Process Insights from Using Group Model Building to
Address Emergency Department Use for Pediatric Asthma
Treatment *
Jill A. Kuhlberg
Washington University in St. Louis - George Warren Brown School of Social Work
Jeanine Arrighi
City of St. Louis Department of Health - Children's Environmental Health
Cindy Pulley
University of Missouri St. Louis Departiment of Education
William Kincaid
St. Louis Regional Asthma Consortium & St. Louis University
Abstract
The rate of emergency department (ED) use for the treatment of pediatric
asthma in St. Louis has been four times higher than the rates in surrounding
counties for over two decades. ED use for conditions that can be better man-
aged in primary care facilities drains hospital resources, and creates significant
stress and strain on families whose children need treatment. Several policies to
address this complex problem involving social, health and health care system
factors have failed to produce any long-term results. Insights into effective
intervention require a deep understanding of the complex system in which the
problem is situated. Thus, to address the problem, we organized group model
building (GMB) sessions to bring stakeholders involved with various aspects of
pediatric asthma treatment, care, and prevention to create a qualitative map
of the structure driving persistently high ED use. Insights from the process
include a shared understanding of the system boundaries, elements and feed-
back structures involved in pediatric asthma exacerbations and care, as well
as an understanding of feedback and focus on endogenous drivers of system
behavior among participants. Using the desired insights to drive the design of
the sessions along with frequent reflection kept the GMB process productive
and responsive.
“Paper presented at the 31st Annual International System Dynamics Conference in Boston,
Massachusetts, USA July 21st-25th, The work reported in the paper was supported by the St. Louis
Regional Asthma Consortium, St. Louis Institute for Medical and Educational R
Social System Design Lab at Washington University in St. Louis. Any errors in misrepresentation
are entirely the responsibility of the first author.
Introduction
Nearly 20% of all children in St. Louis City have asthma. Asthma exacerbations
can be life-threatening when not controlled, and data show that over the past two
decades, children in St. Louis go to the emergency department (ED) for acute asthma
treatment three times more than children in surrounding counties and the rest of
the state (Missouri Department of Health and Human Services, 2010). ED visits
for asthma place a burden on the hospital resources and create stress for families
whose children suffer from these asthma crises. Realizing the scope of the problem,
stakeholders from the St. Louis City Department of Health, hospitals, local health
clinics, universities, school districts, and other non-profit organizations have worked
to address this persistently high ED utilization rate for the treatment of pediatric
asthma, however their efforts have been met with considerable policy resistance, as
rates remain high (Sterman, 2000).
With public health problems like ED use for pediatric asthma treatment in St.
Louis that are complex and dynamic, involve several inter-related factors from social,
health and health care systems, and show strong policy resistance, using a system dy-
namics (SD) approach could generate insights into the structures driving the problem
and potential solutions (Homer and Hirsch, 2006; Hovmand et al., 2012b; Sterman,
2006). More specifically, a group model building (GMB) approach involving partic-
ipants who live and work in the systems driving ED use is necessary for the model
and any of its to be accepted by the stakeholders and the community (Hovmand
et al., 2012b). Where the current strategy in asthma research has been to collect
more numerical data, the real progress can be made with a deeper understanding
of the structure, which lies in mental databases of those embedded in the systems
where these problems are (Forrester, 1980). Using GMB can not only help modelers
tap into the mental databases of stakeholders, but can also enable changes in the
mental models of participating stakeholders (Vennix, 1996).
This paper presents the process and insights from a community initiated and
led GMB project organized to address persistently high rates of ED use for pedi-
atric asthma exacerbations in St. Louis with stakeholders from local departments of
health, school districts, federally qualified health clinics, hospitals and other stake-
holders involved in asthma treatment and prevention. In the first section we will
provide some background on the ED use and asthma problem in St. Louis. Next, we
will describe the core modeling team (CMT) activities and the group model building
session design process. Then, we will present the outcomes from the GMB sessions
and insights from those sessions, and lastly close with comments about how the pro-
cess has guided the St. Louis Regional Asthma Consortiums next steps in addressing
ED use for pediatric asthma treatment.
Problem
The ED utilization for pediatric asthma is situated in a complex system where so-
cial, health and the health care system are intertwined. The physiological event of
an asthma exacerbation, where the airways to the lungs become swollen, restrict-
ing breathing, can be life threatening, and extensive research and data have been
collected on its triggers and correlates. Environmental triggers for pediatric asthma
include pet hair, insects, second-hand smoke, mold, and perfume among others (Mis-
souri Department of Health and Human Services, 2010). Sociodemographic factors
including race/ethnicity and income are frequently cited in asthma research. St.
Louis data shows a disparity between racial/ethnic groups in the prevalence of asthma
in children with more African American children diagnosed with asthma as compared
to white children, and a similar disparity in terms of visits to the ED for treatment—
over 90% of the children making ED visits in 2008 for treatment were African Amer-
ican (Missouri Department of Health and Human Services, 2010). Children who live
in low-income communities have also been found to be at a higher-risk of developing
asthma that those in higher income communities (Miller, 1999).
Several of these factors related to asthma prevalence in children are also related to
ED use and hospitalization for asthma exacerbations (Miller, 1999). However, studies
show that prevalence only explains a small piece of ED use and hospitalization. The
causal mechanisms driving ED use for pediatric asthma exacerbations are not fully
understood (Claudio et al., 2006). In the meantime, the current strategy appears to
be to collect more and more surveillance data, however, these data collected by the
health care systems, public health departments, and other public and private entities
have yet to provide insights that have made change in the ED use and childhood
asthma prevalence in St. Louis.
