Applying Systems Thinking to Improve a Hospital Food Retail Environment

Objective: To identify factors inﬂuencing the availability and sales of healthy food and drinks in a caf (cid:1) e located in a hospital setting in a rural area. Methods: Three online and 1 in-person group model building workshops were conducted with hospital staff members to develop a causal loop diagram. Results: Four areas in the causal loop diagram were identiﬁed, 5 teams were created to implement 15 iden-tiﬁed action ideas, and an action registry was created to track their progress. By May 2023, 4 actions were active, 6 inactive, 4 completed, and 1 abandoned. Conclusions and Implications: The group model building process identiﬁed factors and actions to improve the healthiness of the hospital’s caf (cid:1) e and motivated staff members to act for change. However, progress was limited by staff turnover, recruitment, and inadequate participation from decision-makers. Better leadership and support by senior management can ensure that long-term objectives are achieved and health-ier hospital food environments are sustained.


INTRODUCTION
Unhealthy diets (ie, excessive intake of overprocessed food and low intake of fruits and vegetables) 1 have been recognized as a significant driver of the increase in the number of preventable noncommunicable diseases globally. 2his trend is reflected in Australia, where most people do not meet the national dietary guidelines. 3−6 As a consequence, they experience worse chronic disease profiles than their metropolitan counterparts. 7,8−13 This complex and ever-changing environment includes not only traditional grocery stores (eg, supermarkets or convenience stores) but also food service operations (eg, caf es or restaurants). 14any hospitals and health services have caf es or restaurants available for staff, patients, and visitors. 15,16erefore, these settings have the potential to promote and facilitate healthier food options, thereby improving the health and well-being of all those who use them. 16However, healthy options are not always sold or provided in hospitals and health services, contributing to the chronic diseases they aim to treat. 16,17The state government in Victoria, Australia, has attempted to address this issue through a regulatory mandate (the Healthy Choices mandate). 18It aims to ensure that health services increase the availability and promotion of healthy food and drinks while reducing the less healthy options. 18,19he relationship between individuals and their food environment is multilayered and influenced by various physical, social, economic, and political factors. 14,20It has been suggested that solutions may be more effective when embracing complexity and acting at different levels using multiple strategies 14−16 ; in particular, when engaging with multiple individuals to explore the effects of initiatives aimed at improving food retail environments. 15,16The application of systems science in local settings to address complex problems is gaining momentum. 21This can be done from a range of approaches, such as reframing complex problems from simple, linear causal models to complex models of multiple interdependent and interacting components that change over time and influence one another within a connected whole. 21,22Community-based system dynamics (CBSD) is one approach that builds and extends on socioecological models, highlighting the interconnections and feedback loops among actors, factors, sectors, and levels of a problem, 21,23 allowing the visualization of complex issues and their potential solutions from a nonlinear perspective. 16,24A key aspect of CBSD is community engagement in group model building (GMB) and system dynamics activities, with an emphasis on capability development among participants. 23,25roup model building supports participatory systems thinking by involving stakeholders more directly in the modeling process, such as creating a causal loop diagram (CLD). 21,26asual loop diagrams are used in a variety of ways, such as a way to surface, visualize, and explore mental models. 26They can be used to visually represent the group's shared understanding of the problem and the interconnections in the system of interest. 26Finally, they can be converted to stock and flow diagrams and differential equations to represent sources of feedback underlying dynamic system behavior. 26espite the high need to promote health-enabling food retail environments in hospitals and health settings by embracing the complexity of the problem, 24 there is limited literature exploring food environments from a systems science perspective. 27n addition, to our knowledge, peerreviewed literature related to the implementation of the Healthy Choices mandate with a systems science view is also limited.Therefore, this study aimed to (1) describe GMB participants' shared understanding of the factors influencing the availability and sales of healthy food and drinks in a caf e located in a hospital setting and (2) to report on the progress of identified and proposed actions to improve the caf e food environment.

METHODS
This study was guided by CBSD and was part of an ongoing communitybased participatory strategy to improve the hospital caf e food environment.

Setting
This study was conducted to improve the healthiness of a caf e located in a rural City Council of Victoria, approximately 300 km northwest of Melbourne, Australia.The hospital employs 1,118 team members and provides health care to > 10,000 inpatients, 16,000 emergency patients, and 123,000 outpatients annually. 28This caf e is in the hospital's foyer and provides food and beverages to hospital staff, patients, and visitors.

Participants
Participants involved in the GMB workshops were hospital staff members from diverse areas (eg, health promotion, dietetics, communications, allied health, procurement, management support hospital departments, food service, and caf e staff).Participation in the GMB workshops was open to any staff member working at the hospital.An email with the project details was sent by the health promotion area to all staff members with an open invitation to attend the sessions.For the community presentation, additional flyers were placed at the caf e counter.Participant inclusion criteria were extended to any person who purchases at the caf e (ie, patients or visitors).

