Dr. Tracey England discusses how simulation has been used to support a diverse range of improvement projects across Aneurin Bevan University Health Board.
You'll learn how simulation has helped assess the impact of facility reconfiguration on patient flow, plan the merger of two emergency call center services and to provide better understanding of demand on outpatient clinics.
Tracey has been a mathematical modeler for over 25 years working in both veterinary and healthcare. For the last 7 years she has been a Research Associate (and now Fellow) within the healthcare modeling group of Cardiff University which has enabled her to work with both the Welsh Government and Aneurin Bevan University Health Board.
For the last 4 years she has worked as a mathematical modeler within ABCi, a continuous improvement team within the health board and has worked on numerous projects focusing on discrete event simulation (patient flow, demand and capacity, outpatients and the introduction of ‘111’ in Wales) and forecasting.
Recently she has been involved in a joint project with SIMUL8 on evaluating both the short and long-term bed planning tools they have developed.
Dr Tracey England
Thank you very much, Tom. As Tom has just mentioned, I am going to [00.30] just go over some examples of simulation modeling that we have carried out in the Health Board. ABUHB stands for Aneurin Bevan University Health Board. We are one of seven health boards in Wales, so it is a very different setup to the health boards and trusts within England and Scotland. Hopefully you will find this interesting. If there are any questions at the end then let me know. [01.00]
I would like to introduce the ABCi team, because we are quite unusual. We are actually a mixture of improvement modelers and project managers, so there is quite a mixture of people. I will give a brief introduction into who I am and what my specialities are, and then go through three case studies which will hopefully will give you an illustration of the types of modeling that [01.30] we have been doing lately.
This is the ABCi team. There are 19 of us. I am not quite sure if we are all in the photo. It is quite difficult to get us all in one place at one time. It is headed up by John Boulton, who some of you might know. He is quite a keen SIMUL8 user. The modeling team are labelled. There are four postdoctoral mathematicians. We are headed up by Izabela, who has expertise in [02.00] Excel modeling and also simulation. Doris is more system dynamics. Daniel is mathematical programming, and I typically do simulation and forecasting. In terms of what ABCi does, our remit is to provide capability, innovation and delivery to the Health Board.
For the mathematical modeling team, of which there are four of us, we have developed [02.30] a short program for people within the Health Board so they can use operational techniques, one of which is SIMUL8. We ran one cohort and are starting a cohort this year. Last year we had three SIMUL8 projects that came off the back of the modeling program. We had two in demand and capacity and one in reconfiguration of services [03.00]
So, who am I? I am one of the mathematical modelers. I have been in ABCi for four years. I tend to do discrete event simulation, so I use quite a lot of SIMUL8.
The first case study I would like to introduce is a patient flow [03.30] simulation model. It was set in the Trauma and Orthopaedics department at the Royal Gwent Hospital in Newport. As a bit of background, it is a very busy unit. The royal Gwent is in the centre of Newport. The Trauma and Orthopaedics clinic typically runs 80 fracture clinics in a four week rota, has 24 consultants and runs 207 elective clinics. That is typically [04.00] carpal tunnel syndrome, hands and shoulder and spine injuries, whereas the fracture clinic are more broken arms and legs if you have had an accident.
What the directorate manager wanted us to have a look at was whether the configuration of rooms in the clinic spaces could be altered. At this time there were three clinic spaces, and they wanted to change the configuration of the [04.30] rooms in clinics 1 and 2. The clinicians were quite against the idea, because they thought it would reduce the flow of patients through the clinics, so they wanted to have a look to see if we could evaluate the before and after scenario.
The way we set about the project was by obtaining the physical layout of the clinics for a floorplan. We then shadowed numerous clinics over a three to four month period. We then [05.00] built the simulation model and checked it with the clinician involved, and then we ran some baseline models and presented the results.
This is the floorplan of the clinic. If I use the laser pen I should be able to show you. This was the entrance route that patients would come in. They would go to reception. They would take a seat in the waiting room and then they would be called either to one of these clinic rooms [05.30] or one of these clinic rooms. Typically they are just curtained off cubicle areas. What the department wanted to know was whether these five cubicle rooms could be reduced to four slightly larger ones, and again here whether these could be changed.
It seemed a reasonable project and we thought, ‘It will not take too long. Everything will be fine.’ Then we conducted quite a few shadowing exercises [06.00] and sought expert opinion, and we found out the clinic was a bit more like this. Quite a lot more going on, and rooms were not actually being used as we thought they were being used, and multiple purposes. At that point we decided we were going to do a patient flow model and actually track the patients through the clinics and see what would happen if we reduced the number of rooms from five to four.
