A GIS-Based Decision Support System for Personalized Therapeutic Pathways in Feeding and Eating Disorders: Integrating Social Agriculture and Green Infrastructure into Health-Oriented Spatial Planning

Abstract

Feeding and Eating Disorders (FED) require integrated, recovery-oriented care models that extend beyond clinical treatment and incorporate supportive environments capable of enhancing psychosocial well-being. Within this perspective, nature-based and socio-agricultural practices represent promising yet underexplored therapeutic resources, particularly when integrated into spatial planning frameworks. This study develops and tests a Geographic Information Systems (GIS)-based Decision Support System (DSS) that matches the specific therapeutic needs of individuals undergoing treatment for FED with the spatial distribution and characteristics of green and agricultural environments. The research is based on a case study involving the FED care center “Il Pellicano” in Perugia, Italy. Supply-side data were collected from 65 facilities, including 58 social farms, 6 community gardens, and the center’s private garden. Demand-side data were obtained through a questionnaire administered to patients by healthcare professionals, while supply-side attributes were collected through structured interviews with facility managers. The spatial matching process was implemented in a GIS environment using a non-compensatory multi-criteria approach that integrated thematic activities, spatial and/or organizational accessibility, confidentiality, spatial capacity, and environmental settings. The results reveal a substantial mismatch between demand and supply, with the current territorial system satisfying only 37.67% of expressed therapeutic needs. Sensitivity analysis indicates that the main constraints relate to the limited availability of medium-sized, low-attendance, and freely accessible environments. Beyond the local case study, the proposed DSS provides a transferable planning-support tool for designing personalized therapeutic pathways and strengthening the integration between green infrastructure, social farming, and healthcare systems. The study highlights the strategic role of spatial planning in promoting health equity, social inclusion, and community well-being.

1. Introduction

Feeding and Eating Disorders (FED) are complex mental health conditions characterized by clinically significant disturbances in eating behaviors and in the psychological and biological mechanisms regulating food intake [1,2]. They represent a growing public health concern due to their increasing prevalence, early onset, chronic trajectories, and profound psychosocial consequences [3,4]. Contemporary clinical perspectives increasingly conceptualize health not merely as the absence of disease, but as a state of psychophysical well-being shaped by environmental, relational, and behavioral factors [2]. In individuals with FED, this balance is often disrupted by rigid cognitive patterns related to body image, control, and food intake, frequently reinforced by socio-cultural pressures and digital media environments [4].

The clinical literature highlights the need for integrated, multidisciplinary care models that extend beyond symptom reduction to support long-term recovery, social inclusion, and quality of life [5,6,7,8,9,10,11]. Within this framework, recovery is understood as a dynamic, multidimensional process influenced by clinical interventions and environmental and relational contexts.

In recent years, nature-based interventions have gained increasing recognition as complementary approaches in mental health care [12]. Exposure to green environments has been associated with reduced stress, improved emotional regulation, and enhanced social interaction [13,14,15], while horticultural and gardening activities may foster embodied engagement, responsibility, and connectedness [16]. In the context of FED, reconnecting individuals with food production processes may help reframe maladaptive relationships with food, control, and self-perception.

Within this broader transition toward recovery-oriented, environmentally informed care models, care farming has emerged as an innovative approach that integrates agricultural production with therapeutic, educational, and social inclusion objectives [17,18,19,20,21,22]. These initiatives create multifunctional spaces where health, learning, and community engagement intersect. Evidence suggests that structured engagement with agricultural environments may reduce pathological traits such as rigidity and excessive control in individuals with FED [23]. Community gardens represent a specific form of urban agriculture characterized by collective management and participatory governance [24,25,26,27,28]. They function as nature-based solutions that enhance ecosystem services while fostering inclusion, social cohesion, and well-being [29,30,31].

Despite this growing body of evidence, clinical and spatial planning research have largely developed in parallel. While health studies emphasize recovery-oriented care models, spatial planning research highlights the role of green infrastructure in promoting well-being and equity [32,33,34,35,36] and in regulating ecosystem services to improve the quality of spaces [37]. However, limited attention has been paid to the spatial organization of therapeutic opportunities and to the systematic matching between patient needs and territorially distributed resources [38,39].

