benefits for the staff.

Table 4.

Distance and number of loads saved by using the AGVs per week for all resources in both areas.

Decontamination	Item	Distance (metre)	Loaded Transfer (number)	Loaded transfer (km)	Unloaded transfer(number)	Unloaded transfer(km)
	Basin container	14	132	1.8	396	5.5
	Tray	9	197	1.8	591	5.3
	Total –		329	3.6	987	10.8
Sterilisation	Basin container	10	132	1.3	396	3.9
	Tray	2	197	394	591	1.2
	Total	–	329	1.7	987	5.1

5.2. Load increase

We now investigate the effect of load increase or decrease in decontamination centres. This is of a particular importance when a hospital adds a clinic or an operating room service, which leads to an increase in the load of instruments to be decontaminated. Additionally, there are hospitals that outsource their decontamination services. In these standalone centres, the volume varies depending upon the current number of clients. Volume has also been impacted greatly during the past year, as the COVID-19 pandemic has forced hospitals to postpone non-urgent surgeries, drastically reducing their load of operations. Anticipating a future increase in the number of non-urgent surgeries, we used the simulation model to analyse the impact of increasing instrument loads in both the as-is and AGV scenarios. It is assumed that for each arriving distribution, the number of arriving instruments is augmented in 10% increments, up to 50 %. Besides the average turnaround time for each instrument, the maximum turnaround values are also calculated. This number indicates the longest time it takes to finish the sterilisation cycle for an instrument. Considering the maximum number provides a complementary analysis of the results. The results are presented in Figure 7 and Figure 8.

Figure 7.

Average and maximum turnaround times as-is vs AGVs (left: container, right: tray).

Figure 8.

Average and maximum turnaround times as-is vs AGVs (left: individual, right: basin).

Figure 7 (left) compares the average and maximum values of container turnaround time in the as-is and AGV scenarios for load increases from 10% to 50%. Without AGVs, average turnaround increases from 5 hours (current load) to 15.9 hours (50% increase). When the AGVs are applied, the average turnaround time changes from 3.7 hours (current load) to 10.4 hours (50% increase). In both scenarios, turnaround showed a linear increase; however, when loads are increased by 50%, the application of AVGs reduces the turnaround time by 5 hours compared to the as-is scenario. In order to get a better understanding of the effect of increasing workload, the as-is maximum values are also collected. The maximum value of the container turnaround time ranges from 11.8 hours to 26 hours, while AGVs values range from 11.5 hours to 15 hours. Maximum values are close at the current load, but as the load increases, the value increases drastically in the as-is situation, and less sharply when applying AGVs. The maximum values reinforce the idea that AGV application facilitates load increase. Even with a load increase of just 10%, the average turnaround is still less than the average without AGVs. This demonstrates that applying AGVs provides the opportunity to add 10% to the load while maintaining the current processing rate.

For the trays (Figure 7, right), The as-is average turnaround ranges from 4.9 hours (current load) to 15.3 hours (50% increase). The values are 3.8 hours (current load) to 9.2 hours (50% increase) when applying AGVs. The maximum values range from 9.8 to 19 hours for as-is and 7.8 to 10.5 hours when AGVs are applied. Again, the increase in turnaround is minimised when AGVs are applied. Similar to the results for the container, AGVs make it possible to increase the load by 10% without changing the processing rate compared to the as-is situation. Another interesting observation is that the trays are less sensitive to load increases compared to the containers. This makes sense as the sterilisation process for containers is more complicated.

For the individuals (Figure 8, left), as-is average turnaround time varies from 5.5 hours (current load) to 17 hours (50% increase). The values are 4.1 hours (current load) to 9.2 hours (50% increase) when AGVs are applied. Maximum values range from 12.3 to 20 hours for as-is and 8.8 to 11.5 hours when AVGs are applied.

For the basins (Figure 8, right), as-is average turnaround varies from 3.7 hours (current load) to 5.7 hours (50% increase). The values are 2.8 hours to 6.3 hours when applying AGVs. Maximum values range from 5 to 8 hours for as-is and 4.5 to 7.2 hours when AGVs are applied. Once the load increase reaches 20%, the average AGV turnaround time becomes greater than the as-is average. As basins are in the lowest priority, and the difference between as-is and AGVs is not significant, this is considered an acceptable result. The good results obtained for other instruments when applying AGVs also compensate for the results for the basin.