Typically we split each of the patients’ pathways [06.30] into distinct steps. You can see here we have the nurse calling a patient and then the patient moving into the clinic area, and then the patient going into the cubicle and waiting for the doctor to come whilst the burse puts the notes in the cubicle pockets. With the shadowing exercises we were able to time each of these elements individually and then build up the distributions behind the icons.
In terms of [07.00] seeing the consultant, we timed how long each patient was in with the consultant. As the cubicles had curtains, we sat on the outside and times when patients or doctors went in or out. We did not actually interfere with any of the consulting.
We also looked at physiotherapists and special wounds nurses because they used the cubicles as well as the consultants, so we wanted to make sure that we were getting a full view of how the [07.30] clinics were running.
In the full model, which looks quite complicated, we have an initial streamlined pathway that we would hope each of the patients would go through along here, but quite often what could happen is either at this point or this point a patient might get directed up to the plastering or the X ray room and then have to circle back around and come back in again. There is quite a lot going on. [08.00] We also looked at the special wounds nurses separately and the physios. All the resources are here.
We had quite a lot of information going in on the model, but it was, we hoped, quite realistic and would convince the clinicians that we had had a thoroughly look as to the problem.
We conducted the shadowing exercises between October and January. We also talked to a lot of consultants and nurses and physios to see how they actually conducted each of [08.30] their consultations with the patients.
What we found with running the baseline and the actual scenario is that if we did reduce the rooms it did slow up the clinics considerably, especially when we consultants and physios using the same clinic space.
We noticed that if the clinics were prepared quite well ahead it became more streamlined and the turnaround time in the clinic [09.00] was much shorter, so were able to put in a few behavioral aspects that actually would improve the throughput and the patient experience rather than having to spend £50,000 altering the clinic space.
We found out that shadowing was really important. Although we were not in the consultations, we could see what was going on, and quite often just being there meant that the clinicians would tell us extra bits of information, which was really nice. [09.30] So that is the first case study.
The second one is more of a national program. It relates to 111 call center service coming to Wales. At the time of the study, Wales did not have 111 like England. It was a little bit wary about rolling out the program so decided to do it step by step, health board by health board. This project was actually right at the start with the first health [10.00] board – actually, just ahead of the first health board going live – and analyzing all of their data and producing some simulation models for it.
The first health board that conducted the pathfinder project was Abertawe Bro Morgannwg University (ABMU) Health Board, which is the health board around Swansea. We were asked to look at the call volume data for GP Out of Hours and NHS Direct Wales. The idea was that the [10.00] call volumes from each of those two services would combine when they come through to 111.
The two current systems were run quite differently. NHS Direct Wales runs 24 hours a day, whereas the GP Out of Hours only runs in the evenings and at weekends, and we have different types of staff looking after each service. We wanted to look at the call volumes and look at the staffing rotas as well when we combined the two services. [11.00]
In order to do that we had the data from both NHS Wales for the whole of Wales and the Out of Hours call volumes for each of the seven health Boards. In total, for a year’s worth of data, we were looking at about 800,000 call volume. We initially divided it into the two systems so we could produce simulation models for each. We then developed a combined model, which [11.00] we ran as a baseline, and then we could alter different sections of it to see what staff did what, and what effect it would have.
This is a screenshot of the initial simulation model that we had for NHS Direct Wales. Each of the icons relates to the staff that handled the call at that time and the icons along the bottom are the outcomes, so it may be a call to a GP was needed afterwards or [12.00] the patient was referred to A&E or they could look after themselves at home. Quite a lot of options with NHS Direct.
We had a similar model that we built for the GP Out of Hours Service. Each of those was separate for each health board. We were then lucky enough to have an MSc student over the summer who combined the NHS Direct Model and GP out of Hours into one easy to use [12.30] model.
This side are all the different call volume types that come in from NHS Direct and GP Out of Hours. These are the processes that the calls go through depending on what type of call they are. This is a checking mechanism to make sure we had not lost any calls, and then we have the staffing resources along the bottom, so we could see how utilized each of those staffing resources were under the scenarios.
What [13.00] was really nice about being able to do this is it allowed the 111 team the option of trying out different workforce options and costing them ahead of actually putting the figures into Welsh Government and getting the agreement to go ahead on the pathfinder project and then the actual 111 service in ABMU and now in subsequent health boards.
Just being able to develop a model and visualize it was really, really useful to them.
[13.30] The final case study is quite a new one and relates to two of the modeling fellows that we have trained in our silver modeling course, and they are looking at demand and capacity.
In particular, the Health Board have struggled, really, to understand demand on outpatient clinics; particularly as they have tended to focus on activity rather than the referral demand that comes in. By building this [14.00] simulation model, the modeling fellows have been able to look at referral patterns and they are using it to look at new and follow up appointments and the staff needed and capture the variation from week to week.