Importantly, recovery trajectories may be conceptualized as spatially distributed processes unfolding across multiple scales, from domestic environments to urban gardens and rural farms. This perspective aligns with emerging research in health geography, emphasizing the restorative potential of both natural and built environments [40,41]. Health geography is an interdisciplinary field examining how spatial dimensions, physical landscapes, and socioeconomic environments shape human well-being and disease [42].

However, few studies have operationalized this perspective through spatially explicit decision-support tools. To date, no GIS-based model has been specifically developed to match socio-agricultural environments with the heterogeneous needs of individuals with FED.

To address this gap, the present study develops a GIS-based DSS designed to match therapeutic needs with socio-agricultural and green environments. The system links healthcare services with territorial resources to support personalized, spatially informed care pathways. The following questions guide the research:

(i)

How can a GIS-based DSS support the identification of therapeutic environments tailored to the needs of individuals with FED?

(ii)

What mismatches exist between therapeutic demand and the territorial supply of socio-agricultural environments?

(iii)

What are the implications for spatial planning and health equity?

2. Materials and Methods

2.1. Methodological Framework

Figure 1 illustrates the methodological framework adopted in this study. The research develops a GIS-based DSS designed to spatially match the demand for therapeutic agri-cultural experiences expressed by users of a FED treatment center with the territorial supply of care farms and community gardens. The methodological workflow consists of three main phases: (1) identification and characterization of health demand; (2) mapping and analysis of socio-agricultural supply; and (3) implementation of a GIS-based non-compensatory multi-criteria matching process. This structured approach enables a systematic comparison between users’ preferences and the spatial and organizational attributes of available facilities. The proposed framework integrates spatial analysis and clinical preference modeling within a decision-support framework designed for exploratory and applied planning. Rather than aiming for predictive modeling or statistical inference, the system prioritizes rule-based spatial reasoning to ensure transparency, interpretability, and direct applicability in healthcare planning contexts. This methodological positioning aligns the DSS with participatory and evidence-informed planning tools commonly adopted in health geography [43].

2.2. Study Area and Sources of Data

The demand layer was derived from patients attending an association dedicated to the treatment and psychosocial support of individuals with FED, located in the historic center of Perugia, Italy. The association, “Il Pellicano”, was established in 1997 and supports an average of 175 patients per year from Umbria and neighboring regions. Patients access the center either through referrals from the National Health Service or through private pathways. Most users are young individuals, with female adolescents representing approximately 33% of the total patient population. The center adopts a “help-home” model, defined as a non-medicalized, supportive environment grounded in mutual self-help principles [44]. Its primary objective is to reduce compulsive behaviors associated with FED and to provide practical strategies for long-term management. Activities include assisted meals, food familiarization training, and creative workshops such as painting, sewing, and theatre. The weekly program typically follows structured daily routines, from breakfast to dinner, with supervised meals involving nutritionists and psychologists. Beyond clinical care, the association promotes social inclusion and collaborates with academic institutions, including the University of Perugia and the University of Parma [44]. Since 2018, it has also maintained a private garden used for self-production and horticultural activities, supporting the development of more structured engagement with food-production environments [45].

Given the center’s urban location, the supply layer was constructed by mapping care farms and community gardens within the municipality of Perugia and the surrounding municipalities. The study area includes both urban and peri-urban contexts, with the municipality covering approximately 450 km². Data on care farms were obtained from the official Umbria regional registry, which lists 164 registered care farms [46]. A proximity-based inclusion criterion was applied, selecting facilities located within approximately one hour of travel time from the care center. Among the eligible facilities, 58 care farms met this criterion, while 2 were excluded due to inactivity. In addition, 6 community gardens were identified through municipal and association networks. The private garden of Il Pellicano was also included, resulting in a total supply dataset of 65 facilities.