5.3. Design of experiments

The purpose of this section is to analyse the effect of some factors on the KPIs (turnaround and WIP) so as to test various scenarios in a systematic and comprehensive manner. Five factors in two levels are analysed, making $2^5=32$ experiments. As shown in Table 5, the factors include applying an AGV in decontamination, applying an AGV in the packing area, adding one worker in decontamination, adding one worker in packing, and changing the schedule for instruments arriving from the clinics.

Table 5.

Factors for DOE.

NO	Factors	Value (-)	Value (+)
1	AGV indecontamination area	AGV used in decontamination	No AGV in decontamination
2	AGV in Packing area	AGV used in packing	No AGV in packing
3	Decontamination staff	Current number of staff	Add one worker
4	Packing staff	Current number of staff	Add one worker
5	Clinic arrivals	Current schedule	New schedule

Factors are chosen based on process analysis, interviews with the supervisor of the studied CSD, and the feasibility of the factor. Factors of AGVs in the decontamination and packing areas are considered to analyse the effect of this technology, while highlighting which one seems more effective. The third and fourth factors concern adding one worker in both the decontamination (factor 3) and packing areas (factor 4). Factor 5 tests the scenario of modifying deliveries from clinics. Instruments are delivered from clinics three times daily, arriving at 8 am, 1 pm, and 6 pm. The other alternative considered is to divide the second delivery in two, i.e., one at 11 am and another one at 3 pm. The 32 experiments were tested as scenarios in the simulation model. The results were analysed in Minitab and are described in the following sections.

6. Main effects plot

Results show that the most effective factor is adding an AGV in the decontamination area (Figure 9).

Figure 9.

Main effects plot.

The vertical axis represents the mean of turnaround and the horizontal axis represents the level of factors (−1 if the AGV is not applied and 1 otherwise). The main effects plot shows that applying an AGV in the decontamination area has the most significant effect on turnaround. The mean turnaround for runs without the AGV equals 4.5 hours, while the application of an AVG in decontamination reduces the mean to 3.5 hours. The second most effective factor is applying an AGV in the packing area, but the improvement in turnaround is less compared to the decontamination AGV. The effects of adding decontamination staff and packing staff are similar and weaker than the effect of adding AGVs. Changing the clinic schedule was not shown to reduce the treatment time.

The interaction plot (Figure 10) denotes that there is some interaction between every factor, significant for some pairs and minor for others. AGVs in the decontamination and packing areas reinforce the effect, as expected. Interaction between the AGV in decontamination and the packing worker is more significant than its interaction with the decontamination worker, while the interaction between packing AGV and decontamination or packing staff is on the same level. The interaction plot further implies that interaction between changing the clinic schedule and other factors is unremarkable. The interaction plot reinforces the results of comparable DOE scenarios and provides useful information for decision-makers.

Figure 10.

Interaction plot.

In Table 6, the 32 experiments are listed in columns which are sorted first from lower to higher turnaround and then by ascending WIP. The results show that the best scenario is when applying AGVs to both areas, and adding one decontamination worker and one worker in the packing area while changing clinic arrival. In this scenario, the average turnaround time reduces from 4.7 hours to 2.5 hours and WIP is reduced from 88 to 45. Saving 2 hours in the decontamination process and reducing the WIP by 50%, while considering the number of instruments treated every day, denotes a noticeable improvement (scenario 1). The second-best scenario, which is very similar to the first scenario, is the same combination but without changing the clinic schedule (scenario 2). The next-best scenarios are the combination of two AGVs and adding one worker in the decontamination or packing area (scenarios 3,4,5, and 6). The scenario of adding two AGVs without adding any staff remains a good scenario (scenario 7). Adding two AGVs, as discussed earlier, reduces the average turnaround from 4.7 hours to 3.6 hours with the current clinic schedule, and to 3.8 hours with the new schedule. We can conclude that in case there was a need to reduce treatment time (for instance, when facing an increased workload), it is better to add two workers and two AGVs, while changing the clinic schedule. However, adding two AGVs without adding other options still has a significant improvement on turnaround.

