The Blood Service at the Belgian Red Cross-Flanders is responsible for blood collection in Flanders (Belgium). One of their missions is guaranteeing a constant and sufficient supply of safe blood products. This is a critical public health need, since the blood products can save lives of victims from traffic accidents or in the event of major blood losses in hospitalized patients. The main objective of this project is optimizing the operations flow in donor centers, in such a way that the waiting time for donors is minimized and that the donor center occupation or productivity is maximized. In this case study, the flow of three types of donations is investigated. Blood and plasma are donated in all donor centers (<i>i.e. </i> 11 donor centers in Flanders), while blood platelets are collected in only six donor centers. Based on data collected from the 11 donor centers in Flanders, a simplified simulation model was developed, which can be used to optimize the operations flow based on the expected number of donors and their moment of arrival at the donor center. The simulation model is built in Enterprise Dynamics 9.0 simulation software. The input data in the model are data that have been collected in collaboration with the Belgian Red Cross-Flanders. Different scenarios will be analyzed to gain insight in the impact of small changes in the input parameters on the performance of the flow. In this paper, a gap analysis is conducted to identify extra data needs. With these additional data, a more detailed model can be constructed to test the scenarios, and a dynamic planning tool will be developed to rely on when setting up the capacity of the donor center in order to find a scenario with the most optimized flow.
The Belgian Red Cross-Flanders, part of the international Red Cross and Red Crescent Movement, is an independent volunteer organization with a threefold mission which consists of stimulating self-reliance, providing assistance in the event of emergencies and excelling in blood supply. The department Blood Service is responsible for ensuring a continuous and sufficient supply of safe blood products that will be distributed to the majority of hospitals in Flanders [
The availability of sufficient quantities of safe, high-quality blood is a critical public health need as it is helping doctors to save lives. Matching supply and demand for blood, however, is not straightforward because of several external factors. Seasonality, regional trends and other unforeseen circumstances, like flu epidemics or crises, may impact the availability and need for blood. This matching process is further complicated by the limited shelf life of blood products. As a consequence, a constant supply of blood is needed as well as a good inventory policy in order to reduce blood shortages, which may cause increased mortality rates and hence high costs for society [
In Belgium, 70% of the Flemish people will need a blood transfusion during his/her lifetime, while only 3% are donating. Each year 350,000 bags of blood are donated in Flanders. In the production laboratory, these blood bags are transformed into 600,000 deliverable high-quality blood products which are transferred to the majority of hospitals in Flanders [
The Blood Service at the Red Cross-Flanders aims to improve the efficiency of their donor flow. External factors and variability complicate the organization of donors and resources at donor centers. As a consequence, it is difficult to ensure an optimal utilization of the resources, which may result in queues (i.e. waiting lines) for donors and under- or overutilization of the resources at certain moments of the day. The Blood Service wants to improve their operations with focus on the comfort and satisfaction of donors.
To gain insights in the performance (e.g. queueing problems) of the donor centers, an operations analysis was conducted about the current situation. This in order to build a simulation model. Simulation enables to model the donor flow and test several scenarios based on changing the input parameters, without experimenting or spending any money on the work floor. In this paper, we present the efforts that were done to develop a first simulation model that visualized the donor flow in a reference donor center.
The outcome of the operations analysis indicated that various parameters are influencing the performance of the donor flow. Seasonality in the arrival pattern of donors, the location of the donor center, launching a campaign, the capacity of the resources, the ratio of new and experienced donors, etc. are used as input parameters in a simulation model. The simulation software Enterprise Dynamics 9.0 was used to build a first, simplified model that represents the current donor flow. As reference donor centers for the model, the data from donor center Ghent and Geel were used as input parameters. Based on this model, a gap analysis was conducted to identify additional data needs to build a more detailed simulation model. The final objective of our study is to develop a dynamic planning tool on which the Blood Service can rely on when setting up the capacity of the donor center in terms of both personnel and beds. In the ideal case it would be possible to have a dashboard that can show the optimal organization of resources related to the parameters of the expected situation.
Building the simulation model under the strategy of lean thinking can help to eliminate the waiting times, have a more efficient utilization of the resources, and an efficient use of space. Healthcare quality and costs depend on delivery processes, which often include unnecessary or inappropriate steps that do not contribute to the value of patient care [
The remainder of this paper is organized as follows. In the next section, the donor flow at the Red Cross-Flanders is described. Section 2 focuses on the use of simulation as a tool to optimize the donor flow. An operations analysis was done by the Blood Service, since they were observing queueing problems in the flow. Section 3 provides the process of building the first simplified simulation model and implementing the available data. It is crucial to check for model validation and verification since simulation models are simplifications of reality. The results are provided in Section 4. This section also conducts a gap analysis to identify the missing data and provides recommendations in order to build the dashboard and test different scenarios in the second phase of the project. The conclusion and suggestions for future work are presented in Section 6.