2.3. Data Collection

2.3.1. Origin and Validation of Criteria

The criteria used in this GIS-MCDA protocol were identified through a series of focus groups involving the research team and the clinical staff of the ‘Il Pellicano’ association. The selection was driven by the need to establish two types of functional links between demand (patient needs) and supply (territorial resources): spatial links (encompassing facility size, network distance, and environmental settings) and management/operational links (including specific activities, opening hours, and levels of confidentiality). These criteria were validated against the clinical goals of the treatment center, ensuring that each parameter directly relates to known psychosocial vulnerabilities of FED patients, such as social anxiety (confidentiality) and the need for perceived autonomy (organizational flexibility).

2.3.2. Questionnaires to Evaluate Demand and Supply

Data were collected using two structured questionnaires designed to capture both demand-side preferences and supply-side characteristics. The first questionnaire, targeted at Il Pellicano users, included 21 questions on preferences for agricultural environments and green spaces. It was administered face-to-face by healthcare staff during therapeutic activities. The second questionnaire was addressed to managers or owners of the identified facilities and included 16 questions focusing on environmental, organizational, and functional characteristics. The research team conducted these interviews by telephone. Participation in both surveys was voluntary.

Both questionnaires included six matched items designed to enable direct correspondence between demand and supply according to the following criteria: spatial accessibility, level of confidentiality, spatial capacity, thematic activities, organizational accessibility, and environmental settings. For each criterion, response categories were harmonized to allow direct one-to-one matching between user preferences and facility attributes (

Table 1. Demand–Supply Correspondence Criteria.

Criteria	Demand-Side	Supply-Side
spatial accessibility	Preferred distance	Actual accessibility class
	Walking distance < 300 m	High accessibility (<300 m)
	Walking distance 300–10,000 m	Moderate accessibility (300–10,000 m)
	Distance > 10 km	Low accessibility (>10 km)
level of confidentiality	Privacy needs	Social exposure conditions
	High need for privacy	Low-attendance environments
	Moderate need for privacy	Moderately attended environments
	Low need for privacy	Highly attended environments
spatial capacity	Preferred size	Facility size
	Small spaces (<1 ha)	Small-sized facilities (<1 ha)
	Medium spaces (1–2 ha)	Medium-sized facilities (1–2 ha)
	Large spaces (2–5 ha)	Large facilities (2–5 ha)
	Very large spaces (>5 ha)	Very large facilities (>5 ha)
thematic activities	Activity preferences	Offered activities
	Nature and biodiversity	Nature and biodiversity
	Agriculture and farming	Agriculture and farming
	Food education	Food education
	Environmental awareness	Environmental education
	Animal-related activities	Animal-related activities
organizational accessibility	Access preferences	Access and use conditions
	Open access at any time	Freely accessible
	Access by reservation	Reservation required
	Scheduled access	Time-slot access
	Flexible/adaptable arrangements	Flexible access arrangements
environmental settings	Preferred setting	Spatial configuration
	Open and visible environments	Open/exposed settings
	Quiet and secluded environments	Quiet/secluded settings

).

The selection of these criteria was grounded in both the clinical experience of the “Il Pellicano” center and the literature on health geography.

Spatial accessibility was operationalized using network analysis in QGIS (Shortest Path tool) based on freely downloadable OpenStreetMap road data downloaded from Geofabrik (https://www.geofabrik.de/data/download.html). Unlike Euclidean distance, network-based measures provide a more realistic representation of travel effort and accessibility conditions. Each facility was assigned to an accessibility category based on the minimum travel distance to the care center. Previous studies have demonstrated that the choice of GIS technique can significantly affect the outcomes of green-space accessibility analyses [47]. For this reason, network analysis was adopted instead of straight-line distance calculations. In the present study, the significance of this criterion is not merely logistical. It also represents the physical act of moving from Il Pellicano association toward a farm environment. This process contributes to the spatialization of the therapeutic pathway, symbolizing the transition from treatment to recovery within the territory [43].

