Using Simulation for Facility Planning in Healthcare

VII. Case Studies at Memorial Health System

1. Overview

These are case studies showing where Memorial Health System has gone through the four process steps. The first is [15.30] non-clinical and the other looks inside the hospital from a clinical operations standpoint. I will take you through how we considered the four steps, as well as some of the results from our SIMUL8 simulations.

2. Front Circle Drive

We built a new 114-bed patient care tower at our 500-bed [16.00] flagship hospital in Springfield. Along with this came a new front entrance and drive for access to the hospital, where we had patients, visitors and delivery people coming in and out on a daily basis throughout the day.

First, we wanted to know our inquiry, looking at the best way to set up our circle drive to avoid congestion and maintain safety for employees and patients being discharged from current traffic volumes. [16.30] This might seem like a fairly simple ask, where you look at it and see people come in and out, drop patients off and keep going. That said, as we looked at our process, we also have complimentary valet service where visitors and patients are discharged and helped to their vehicles. We wanted to look at avoiding congestion, or overbuilding [17.00] with too much space for cars to come in and out, but ultimately making it so that patients could be safely loaded and could cross the new circle drive while avoiding accidents.

When we looked to know our space, previous to construction we had two lanes for traffic throughput. This was a unique simulation for me, because when I came in it was before I had been introduced to SIMUL8, so my co-worker [17.30] Lance had built this simulation whereas I managed multiple ancillary departments, including that valet staff. I was able to come at it from a stakeholder perspective, while Lance approached building this simulation from an operations improvement perspective. I was able to see both sides of the coin throughout this process.

Lastly, we went to know our data. This was another unique experience because [18.00] we did not have a huge bucket of data to pull things from. All we had to depend on were some estimates from the valet staff. We knew how many vehicles we were parking, we had a rough estimate of how long it took to park them and where we were seeing vehicles from, but we were in a situation where we did not have a big amount of data.

I will now bring Lance in to [18.30] walk through how we were able to overcome some of these obstacles and the considerations that he used to build the simulation, so we could meet the inquiries and questions that were originally posed.

Lance Millburg

As we were building the simulation and going back to the four steps, know your inquiry means not only knowing the questions you are trying to answer, but knowing your audience and understanding their objectives to assess the level of rigor [19.00] that you have to apply. If you have done this type of work and are working in an academic setting where the amount of accuracy is paramount, that is a different conversation to working on an operations perspective where you can allow some level of margin for error on some of these things, as long as there are statistically significant differences in the data that you are looking at.

From that perspective, as we are building this out, knowing that our data is limited [19.30] and understanding that we are doing this from an operations perspective where we have a little bit of margin for error to work within, was informative to the types of distributions that we could do. The other part of knowing your inquiry is understanding what and how people will be able to consume. We had a great deal of visual and we built out an Excel front-end for this as well, where people could manipulate the variables and have the simulation [20.00] run without having to have an intimate knowledge of how to do that. One of the goals for this in terms of knowing your inquiry was about how we would present the data to our stakeholders and that visual representation of it would be important.

In understanding our data, we knew that we had some level of observation data, so we knew that we were never going to be working with a lot of distributions such as the log normal or some of the other distributions. Normal distribution was not something that we had an opportunity to use with our data sets, but we understood that it would be more around time estimates, which was where we utilized triangular and uniform distributions. For those that have worked on project management, the triangular is similar to the [inaudible] analysis. [21.00] If you have multiple conversations with multiple stakeholders you start to be able to form a good, relatively accurate, distribution. What is the best-case scenario, the worst-case scenario or the most likely? If there is no most likely then you will be using a uniform distribution. If they give you three options, then you begin to build it from a triangular perspective. That is how knowing our data in this situation, plus knowing that we had some challenges [21.30] to our data set, informed the types of distributions that we used.

With this simulation, we knew that our objective was to keep patients’ vehicles off the main road, which is First Street in Springfield. It was important to prevent traffic from backing up onto the main road, which had happened in the past. [22.00] Another objective was around utilizing it in a way that would not negatively impact the flow of our concierge service and valet parking. Through several iterations within this simulation, we came up with not using the middle lane for anything other than an ability for people to flow through. We recognized that we wanted to keep the lane [22.30] closest to the hospital as open as possible, so that would be for patients that were going to be parking for five minutes or less, which is the operational guidelines that the valets are working under. They meet the patient and if it is going to be a longer stay or a longer parking, if they need to go up and get the patient, or they need to have a conversation with the provider or for dropping something off that might take more than five minutes, [23.00] that is all in the outer lane. We also knew that we had commercial traffic in addition to patient traffic, such as couriers and other things. These are shown by the trucks.