Group Model Building Workshops
Three GMB workshops were conducted between September and November 2021 with key stakeholders via Zoom. 29In May 2022 (when coronavirus disease 2019 restrictions allowed group gatherings), a fourth workshop was conducted with the community, in which the resulting CLD was validated face-to-face to ensure accurate representation.The modeling team, comprised university researchers, adapted scripts for the GMBs from Scriptapedia 30 (GMB Scripts 31 ; Supplementary Material) and proceeded as shown in Table 1.The GMB workshops explored factors influencing the availability and sale of healthy foods and drinks at the caf e.This information was organized into 4 themes using the Systems Thinking for Community Knowledge Exchange tool 32 to create a CLD.The GMBs were facilitated by researchers with expertise in CBSD and GMB procedural training (ie, meeting convener/closer, modeler, facilitator, note takers, and debriefer).All workshops were audio-recorded to aid in the verification of information after each session.In 2023, researchers revisited and updated the CLD.This map was presented to the hospital team members for validation in May 2023 and was adjusted according to their feedback.

Analysis
Group model building participants were classified under the Extended 4 Voices of Design Framework to show the different voices represented during this process. 33,34The Extended 4 Voices of Design Framework proposes that it is crucial to identify different stakeholders and connect people who want to do something with those who can help to do it. 33It is important to ensure the initiative is context-sensitive and could work; this way, an ecosystem for change is created to gain traction on an issue and create change. 33,34The different voices proposed in this framework are 33 (1) the voice of intent is represented by motivated and passionate people (ie, champions) or authority to drive change, (2) the voice of experience corresponds to end-users and their lived experiences (eg, clients, caf e staff), (3) the voice of capability considers the resources that stakeholders can provide (ie, financial resources, time, labor, knowledge, space), and (4) the voice of design considers operative people that will connect the different voices and implement and manage initiatives.Under this view, it was considered that 1 person could represent > 1 voice (eg, health promotion, which is the voice of intent and design).
An action register was developed and used by researchers to keep track of all identified actions.It comprised (1) theme area targeted by the action idea, (2) action idea, (3) factor linked to the action idea, (4) implementation timeframe (ie, short, medium, and long-term), (6) action status (ie, active, inactive [never implemented], abandoned, or completed), and ( 7) conditions for action implementation on the basis of the capabilities, opportunities, and motivations framework. 35Capabilities refer to the attributes of the person, including knowledge and skills.Opportunities are characteristics of the external environment, including time and resources.Motivations are attributed to goals and intention to act on a specific implementation.This framework provides a way to identify the interaction between components that potentially can influence an implementation.
For example, opportunity can influence motivation and capability. 35These attributes were identified in the individuals who conformed to the implementation teams.Researchers updated the action registry monthly according to the implementation teams' updates provided via email or meetings.Action ideas were then integrated into the final CLD by linking each action to its linking factor; this enabled the visualization of the influence of the action ideas.

Ethics Approval
The research was approved by the Deakin University Human Ethics Advisory Group (reference no.HEAG-H23_2022) after a full ethical review process.All participants provided informed written consent to participate in this study.

RESULTS
Participants involved in the process belonged to diverse hospital areas, such as health promotion (n = 1), dietetics (n = 3), communications (n = 1), management support (n = 2), allied health staff (n = 4), procurement (n = 1), hospital food service (n = 1), caf e staff (n = 2), researchers (n = 4).Additional participants that were only involved in the CLD presentation (May 2022) consisted of other hospital staff (18%) and community (n = 4), such as patients and visitors.Each GMB had 14 participants on average; however, not all participants attended every workshop because of other commitments.Table 2 shows the distribution of voices among the GMB participants according to the Extended 4 Voices of Design Framework. 33Because of the extra community session, end-users were the most prominent voice of experience (34%).They were followed by knowledge as the main voice of capability (22%) and motivation to take action from the voice of intent (15%).Authorities to drive change (2%) and financial resources (2%) were the less active voices involved.
The evolution of the CLD developed by the hospital team members is shown in Figure 1.Participants described different factors influencing the type of foods and drinks offered at the caf e, noting the lack of space as the major limitation for improving the variety of healthy and visually appealing food.
Factors provided by participants were organized and clustered in 4 color-coded themes (key areas) described in Table 3.
Participants identified and prioritized 15 action ideas according to their perception of feasibility (eg, power to act) and time.For example, participants considered a priority the increase in the availability of healthy drinks and to develop a health promotion campaign to promote the changes within the caf e. Proposed actions were integrated into the final CLD and linked to the factor that had the most impact; this means that implementing the action would directly affect 1 part of the system, affecting other parts of the CLD. Figure 2 shows the final CLD with the action ideas linked to the most relevant factor.Actions were colorcoded according to their implementation time frame: green (short-term; 40%, n = 6), yellow (midterm; 14%, n = 2), and red (long-term; 46%, n = 7).
Five implementation teams were created to act on the "quick wins," such as actions that could be implemented in the short term to keep the team's motivation and momentum going. 24Each implementation team developed an implementation plan to act on the proposed actions.For example, the team dedicated to the organizational support theme developed a communication and marketing campaign plan that involved a Caf e Naming competition.The development, implementation, and results of this activity have been reported elsewhere. 36able 4 shows the latest status of the actions (July 2023), with 4 actions (27%) active, 6 (40%) inactive, 1 abandoned (6%), and 4 completed (27%).Further description of the Table 2.