Confidentiality reflects the degree of social exposure associated with each environment. Demand-side responses captured users’ privacy needs, while supply-side classifications reflected average attendance levels and perceived social intensity. According to the staff of Il Pellicano, this criterion is particularly relevant for managing social anxiety and environmental sensitivity among patients. Bi et al. [48] underline the high value of urban open spaces that ensure privacy, thereby improving citizens’ well-being.

Spatial capacity refers to the physical size of the available environment. Users expressed preferences regarding the dimension of spaces. At the same time, facilities were classified into four size categories based on total surface area: small (<1 ha), medium (1–2 ha), large (2–5 ha), and very large (>5 ha).

Thematic activities represent the main experiential and educational opportunities offered within each facility. Both demand and supply were categorized into comparable thematic domains: nature and biodiversity; agriculture and farming; food education; environmental awareness; and animal-related activities. Multiple selections were allowed on both the demand and supply sides. The classes correspond to those used in the regional register of care farms [46]. This criterion is particularly relevant because different activity themes may counterbalance specific pathological traits commonly associated with FED, such as rigidity and excessive control [17,23].

Organizational accessibility refers to the operational conditions that regulate access to and use of facilities. Demand-side responses reflected users’ preferences regarding flexibility and access arrangements, whereas supply-side responses described the organizational models adopted by each facility. Categories included freely accessible facilities, reservation-based access, scheduled time-slot systems, and flexible or adaptable arrangements.

Environmental settings describe the spatial and perceptual characteristics of the facility context. Respondents expressed preferences for either open and visible environments or quiet and secluded settings. Facilities were accordingly classified according to their prevailing environmental configuration, while mixed settings were included where appropriate by assigning both categories.

2.4. Data Preprocessing, Coding, and Variable Normalization

All questionnaire responses from both patients and facility managers were systematically preprocessed prior to spatial analysis to ensure consistency between the demand-side and supply-side datasets. The preprocessing phase involved three main steps: coding categorical responses, harmonizing response scales, and handling neutral or “no preference” responses.

First, all questionnaire items were transformed into structured categorical variables. Each response option was assigned a unique numerical or symbolic code to enable integration into the GIS database. For instance, spatial accessibility preferences and facility accessibility classes were encoded using ordinal categories reflecting increasing distance ranges, while privacy and environmental setting variables were coded according to increasing levels of social exposure or openness. Thematic activity preferences were treated as multi-label binary variables, allowing multiple selections for both respondents and facilities.

Second, a harmonization procedure was applied to ensure direct comparability between demand-side and supply-side attributes. This involved aligning semantic categories across questionnaires so that each demand-side preference corresponded to a supply-side attribute with an equivalent operational definition. Where necessary, qualitative descriptors were standardized into consistent categorical classes (e.g., “quiet environments” and “secluded settings” were merged into a single category representing low-social-exposure environments).

Third, responses indicating “no preference” or “indifferent” choices were treated as non-restrictive conditions within the matching procedure. Specifically, these responses were excluded from the constraint set associated with the corresponding individual, allowing the system to consider all available categories for that criterion. This approach was adopted to avoid artificially constraining the solution space and to reflect the absence of explicit user requirements rather than missing data.

No statistical normalization procedures were applied, as the analytical framework is based on a non-compensatory and rule-based matching logic rather than on aggregated utility functions. Consequently, variables were treated as categorical constraints rather than continuous indicators requiring scaling. Finally, all coded variables were integrated into a spatial-relational database in QGIS, where they were linked to georeferenced supply locations and individual demand profiles. This structure enabled the implementation of SQL-based queries for multi-criteria spatial matching.

2.5. Development of the GIS-Based Decision Support System

The DSS was implemented in QGIS (version 3.40.6) software. All facilities and the demand origin were georeferenced and stored within a spatial database. Questionnaire responses were encoded as attribute data linked to each spatial entity. The matching procedure was implemented using SQL-based queries within the GIS environment. Each individual demand profile was compared against all supply records. A facility was considered suitable only when all selected criteria were simultaneously satisfied. The system is defined as a non-compensatory Multi-Criteria Decision Analysis (MCDA) model, in which each criterion functions as a mandatory constraint with equal weight. No normalization or weighting procedures were applied, as the objective was not utility maximization, but rather constraint satisfaction.