Finally, those in the United States know that our traffic laws allow for turning right on red or to turn right out of a right-turn lane in many circumstances. There was a question [23.30] about whether we put that straight to go across traffic in the right-hand lane or the left-hand lane and we were able to model down to that level of detail based on observation data and conversations with our subject-matter experts. Using a straight-right combination would have led to more backing up of traffic because the majority of people were trying to turn right and the fact that they can do so legally meant that things would back up [24.00] versus a lot less traffic turning left to cross traffic or go straight. Based on the simulations and results, we decided that it made a lot more sense for us to have the left-straight combination out of the circle drive.

Another thing that was nice about this simulation from a visual perspective is that SIMUL8 allows you to change the icon based on distribution, so the trucks shown are a truck variable [24.30] and the cars are a passenger variable. We could change the distribution of commercial traffic versus patient traffic and see how that would impact it, not only from a numbers perspective but visually as well. That was helpful in giving multiple ways of interpreting the information for our stakeholders, especially those that are more visual.

Graham Prellwitz

Thank you, Lance. I know that I can speak from an operational perspective as well. [25.00] We identified that many of those vans were represented by flower, food or equipment delivery, and based on this simulation, we decided to move as much of that routine delivery to a different entrance of the hospital to ensure that we left as much space as possible for our patients and for visitors to be able to use the main lobby of the hospital. There were some operational repercussions [25.30] that also came from this simulation, where we able to identify how we could best serve our patients.

3. Hospital Outpatient Diagnostic Testing Centre

The second case study looks at one of our small critical-access hospitals. In the United States we have these 25-bed hospitals located in rural areas. One of them wanted to rebuild a portion of their hospital [26.00] to centralize five different testing areas that were previously scattered throughout the building, including infusion services, pulmonary function testing, sleep lab, the radiology exam area and an area for general exams. They wanted to bring that together to one 2,600 square foot space, centrally located towards the front of the hospital for patients that were coming in and out for these outpatient diagnostic tests.

We started [26.30] with knowing our inquiry and were posed with looking at the space needed to be dedicated to each of these testing areas, given current volumes, to avoid long wait times for patients and any throughput issues. Then we looked at whether we would be able to register patients in the room without causing significant issues for the throughput of patients. I will talk about this later, but this came from [27.00] a decline in patient experience scores, linked to the registration process. We used to have the traditional registration process, where a patient comes in, checks in at a check-in desk, receives papers to fill out and after 15-20 minutes is called back to the procedure room. We wanted to see if there was any way to alter this to improve patient experience [27.30] as they moved through the process.

For knowing our process, we had the more traditional waiting room set-up but wanted to look at moving the registration portion to the care room where it would be more private and have the ability for the registrars to go to the patients rather than vice versa. Secondly, we wanted to look at [28.00] our current scheduling for our sleep lap studies. We found that that was inefficient and that came out when looking at our current and future state process maps, where we noticed that there was a long lag that was resulting in cancellation or delays. We wanted to focus on that to see if there was an opportunity to drive volumes as well as revenues by getting more patients in for those studies. [28.30]

In knowing your space, we were posed with a general outline or blueprint of what they were looking at. From looking at our square footage and the regulatory standpoint of how big the rooms needed to be, we knew we had approximately 15 rooms to be divided among the five services.

Knowing our data presented another unique situation where we were building the simulation and realized we had small hourly volumes, [29.00] where some of the procedure areas might only see one to two patients a day or even a week. We had to look at this from a different angle and build into the simulation that we might not see a patient every day but leverage that to make it as realistic as possible as patients moved through the area. Again, I will turn to Lance, and how [29.30] he was able to use some of his SIMUL8 and statistical knowledge to build this data into the simulation to represent appropriate hourly volumes and patient arrival rates.

Lance Millburg

When you look at this simulation, there is a great deal of different, disparate patient populations in one particular area. It would have been easier if there had been one specific situation where we had a homogenous amount of patient population, [30.00] because there is some standard functionality in SIMUL8 where you could manage those distributions. We had phlebotomy patients, of whom they see several dozen a day, with one to two sleep lab patients a night and certain other pulmonary function testing or other infusion services, where they maybe do one or two a week, which create different problems [30.30] because of the disparate patient populations coming in.

Part of the nice thing and part of the challenge of SIMUL8 is the fact that you can make the simulations as simple or as complex as you need them to be. For those of you that are new to SIMUL8, we had a webinar last year on some of the steps that we take to build out a program and I would encourage you to go through the SIMUL8 archives [31.00] and look at that if you are interested. One of the things that we talk about is to make sure that, as you learn and build out your SIMUL8 program, you identify some folks that can take advantage of some of the more advanced capabilities of SIMUL8 because it is a fantastic program that allows you to get into the details.