Several programs and policies have been implemented to address ED use in St.
Louis by various stakeholders from the health department and hospitals to schools
and other non-profits. Awareness campaigns about the severity of asthma as a
chronic disease, asthma management plan promotion, and distributing information
of about asthma triggers are all strategies that have been tried but resulted in little
or no impact on reducing ED use.
From a SD perspective, this dynamic problem of ED use exhibits a strong policy
resistance, as its multiple interwoven parts are interacting to compensate for any
of the shocks that concerned stakeholders are introducing into the system to create
changes (Forrester, 1987; Sterman, 2000). And, while the focus has been to gain a
better understanding of the problem by augmenting the numerical database about
ED use and related factors, a deeper understanding of the structure driving ED
use is more likely to come from the stories of those embedded in the system itself
(Forrester, 1980). By tapping into the mental databases of stakeholders working in
asthma treatment, prevention and related fields using group model building, we can
begin to unpack some of the causal mechanisms behind the correlation-based research
efforts , create a shared vision of the problem across the diverse stakeholders involved,
and and moved towards effective intervention.
Project description
Core Modeling Team Formation
The initial idea to use an SD approach to pediatric asthma utilization came out
of one core modeling team member’s work in the health department and a week
long exposure to systems thinking in public health training, where she developed a
causal loop diagram that showed a counter-intuitive system archetype, “shifting the
burden,” to characterize the problem (Senge, 1990; Lane, 1998) (Figure 1). After
enrolling in an SD course, followed by a GMB course at a local university, she had
conversations with the St. Louis Regional Asthma Consortium (STLRAC) about us-
ing GMB to address the persistently high ED use for pediatric asthma exacerbations—
one of the city’s most pressing and complex public health problems. The goal was
to go into deeper detail into the causal mechanisms underlying the general structure
with community stakeholders (Vennix, 1996). She approached the Social System
Design Lab (SSDL) at the George Warren Brown School of Social Work at Wash-
ington University and after several conversations, decided that GMB would be both
a an appropriate and useful approach to addressing pediatric asthma treatment in
emergency departments.
The grant mechanism that supported this project (through the St. Louis Institute
for Medical Education) emphasized the participation of students and the provision of
practice with methods useful for studying and understanding public health problems.
This unique focus allowed the core modeling team, which was assembled after the
support was secured, to build in and spend considerable time reflecting on the GMB
process and learning, giving many CMT members the opportunity to participate in
the various roles during GMB s
The CMT was comprised of eight diverse stakeholders interested in asthma, public
health and/or GMB and SD. The team’s experience in SD and substantive area of
pediatric asthma ranged from some with extensive knowledge about asthma and new
to SD and GMB, to others with experience in SD and GMB and little knowledge
about pediatric asthma, and one member (the project’s community leader) falling
in the middle of those extremes. Members also had a diversity of work experience,
with one member being a trained physician with extensive experience working with
asthma patients and research in St. Louis, another from the division of Children’s
Environmental Health at the St. Louis City Department of Health, three doctoral
students (one in each social work, public health, and educational technology), and
three master’s level students (one from public health, one in social work, and one dual
degree social work with public health). Although some core-modeling team members
had previously worked together on other projects, the entire group assembled to work
as a team for the first time on this particular GMB project.
Quick FKES
Rloscue meds,
ED vans,
Hospaalzations
PROBLEM
SYMPTOM SIDE EFFECT
Urconmeties Mecscal costs,
porustens Quasny of tte
‘ata mparments
exacerbatons
FUNDAMENTAL ISSUE
‘Asthma Care
Figure 1: Initial “30,000 ft view” CLD of the ED use for pediatric asthma problem in
St. Louis. Bl: Children go to the ED for treatment when they have an exacerbation in a
B2: Children can also have management plans that can reduce exacerbations, but
these take time to develop. R1: The more children use the ED for asthma exacerbation
treatment, the less resources are available for longer-term asthma care programs, the more
exacerbations there are being treated in the ED.
crisi
Either the CMT held a meeting or a GMB session every two weeks beginning
in September 2012 through mid-December 2012. The CMT met for three sessions
before the GMB sessions with community stakeholders were scheduled. During these
sessions CMT members shared information with each other about the problem of ED
use and childhood asthma, GMB and SD. Scripts like “Hopes and Fears” allowed
CMT members to share feelings and expectations about the project and their in-
volvement. With such a diverse group on the core modeling team, this script proved
to be very helpful in normalizing fears, as well in understanding what each member
hoped to gain from the project.
The scope and goals for the project were also defined by the group in these first
sessions. Several potential options were discussed that varied in terms of types of
insights we were hoping to generate, types of models produced (e.g., simulation or
qualitative models), types of stakeholders to include, and number of sessions to hold.
Based on the projects timeline, expertise of the core modeling team members, and
a
previous work on the problem, the core modeling team aimed to produce a qualitative
map of the structure driving ED use for pediatric asthma from the perspective of
the stakeholders working in the communities in hospitals, clinics, schools, and other
programs. The insights sought were: 1) to engage participants in understanding
that the ED use problem was embedded in a system, 2) to have stakeholders identify
what parts in the system are relevant to include in a model targeting the reduction
of ED use, 3) to create a shared vision of the structure (i.e., causal links between the
relevant variables) between participants in several participant stakeholder groups.