The model is expressed as Equation (1):

$$\mathit{S}={\displaystyle {\bigcap }_{\mathit{i}=1}^{\mathit{n}}}\left({{\mathit{w}}_{\mathit{i}} \mathbf{\ast } \mathit{C}}_{\mathit{i}}\right)$$

(1)

where S represents the set of suitable facilities obtained through the intersection of n criteria (Ci). Each criterion is associated with a weight (wi) equal to 1 and functions as a hard constraint.

This methodological choice was motivated by the clinical nature of the application. In therapeutic contexts related to eating disorders, certain environmental conditions cannot be considered mutually compensable. As argued by Zhang [49], compensatory logic, where a high score in one dimension offsets a deficiency in another, is often undesirable in environmental and health-related contexts. In FED therapy, a strong need for privacy cannot be “compensated” by high accessibility; each criterion must therefore serve as a strict constraint to ensure clinical appropriateness and psychological safety. Following the “ordinal input for cardinal output” approach described by Greco [50], the system transformed subjective preferences into spatial variables to identify compatible therapeutic settings without generating undesirable trade-offs among different psychological needs. Consequently, the procedure followed a constraint-satisfaction logic commonly adopted in GIS suitability analysis and non-compensatory MCDA approaches.

Finally, the system was exported to a shared Google My Maps platform to support interactive consultation by healthcare professionals and, where appropriate, patients. The platform enables users to visualize the spatial distribution of facilities, access descriptive information and images, and facilitate the collaborative design of tailored therapeutic experiences. When no site satisfies all constraints, users can progressively relax criteria according to their preference hierarchy, thereby identifying acceptable alternatives while preserving their agency in the decision-making process.

3. Results

3.1. Demand Analysis

Of the 175 patients supported annually by the association, 24 participated in the study, representing 14% of the total population. Although the sample size is limited, it provides exploratory insight into heterogeneous preferences within a real-world clinical context. Given the small number of respondents, percentages should be interpreted as indicative tendencies rather than statistically generalizable findings. The age distribution shows a predominance of young adults: 50% of respondents were aged 18–25, 25% were aged 26–35, and the remaining 25% were aged 36–45. This reflects the typical demographic profile associated with FED prevalence and suggests that the observed preferences are particularly representative of younger individuals engaged in recovery pathways.

Table 2. Demand-side criteria (patients’ preferences).

Spatial Accessibility	Demand (%)
Walking distance < 300 m	0.00
Walking distance 300–10,000 m	8.33
Walking distance > 10 km	16.67
No preference	75.00
Level of confidentiality
High need for privacy	12.50
Moderate need for privacy	33.33
Low need for privacy	25.00
No preference	29.17
Spatial capacity
Small spaces (<1 ha)	4.17
Medium spaces (1–2 ha)	29.17
Large spaces (2–5 ha)	16.67
Very large spaces (>5 ha)	33.33
No preference	16.67
Thematic activities
Nature and biodiversity	43.00
Agriculture and farming	43.00
Food education	27.00
Environmental awareness	28.00
Animal-related activities	4.00
No preference	0.00
Organizational accessibility
Open access at any time	50.00
Access by reservation	8.33
Scheduled access	8.33
Flexible/adaptable arrangements	0.00
No preference	33.33
Environmental settings
Open and visible environments	29.17
Quiet and secluded environments	37.50
No preference	33.33

summarizes the demand-side results, highlighting a differentiated preference structure in which some spatial and organizational attributes are clearly prioritized, while others display high levels of indifference.

A first relevant finding concerns spatial accessibility. A large majority of participants (75.00%) reported indifference regarding travel distance to therapeutic environments. Only a limited proportion expressed specific distance constraints, suggesting that proximity is not perceived as a primary determinant in therapeutic site selection. This result indicates that experiential and qualitative dimensions of place may outweigh spatial proximity in shaping therapeutic preferences.

