Modeling to Learn 3.7 Consult Manual
2024-11-07
Chapter 1 Why use Modeling to Learn
Click on the hyperlinks or video thumbnails to watch the videos. Transcripts of the videos are also provided.
1.1 How might Modeling to Learn help?
You might wonder What is Modeling to Learn, why use it, and how might it help? We often don’t understand what is driving our care problems over time. Modeling to Learn helps to upgrade the decisions you make all day, every day, because the decisions made most frequently will be the most powerful for improving our care. We’re introducing Modeling to Learn, which has been developed and evaluated in the VA for over 9 years. Modeling to Learn is a way to upgrade care decisions by understanding the dynamics of common care problems over time. Are you very, very busy? Modeling to Learn includes step by step guides with gifs showing each click. These guides are available at mtl.how. Are you unsure what data to use, how it’s defined, and are you wishing it was current enough to guide your care decisions? Modeling to Learn includes a transparent, real-time locally defined data user interface available at mtl.how/data, where you specify the clinics that define your data views. Are you unsure what is realistic for improving care locally, given your patients’ needs and your existing staff and resources? Modeling to Learn also includes a simulation modeling interface available at mtl.how/sim, so you can try out new decisions before you implement them in the clinic. We all want to meet Veterans needs for timely, high-quality care. So, what gets in our way? Watch that video to find out.
1.2 How can a Modeling to Learn consult help?
Modeling to Learn improves visibility and provides new insights into how common care problems persist over time. How? Modeling to Learn is based on over 60 years of scholarship known as participatory system dynamics. For this reason, we call ourselves Team PSD for Team Participatory System Dynamics. The Data User Interface and Simulation User Interface comprise two versions of Modeling to Learn. The data-only version is known as Modeling to Learn Red; Modeling to Learn Blue ads participatory learning from simulation. Team PSD supports MTL Red and MTL Blue and is carefully evaluating how each works to support VA in meeting Veterans needs. Learning from simulation can help us to place a better initial bet on what is likely to work locally by evaluating alternative decisions via simulation before we implement them in the real world. MTL Red tells us where we’ve been over the last two years based on the clinic selections made to produce the patient data reports and team trends or visualizations. Many staff report that viewing the Data UI real-time patient data tabs or the team trends is efficient and encouraging. The data tabs help with clinical decision making. The visualizations of team trends provide leading indicators that improvement efforts are paying off, which can be validating. Why is Modeling to Learn useful when we have critical staffing and hiring needs? Watch that video to find out.
1.3 What gets in the way of meeting patients’ needs?
We all want to meet Veterans’ needs for timely, high-quality care. So, what gets in our way? Modeling to Learn is a way to upgrade care decisions by understanding the dynamics of common care problems over time. Demanding clinical days mean we don’t often have bandwidth to gain new insights, particularly in our understanding of interdependent clinic dynamics over time. We’re often flying blind to the impacts that our daily decisions have on the overall community that relies on us. As clinicians, we look at the patient in front of us and decide when we think they should be seen again. But have you ever been told you cannot see a patient as soon as you would like due to the constraints of the clinic? For example, an evidence-based course of care requires starting treatment without delay and keeping the Veteran engaged in a therapeutic dose of care over time to meet their needs. If we only emphasize Veterans starting care, but the time between visits extends way out, this interferes with evidence-based care and Veterans getting better. How can we get more Veterans better without adding hours in our day? Watch that video to find out.