Although the CMT understood that that SD simulation modeling can provide some
deeper insights into system behavior (Homer and Oliva, 2001), and that qualitative
models can often misrepresent feedback (Richardson, 1996), for the types of insights
that they hoped the project could gain, qualitative diagrams that could reflect the
mental models of diverse community stakeholders was central to values of the CMT
and the project goals. Future work towards a simulation model could come following
this first phase of the project, however the scope of the activities planned for this
project were agreed to be in line with the insights sought.
The CMT decided on using a scripted approach to the GMB process (Hovmand
et al., 2012a), and decided on a total of four GMB sessions. During a planning
meeting, the CMT created a process map to organize how the potential sessions and
project would unfold, including the number and types of participants, and session
outputs (for an example of an early version of a process map see: Figure 2). The
following section will describe in more detail the structure and activities of the GMB
sessions.
Few case studies exist that describe how GMB sessions are organized and un-
dertaken, and none that the authors could find described in-detail this process on
a community initiated and led GMB project. Luna-Reyes and colleagues (2006)
describe their process of designing GMB sessions by describing the scripts they de-
veloped and used. The objective for each script was detailed, followed by a summary
of how the script was carried out (process), and was followed by a critical assess-
ment of the script in addressing the desired objective. Following this framework, we
will describe our GMB sessions script-by-script, including discussion of objectives,
process and assessment.
GMB Session 1: Professionals from a Patient Referral Task Force and Managed Care
The first GMB session included 21 participants (plus the eight core modeling team
members), who were professionals working in health clinics, hospitals, managed care,
and emergency departments in St. Louis. Inviting professionals from the along the
continuum of childhood asthma care was the plan for this first session to take a first
pass at some of the relationships driving ED use. During this session, a short verbal
introduction to the problem was given, and a reference mode showing the desired
and feared trajectories of ED use for pediatric asthma exacerbations drawn on a
whiteboard. Two scripts, “Behavior Over Time Graphs” (BOTGs) and “Structural
2012
Planning &Training(Sep. 14) | Planning & Planning & ‘Model od it | Transfer of
Training Training Formulation | Formulation | Formulation | Formulation | Ownership
(5p.25) (0.9) (oet.23) {Wow 6) (Wow. 13) (Wow. 15) (ee. 4)
Community | Rindersot | i peda
ieee eres tree
—
ht
L
[35
Home!
Families
—
Policy
Makers
ele) ete 1
core
‘Modeling
"EB & fa
Figure 2: Example of a process map created at CMT meeting, outlining the GMB sessions,
participants, and outputs.
Elicitation”, were used to structure the rest of the session.
Script 1: Behavior Over Time Graphs
Objective This script has several objectives: 1) orient participants to the dynamic
problem that is the focus of the session, 2) elicit variables related to the problem, and
3) encourage participants to think about the problem of ED use for pediatric asthma
exacerbations as a dynamic problem and in the process, understand some of the key
variables and relationships related to problem. This exercise has an important role
in giving all participants an opportunity to speak or (when working in pairs on the
task) have their voice heard at least once in front of the entire group.
Process The facilitator encouraged participants to hand-draw “behavior over time
graphs” (BOTGs) of variables that “cause or are affected by ED use for pediatric
asthma exacerbations.” The facilitator drew one herself in real-time, talking out loud
as she describes the components each graph should include: the x-axis for time (any
time scale and horizon the participant chooses), the y-axis for the variable chosen
(percent of children hospitalized, number of trips to the hospital per month, etc.), a
trend line that shows past behavior as the participant understands it to be, a dotted
line indicating the present time, and desired and feared trajectories. Drawing this
in real-time instead of having one example that had been pre-made has increased
the success of participants’ creating their own graphs with all of the useful elements
clearly included.
The example the facilitator drew had nothing to do with the topic of the GMB
session so as not to influence the participants, and was on a time scale that was
probably also not one that the participants would choose (the particular example
had something to do with stray dogs beginning in the year 1713 and going until
2080). Our experience is that the example the facilitator chooses can heavily shape
the participants’ graphs. If the example shows an upward trend in a variable in
the past, and an increasing upward trend as something feared, and a negative trend
as desired, many participants will follow suit framing all of their variables that way.
We have thus, created graphs that have unrealistic oscillations (sometimes with curly
loops), have time horizons that go centuries into the past and future, and use variables
like stray dogs in the city and price of gold.
After walking through the example, the facilitator gave the participants fifteen
minutes to create as many graphs as they could. Since the group was large (more
than 10), we asked participants to pair up and create graphs together. As things
slowed down, the facilitator asked the participants to order their graphs with the one
they liked/felt most strongly about on the top. Then each pair shared their graph
with the facilitator and group, telling the story behind it. The facilitator began with
a graph that was especially clear and included all the elements the example graph
included. A “wall-builder” then taped the graphs to the wall as they were shared,
organizing them into thematic groups as they emerged.
Participants shared until there are no graphs left with new stories. Many par-
ticipants continued making graphs if something came to mind while others were
presenting, and were encouraged to do so quietly. The wall-builder then shared the
groups she constructed and asked for feedback from the group on the groupings,
asking if any categories need to be changed or any graphs needed to be moved.
Assessment The participants who work in several sectors associated with pediatric
asthma in St. Louis agreed that focusing on ED use for pediatric asthma exac-
erbations was an important problem to come together to address. Ten pairs and
one group of three generated 25 unique BOTGs, that the wall builder grouped into
four groups: environment, issues related to parents, access to medications and insur-
ance, and quality medical care. The participants seemed engaged in the creating the
graphs (many drawing more even as the sharing started), and commented both on
how both they had not thought of some of the variables others had, and how others
were “exactly what we had drawn.”