The operations or donor flow is divided into five steps. The sequence of these steps is similar in the 11 donor centers of the Red Cross-Flanders, but they differ in the types of donations that can be made. In the donor centers, mainly three types of donations are collected. Blood and plasma can be donated in all donor centers, whereas blood platelets in only six donor centers. In this case study, a simplified simulation model is built by analyzing two reference donor centers. In Ghent, the three types of donations are collected, while the donor flow in Geel is not equipped to donate blood platelets.
In this paragraph, the different steps of the donor flow, displayed in
The healthcare sector is becoming more competitive. Hospitals are aiming at delivering efficient and effective healthcare which requires high quality medical care. “In healthcare, efficiency means a better allocation of scarce resources which will result in a higher overall quality of healthcare” [
The main objective of this project is improving the operational flow in a donor center by using simulation in order to find the optimal flow in different situations. An efficient donor flow can be obtained by optimizing the utilization rate of the resources and decreasing the waiting times for donors which will result in the collection of safe, qualitative blood products for saving patient lives. However, the results of the operations analysis (see Section 1.1) showed that various parameters are influencing the performance of the donor flow. A simulation model enables the Blood Service to change the input parameters and answer what-if questions to find the optimal flow in each situation. Moreover, the visualization of the donor flow, which can also be displayed in 3D, is a strong advantage when using simulation to convince medical experts of the benefits.
At first, simplified simulation model, corresponding to the current donor flow, is built in cooperation with employees and medical experts at the Blood Service. Involvement of the personnel in the healthcare sector is vital so that the model under development can be validated based on their understanding of the operations flow [
Keep in mind that simulation models are simplifications and that sometimes it might be difficult to guarantee their validity. It is crucial that the model reflects the behavior of the real donor flow. Model validation and verification (see Section 3.4) will check for this. In order to build a valid simulation model, data should be implemented in the model. The Blood Service is collecting data in a blood information system (CTS Serveur software package by Haemonetics). By applying goodness-of-fit tests, probability distributions are selected that best fit the data. However, lack of data is a well-known problem in healthcare. Additional data should be collected to obtain accurate simulation results in order to develop the dynamic planning tool.
The core of the model to build is to be found in the operations layout (summarized in
The Blood Service is recording data about the donor flow in the blood information system (i.e. CTS Serveur software package). These data are converted to useful information reports using a business intelligence database (cognos database), in which the number of arriving donors per hour and per day can be retrieved, as well as the time the donors are spending in the donor flow (i.e. end-to-end donation time). However, only two time registrations are shown in this database: (1) the moment at which the medical questionnaire is printed at the registration desk, which is assumed to be the start of the donor flow (step 1), and (2) the start of the actual donation (step 4), which is recorded at the moment of scanning the post-donation card in the blood information system during the first minutes of the donation. This second registration moment is not pre-defined and hence not accurate. For the donation of blood, this registration should happen prior to the start of the actual donation, while plasma and platelets donations already start and the donation is registered during its first minutes. The remaining duration of the actual donation is estimated by medical experts to be 6, 40 and 75 minutes for the donation of blood, plasma and platelets respectively. Furthermore, the blood information system provides important data on the busiest times (i.e. peak hours) and least congested times (i.e. trough hours) of the day. At these times, the arrival pattern of donors will be different. Changing this parameter in the simulation model will affect the performance of the donor flow. The objective of this project is to find the most optimal flow, such that the Blood Service is capable of adapting the input parameters in real-time to the arrivals of donors by optimizing the organization (utilization rate of the resources) of the donor center. As a result, the waiting times should decrease.
As mentioned in paragraph 1.1, Ghent and Geel are serving as the two reference donor centers in this study. Data from available manual data collections in the 11 donor centers can be used to be implemented in the simulation model. Some data, however, are center-specific, such as arrival times of the donors. In the first, simplified simulation model, the exact arrival times were implemented in order to mimic the behavior of the real donor flow. Later on, probability distributions will be selected that best fit the arrival data. Ghent is one of the largest cities in Flanders. The donor center in Ghent is equipped to collect the three types of donations. In 2015, 24,304 donations were collected, which is higher than in an average Flemish donor center. They collected 18.6% of the blood donations, 13.8% of the plasma donations and 22.5% of the platelets donations across all donor centers in Flanders. The donor center in Ghent was also selected because it serves as a reference center in a pilot study, approaching this project. The donor center in Geel is representative for all donor centers that are not collecting platelets donations. In 2015, a total of 7,804 donations are collected, which can be split up to 4.5% of the blood donations and 6.6% of the plasma donations. It is assumed that the donor flow works similarly in the 11 donor centers in Flanders, except for the type of donations they are collecting.