1.4 If we keep making the same care decisions, will things get better or worse?
What if we keep making the same care decisions? Will things get better or worse? This is almost a trick question. If we keep making the same care decisions, more than likely things will stay the same in the clinic. Very rarely do clinical teams not have any local improvements that they would like to make. So, what do we mean by staying the same? We know that things in the clinic are not static but are always dynamic and always changing. We’ll describe how start delays affect treatment decisions to explain what we mean. The Modeling to Learn Simulation User Interface at mtl.how/sim is designed to help a team see what is likely to happen over the next two years if they continue to make the same daily clinical decisions that were made over the past two years. We call this the base case of no new decisions. Any number of feasible alternative decisions a team or site could make can be compared against this base case to find the best options for improving care over time. Because feedback dynamics produce non-linear trends or system behaviors, this means even the base case run can be surprising for some teams. The base case may show even more undesirable likely futures which teams would need to make new decisions to avoid. An example is the way start delays affect treatment decisions through balancing feedback. As the number of patients waiting to start a specific service grows, clinicians must adjust and try to find another way to meet their Veterans’ needs. Over time, as the clinicians learn about a local delay, it starts to affect what they think the most ethical, clinically appropriate treatment for the Veteran will be. An analogy from daily life will help. If there’s a traffic jam outside your VA, you may try to find another route home. As more drivers choose alternative routes, traffic returns to normal. Of course, when drivers don’t know about the jam, it will get worse over time. And whether drivers choose alternative routes or not—with balancing feedback like a traffic jam, where more cars are getting on the highway than are getting off—cars will build up and slow down the rate of traffic for everyone, whether we like it or not. We have these kinds of experiences in our clinics, too. Balancing feedback can be a blessing or curse. A blessing in that the balancing feedback causes the clinical system to return to its status quo. A curse if the status quo is undesirable for quality of patient care or provider quality of work life. When you start Modeling to Learn, it is surprising to see how powerful feedback is in driving care dynamics in any clinical care setting—VA or non-VA, VAMC or CBOC, urban and rural, BHIP, or PCT. General system insights, like the link between balancing feedbacks and nonlinear system behaviors, can be understood quickly from Modeling to Learn. You can understand exactly how many patients will be waiting and when. And then discover options for addressing the number of patients waiting for specific services with your existing staff mix. If you’d like specific recommendations for your local site or team based on these dynamics, request a Modeling to Learn consult or use the guides and videos at mtl.how. Clinical oscillations can really whipsaw clinicians and patients around, which is stressful and can increase potential clinical risks. If the number of patients waiting to start grows for a long time, it’s overwhelming. But it can also be exhausting when things go up and down quickly and clinics feel like they’re chasing their tails trying to find improvements. Especially given these dynamics can vary service by service, even within the same team. If you want to understand this principle, go to mtl.how/sim and check out the Team Care module. Or are you interested in another common balancing care problem, how the appointment backlog extends the weeks between visits? Watch that video to find out.
1.5 Why is Modeling to Learn useful when we have critical staffing and hiring needs?
Leaders and clinicians often ask how can Modeling to Learn be useful when facing critical staffing and hiring needs? With limited staff coverage, the need to hire is priority number one, but even then there isn’t a magic wand that produces staff where they may not exist today. What could help now? From the beginning, Modeling to Learn prioritized evidence-based episodes of care within existing staff time. When Modeling to Learn launched nationally in the VA in March of 2020, it was to empower staff to find the highest yield local improvements without asking staff to do more with less. As a result, when staff coverage may not change quickly, Modeling to Learn will find options for improving quality of care for Veterans and quality of work life for providers. How? What does Debbie mean by high yield local improvements? Because evidence-based behavioral healthcare is delivered over time, there are many possibilities. Small decisions made all day, every day by the clinicians, when compounded over time, can be surprisingly powerful. Think of your savings account or your waistline. In Modeling to Learn, we look for the lightest clinical lift teams and VAs can make that have the biggest payoff for Veterans in terms of timely, high-quality care. We work hard to avoid big difficult changes with limited benefit. When working with the Modeling to Learn-read data user interface, clinical teams are often motivated when they see trends that reflect hard-won efforts to implement high quality episodes of care which may not show up in other data systems for some time. The Modeling to Learn Blue simulation user interface saves staff time because alternatives can quickly be assessed during a modeling consultation. Change is hard, but we no longer have to learn by trial and error, wearing out already burdened staff. Does that sound too good to be true? If so, you may be wondering about examples of Modeling to Learn use cases for the pain points you face in your specific team or program. As an example, how does Modeling to Learn benefit substance use disorder, or SUD programs? Watch that video to find out.