Professionals in the health and social service fields are bombarded by statistics
that stratify individuals by race/ethnicity, gender, socioeconomic status and several
other “status” variables. Correlation-based thinking is not what is needed to create
Figure 3: Photos from first GMB session. (Clockwise from top right:) Participants creating
BOTGs, Wall builder grouping graphs, participants choosing important variables, modeler
during structural elicitation script.
CLDs or SD models, and the BOTG script helped to shift participants away from
thinking in terms of correlations and “sociodemographic predictor variables” and
towards variables that demonstrate change over time. The group seemed comfort-
able thinking about the operational causes and effects of the ER use, and no one
questioned the use of the word “cause” or “effect” (as can happen when participants
are used to using the word cause with much caution), or asked “What about race?”
during the script.
Script 2: Dots
Objective The objectives of this simple script are to get an idea of the variables
the participants find to be the most essential to understanding the ED use problem
and to reduce the number of variables from generated in the BOTG script to a more
manageable number (five or six) before moving into the structural elicitation script
that follows.
Process Using small sticker dots (three for each participant), the group voted for
which BOTGs they think were the most essential to consider when thinking about
ER use for pediatric asthma. They could place their sticker dots on any of graphs,
9
and could have even given multiple sticker dots to graphs they found to be especially
important. The facilitator assured the group that these variables were just a starting
point, and that more will surely emerge from the conversations in the next part of
the session. Participants were invited to walk up to the wall to place their stickers
before taking a small refreshment break.
The facilitator for the structural elicitation script, modeler, wall builder and pro-
cess coach (in charge of things related to time management and making observations
about interactions and dynamics within the session) met to discuss graphs that were
receiving the most votes. After all the participants had cast their stickers, the mod-
eler for the next script writes the names of the top four variables scattered on the
white board, along with a variable of ED visits. The dots script took about 6 minutes,
and was included in a ten-minute break for the group.
Assessment This short script ran smoothly and maintained a positive environment
in the convening room. Participants discussed the process with others at their table
and at the wall while making their choices. These conversations continued throughout
the break and the general energy in the room was buzzing and positive. Sixteen of
the twenty-five BOTGs received at least one sticker vote, and four had more than
six stickers attached to them: referrals to primary care providers after ED visits,
improper use of inhalers, ED discharges with no prescription or plans for asthma
control, access to primary care providers office, and relationship between patients
and their primary care providers. These were used to start the structural elicitation
following the break.
Script 3: Structural elicitation
Objective In the structural elicitation exercise the objectives are to make a first cut
at the causal mechanisms driving the dynamic problem. One insight the script aims
to foster is that there is complex system driving the problem, with many components,
all moving together. At a deeper level, the idea is to move participants from a creating
a“dead buffalo” diagram where several variables are directly causing one outcome, to
one that shows operational causal linkages and feedback loops that could be driving
the problem (Sterman, 2000).
Another important objective in the script is accommodating the inclusion of
several perspectives in the CLD. Especially at the beginning of a GMB project,
where stakeholders are from several different sectors, and have different power in their
professional life, accommodating these diverse perspectives is important in team and
trust building in the GMB process.
Process During this script the facilitator, first gave an example of how the group
would show how causal linkages would be depicted on the whiteboard, and gave an
example of a balancing and reinforcing loop using an example based on the stray
dogs variable (used as the example in the BOTG script).
10
The facilitator then explained that the group would draw the same types of causal
linkages between the variables they had identified in the BOTGs and that they were
free to include other variables as they went along. The first link drawn described
that as the quality of the relationship patients have with their primary care providers
(PCPs) increases, their use of the ED decrea: Others discussed how discharge
plans for patients could also decrease their use of the ED in the future, especially
when there is communication between the ED and the patients PCPs. Both of these
linkages were influenced patients use of controller medication. Participants told
stories about how patients knowledge of how and when to use controller medications
can and appointment availability at PCPs affect their decisions to go to the ED for
treatment.
Conversations and links were rapidly being suggested from the participants and
the facilitator and modeler worked together to ensure that the stories told were re-
flected in the structure drawn on the board, which proved challenging (as the modeler
was having to draw variables and links as they were quickly being suggested for all
to see on a wall of dry-erase boards). Participants included variables that discussed
in the earlier BOTG script and integrated others not previously mentioned easily.
Some variables that seemed to be outside of the model boundaries like poverty were
suggested, and while the group acknowledged that addressing poverty was beyond
the scope of the model, for the time being, participants felt more comfortable includ-
ing it and “social determinants of health”, as long as the model was going to be a
“rough cut” and "first draft”.
The balancing loop between ED use and asthma exacerbation was identified and
discussed in the group, but the early stories shared did not describe any other feed-
back loops (Figure 4). The facilitator guided the participants attention to the“dead
buffalo” nature of the emerging diagram, and asked participants to focus on rela-
tionships between variables that were already on the board. The participants stories
began linking more variables, but no one was describing how ED use affected any-
thing in the model. The facilitator asked the participants to consider what variables
it affected, but time was running out and while the interest in CLD building appeared
to be in full swing the session had to end. The CMT met following the session to
debrief.
Assessment While this session did produce a preliminary causal map, participants
wanted more time to add more of their experiences to the diagram. This was ad-
dressed by discussing that nothing about this diagram had to be final or perfect and
that the model building is an iterative process. However, the CMT reflected on these
reactions from the participants and the lack of time to focus on identifying feedback
loops in the planning for the next sessions, both in terms of time and agenda setting
and in the way the Structural Elicitation script was carried out. The CMT entered
the diagram created in the first session into Vensim and redistributed it to all the
participants of that week along with some photographs from the session, both so they
could think about the diagram and other relationships they might want to discuss
11
ER Utilization for Pediatric Asthma Treatment Model (Version 1: October 23, 2012)
This project is led by the St. Louis Regional Asthma Consortium with funding from the Institute of Medical Education and Research and in
collaboration with the Social System Design Lab at Washington University in St. Louis.
