Face the Fight Focuses on U.S. Veteran Suicide with Help From Stella® Architect

David Rozek

U.S. military veterans embody the qualities of service, strength, and resilience. While most veterans are thriving today, still too many U.S. veterans die by suicide. According to the U.S. Department of Veterans Affairs, suicide is the second leading cause of death among veterans under 45 years old, and more than 125,000 veterans have died by suicide since 2001.

Face the Fight, an initiative supported by corporations, philanthropic foundations, non-profits, government liaisons, and veteran organizations, is focused on raising awareness of veteran suicide, the complex issues around it (e.g., the stigma around mental health and risk factors such as homelessness), and the interventions that reduce risk. Launched in June 2023 by USAA, Humana Foundation, and Reach Resilience (an Endeavors Foundation), Face the Fight is committed to building on advocacy work, forming a coalition of like-minded companies and organizations to bring together collective resources to scale evidence-based interventions and drive meaningful change. Face the Fight is supported by scientific experts, clinicians, and data analysts at The University of Texas Health Science Center at San Antonio (UT Health San Antonio).

“Veteran suicide is a very wicked, multi-faceted problem,” says David Rozek, PhD, ABPP, Senior Scientific Advisor to Face the Fight and Associate Professor at UT Health San Antonio. While Post Traumatic Stress and other mental health disorders do impact many veterans, nearly half of people who die by suicide do not have a diagnosed mental illness. Rather, high rates of veteran suicides (currently 71.8% higher than the rate of suicide in the general population) can be attributed to a mix of complex issues including stigma, employment stressors, housing and financial stressors, legal issues, relationship stressors, and health challenges.

Rozek, his colleague Katy Dondanville, PsyD, ABPP, UT Health San Antonio, and Karim Chichakly, Co-President, isee systems are using system dynamics and Stella Architect to understand how various interventions work together to prevent veteran suicide. “You can’t just look at one cause of suicide or suicide intervention to understand the whole problem,” says Rozek. “Using Stella Architect, we’ve created a model that can be continually modified to help Face the Fight understand the whole system that impacts veteran suicide.”

The modeling team’s first challenge was finding the data needed to describe the veteran population and current intervention outcomes. “Some data is just very hard to find,” says Rozek. “Clinical trials are generally focused on interventions delivered in health care settings and can give us baseline effectiveness rates. They often describe ideal-world, not real-world, conditions. We also have little data on gatekeeper training and how the training translates to actual delivery or use of the training.”

Due to the inadequate data available, a team with expertise in suicide and the veteran population helped develop the model’s initial data sources, many times using the best data available and the best expert judgment for a model parameter. “We created a distribution of low, medium, and high distress levels to measure risk for suicide,” says Rozek. “Then we built a baseline model.”

The baseline model used the veteran population (including all branches of service and both combat and non-combat postings) and applied no interventions to get the baseline for lives lost to suicide. “Once that model was built, we could add interventions,” says Rozek. Comparing the baseline to the number of lives lost with interventions yields the lives saved through the interventions.

Veteran population divided into distress categories

Understanding suicide intervention efficacy demands a system dynamics approach. “Most interventions work with, or lead to, others,” says Rozek. “Suicide screening is a good example. When we go to the doctor or even the dentist, we’re asked about depression and self-harm or suicide ideation. Screening for those things is an intervention, but it’s one that is meant to lead to another, like treatment. So, if we want to see if suicide screening saves lives, we have to model its relationship to treatment or other interventions.”

The core team of Rozek, Dondanville, and Chichakly assembled a group of 44 expert advisors that reviews the model as new inputs and variables are added. “The Scientific Advisory Committee (SAC) takes a broad view of the model to help us validate our work, think through data issues, and test variable relationships,” says Rozek. A smaller working group within the SAC meets quarterly to address questions and problems that arise as the model is built and run.

“Stella Architect allows us to share our modeling work with Face the Fight partners and advisors around the world,” says Rozek. “All they need is web access. They’re all experts in aspects of veteran suicide but most of them are new to system dynamics and modeling. They don’t need any training to review and run a Stella Architect model. And we don’t have to worry that someone will change or break the models we share.”

Face the Fight and the SAC questions and feedback inform model improvements. Rozek is able to make small changes to the model, but he and Dondanville rely on Chichakly to make more significant improvements. “Karim is amazing,” says Rozek. “He has so much model building experience and expertise. His ability to translate the issues surrounding veteran suicide to the model has been instrumental.”