Table 6.

Scenarios tested in the simulation.

NO	AGV decontamination	AGV packing	Staff decontamination	Staff packing	Change in clinic schedule	Turn around average	WIP
1	+	+	+	+	+	2.5	45
2	+	+	+	+	-	2.7	44
3	+	+	+	-	+	2.8	44
4	+	+	-	+	+	3.1	45
5	+	+	-	+	-	3.2	44
6	+	+	+	-	-	3.2	46
7	+	+	-	-	-	3.6	51
8	+	-	+	+	-	3.7	55
9	+	-	-	+	-	3.8	54
10	+	+	-	-	+	3.8	55
11	-	+	+	+	-	3.9	61
12	+	-	+	+	+	4	64
13	-	+	+	-	-	4	67
14	-	+	+	+	+	4.1	74
15	+	-	-	+	+	4.1	75
16	+	-	+	-	+	4.2	61
17	+	-	-	-	-	4.2	69
18	-	+	-	+	+	4.2	88
19	-	+	+	-	+	4.3	58
20	-	-	+	+	+	4.4	83
21	-	+	-	+	-	4.5	65
22	-	-	+	+	-	4.5	72
23	+	-	-	-	+	4.5	82
24	+	-	+	-	-	4.6	65
25	-	+	-	-	-	4.6	74
26	-	+	-	-	+	4.7	71
27	-	-	+	-	-	4.7	80
28	-	-	-	+	-	4.7	82
29	-	-	-	-	-	4.7	88
30	-	-	+	-	+	4.8	75
31	-	-	-	+	+	4.9	89
32	-	-	-	-	+	5.2	90

The next two scenarios with good results include adding an AGV in the decontamination area, in combination with adding two workers (scenario 8) and adding a decontamination AGV in combination with one packing worker (scenario 9). It means in case of limitations on the purchase of AGVs, it is better to add an AGV in the decontamination area and it would be good to consider adding staff in packing to balance the rate of work in these two areas. The impact of adding two AGVs and changing the clinic arrivals (scenario 10) is similar to adding only one AGV and one staff member as it shows that changing the clinic schedule does not make for a good combination in this scenario. The next-best scenario is to add an AGV in the packing area and one worker in each area (scenario 11). Adding an AGV in decontamination and two staff members, while changing the clinic schedule, is in the rank of 12 in the ascending list. If adding only one worker and an AGV are added in the packing area, it is better to add one decontamination worker to maintain balance (scenario 13). Another noticeable result is if only one AGV is purchased, the decontamination AGV would be more efficient (scenarios 15 to 26). Scenarios 26 and 27 consider adding only one staff member to the as-is situation, with the result being only slightly different from as-is (scenario 28). Finally, the results show that changing the clinic schedule without adding AGVs or staff increases turnaround time and is not recommended.

7. Discussion and managerial insights

The results of this study confirmed the reduction in waiting time, turnaround time, and WIP when applying AGVs in a CSD. Moreover, the simulation revealed the distance of transfer and amount of loading and unloading that could be saved via transfer automation, confirming the ergonomic benefits of applying AGVs (Benzidia et al., 2019; Chikul et al., 2017). In the centre studied (CHUL), the benefits also included reducing chaos in the system that happens because workers face constant distractions when attempting to manage the racks.

Besides the benefits and many advantages of AGV implementation in hospitals, there are some challenges. Investment costs for buying and implementing AGVs are high. However, when considering the costs and benefits, long-term and intangible benefits should be taken into account (Chikul et al., 2017). While the managers must be convinced to make the initial investment, the acceptance and involvement of working staff is another issue. Staff might be resistant to new technology, or afraid of job loss; however, the collaboration of human operators is essential for the success of automation plans. Another barrier to consider concerns the physical structure of the hospital and limitations in layout, especially in older buildings (Benzidia et al., 2019). In the case of the CHUL, lack of staff and the high number of transfers with loaded racks motivated workers to accept the application of AGVs.