The model came off the back of one that has been developed in the Health Board for ophthalmology. The modeling fellows are still in the pilot stages. They are focusing on one specialty. They have [14.30] developed the discrete model and it is using daily or weekly data depending on what the service wants to find out. In terms of the weekly model, they have also incorporated a forecasting model so that they can then forecast as well as using the simulation to show what is going to happen to demand in the future.
This is just a screenshot of one of the subspecialties they have been working on. Essentially a booking will come in and that will generate [15.00] a new appointment for this subspecialty, arthroplasty, and then, depending on whether the patient attends their appointment or cancels it or does not attend or is discharged, it may go up into a follow up pool, and there are different outcomes for that, as well.
Quite often in outpatients, a patient may cycle through the follow up loop several times. This is why it is a useful to have simulation, [15.30] because one thing that has not been captured in the past is how many follow up appointments a patient might need off the back of a new appointment, and this is allowing the service to do that as well.
The modeling fellows have found it is allowing them to understand the referral demand from each subspecialty. They can see the variation, and in the plot below you can actually see the forecasting model that they have produced. [16.00] At the moment, over a three year period, it is out by about six referrals on about 36,000 datapoints, so although it does not look as though it is matching exactly here, it is actually doing really, really well, so they are pleased.
Just to pull some thoughts together, I think we have found that SIMUL8 has been really useful for looking at very large datasets and building [16.30] models off the back of those. We have also found it very useful for when we have conducted small shadowing exercises, and then we can build a model that reflects what we are seeing there. The visualization for clinicians and managers has been really useful, because they can see what is happening in their clinics in a model and then are more likely to believe the results.
We have done quite a mixture of projects: high level models and ones that are looking at patient flow.
In the next few months we are looking at a new hospital build, and we are hoping to use SIMUL8 for part of that. We are also looking at a full cancer pathway for lung cancer. That is essentially what we have got planned and are using SIMUL8 for, so it has been an incredibly useful tool for us in the Health Board.
It seems like you are lucky enough to work in a very talented team [18.00] with a combination of lots of different analytical skillsets. What goes into the decision criteria for you to decide to use SIMUL8 over other potential tools? What kinds of things do you look for in a project that would lead you towards SIMUL8?
Dr Tracey England
Sometimes it is a case of data quality. Sometimes it is the user specification in that they want the visual side of [18.30] things so they can present to other people; they what the ‘what if’ scenarios. Occasionally it may be just how complicated something is and whether you can break it down into smaller parts. It varies, really. The piece of work we are thinking about doing for the new hospital build is an area where we were really trying to work out which is going to be the best option on the modeling techniques, and we came up with quite a few. [19.00] We still have not decided which one we are going to go for yet.
Okay, great. One other question: I can see that you use simulation well to answer the main questions that you have in a lot of your projects, but you also touched upon some additional benefits that came out of running a project with simulation. For example, the need to maybe shadow clinicians and things like that [19.30]. How would you always run a project to make sure that you are getting those kind of additional benefits? Would you recommend something like shadowing all the time when it is appropriate, even when you have data, for example?
Dr Tracey England
I think it is very useful to try and get at least one or two clinicians on board right at the start, and quite often they will say, ‘Come and have a look at the clinic,’ so even if [20.00] you are not formally shadowing, it is good to go and have those discussions, because quite often the data might tell you one thing, or you think it is telling you one thing, but when you actually see it in action it is very different.
It is always worth doing that shadowing side of things, because things come up that might explain where there is a blip in the data or why something just seems unusual. For instance, when we were doing the patient flow [20.30] model for the trauma clinic they actually had a nasty road traffic accident on that day, and that department would have been closed for outpatients and then would have accepted patients from the road traffic accident. That would have been a very different set of data than usual. If we had not been shadowing, we would not have known that.
I think that would be I have seen as well. [21.00] It is always good to be able to capture information that is going to help you feed all the inputs into the model, and all the variation that we know can happen in real life to get that valid representation.
Excellent. Thanks again to you, Tracey, for presenting. Thanks, everybody, for listening.
We hope you enjoyed the webinar session.
If you'd like to find out more and to talk about how SIMUL8 could be used in your healthcare projects, please get in touch.Contact us
Using Simulation for Facility Planning in Healthcare
Improving patient flow at Johns Hopkins
See the features of SIMUL8's healthcare simulation software
View videos of SIMUL8 and learn about the benefits of simulation.
If you're considering simulation for your next project, our team are happy to help.
Explore SIMUL8 Professional's features with a free 14 day online trial.