DISCUSSION
This research brief describes the use of a CBSD approach to developing a CLD through GMB workshops, representing the factors influencing the availability and sales of healthy food and drinks at a caf e located in a hospital setting.The CLD highlighted 4 thematic areas in which actions could be implemented to improve the caf e's food environment.The GMB technique was used to develop this CLD, identify action ideas, and establish implementation teams.The success and realization of many action ideas were due to the combination of capabilities, opportunities, and motivations in the implementation teams.In the CLD, different types of actions were identified.Among them, structural actions such as expanding or relocating the caf e and expanding the caf e food preparation area were recognized.Procedural actions (eg, disconnecting the caf e food from the inpatient menu) were deemed to have the greatest potential for impact.However, despite their potential, these actions had fewer conditions to be implemented, and no implementation teams were formed to advance them.This research brief contributes to the limited empirical evidence on the use of GMB to identify factors that could improve hospitals' retail food environments. 21,24Our study builds on previous efforts to implement tools to track the implementation of action ideas from GMB workshops, 37 which can help to systematically capture efforts led by community members (eg, employed team members).GMB practice can motivate stakeholders to work together to bring about change at the community level. 21Our results show the active involvement of implementation teams despite the implementation barriers (eg, staff turnover, limited support, and competing interests).Previous research has shown that the key factors for successful implementation include having clear practice objectives, receiving widespread support across the system, and ensuring that monitoring and reporting requirements are in place. 38he hospital was merged into a bigger health care group, which may have led to a high staff turnover; staff who left were not replaced, and positions and roles disappeared.As a result, a fragmentation of implementation teams occurred, 39 and it was not possible to find additional staff members to join the implementation teams because of staffing gaps, workflow disruptions, or lack of time and leadership.Although implementing actions depends on strong implementation teams, 37 unexpected factors in the hospital setting can hinder the implementation of actions, resulting in a loss of momentum and motivation to implement action ideas.Our findings echo previous research that has shown the need for trained staff to support and sustain implementation. 38,40ur findings showed capable voices that contributed with knowledge and human resources, but the voices of stakeholders that could contribute with the resources and power to enable change were limited.This finding reinforces the importance of having higher decision-level stakeholders to achieve the structural changes required for stronger sustainable outcomes. 38,41Previous research has reported that leadership support can enable the implementation of healthy food service guidelines. 40,42However, the lack of  1. Infrastructure and staff resources (blue) referred to the hospital kitchen and caf e's capacity to prepare healthier options The existing infrastructure within the caf e space, including food preparation space, food storage, and equipment availability, are some of the biggest barriers to increasing the caf e healthy food preparation capacity and achieving significant and sustained change The caf e dependency on the hospital kitchen is currently contributing to the higher availability of less healthy hot food (red and amber) options within the caf e according to the Healthy Choices guidelines b The hospital kitchen mentioned that because of staffing issues, they are unable to provide food to the caf e that is different from the inpatient menu (except in a very limited capacity, such as cooking roast pumpkin for the caf e to make a salad) Participants considered that if the kitchen's healthy food preparation capacity is increased by hiring more skilled and trained staff, the availability of healthy foods at the caf e could be improved 2. Food supply (green): Considered product availability to prepare healthier options The kitchen provides, on average, 4−6 different meal options per day from the inpatient menu.According to the Healthy Choices guidelines, b most of these options are coded red; many are amber, and very few are green The caf e healthy food preparation capacity depends on the availability of foods, which depends on food supplier access Food and beverage supply availability beyond the area is limited (ie, fewer healthier drinks and snacks-food providers report that it is often not financially viable to supply to Horsham) 3. Organizational support (orange) includes factors that can increase the staff's acceptability of healthier options Required support from leadership across multiple levels of the organization to champion food changes at the caf e Participants considered that communicating and promoting changes could lead to increased team support, readiness to change, and eventually more acceptability of the healthy caf e menu options 4. Organizational finance (yellow) is related to financial support required to make changes and possible monetary returns The caf e depends on the kitchen budget-this has benefits ordering but limits innovative opportunities or empowering decision-making by caf e staff/management Organizational finance could be directed to staff resourcing and increasing the caf e physical space in the future a Descriptions capture additional information from participants' discussions during the group model building sessions that helped to organize and theme factors; b Healthy Choices is a framework for improving the provision and promotion of healthier foods and drinks in key settings within Victoria, Australia.The Healthy Choices: Food and Drink Classification Guide categorizes foods and drinks as green, amber, or red on the basis of their nutritional value. 19esources and prioritization to improve the healthiness of food environments in hospital settings is considered a barrier. 15,17he developed CLD provided a pathway for decision-makers to support actions toward achieving a healthier food retail environment within the hospital that could benefit the health and well-being of staff and the broader community.The limitations of the methods and approach used relate to the limited ability of GMB applications to improve the food environment within hospital settings.Therefore, comparing this implementation of Healthy Choices in hospital food outlets is difficult. 42,43The CLD and actions represented in this study may not apply to other hospital food retail outlets.In addition, this study was conducted in a hospital setting in Australia.Thus, we cannot generalize that our results are relevant to the health care structure of other countries.More research is needed to demonstrate the behavioral impact of initiatives targeting the food environments.