—
Poverty + Social determinants Cat
} — oustnites
Housing Quality Multigenerational
cohabitation
i Asthma crisis
Provider's
OO Chon + effective teaching
siree wna.
Smoking head of
households © EReommunication ‘Asthma education
Ye ~\) + Fa
+}
Capacity of PCP
- Z LV
WR Rer’ wee ie \ Z|
sthma as chronic
condition
Urgent access |” +
to \
Funding for
prevention
Quality ED ee
discharge plone Taso halen
Routine access to
care ot PCP office
‘Young parents z 3
relationship with a
Figure 4: Handout created for participants from the Integrated Health Network.
and to remind them that they were invited to future GMB sessions (Figure 4).
GMB Session 2: Community health workers
This second GMB session included 16 stakeholders the core modeling team character-
ized as community health workers. These participants were professionals including
school nurses, managed care providers, wellness and tobacco cessation coaches.
Script 4: Structural elicitation
Objective The objectives for the structural clicitation script are the same as in the
previous section.
Process Learning from the timing problems in the first session, this session began
with the structural elicitation, with the same variables the first session stakeholders
had chosen (cutting out the BOTGs and Dots scripts that were included in the
previous session). Also in this session, instead of drawing the CLD on the board, a
12
modeler created the diagram in real-time in Vensim, the modeler-facilitator led the
conversations informing the structure for approximately 90 minutes. Participants
drew causal links between the initial variables, quickly nominating other relevant
variables.
Since time seemed to constrain the identification of feedback in the previous
sion, the process coach gave a signal to the modeler-facilitator to begin a shift in
identifying feedback loops with 30 minutes remaining for the script in this session.
As the nomination of links began to slow, the modeler-facilitator asked for any addi-
tional variables that participants felt needed to be represented that were not already
included before handing the facilitation to the process coach. The day’s accomplish-
ments were then reviewed and situated within the context of the modeling project
scope, and participants were given a chance to reflect on their experience in the
process and share their observations. CMT members debriefed following.
S-
Assessment Participants seemed much more satisfied with the modeling process
during this session. Their contributions to diagram were rapid at first and seemed
to slow down towards the end of the time allotted for the script, instead of being cut
off by time restraints as in the previous session. Still, feedback relationships were
not drawn.
The CLD was cleaned in Vensim and sent to participants the following week
(Figure 5). Core modeling team members worked to integrate the CLDs and clean
them for presentation in the following GMB session. Emphasizing and identifying
feedback in the diagram was set as the goal for the third session as both CLDs had
several links leading directly to“asthma crisis” and “ED use”.
GMB Session 3: Combined Stakeholder Groups
This third session had thirteen participants. Stakeholders from the first two sessions
were invited to attend if their schedule allowed, and two participants attended for
their first time. The session lasted 90 minutes, and the script for the session was
used as a result of the observations and CMT reflection that time needed to be taken
to pull out more feedback from the stories of the participants.
Script 5: Merging maps and closing loops
Objective The objective of this script, was two-fold: 1) bring diverse groups of
stakeholders together to make steps towards a qualitative diagram that reflects the
experiences of all those who worked to create it, and 2) work with the participants
to identify and capture relevant feedback loops in the structure. Special time was
taken to ensure that variables with the same name in different diagrams had the
same meaning in the shared diagram.
ER Utilization for Pediatric Asthma Treatment Model (Version 1.2: November 6, 2012)
This project is led by the St. Louis Regional Asthma Consortium with funding from the Institute of Medical Education and Research and in
collaboration with the Social System Design Lab at Washington University in St. Louis.
exposure 19 doctor's attention To
tection nen “eer necctue
andions cess ON
crgert are
riggers
parent's or family
time off work
care at POP office \
Spacers DME
\ overuse of short
term medication
transporttion
provider‘s eget parents or 7
effective teach compliance with legal guardian _~
ee ing a a
= asthma treatment
Figure 5: Handout created for participants from the second GMB session with community
health workers.
Process The modeler facilitator reviewed the process from the last two sessions
and explained the tasks at hand for the current session. As the two CLDs had been
integrated off-line by the CMT, some time was spent explaining that process and any
questions or initial reactions were discussed before the work on the model began.
Then, the participants worked in pairs to identify links that they believed exist
between variables already in the diagram, but that might not have been discussed
in the previous ms. Each participant had a paper copy of the current version
of the CLD and a blank paper to note the variables, relationships, and polarities
that they considered relevant but missing from the diagram. The modeler facilitator
asked them to make note if the links they were making closed any loops. After about
10 minutes of working independently, the partners began to share the links they
wanted to add, and the stories that accompanied them. While participants shared
the stories, a modeler added the links in realtime in Vensim, projected on a large
screen, in a similar way as had been done in the Structural Elicitation script in the
second session. The facilitator highlighted feedback and delays when possible.
Assessment This session made steps in moving the diagram from one of correlations
and linear thinking to one of operational mechanisms and feedback. Participants
were able to remove some variables that no longer seemed relevant, and add in
intermediary variables that made up the causal structure of the system. Those who
had not seen Vensim before reacted very positively to the real-time modeling. At
the end of the session, participants appeared energized after having made another
“cut” at the model and were anxious to see what it would look like cleaned up with
fewer crossing lines and the removal any unnoticed redundancies. They trusted the
CMT to do this offline after understanding and seeing the output of this process
several times by this time in the process, which showed a high level of trust between
the stakeholders and CMT members. The session ended as the previous sessions
did, with a review of the session and discussion of what to expect in the following
meeting. The CMT met following the sessions to debrief and discuss preparations
for the final meeting.