1.6 Why is the MTL Red Data User Interface useful?
Why is Modeling to Learn Red useful and how does the Modeling to Learn Data User Interface provide new insights? The primary value of MTL Red is its power to efficiently query the VA Corporate Data Warehouse directly, and we’ve come a long way over the years. Yes we have. When we first began, we were using an Excel workbook so that frontline teams could carefully review the clinic selections that define their Modeling to Learn team data sets. Fast forward to the present and now teams have real time data available to them from within the VA domain from any computer with PIV badge access. Since the Data UI includes PHI, if you go to mtl.how/data, you will be able to see the same data you have permissions to access in the electronic health record. But given that clinicians, managers, data leads, quality improvement staff, evaluators … well, basically everyone is so busy, no one has time to review another data dashboard unless it offers something of really high value that distinguishes it from other resources. When you request a Modeling to Learn Red consultation, we work with your team or VA to explore care data that can address your locally identified priority. Clinics or scheduling grids are often changing. For that reason, we wanted a user interface where clinic selections can include active or inactive clinics over the last two years. That way, the team can filter the information to find the most appropriate clinics to include in their data set to gain new insights. And in Team Flow, clinic selections can be used to evaluate transitions between an episode of care in one team and the start of another episode of care in a higher or lower intensity care setting. Modeling to Learn Red also enables zooming in to check on the care of an individual patient at the start of the clinical day or during case reviews at a team meeting. But with the MTL Red Data User Interface, you can also zoom out to view teen care trends, bringing patient-level care coordination and trend-level process improvement decisions together. And that’s where things start to get interesting. Based on the clinic selections, the next set of tabs to find local data values for common care problems, including care coordination, psychotherapy, medication management, team care, and team flow. Each tab features simple definitions of how data were estimated for the common care problem. Detailed definitions with technical specifications are also provided to allow valid comparison of these data to other VA dashboards. That said, a focus on data details could be frustrating and add limited value. Modeling to Learn emphasizes understanding system problems in care flow over time. These care flow problems can be defined accurately with just five key time-based variables that drive care quality. How do five key variables drive care quality? Watch that video to find out.
1.7 Why is the MTL Blue Simulation User Interface useful?
Why is Modeling to Learn Blue useful? Well, why is it wisest to focus on the dynamics of care over time? The short answer is that clinical and improvement teams cannot adjust one part of the care equation without everything else changing. In Modeling to Learn Blue, we zoom out to see how care variables are locked in relationship with one another over time. The key variables that define either a poor quality or high quality episode of care must be understood together. Building from MTL Red, you can export your local data set created in the Modeling to Learn Data User Interface at mtl.how/data. Then, if you navigate to mtl.how/sim, you can find the Modeling to Learn Blue Simulation User Interface, which is a dynamic and interactive way to understand why problems with care coordination, medication management, psychotherapy, team care, and team flow persist over time. The Simulation User Interface is a way to see how adjustments in one part of an episode of care explain subsequent impacts in the care system. For any of the common care problems, the simulation saves you time and energy by accounting for the local new patient start rate in patients per week and the local appointment supply and appointments per week. The simulation also keeps track of the local new patient wait time in weeks, time between visits and weeks, and the engagement duration over time, again in weeks. All are calculated for you automatically in the Data User Interface, but their interdependence is accounted for in the Simulation User Interface. The Modeling to Learn Blue Simulation User Interface empowers teams to avoid ineffective strategies because you very quickly learn to develop new insights that would be inefficient, if not impossible, to figure out in your head or by hand. Learning from simulation is designed to help upgrade local decision-making. Teams develop new rules of thumb and insights in which the dependent dynamics among these variables that define care are all taken into account. With a Modeling to Learn consult, we come alongside with partners mid stride in their daily clinical activities who may have limited insight into what is likely to happen over the near future if they keep making the same decisions every day. With Modeling to Learn Blue simulation learning, sites and teams can safely see the impact of new decisions while building new capacities for systems thinking. Why is applied systems thinking more likely to help us avoid costly mistakes? Watch that video to find out.