The available data at the Red Cross-Flanders, collected either by the blood information system or manually by the employees, are implemented in a first, simplified simulation model. These data represent times of the day and should be converted to times expressed in minutes. For example, 9:05 is equivalent to 5 minutes, because we only take into account the opening hours (i.e. 9 am till 7.30 pm) of the donor center. Another example, 15:45 can be expressed by 405 minutes (15 *60 + 45 − 9 *60). To capture variability in the data, AutoFit is used to find the probability distributions that best fit the observed data. AutoFit is a goodness-of-fit test integrated in Enterprise Dynamics, which determines whether a certain data set can be represented by a certain distribution based on shape, mean and standard deviation. By default, the Anderson-Darling test is used [
The simulation model, as shown in
Donor flow | Datapoints | Distribution (minutes) |
---|---|---|
Registration desk | 53 | Erlang (2.55, 4.00) |
Medical questionnaire + waiting | 26 | Weibull (5.06, 1.80) |
Doctor | 574 | Weibull (3.91, 2.50) |
(Queue before plasma) | 22 | Erlang (3.79, 2.00) |
(Queue before platelets) | 14 | Erlang (6.31, 1.00) |
Actual donation blood | 0 | Normal (6.00,1.00) (assumption) |
Actual donation plasma | 42 | Logistic (42.82, 9.30) |
Actual donation platelets | 28 | Normal (77.91, 66.59) |
The basic atoms used in this model are Source, Queue, (Multi-) Server and Sink atoms. The simulation model can easily be adapted to a flow without platelets donations, as these are only collected in six donor centers.
The donors arrive at the donor center according to a random arrival pattern. Data regarding arrivals at the donor center in Ghent were used and are converted to data lines in the ArrivalList atom. This atom creates donors based on the pre-defined list, which contains data on the arrival time, a name, the arrival quantity and a channel through which the incoming donors are sent by using three labels. As such, a distinction can be made between the three types of donations. When entering the donor center, the donor may have to wait in a queue before being served by the person at the registration desk (i.e. entering queue). At the registration desk, which is represented by a Server atom, one employee is occupied with registering all arriving donors. On average, this activity takes 2.55 minutes and AutoFit suggests an Erlang distribution (see
In the first model, Queue atoms are used to model waiting lines. However, the Red Cross-Flanders collected data on the time that donors are spending in a queue before being assigned to a bed in step 4. Therefore a second simulation model has been developed in which the queues are replaced by Multi-service atoms, representing the observed waiting times. By comparing the two simulation models and the observed data, one can check whether the first simulation model is valid (see paragraph 3.4). The observed data suggest an Erlang distribution with an average waiting time of 3.79 minutes for plasma and 6.31 minutes for platelets. For blood donations, there are no data available on queueing. The first simulation model suggests that this queue is always empty. Hence no Multi-service atom has been included in the flow of donors in the second simulation model (see
When developing a simulation model, it is crucial to do a validation and verification. The Red Cross-Flanders will make important decisions based on the results of the model, which will affect the employees and the donors. Model verification is often defined as “ensuring that the model and the implementation are correct”, while model validation determines whether “the model within its domain of applicability possesses a satisfactory range of accuracy consistent with the intended application of the model” [
The simulation model is verified by investigating whether the model is built in the right way. The models in
Model validation is important to ensure that the right model is built to meet the intended purpose of the project. Three tests were executed to check for model validation.
In the first test, the ArrivalList atom was observed to make sure that the donors arrive at the donor center according to the observed data. The second validation method monitors the Sink atoms according to the principle “what comes in, goes out”. The 112 incoming donors also leave the model, labeled by the three types of donation. By using the ArrivalList atom, label values were assigned to the incoming donors referring to the type of donation. A third test is performed to check whether the model corresponds to reality by comparing the average end-to-end donation time (i.e. the time a donor spends in the donor flow).
This section presents the results obtained by running the first simulation model. Based on
A first method for interpreting the results is tracking the information on the atoms while the model is running in order to check if the model is working logically (i.e.