One of those functions is something called Visual Logic, which is a way to build custom programing that [31.30] is a step above the things that are normally provided, although SIMUL8 does a fantastic job of giving you the majority of tools that you need. It works by having the next patient or work item scheduled at the time [32.00] that a work item comes into the work stream. For example, if we set up 2,400 minutes or a day and a half as the average time between two work items, then it runs the risk that SIMUL8 will never pick up that next patient, depending on how the clock was set up.

There is a great deal of detail that goes into this, but the point is that you can set SIMUL8 up through [32.30] Visual Logic to force it to look at things from a probability perspective as opposed to a specific timing perspective. Instead of saying there were 2,600 minutes between patients, it was about saying there was a 15% probability of a patient being seen today. That shift in how SIMUL8 works with the model allows you to become much more accurate in getting the patients to come through. [33.00] There are ways that you can go about making SIMUL8 look at things from a different statistical perspective through Visual Logic and I would encourage anyone wanting to take SIMUL8 to the next level in your organization to find the person or resource that can invest the time in learning the Visual Logic functionality and allowing that customization to come into place with the simulation.

The simulation is a great example [33.30] of how you can have multiple different disparate distributions coming in from the work items that are being seen on an hourly or even a minute basis and those patients or work items that come across a couple of times a day or once or twice a week. You are still able to work within the same confines of the model but treat those different populations in a [34.00] different way. Again, for those of you in healthcare, in the United States we have something called STEEEP: safe, timely, effective, efficient, equitable and patient-centred. These customizations are key to being able to get into things that are patient-centred and to customize the care to the types of patient populations that you are working with.

To wrap up, [34.30] this is about knowing that there are much more advanced simulations that can be built in SIMUL8 than even the excellent standard functionality allows. I do not want to take away from that, because there is an awful lot that you can do with the standard – outside of the Excel interface, that entire circle drive simulation was standard functionality – but you can also jump into something like this [35.00] where it is much more unique.

Another thing we did in this simulation was build-in revenue and cost, taking advantage of the standard functionality that SIMUL8 has for giving a standard revenue and cost to a work item. That meant that we could test not only throughput but also the margins and volumes to find the break even on the project, where you need to have a certain amount of patient volume before you cover your variable costs. [35.30] For those of you with more of an accounting background, there is a lot that you can do with SIMUL8 to offer that type of information to stakeholders so that you are not only catering to the operational folks that are trying to make this work from a day-to-day perspective, but you can also offer information to the financial folks. You can show testing of contribution margin or some of the financial ratios to see the true ROI and the net present value on [36.00] these types of projects and whether we can make projections on growing volume and its effect on the financial projections of a project.

4. Results

Graham Prellwitz

Thank you, Lance. Continuing on that point, with the results that we saw from the registration staffing model, we were able to identify that we would not slow down the throughput of patients. We were able to put one receptionist upfront to greet [36.30] patients and do a quick check-in before leading them back to the room where they were being provided the care. Then we were able to build in registrar staffing that would float through each of the units that were previously in five different areas of the hospital to being centralized in one area.

We were also able to recommend the number of rooms per specialty in order to maximize [37.00] space utilization. From looking at square footage layout we had a cost estimate of around $500 per square foot to build this out. The average room was around 163 square foot, so we were looking at about $82,000 to build. We wanted to make sure we were not overbuilding but also that we were not underbuilding. From looking at current volumes and patient flowing throughout, we were able to identify that we had [37.30] a primary exam that would be used during the day for some of the more general exams and were able to create it as a hybrid to be used as a sleep room during the evenings when the sleep lab studies were being done.

Lastly, we identified that our sleep lab scheduling process was fairly inefficient, and we had some opportunities there. As they are currently building out the physical space, [38.00] they are also launching a quality improvement project that has a projection of potentially $100,000 annual increase in gross revenue that could come from dealing with some of these issues that we see in our sleep lab process. We could grow volumes and decrease our cancellation and no-show rates to experience growth in volumes and revenues.

Those are the two examples [38.30] that I wanted to bring to you. I want to thank everybody online, including SIMUL8, as well as Lance, for listening as we walked through our process. I will open it up now for any questions. My contact details are listed onscreen so if in the future you want to know how we used some of the aspects of SIMUL8 or our process, please feel free to reach out with any questions. [39.00]

Questions and Answers

Tom Stephenson

Excellent. Thanks again to Graham and Lance. It is good to see how you have built up a methodology that will help ensure success throughout any simulation project and it is good to see that applied to such a varied range of projects. I think you have done a great deal of good in utilizing the flexibility and the power of SIMUL8.