An additional core modeling team mecting was planned to go over the notes from
the previous sessions to clean up and simplify the CLD, and identify any thematic
subsystems in the model. Revisiting the stories of the participants and using the
knowledge of the core-modeling team members with experience in asthma treatment,
the CMT was able to remove several direct links from variables to ED use that
had previous versions looking like “dead buffalo diagram and reflect them more
appropriately through intermediary links (Sterman, 2000). The model refining was
done as a group, and creating a cleaner visual representation was then done outside
of the meeting between two core modeling team members.
GMB Session 4 (Integrated CLD Presentation): Combined Stakeholder Groups
Script 6: Transferring Ownership
Objective The objective of the Transferring of Ownership script is to review the
GMB project goals and the work accomplished through the GMB sessions, go over
the current version of the model, and discuss the next steps for the project. Both the
facilitators and participants discuss how they might use the model moving forward.
Process The community facilitator opened the session with a presentation reviewing
the motivation for the use of GMB to address pediatric asthma treatment in EDs in
St. Louis, which began with a brief overview of numeric data available on rates of ED
use and hospitalizations over time, the initial “shifting the burden” qualitative model,
and discussion of how coordination across several organizations and individuals was
needed to make any headway on reducing the rates. The review of the background
and motivation for the project was followed by a review of the work of the previous
GMB sessions and work done by the CMT betwe: ms, these included pictures
of the models at various stages, and photos of the participants at work during the
sessions and CMT meetings.
on se
‘sTappoyayeys AIUMUIUIOD YIM WoIsses GPT [PUY oy} ye poqoseid GTO :9 oS
au0y
‘ayy ul Supjows
sepipiqiow-o2
pu
2y7140 AenD
Henb jeUIWLOI A? ———
uysnoy
sosinu jooyas
2B Jooy>s yaym uone>tununLo:
ee soubpuane
ere 7 Jooups
——
6
ooyos 1e
eld uome ewupse
souelnsuy
‘axon 0} 559290
eld uome
sysin weuniedep
ewyise aw91in>
Ao batt
‘anosau jo asn
syeayL
a eel spun ewngse 8
Sanprquow-oo'
eungse Bussasppe 5
quawa8euew pue : wone2ppew/sa) 5
suoneajeyds
ws Pres emjendsoy i
quaussasssy [eIPaW, Bone ba
= se ate> jo swouddis0>
jain 03 59200
f “ woupu0. eM <= juan 01 eugse 40 uypury
uo si ti054 - 2wuouyp se ewyse :
euiyisejo sn ir Jouopdanied sauaned
a
Apuauyewse 4
+ a7
swawow a
‘aigey>eas eunse soypeon euiyase
a
quawssasse
Josauos ewyase /moy:
uogespaw yead auanos
16
The presentation part of the script closed with a discussion of the current version
of the CLD (Figure 6), which had been cleaned up between the previous GMB session
by the CMT as a group. Six sectors of the model came out in the cleaning process,
and those were highlighted in the presentation: education, threats to asthma exacer-
bations, quality of life, environmental triggers and co-morbidities, medical assessment
and management, and medication. These sectors and corresponding color-shadings
came out of the model cleaning process and thus some time was gone over discussing
that process with the participants.
The community facilitator went over feedback stories and loops that had been
told in previous sessions and some new feedback loops had been identified in the
model cleaning process. While going over these sectors of the model, the community
facilitator asked the participants if the diagram reflected what they experienced or
had told in previous sessions so as to ensure the cleaning process did not miss or
mis-represent stories or relationships shared throughout the project. Participants,
particularly those for whom it had been their first session had questions about the
relationships that others who had been in the sessions answered.
Ideas about how the CLD could be used in the participants’ work and how it might
be used in the future by continuing the modeling process were shared. Participants
were asked to think about and share where their organization’s work or their own
position’s fits into the diagram, guided by prompts such as: what if any gaps were
evident that could be filled, what new links needed to be created or removed, and
how could participants (and their organizations and programs) use the CLD in their
work. The session ended with these conversations in a large group with participants
also adding what they had taken away from the process. As with the previous GMB
sessions, the CMT debriefed following the session.
Assessment The tone of the room was positive during this final session. Although
the majority of the participants had participated in at least one of the previous
sessions, for some it was their first meeting. Having new participants at a session
where the majority of the work has been completed (at least for this phase of the
project) can be challenging as they are not familiar with the GMB process and SD
and can come with questions that come from that unfamiliarity. In one case, a
new participant was eager to find which “one thing” could be changed to fix the
problem, and in this instance participants were able to explain that finding that
“one thing” was not the purpose of these sessions, and that maybe there was not
any one solution. The CMT reflected that the participants’ ability to explain this
showed a strong understanding and engagement in the project.
It was also considered positive that stakeholders were eager to continue in upcom-
ing GMB processes. The community facilitator explained that possible next steps
for the project as organized by the Regional Asthma Consortium and St. Louis City
Department of Health included working on building a simulation model in additional
GMB sessions, involving other groups of stakeholders in adding to the qualitative
model, and/or taking the model “on tour” to several asthma focused task forces and
17
interest groups to get more feedback and interest in participating in future modeling
efforts. Many participants reiterated that it would be extremely useful to understand
the experiences of the children with asthma and their families using GMB that they
felt uneasy trying to include in the CLD they had worked to create.