For the procedural products and services provider, it was a first experience in applying simulation modelling as a decision support tool for analysing a system and its productivity. They found it a very useful tool to be applied in the future to facilitate their decision-making process and their communication with their clients. After the research, they started using the results of the study to simulate various scenarios such as adding different numbers of AGVs to the decontamination centres and analyse the changes in KPIs. They then presented the results to their clients to show how these scenarios could affect their system and which scenario would be more suitable for them. Managers also started using the simulation model as a Research and development (R&D) tool for generating new ideas for product development. For CHUL, the simulation tool appeared as an efficient way to analyse and test various scenarios before making changes in their real system as these changes may be very controversial and time-consuming. Also, as it is sometimes difficult to convince the workers to adapt to a new routine and to measure the effect of a change on the system productivity, the simulation model was very well accepted by the manager of the CSD. Concerning the implementation of the AGVs in the CHUL’s CSD, a challenging phase was the time needed to completely instal and apply the new AGVs in full efficiency. However, training sessions and user-friendly design for AGVs contributed to turn this project into a success. After getting used to the AGVs, the staff even found it difficult to get back to the previous system when AGVs were out of order because of technical issues. The lesson learnt from this study revealed that reducing internal transfer has a significant effect on the productivity of the CSD system. Another noticeable point to be considered is that only adding more staff or AGVs is not the best method to improve the system and it might even cause undesirable results such as increase in the total processing time and number of WIP.

In theory, there is an advanced perspective about automation in healthcare systems, such as smart hospitals and applications of industry 4.0. However, in practice, there are many barriers to implementing automation solutions, especially in second-level services such as sterilisation that does not have direct contact with patients. The generic simulation model developed in this study therefore becomes a valuable decision support tool to analyse various aspects of applying AGVs in CSDs, and to measure the real value such a technology could bring to the system.

8. Conclusions

This study presents a detailed simulation model to analyse the effects of using automated transportation in a CSD. The methodology is to test the scenario of adding AGVs and compare the results. This is combined with sensitivity analysis and DOE, to provide more comprehensive analysis of effects. Based on turnaround, queue time and WIP measures, results show a significant reduction in all indicators, except the queue time for washers. According to the load increase, applying AGVs enables the system to increase capacity by around 20%, without a significant increase in treatment time. Moreover, analysis of DOE implies that the best scenario is to apply one AGV in decontamination and one in the packing area or, in case of a limitation on the purchase AGVs, applying one AGV in the decontamination area would be most efficient. The main contribution of this research is to provide a simulation model for automation in a CSD and using AGVs in this system. An additional contribution for the CSD is a tool that can be used to analyse the performance of a system and test various scenarios. The contribution to the industry is that the model can be modified and used as a decision-making tool by demonstrating the impact of using AGVs in a given system, the number of AGVs that would be needed, and the potential benefits expected for patients. Moreover, the potential increase in capacity for the centre by using AGVs in emergent situations, such as the Covid-19 pandemic, could be considered in the final decision.

Although the results are promising, there are some limitations to this study. The potential for unexpected events such as machine failure or missing staff were not considered. Secondly, since the introduction of new technologies in the workplace comes with its challenges, aspects such as learning curves or reluctance to change were not analysed. Moreover, this research does not cover the economic analysis of investment return for hospitals implementing AGVs. Although this aspect would be interesting for further study, hospitals might invest in this technology for the long-term ergonomic and safety benefits rather than a short-term return on investment.

The study could be extended to include other aspects of automation and connectivity in CSDs, to implement industry 4.0 infrastructures. According to Molino et al. (2020), industry 4.0 is based on concepts such as cyber-physical systems, the internet of things, digitalisation, smart technologies, automation, and robotics. In this context, autonomous systems act as enablers for the core concepts and vision of industry 4.0. As AGVs are integrated with other machines in the system, the combination of RFID, information systems, and the internet of things would provide infrastructures for implementing industry 4.0 in CSDs.

Disclosure statement

No potential conflict of interest was reported by the author(s).