Privacy preferences show greater variability. While 33.33% of respondents indicated a moderate need for privacy and 12.50% a high need, 25.00% reported a low need for privacy, and 29.17% expressed no preference. This distribution highlights the coexistence of diverse psychosocial needs, consistent with the heterogeneity of recovery trajectories in FED contexts.

Spatial preferences indicate a clearer tendency toward larger environments. Very large facilities (>5 ha) were preferred by 33.33% of respondents, followed by medium-sized spaces (29.17%). Small spaces (<1 ha) were rarely selected (4.17%). This pattern suggests that users may associate larger environments with greater freedom, diversity of activities, and reduced social pressure.

Thematic preferences show strong convergence toward agriculture and nature-based activities, each selected by 43.00% of respondents. Environmental awareness and food education were also relevant, whereas animal-related activities were marginal (4.00%). This finding suggests that direct engagement with land-based and productive processes plays a central role in therapeutic imaginaries and recovery expectations.

Organizational preferences highlight a strong demand for autonomy. Half of the respondents (50.00%) preferred freely accessible environments without time constraints, while fewer participants selected more structured access models. The relatively high proportion of indifferent responses (33.33%) suggests that organizational format is secondary for a subset of users.

Environmental preferences are more balanced. Quiet and secluded environments were preferred by 37.50% of respondents, while 29.17% preferred open and visible settings. One-third of respondents expressed no preference, further reinforcing the heterogeneity of environmental needs.

Overall, the demand analysis reveals variability rather than uniformity, underscoring the need for individualized spatial matching strategies rather than standardized therapeutic solutions.

3.2. Supply Analysis

Table 3. Supply-side characteristics (facilities).

Spatial Accessibility	Supply (%)
High accessibility (≤300 m)	1.54
Moderate accessibility (300–10,000 m)	3.08
Low accessibility (>10 km)	73.85
No response	21.54
Level of confidentiality
Low-attendance environments	0.00
Moderately attended environments	9.23
Highly attended environments	66.15
No response	24.66
Spatial capacity
Small-sized facilities (<1 ha)	10.77
Medium-sized facilities (1–2 ha)	4.62
Large facilities (2–5 ha)	12.31
Very large facilities (>5 ha)	58.46
No response	13.85
Thematic activities
Nature and biodiversity	100.00
Agriculture and farming	100.00
Food education	45.00
Environmental education	8.00
Animal-related activities	33.00
No response	0.00
Organizational accessibility
Freely accessible	9.23
Reservation required	7.69
Time-slot access	10.77
Flexible access arrangements	47.69
No response	24.62
Environmental settings available
Open/exposed settings	50.77
Quiet/secluded settings	23.08
Mixed environmental settings	4.62
No response	21.54

presents the main characteristics of the territorial supply.

Spatial accessibility shows a strong concentration of facilities in peripheral areas: 73.85% are located more than 10 km from the therapeutic center. Only a minimal proportion (4.62%) is located within 10 km. This confirms a structural spatial separation between the care center and socio-agricultural resources.

Privacy conditions reveal a marked imbalance. Most facilities (66.15%) exhibit high attendance levels, while none are classified as low-attendance environments. This indicates a system primarily oriented toward socially active contexts, with limited availability of more secluded settings.

Spatial capacity is dominated by very large facilities (58.46%), whereas medium-sized (4.62%) and large (12.31%) spaces are underrepresented. This suggests a polarized spatial structure characterized by the prevalence of large rural farms and a limited availability of intermediate-scale environments.

Thematic activities are widely available across the core domains: all facilities offer nature and biodiversity activities as well as agriculture and farming activities. However, environmental education (8.00%) and animal-related activities (33.00%) are less frequent, indicating only partial thematic diversification.

Organizational models show greater variability. Nearly half of the facilities (47.69%) offer flexible access arrangements, while fully open-access models are less common (9.23%). This suggests the presence of latent organizational adaptability that is not yet fully aligned with user preferences.