Type of donation | Reality (observed data) | First model (queue atom before actual donation) | Second model (multi-service atom before actual donation) |
---|---|---|---|
Blood | 17 minutes (estimation of 6 minutes) | 21.2 minutes | 21.3 minutes |
Plasma | 57.7 minutes | 60.7 minutes | 65.3 minutes |
Platelets | 95 minutes | 91.6 minutes | 98.5 minutes |
model verification). The ArrivalList atom represents the number of arriving donors and the Sink atoms mention the number of donors, separated by the type of donation, who leave the flow and have a drink in the donor corner. Furthermore, the Server atoms indicate the utilization rate of the servers, which is the ratio of the busy time and the total time. For example, the doctor is examining donors for 60.87% on average of his/her time during a simulation run of 12 hours (see
Results can also be obtained by connecting monitors (e.g. bar graphs, status pies, etc.) to atoms. As shown in
Thirdly, results can be interpreted by retrieving a summary report (see
times for donating blood or platelets. In the observed data, however, the waiting times were longer, as measured in the second simulation model.
The three above mentioned methods are particularly useful to display results directly during the simulation run, but these are less appropriate when making decisions based on the results of the model. These techniques are primarily utilized for building and testing the model (i.e. model verification and validation in section 4.4), while the fourth measuring technique, experimentation, is used later on in the process when the model is more reliable. In this case, an experiment will be conducted in which performance measures (e.g. maximum waiting time in each step, maximum utilization of beds, etc.) can be defined. The experiment was set to execute ten separate runs of 12 hours.
A donor is spending on average 52 seconds in the entering queue. In peak hours, however, the waiting time can increase to 6 minutes on average for the ten observations. In the donor centers, five spots are arranged for completing the medical questionnaire, while only 1.3 of the spots are used on average. When observing the queues before being assigned to a bed, donors are spending on average 1.3 minutes in the waiting line for donating plasma, while they do not have to wait to donate blood or platelets. During peak hours, donors are waiting maximally 13.3 minutes for plasma donations. In the second simulation model the donors are forced to wait for some time (see
Atom | Average/Maximum | St. Deviation | Lower Bound (95%) | Upper Bound (95%) | Minimum | Maximum |
---|---|---|---|---|---|---|
Queue Registration (s/min) | 52 s/6.2min | 2 s/23s | 51 s/5.9min | 53 s/6.5min | 48 s/5.6min | 54 s/6.9min |
MQ + Waiting (spots) | 1.3/5 | 0.01 | 1.3 | 1.3 | 1.3 | 1.3 |
Queue Blood (min) | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
Donation Blood (beds) | 0.3/2.1 | 0.01 | 0.3 | 0.3 | 0.3 | 0.3/3 |
Queue Plasma (min) | 1.3 min/13.3min | 36 s/4.1min | 51 s/10.4min | 1.7 min/16.2min | 24 s/6.5min | 2.4 min/19.6min |
Donation Plasma (beds) | 3.8/7 | 0.11 | 3.7 | 3.8 | 3.6 | 3.9/7 |
Queue Platelets (min) | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
Donation Platelets (beds) | 1.9/4 | 0.05 | 1.8 | 1.9 | 1.8 | 1.9/4 |
personnel occupation (i.e. human resource team). Finally, the actual donation time differs between the three types of donations, as well as the number of beds. The blood donation atom has a capacity of three beds and an average content of 0.28 beds utilized. Plasma donation has a capacity of seven beds in the donor center, while the average content is only 3.8 beds utilized. For the donation of platelets, four beds are available in the donor center from which 1.9 beds are utilized. This result illustrates that, on average, there is underutilization of the beds. Although, in peak hours, the beds are maximally utilized.
The results of the experiments show that simulation is useful for a better understanding of the current donor flow. Moreover, the visualization of the flow can be a useful tool to convince the medical experts of the improvements that can be made by adapting the capacity of the donor center to the expected number of arriving donors, and hence decreasing the waiting time. By analyzing the results obtained in the two simplified models, recommendations can be made. The purpose of these recommendations is identifying the gaps of information: in order to build a more detailed simulation model for testing different scenarios and optimizing the donor flow, more data will be needed. In this gap analysis, it will become clear which additional data should be collected or extracted from available reports within the Red Cross-Flanders. When more data are available, suggestions for improvements can be made in several realistic scenarios by changing the input parameters.
Donors arriving at the donor center wait in the entering queue. However, some of them have an appointment for their donation and do not have to wait in this queue. When donating plasma or platelets, it is recommended to make an appointment because these donations have a longer duration. To implement two queues (i.e. for donors with/without appointment) in the model, data on the percentage of donors that have an appointment should be collected. This information will also be very useful for the sche- duling of personnel as it creates a better knowledge on the inflow of donors to the donor centers. However, as donating blood is a voluntary activity which can be done when it suits the donor best, one should be careful to instigate donors to make an appointment. Another recommendation could be distinguishing between new and experienced donors. A new donor is spending more time at the registration desk than an experienced donor, because he/she receives more information and asks questions about the donor flow. Analyzing the end-to-end donation time of these two types of donors will indicate whether it is advantageous to make separate queues. It is important to collect new data for the time spent at the registration desk in order to build the more detailed model, because the currently observed data are not very reliable as they represent the time that the medical questionnaire is printed, which is often at the end of the registration step.