Insights from the GMB Sessions
The goals of the project as defined by the core modeling team were to develop
certain levels of system insights with stakeholders through GMB. Through GMB,
stakeholders and CMT members developed insights about the system boundary,
structure, operational mechanisms, and feedback. The core modeling team also
learned from the process of GMB especially about the importance of insight and
project goal-driven design of the sessions and process oriented reflection amongst
CMT members.
Complex System Insights
One of the defined goals of the project was to develop the insight that the ED use for
pediatric asthma exacerbations problem was embedded in a complex system. And
while this seems intuitive to system dynamicist, mental models of participants early
on often focused on one variable being a “the main cause” (i.e.,“parents dont see
asthma as a chronic condition” , “primary care doctors cant schedule patients when
they need it”, “medications are too expensive”). Very quickly participants began to
see that it was +” any of these causes but the interaction of all of these things.
Participants found the value in mapping the structure and seeing the complexity
represented visually.
Boundary Insights
Stakeholders spent time discussing the model boundary, especially as it pertained to
the inclusion of poverty and insurance. The discussion of how the model boundary
and the scope of the modeling project would change in order to make poverty en-
dogenous to the model was a helpful one in that it appeared to be empowering to
participants since the focus was “on things what we have control changing” and not
feeling overwhelmed by one variable that seemed to have relationships to everything.
This was also a key variable in conversations about causal vs. correlation-based re-
lationships in SD. In the end, certain aspects of poverty like available income were
included in the model, whose meaning was negotiated by participants, and which
seemed to be more on the level of aggregation of other aspects of the CLD and
within the boundary.
The boundary of the model was questioned frequently when participants had con-
versations about the inclusion of the child patient/family perspectives. Participants
felt uncomfortable about making assumptions about patients and their families, and
18
so the participants were reassured that these sessions were preliminary sessions to
start to understand the issues, and that they were not responsible for expanding the
boundary to include every aspect of the problem. The participants and the CMT
did express that future modeling processes must involve patients and families.
Structural Insights
Creating the space for several stakeholders to share their mental models about the
causal linkages between several factors in GMB sessions did not come without dis-
agreements amongst participants. In every session, several causal links were debated,
and in come cases, quite heatedly. In these cases, where the participants mental mod-
els were not the same, structure was agreed upon that could show several points of
view (often by adding intermediary variables) or a structure was agreed upon that
reflected a consensus opinion of the situation. Having the CLD as a focus of atten-
tion shifted the discussions from having a blaming tone or alienating the voices of
participants offering alternative perspectives.
Insights Moving Toward Operational Thinking
The ” Behavior Over Time Graphs” script in the first session was useful in setting the
tone for participants to think about variables that explicitly cause or are affected by
ED use for pediatric asthma exacerbations, away from variables used frequently in
correlation-based research like race/ethnicity and gender. During structural elicita-
tion scripts in both sessions, when variables like these were brought up, participants
were pushed to describe what they thought were the causal mechanisms under that
named correlation. In the first two sessions, a variable concerning the age of the par-
ents of children with asthma was explained by different participants as being related
to developing and sticking to asthma treatment plans.
This one variable had several stories beneath it, including ones that described
younger parents lack of time, shorter attention spans, different understandings of
responsibility, lack of knowledge about the gravity of asthma as a health concern,
among others. Participants appeared to be challenged when asked to present these
mental models, but normalizing the experience by sharing that these CLDs require
this operational perspective created space for participants to explore their own mental
models and question and compare theirs to those of others.
Feedback Insights
The importance of the feedback was emphasized throughout the GMB sessions and
participants had no trouble relating to reinforcing loops and balancing loops as they
were presented, placing them in the context of asthma treatment. Although feedback
loops were identified as they were made in all three working group model building
sessions, an important insight came at the third session when a new participant pro-
vided new information about a hospital program designed to give more information
19
to patients about asthma management. If the goal was to shift the burden away from
the EDs and hospitals, creating a program that addressed more of the patients needs
at the hospital or ED, might only strengthen the loop of families taking their children
to hospitals for treatment. The relationship of feedback to unintended consequences
of policies became very real to participants and CMT members.
Complex Problem Solving Insights
In the final session, where the integrated and simplified CLD was presented to stake-
holders, some newcomers were struck by the complexity of the diagram, and were
anxious to know where to intervene. The city had in past years reduced a major
problem with lead poisoning, and one participant wanted to know where in his words
“removing lead paint in the windows was in this diagram”. Using the model, the
facilitator could explain how although lead poisoning and asthma are both pub-
lic health problems and both most likely to occur in low-income communities of
color in the city, the mechanisms are quite different. Although lead poisoning was
a complicated problem, childhood asthma and ED use are more complex. Just as
lead poisoning required a coordination of efforts to address, these ED use for asthma
problem requires that as well, but after the complexity of the problem and important
feedback loops are understood. The participants involved in the modeling sessions
seemed to understand that there was no solution, like “removing windows” for the
asthma problem.
Process Insights from the Group Model Building Project
A core component of group model building is co-creation with stakeholders not just of
models or qualitative diagrams but also of the insights throughout the process. This
was certainly apparent in the current GMB project. Members of the CMT, with their
combined domains and levels of expertise, might have just as easily created a model
that looked in the end very similar to the one created by participants in modeling
sessions, however, the insights would have remained with just them. For complex
problems like ED use for pediatric asthma exacerbations, intervention strategies will
require the coordination of several stakeholders, and that requires a shared vision of
the problem, its drive, how certain goals can be achieved and why others might fail.