IMPLICATIONS FOR RESEARCH AND PRACTICE
Creating health-enabling food environments is a complex task that may require collaborative solutions and a perspective that does not strictly follow a linear, step-by-step progression.The GMB process helped identify key areas (ie, infrastructure and staff, food supply, organizational support, and finance).In addition, it supported the identification of actions (eg, a mass media campaign, a review and modification of menu options, and support from local producers) that could improve the availability and sales of healthy foods while motivating staff members to act for change.However, challenges such as staff turnover, insufficient recruitment of new staff, and inadequate participation from higher-level decision-makers have impeded progress in enhancing the hospital food environment.Within a hospital setting, in which staff turnover rates tend to be high, it is crucial to have committed individuals who strive to create a workplace that prioritizes health and wellness, thus enabling sustained improvements.We consider that the ongoing recruitment of "champions," whether they are staff or community members, would be beneficial to keep the momentum going toward incremental changes.Alternatively, hospitals could have a position explicitly working to create a health-promoting workplace.In addition, it is imperative that senior management offers leadership and supports potential innovations to improve the food environment.As a result, long-term objectives are achieved, and new opportunities to maintain a healthier food environment are identified.a Capability (attributes of the person, including knowledge and skills), opportunity (characteristics of the external environment, including time and resources), and motivation (goals and intention); b As classified by the Healthy Choices: Food and Drink Classification Guide. 19

a
Representation of Voices in the Group Model Building Workshops According to the Extended 4 Voices of Design Framework Hospital staff refers to hospital staff, regardless of their working area, that did not attend any of the initial 3 group model building workshops; b Community refers to visitors (eg, patients or family members and the healthy eating advisory service visitor).action ideas is presented in the Supplemental Table.

Figure 1 .
Figure 1.Evolution of the Causal Loop Diagram Throughout the Group Model Building Workshops.

Figure 2 .
Figure 2. Integration of Action Ideas to the Causal Loop Diagram Developed Through Group Model Building Workshops.

Table 1 .
Group Model Building Process November 2021 Group review and validation of second CLD and theme identification Identification of action ideas that would increase the availability and sale of healthy food and drinks at the caf e Prioritization of action ideas on the basis of their feasibility and perceived impact Establishment of 5 implementation teams After workshop 3 Using the final CLD as a logic model for implementing action ideas Community presentation: May 2022 Project introduction and presentation of the third CLD Identification of missing factors and action ideas through a world caf e dynamic, b with the community After community presentation Project introduction and presentation of the third CLD Identification of missing factors and action ideas in the community presentation CLD update presentation: May 2023 CLD update presentation to the working group CLD indicates casual loop diagram; STICKE, Systems Thinking in Community Knowledge Exchange.Detail of the activities can be found in the group model building scripts (Supplementary Material); b Simple, effective, and flexible method for hosting large group dialogue. a

Table 3 .
Description of Themes Identified in the Causal Loop Diagram Developed Through Group Model Building Workshops Carmen Vargas and Jillian Whelan are supported by the National Health and Medical Research Councilfunded Centre of Research Excellence in Food Retail Environments for Health (APP1152968).The opinions, analysis, and conclusions in this paper are those of the authors and should not be attributed to the National Health and Medical

Table 4 .
Tracking Registry of Action Ideas Generated by Group Model Building Participants and Components that Influenced the Implementation