1.8 Why is applied systems thinking more likely to help us avoid costly mistakes?
Why is applied systems thinking more likely to help us avoid costly mistakes? Unless we understand interdependent system effects over time, we’re very likely to choose ineffective strategies for our clinical work. And when chronic impairment, relapse, suicide, and overdose are the critical concerns for our patients, we need system insights about how to best organize our clinical practice. Applied systems thinking includes understanding complexity, feedback, and system behaviors over time. By complexity, we mean understanding the relationship or the interaction among two or more variables, such as wait times and the improvement rate. By feedback, we mean moving beyond simple cause and effect events to consider how some effects are reinforced and get stronger over time and other effects are reduced over time. To explain feedback using David’s example, it’s not just that if you have long wait times then the patient improvement rate will go down because patients aren’t starting care. It’s also that as the improvement rate goes up, then at a system level wait times will go down as Veterans get their needs met and graduate from care. Our systems are perpetually causing themselves over time through feedback. Causal feedbacks explain trends over time, called system behaviors. Given that feedback can be a reinforcing effect or a balancing effect over time, Modeling to Learn provides better insights about short- and long-term understanding of change over time. This can include decisions that may make things worse before they get better, or may make things better before they get worse. For example, longer wait times for new patients may reduce their improvement rate in the short term, but if this enables higher quality care engagement for them and existing patients over time, then more Veterans can get better and graduate from care and wait times will go down for the entire community of patients. It is through improving our systems thinking that Modeling to Learn really helps us move beyond making costly mistakes. If you’d like to learn more about how Modeling to Learn helps us improve the dynamics driving where Veterans get stuck or drop out of care, watch that video to find out.
1.9 Why is Modeling to Learn able to provide new insights?
Why is Modeling to Learn able to provide new insights about where Veterans get stuck or drop out of care? Because the Modeling to Learn Simulation User Interface depicts the states of care as stocks based on the flows in and out of that state. Take for example, patients waiting to start. We depict patients waiting to start as a stock, a rectangle with an accumulated volume indicator. The number of Veterans waiting to start in a typical week is determined by the difference between the number of Veterans that flow into that state and the number of Veterans who flow out of that state each week. The rate dials depict the flows in and out of stocks, and the primary rates in Modeling to Learn are patients per week, appointments per week, and episodes of care per week. The initial values for any team or site depicted in the Simulation User Interface are based on the last two years of data exported from the Data User Interface. The important idea is that in any dynamic system, things get to be the way they are over time. Things either stay the same or improve decision by decision over time. For example, in a clinic or site, the number of Veterans engaged in a given state of care, such as medication management or psychotherapy, reflects decisions that lead to accumulation in that state over time. The dynamic stock and flow diagram provides a way of understanding where Veterans get stuck. As clinicians know, often there is more than one way out of a state of care. With psychotherapy, you may flow from the first session to the second session, but you may drop out at any point in that flow. Why does Modeling to Learn emphasize flows through care over time? Watch that video to find out.
1.10 Why does Modeling to Learn emphasize flows through care?
Why does Modeling to Learn emphasize flows through care over time? Our goal is to ensure Veterans flow through our addiction and mental health care services to recovery. A basic idea is that care occurs over time. Flows are powerful because they express a rate of change, which means they give us key information about where we will be in the future. Imagine your next trip in the car. We need the miles per hour to know when we will arrive at our destination. But knowing the number of miles to our destination is not enough. We must also know how fast we’re traveling to know when we will get there. In other words, miles per hour. The same is true in our clinical care. It’s not enough to know how many new medication management appointments we have on the books today. We need to know how quickly those appointments were added to the books or the booking rate, in appointments per week, as well as how quickly those appointments are completed or the completing rate, in appointments per week. When the inflow is greater than the outflow, the level of the appointment or patient stock will rise. When the outflow is greater than the inflow, then the level of the appointment or patient stock will fall. Opening the Modeling to Learn Simulation User Interface at mtl.how/sim, you can see where all the Veterans are accumulating in stocks. You can understand the inflows and outflows for each stock and which accumulations are likely to get better or worse over time. Many clinical teams talk about feeling inundated by the addiction and mental health needs of the communities that rely on them. They can feel like they’re drowning, providing the volume of care needed, especially when short staffed. We understand this kind of flooding from the natural world. To stay safe, it’s critical to know how fast the waters are rising and where. We understand this by understanding flows. You may relate to this feeling of being overwhelmed and want to know more about how Modeling to Learn can help. For example, how does Modeling to Learn help improve medication management? Watch that video to find out.