Face the Fight is already learning from the model. “We have projected 3,500 veteran lives saved from Face the Fight interventions through 2032, which is truly remarkable. We also can see that veteran suicide prevention is facing capacity issues,” says Rozek. “If there aren’t enough people and organizations trained in or offering needed interventions, lives can’t be saved. So, the pool of clinicians must be expanded, and at-risk veterans need greater access to non-clinical interventions like career transition support, food programs, financial advice, housing assistance, and peer-to-peer advising and counseling.”

bCBT (brief Cognitive Behavior Therapy) intervention path (other paths from Screening not shown)

It’s also clear that answering programmatic questions will require additional data. “We need to break down the veteran population by gender, race, and ethnicity,” says Rozek. “The increasing rate of suicide among female veterans, especially suicide by firearms, raises many questions. What is different about that group? What interventions work best for women? We also see that Alaskan Natives and American Indian are two of the highest risk groups. Are there cultural differences that interventions have to consider?”

The model is already being used to support funding decisions. “In 2024, Face the Fight is deploying its third and fourth round of grants.” says Rozek. “Using the model, we can see the impact of each proposal and whether it meets stated goals. Face the Fight can also check for saturation and make sure that funding is spread out across needed intervention programs.” The first model update will be summarized and published in Face the Fight’s inaugural report in June 2024, in coordination with the one-year anniversary of the initiative’s public launch.

However, the model remains a work in progress. “It will continue to change as new data becomes available,” says Rozek. “As we run the model, we and our advisory groups will test new hypotheses. The science is always changing.”

Using Stella® to Support Corporate Strategy Development

Throughout a very successful history of rapid growth, a key biopharma company, with leading positions in multiple product platforms, has continually faced questions such as:

• For how long can their existing core business continue to grow at a sustainable rate?
• How much to invest in their core platforms versus expanding into adjacent areas?
• How to trade off capital, R&D, and acquisition expenses?
• How to ensure that planning activities are robust to changes in the competitive and regulatory landscape?

These questions are addressed regularly using a variety of tools and techniques that typically combine informed qualitative discussion with quantitative tools looking at specific elements of a strategy being considered, generally in the form of business cases.

In 2021, working with isee system and Stella, the company set out to address these perennial issues using a comprehensive, but very high level, system dynamics model of their business to study how it might evolve over the next 40 years. A very high-level overview of the resulting model is shown in Figure 1.

Figure 1. High level overview of strategy model

The outcomes of the model were presented not only as year-by-year value projections, but also as ranges of values as shown in Figure 2. This allowed the conversation to focus on the reasons for the shifting of the median and distribution of values in addition to the values themselves.

Figure 2. Illustrative outcomes not to scale.

Because of the commercial nature of this work, it is not possible to provide any substantive details on the model itself. We can, however, reflect on the experience of working with such models.

The project lead from the biopharma company engaged with the team from isee systems to develop, use, and communicate the model and its results. We put some questions to the project lead about how that worked and, most importantly, how it differed from the more commonly used tools for strategy development.

Questions:

1. What was your impression of what value this work might bring when you first got to see a Stella model being demonstrated?
   
   We could quickly see the value for creating multiple, very rich, scenarios that could help us see a range of outcomes over different time periods. As we used the model more, we were able to look more closely around how the interactions within the model generated the outputs we were able to see. This gave us a deeper understanding of how our choices could play out in the context of different uncertainties.
2. What surprised you most about the model that was developed for this project?
   
   What surprised me most were the insights it provided around the sensitivity of different levers we could modulate. It helped us to identify a few areas where uncertainties were not all that meaningful. It also identified areas where we needed to focus and provided rationale for “no regrets” moves that work in any given environment.
3. Did having the model make it easier or harder to explain strategic alternatives that you would likely have discussed anyway?
   
   The model gave us a new way to look at questions we may not have discussed before. This is especially true when we pushed analysis beyond the model itself, for example by using PowerBI visualization and doing optimization experiments.
4. How did you become comfortable with the learnings and insights that came from the model?
   
   We had a team-based, cross-functional group that met regularly to design, validate and test scenarios using the model. The project was supported by a senior leader who provided guidance, and we met periodically with other senior leaders to demonstrate the model, discuss outputs, and iterate our story.
5. Without going into any commercially sensitive specifics, can you tell what you found to be the most intriguing insight that came out of the modeling work?
   
   We were able to clearly see areas of our business with durable prospects, while also showing areas that were sensitive to disruption. This helped us to come up with two significant changes to our long-term investment strategy that we believe will drive our next chapter of efficiency and growth.
6. Were there any areas, again not going into any commercially sensitive specifics, for which you feel the modeling work fell flat?
   
