Inside certain computers at The Children’s Hospital of Philadelphia are active microcosms in miniature of many of the hospital’s units. These computer models are simulations, programmed for the purpose of quality improvement. Manipulating these models can identify opportunities and aid planning for systematic improvements.
Building miniature hospital units in silico is all in a day’s work for T. Eugene Day, DSc, program manager for health systems in the Office of Safety and Medical Operations at CHOP and a principal investigator who publishes his simulation-based findings in medical and quality improvement journals.
Since many in the healthcare world may be more familiar with “simulation” in the form of clinical skills training scenarios that use manikins and actors as patients, Cornerstone chatted with Dr. Day to learn more about his recently published papers, and about how he uses his specialty, a form of systems engineering called discrete event simulation, to model and inform improvements to patient care. The edited conversation follows below.
What do your simulations look like?
Simulated cohorts of individual patients present for care at a simulated clinic or healthcare delivery facility and then go through the processes of care and have their own simulated outcomes. We lay out a blueprint of the hospital and we have graphical icons of each individual simulated patient, provider, and location. If you’ve ever played the computer game Sim City, it’s sort of like Sim Hospital. It’s like playing a videogame. A bunch of clinicians in surgery here at CHOP talk about it as “flying babies” because they can see the icons, which look like little pediatric patients, flying through the facility from location to location. And, of course, we can accelerate this so it moves very rapidly. We can simulate years’ worth of data in a few minutes.
We don’t have an “average patient” in the simulation. We use the parameters from an aggregated set of de-identified real-world patient data, and create a statistically identical simulated dataset. That dataset becomes an entire simulated cohort of patients, each with their own individualized data fields such as an age, a sex, and a reason for their medical visit.
How do you build a simulated healthcare facility with processes that behave like the real thing?
Building a simulation starts with several years of training in computer science. The fundamental methodology underlying this is called object oriented programming.
After that, the first thing you have to do is go and live in the system. You interview all the people who work in the system. When I was modeling the emergency room, for example, I was here nights, weekends, and during regular days, to see how the system behaves.
Then you build a flow chart that shows what path a patient follows through a system, what the doctors do, and what the nurses do.
Next, you build a systemic decomposition to break apart the system into each of its constituent elements: The resources are one element, including anything you have, whether it’s human, such as a doctor, or a tool, such as an EKG machine, that is required to do the work of the system. Another element is locations, which are both the places where things happen, and also virtual locations like email inboxes. The third element is the entities, the things upon which we do our work, which are typically patients, but also things like computer records or EKG sheets.
Once we have that decomposition, we have to build in a flow to answer the question: “How do entities consume resources at locations, and then proceed from one location to the next, until they have completed their entire circuit through the system?” And then: “How do resources search for new tasks?” Once you have that flow and the decomposition, you code that into a commercially available software suite. Then you have to validate your model, which can be done in a variety of ways.
When you have built and validated a simulation, how do you then use that model of a clinical system to improve patient care?
One of the projects when I first arrived at CHOP was to better manage our postsurgical patients to have them recover in surgical beds rather than medical beds. It’s well known and published that postsurgical patients that are recovering in surgical beds rather than medical beds have fewer complications. We were able to increase the proportion of our patients who were recovering in surgical beds from about 79 percent to about 93 percent. It’s distinctly plausible that that improves outcomes as a result of that systemic intervention, not a medical intervention.
As another example, one of the five procedure rooms in the Cardiac Operative and Imaging Center, called the Hybrid Suite, had to be closed for renovations for six to eight weeks. It’s called the Hybrid Suite because it can act either as an operating room or as a cardiac catheterization lab. They wanted to know, can we perform the same volume of cardiac catheterizations and cardiothoracic surgeries in only the four remaining rooms? And if we can, what do we have to do to make that happen?
So we used an existing simulation, which we had previously developed for another quality-improvement question, to test that theory. First of all, is it even possible?
Simulation allows us to test what are sometimes grandiose attempts at systemic level outcomes, against the constraints within the system. It could have told us, for instance, that running the same number of patients through with fewer rooms might only work if you had 29 hours in a day.
We actually found that we could do it with a lengthening of the day of only about an hour, so long as a we were able to reduce our turnaround time between procedures, and to start earlier in the day so that we didn’t bump up against shift changes and that kind of thing in the evening. We had the actual time and personnel necessary to do all the procedures.
But simulation doesn’t tell us how to achieve that improved time efficiency. For that, our performance improvement group led what we call a rapid cycle improvement event. We engaged nursing leadership, anesthesiology, and cardiology, into redesigning some processes to get things happening earlier in the day and to reduce turnaround time.
We were completely successful. During the equivalent period the year before the renovation, when we had five rooms open, we had done 135 catheterizations. During the period of renovation with one of the rooms closed, we did 138. We did that with no increase in nurse overtime, and no family complaints. It was a very successful project. We wrote that up, and it is appearing in the Joint Commission Journal for Quality and Patient Safety, which I’m very pleased about because it’s really one of the bigger journals in my field.
Can you tell us about your motivations for publishing this type of work in journals when you don’t have the “publish or perish” pressure of a faculty member?
The Children’s Hospital of Philadelphia is well known as one of the best places in the world for a child to come and get medical care. Our safety and the great work we are doing in quality improvement are less well known. I believe that publishing in medical and quality literature is valuable to getting that information to the public.
But the biggest reason is that I believe that publishing and disseminating this work and helping other institutions to do what we do, can improve outcomes for people and improve access to care. I like to say that my long game is public health. Being able to access medical care is one of the primary challenges of our health system in America, whether that is because it is overburdened or whether that is because it is prohibitively expensive.
I believe that the appropriate application of systems engineering methods in healthcare can relieve both of those problems. It can allow more members of the public to engage with those services, it can reduce costs, and I think it improves health.