A third recommendation to the simplified model is splitting the Multi-Service atom “MQ + waiting”, which consists of answering the medical questionnaire and the waiting time in front of the medical examination at the doctor’s office. In the current model, these two activities are merged together because the available data include the time at which the donor starts filling in the questionnaire and the time at which the donor enters the doctor’s office. This creates a bias in the waiting time in our model. The Multi-Service atom for completing the questionnaire should therefore be followed by a Queue atom, and data should be collected on the duration for completing the questionnaire. In this way, the end-to-end donation time could be reduced and a potential bottleneck could be identified.
The number of personnel varies during the day because of lunch breaks, phone calls, administrative tasks, bathroom breaks, etc. As a consequence, the employees are not available to serve the donors for 100% of their time. The utilization rate of the resources in the first model is not accurate because the available staff has been ignored. In the more detailed model, the time that employees are not working with donors directly (i.e. down time) should be included in the model. In Enterprise Dynamics, additional staff (Human Resource Team) can be added to analyze the impact of changing the capacity of the personnel. In the actual donation step, the nurses and assistants should be assigned to the right beds and donors as this might impact the queue in front of the actual donation. For example, when a donation bed is available, the donor goes to this bed in the model. In reality, however, a nurse is needed to assist the donor to the bed and to prepare the bed for the next donation. Furthermore, it is recommended to collect additional data in the actual donation step. The durations for blood, plasma and platelets donations are ending when the donation set is removed from the donor. Hence, no data are available on the times that the donor leaves the bed (and the bed becomes available for preparing it for the next donation), which is needed to maximize the utilization of the beds.
At the doctor’s office, the donor selection includes an eligibility assessment involving a medical questionnaire, an interview and physical examination (pulse, weight and blood pressure). Recently, a hemoglobin screening was implemented for new donors and donors who had a low hemoglobin level the last time they donated. The Server atom at the doctor’s office in the simplified model contains data in which the hemoglobin screening is included versus not included. In order to obtain a reliable model, the actual percentage of hemoglobin screenings should be compared with the ratio of screenings in the observed data. Another recommendation when building the medical examination atom is to make a connection to the postponed donors Sink atom. These donors do not fulfill the health requirements and are refused to continue the donation (i.e. they leave the flow after step 3). It is important to identify the postponed donors, because they increase the end-to-end donation time of all donors as they also utilize the donor flow until the medical examination. To be able to dispose the right amount of donors, data should be collected to indicate the disposed percentage in the atom.
By collecting the additional data and implementing them in the model, the model will become more reliable and potential bottlenecks may be identified. Different scenarios, such as building two queues (appointments versus no appointments), changing the capacity of the resources, etc. can be tested and compared. In the end, a dynamic planning tool will be developed based on which the Blood Service can adapt the capacity of the donor center to the expected number of donors in order to get the optimized donor flow which decreases the waiting times for donors and increases the productivity of the resources.
The overall aim of the research presented is to optimize the operations flow at the donor centers of the Red Cross-Flanders in order to improve customer (donor) service and the donor center productivity or utilization. Since this project is still ongoing, this paper only presents a first simplified simulation model, which visualizes the current flow at the donor center. A gap analysis is conducted in which the missing data are identified that are needed to develop a more detailed simulation model. Several recommendations are made which will be tested in different scenarios in order to find the optimized flow at the Red Cross-Flanders.
Decision makers can make better and less risky decisions regarding changes in use of human and material resources on the knowledge created by simulation experiments. There are some limitations in our approach. Models are simplifications and sometimes it is difficult to guarantee that they will be valid. Simulation calls for special expertise and a detailed knowledge of the system being depicted is necessary. In the future, a dashboard will be developed, based on the scenarios tested in the detailed simulation model, which the Red Cross-Flanders can rely on when setting up the capacity of the donor center in terms of both personnel and beds.
Moons, K., Berg- lund, H., De Langhe, V., Kimpe, K., Pin- telon, L. and Waeyenbergh, G. (2016) Opti- mization of Operations by Simulation―A Case Study at the Red Cross Flanders. Ame- rican Journal of Industrial and Business Ma- nagement, 6, 1001-1017. http://dx.doi.org/10.4236/ajibm.2016.610096