   The model uses a lot of underlying data and relies on critical assumptions from subject matter experts (SMEs) for interactions and their magnitude of impact. This reliance on informed assumptions and the long-range nature of the model means that outputs roughly approximated other enterprise workflows (for example, financial plans) but were not identical. Colleagues who worked closely with the model over time were able to understand the “roughly right” outputs and implications that help us make strategic choices. Others, looking at more tactical choices (e.g., Project X vs Project Y), were not always satisfied with the model outputs.
7. How would you compare the amount learned during model development relative to model use?
   
   We learned a considerable amount in both phases. For the model development, we worked with SMEs from each of our major functional areas. This helped us turn our mental models into a single digital twin model. When the model was validated, we were able to study increasingly sophisticated scenarios to see the impact of our choices against different uncertainties. Later, when we had much more experience with the model, we asked questions in other ways, which helped us think about optimal investment choices that could achieve different outcomes.
8. Can you list a few pros and cons of this type of strategy work versus what you have done more traditionally?
   
   When thinking about cons, it mainly has to do with time and investment of working with the model itself. Given the complexity of this type of project, there needs to be critical input from around the enterprise on a consistent basis. The pro is that you can uncover fundamental insights for the enterprise, looking from the top-down at demand and competition, along with investment and allocation choices that can help the user to understand strategic options in a volatile and uncertain world.

On the Road

Bergen, here we come! We are gearing up to attend this year's ISDC in Norway. Co-presidents Karim Chichakly and Bob Eberlein, lead software developer Billy Schoenberg, and customer service specialist Hilary Allen will be there in person! It will be great to catch up with many of you who we have not seen in far too long. Please stop by our booth or catch us during breaks; we love hearing what you are doing with the software. Also, catch our presentations – we have quite a few this year! Karim will present An Ecosystem model on Post Mining land use and Reducing Veteran Suicide Risk in the USA. Billy and Bob will present Presenting Uncertain Scenarios and Evolving Program Construction as the System Dynamics Conference. Billy will also present Teaching with Loops That Matter in the classroom and Towards a ‘fit for purpose’, fully coupled, integrated, and interdisciplinary world-Earth model- FRIDA Version 1.0. Billy is also a co-author of the presentation Modeling the global macroeconomic system for the next generation world-Earth models.

The South African Chapter System Dynamics Competition

The seventh annual South African Chapter of System Dynamics competition has finished its first stream. To encourage learning and group dynamics, this year’s learning lab is much more focused on collaborative learning than being a competition. The first stream focused on the Fish Banks case study. There were seven participants, and the judges were delighted with the innovative policy investigations they produced. The winner for the first stream was Jurgen Olivier for his ingenious thought process of using one policy (taxes) to fuel another policy (fish regeneration). Jurgen currently works with the Secretariat of South Africa's Presidential Climate Commission (PCC), which engages with various modeling communities focused on understanding South Africa's possible Just Transition pathways to a low-carbon economy.

The second stream focuses on predator-prey dynamics and will begin in June. If you are interested in participating or would like to learn more about the competition, please visit https://systemdynamics.org.za/system-dynamics-learning-lab-24/.

Systems Thinking at Large

Systems Thinking at Large is back! It's been some time since we released a model. This new model focuses on beavers. As the western United States continues to search for solutions to devastating droughts and wildfires, researchers have proposed reintroducing beavers to their former habitats. Beaver-engineered environments have consistently been shown to increase drought resistance and act as a wildfire break, but can they slow climate change, too? Similar to our previous model on the wolves of Yellowstone, this model uses storytelling to examine the benefits of beavers.

Customer Support Hub

Exciting news! We have a new webpage designed to help you quickly find answers to your customer support questions. This page collects our most frequently asked questions for both customer and technical support, and offers resources for every stage of the modeling process. Just think of it as a quick reference page and stand in for when we are out of the office! Check out the Customer Support Hub to see for yourself.

Software Update

Our most recent software releases introduced more innovative new features, from small-scale improvements to large-scale innovations. Highlights for version 3.5.1 include display labels for array elements, new builtins to simplify the parametrization of nonlinear relationships, and the option to search by units in the Find window. Versions 3.6 and 3.6.1 added variable tags (to easily group and organize variables), the ability to import files from the interface, the option to ghost modules, and the allowance of explicit dimension names in array operations. We also made major performance and stability upgrades to the interface. For a complete list of new features, please check out our feature updates.