We have one question so far. [39.30] Do keep asking questions if you have any more. The first is, ‘In terms of the project – I guess any of the projects – were patients involved in any of the decision making or did they do any observations of the traffic flow?’

Graham Prellwitz

We utilize simulation a great deal at Memorial and we have a dedicated building [40.00], the Memorial Center for Learning and Innovation. We have two different types of panels that receive patient feedback from various quality initiatives or operations improvement projects that we have worked on. We use that and meet with patients around once every two months, these are volunteers that have either been patients in the past or are community stakeholders.

Some of the simulations have been brought to these patients for them to look at and see [40.30] how we are not only utilizing simulation for bedside care but also for the facility planning standpoint. It gives us a great opportunity to bounce that off people who are outside the organization or might come at the issues that we are experiencing with a different mind frame. We have found that platform to be effective in reaching out to the general public.

Lance Millburg

If you go back to some of the points in the steps, knowing your data and knowing your inquiry [41.00] are important. When it came to the critical-access hospital project, there were constraints presented to us even before we started the project. Leadership at that hospital had reached out to their patients and understood that some of the things that were bothering the patients were the amount of walking they had to do between different tests and the amount of confusion to try and find four different places when they are trying [41.30] to get four different tests done to be ready for their surgery tomorrow.

As Graham pointed out, I think it is great to bring the actual simulations and information to the patients, but I would also encourage you to try to have an understanding of what the constraints are from your patients before you start the process of modeling because you want to make sure that you are taking their thoughts and information into consideration [42.00]. Graham had a slide with the different pieces of data, and constraints was one of those. I would consider patient satisfaction to be an informative source for those constraints. Try to make it a part of your process to even reach out to patients in the initial forming stages and have them as part of the constraints that you build into your simulation.

Tom Stephenson

Great, thank you. One more question has come through [42.30] asking how many projects in total you have done?

Graham Prellwitz

To give some additional history and the webinar that we did last year goes into some of these details, we started using simulation in 2011 and 2012 but we recognized that we were not getting the full power [43.00] and return on investment through the Balkanised approach that we had taken with our use of SIMUL8. In 2015, it was my black belt project in Lean Six Sigma to bolster our use of SIMUL8 and at that point we saw a huge explosion of different projects. Over the last two years, I believe the total is somewhere in the 30 to 40 range of different models that we have built out and different projects that we have done. [43.30]

These range from the use of different resources in the emergency department with CT scanners and evaluating if we can improve our processes of care to get better utilization because that equipment is not cheap. We have done things on home health and had some good projects around structuring the teams that go out to our home health patients. We reviewed where it made sense [44.00] to put our borders for how far we go out based on the cost of care with RNs versus LPNs, which is a different designation of clinical expertise in the nursing perspective. That helped us understand drawing borders for different teams and taking into consideration how far the different towns were.

We have done projects in our ORs, specifically on block scheduling [44.30] and we understand how difficult that is but were able to do ‘what if’ scenarios of giving one physician more block time than another based on the types of procedures that they do, their outcomes levels and the costs and contribution margin to the hospital. We have used SIMUL8 to do those ‘what if’ scenarios and show the true impact to the organization. Things like the circle drive and the facility things that we have talked about today, in addition, when we opened our tower, [45.00] we did a project on the square footage that nurses were covering. I believe it was triple so that is a lot of space for a nurse to cover. We looked at assigning nurses to where they will not be impacted by the additional square footage.

Those are some ideas that we have done out of around three or four dozen projects associated with SIMUL8. We have built it into our [45.30] methodology and governance process to where when a new Lean Six or facility project comes in, we ask if they feel that there is the possibility of using SIMUL8. It is built into our governance and how our belts are trained. All black belts are trained in the basics of using SIMUL8, so getting more people using the product is key. That is a long answer [46.00] to say we have done three or four dozen, with some additional context to the breadth of experience that we have had with this product.

Tom Stephenson

Excellent, thank you. It is great to see such a wide variation of processes that have been simulated and such value that you have from it. We have some additional questions on the chat panel, but I think it is probably easiest if we come back to these and carry on the conversation. [46.30] We will make sure that we post any of the questions on the SIMUL8 Health LinkedIn group. Also, there will be a recording available at simul8healthcare.com and if you have any further questions – can people contact you?

Lance Millburg

Absolutely.

Graham Prellwitz

Fantastic. Yes, of course.

Tom Stephenson

Brilliant. Thanks everybody for listening [47.00]. Thanks again to our presenters for a great presentation and tune in again next month.