Steps towards intervention, thus, are facilitated through GMB more than through
other model building strategies that do not include community stakeholders from the
beginning.
The design of GMB sessions should be driven by the project goals and insights,
but should also be flexible. Process maps like an early one from this project shown
in Figure 2, are invaluable tools for organizing and viewing the scope of a project
and how components work towards the project goals. In designing these sessions,
practical constraints like the availability of participants in terms of time and available
meeting spaces to adequately accommodate participants and goals for the session
20
influence the kinds of scripts that you can use to meet session goals. While process
maps for the entire project are helpful to make in the planning phases, adjustments
can be made as needed as goals evolve, things can be scaled up or down, or change
in other ways. Our process map changed several times responding to changes in
varying scope. Our earlier stages of planning had more stakeholder groups and
then reduced in scope, went between separate meetings for different stakeholder
groups to careful mixing of stakeholder groups, planned for using the same scripts in
initial sessions with stakeholders and based on session reflection changed things for
subsequent sessions. However, despite changes in the process, the idea of designing
these sessions with a common goal in mind kept the project on track and sessions
productive.
Reflection or debriefing after GMB sessions serves many important purposes.
Through group reflection the CMT could give each other feedback and discuss what
went well and what needs work both on the modeling team member level and the
session level. These post-session discussions helped to inform adjustments in the
process map, CMT roles in subsequent sessions, and helped members improve their
GMB practice, which was especially useful as there was an expressed interest in
CMT members to build their own capacity in group modeling building and system
dynamics.
Conclusions and Next Steps
The persistently high rate of emergency department use for the treatment of pediatric
asthma in St. Louis is a complex problem caused by the interaction of social, health,
and health care system factors. Although several organizations have implemented
programs aimed to reduce the use of the emergency departments for conditions that
could be managed by the patients and families, they have not seen their desired
results. This project, using GMB with stakeholders who work in the community,
hospitals, clinics, schools and non-profits was a first step in understanding the drivers
of this problem and developing a shared vision amongst stakeholders for what the
problem is and how what they do at their organizations affects other parts of the
system and the problem. This project was was uniquely both community-initiated
and community-organized. Even with no experts in GMB or SD on the CMT and no
simulation model, several important insights were gained in the process of creating
this iteration of a CLD that were shared amongst CMT members and stakeholder
participants.
The St. Louis Regional Asthma Consortium, St. Louis City Department of
Health, and other core modeling team members have taken this initial causal loop
diagram “on tour” to highlight the productive work towards understanding the prob-
lem accomplished in the sessions in a short amount of time and to receive feedback on
what was done and engage more stakeholders. In the spring of 2013, CMT members
worked with students from a GMB course to organize sessions with children who have
21
asthma and their families— the next step that was most strongly advocated for by the
stakeholder participants themselves. These sessions kept with insight-driven design
process and continued with the work started in these first sessions by reflecting the
patient stories in a CLD. Work is now being done to integrate these CLDs and begin
ons that work towards producing insights that can be gained through
planning s
simulation.
22
References
Claudio, L., Stingone, J. A., and Godbold, J. (2006). Prevalence of childhood asthma
in urban communities: the impact of ethnicity and income. Annals of epidemiology,
16(5):332-340.
Forrester, J. W. (1980). Information sources for modeling the national economy.
Journal of the American Statistical Association, 75(371):555—-566.
Forrester, J. W. (1987). Lessons from system dynamics modeling. System Dynamics
Review, 3(2):136-149.
Homer, J. and Oliva, R. (2001). Maps and models in system dynamics: a response
to coyle. System Dynamics Review, 17(4):347-355.
Homer, J. B. and Hirsch, G. B. (2006). System dynamics modeling for public health:
background and opportunities. American Journal of Public Health, 96(3):452-458.
Hovmand, P. §., Andersen, D. F., Rouwette, E., Richardson, G. P., Rux, K., and
Calhoun, A. (2012a). Group model-building ‘scripts’ as a collaborative planning
tool. Systems Research and Behavioral Science, 29(2):179-193.
Hovmand, P. S., Nelson, A., and Carson, K. (2012b). Understanding social determi-
nants from the ground up. In Society, S. D., editor, International Conference of
The System Dynamics Society.
Lane, D. C. (1998). Can we have confidence in generic structures? Journal of the
Operational Research Society, pages 936-947.
Luna-Reyes, L. F., Martinez-Moyano, I. J., Pardo, T. A., Cresswell, A. M., An-
dersen, D. F., and Richardson, G. P. (2006). Anatomy of a group model-building
intervention: building dynamic theory from case study research. System Dynamics
Review, 22(4):291-320.
Miller, R. L. (1999). Breathing freely: the need for asthma research on gene-
environment interactions. American journal of public health, 89(6):819-822.
Missouri Department of Health and Human Services (2010). Asthma in St. Louis
City. Missouri Department of Health and Senior Services.
Richardson, G. P. (1996). Problems for the future of system dynamics. System
Dynamics Review, 12(2):141-157.
Senge, P. (1990). The fifth discipline. New York: Currency Doubleday.
Sterman, J. D. (2000). Business dynamics: systems thinking and modeling for a
complex world, volume 19. Irwin/McGraw-Hill Boston.
23
Sterman, J. D. (2006). Learning from evidence in a complex world. Journal Infor-
mation, 96(3).
Vennix, J. A. (1996). Group model building: Facilitating team learning using system
dynamics. J. Wiley.
24