Environmental settings are predominantly open and exposed (50.77%), whereas quiet and secluded environments are less common (23.08%). Mixed configurations are relatively rare, suggesting limited diversification in environmental design and spatial organization.

Importantly, 89% of facilities expressed willingness to develop dedicated programs for individuals with FED. Despite this high level of declared openness, only one facility currently implements structured interventions, revealing a substantial implementation gap between institutional willingness and operational practice.

3.3. Demand–Supply Matching

The GIS-based non-compensatory matching process identified full correspondence between demand and supply for 37.67% of participants. Given the model’s strict constraint-based logic, this value should be interpreted conservatively. Any mismatch in a single criterion results in exclusion from the full-match category. More specifically, the 37.67% value should not be interpreted as the proportion of therapeutically usable environments overall, but rather as the proportion of cases achieving complete compatibility under a strict non-compensatory matching logic. The results indicate that complete alignment between patient preferences and facility characteristics remains limited, reflecting both the specificity of user needs and structural constraints within the territorial supply system.

Case-based analysis illustrates this variability (Figure 2). Patients P1 and P8 identified only one compatible facility (“Orto di Montemorcino”). Both expressed a preference for large, secluded environments with agricultural and environmental thematic orientations, while differing in their secondary thematic priorities. In contrast, Patient P3 identified four compatible facilities, reflecting a less restrictive preference profile and greater alignment with the available territorial supply. Figure 2 illustrates these examples and highlights how different preference structures generate substantially different levels of spatial compatibility, ranging from single matches to multiple alternatives.

Beyond full matches, the system also reveals patterns of partial compatibility, indicating that although no single facility satisfies all criteria simultaneously for many users, several facilities still meet subsets of therapeutic needs. This suggests the existence of latent therapeutic opportunities that are not captured through strict Boolean filtering.

Figure 3 shows how the DSS appears to patients within the My Maps environment. Users can select each facility to access photographs, brief descriptions, and information on the site’s main environmental and organizational characteristics. The linked map can also be accessed through Google My Maps mobile application (web-based service, continuously updated by Google), allowing users to locate facilities at any time. Furthermore, users can choose which selection criteria to delete and generate a new result based on matching.

Overall, the results highlight three main analytical insights: (i) the heterogeneity of therapeutic preferences among individuals with FED; (ii) the structural characteristics and current limitations of the territorial supply system; and (iii) the usefulness of GIS-based multi-criteria matching in revealing both compatibilities and gaps within health–environment systems.

3.4. Sensitivity Analysis

Given the non-weighted structure of the model, sensitivity analysis was conducted by progressively removing individual constraints and evaluating changes in (i) the percentage of satisfied patients, and (ii) the average number of matching facilities per patient.

This approach allows assessment of the relative restrictiveness of each criterion (

Table 4. Sensitivity analysis results.

Match Without [X] Criterion	Satisfied Patients (%)	N. of Facilities (Mean)
None	100.00	65
Organizational accessibility	62.67	7.5
Spatial capacity	58.50	4.0
Level of confidentiality	50.17	4.4
Environmental settings	50.17	2.8
Spatial accessibility	46.00	2.3
Thematic activities	46.00	3.5
All criteria	37.67	2.3

). The average number of matching facilities was calculated only for patients with at least one compatible facility identified.

When no constraints are applied, 100% of patients match all 65 analyzed facilities. As constraints are progressively introduced, the matching rate decreases, confirming the cumulative restrictiveness of the multi-criteria system.

The most restrictive criterion is organizational accessibility. When this constraint is removed, the proportion of satisfied patients increases to 62.67%, while the average number of matching facilities rises significantly to 7.5 per patient. This finding indicates that temporal organization and access conditions constitute the main limiting factor within the current territorial supply structure. Spatial capacity also exerts a strong restrictive effect. Removing this criterion increases the proportion of satisfied patients to 58.50%, suggesting that facility size plays a relevant role in limiting compatibility and that users’ sensitivity to perceived spatial adequacy reflects this. Confidentiality and environmental settings exhibit intermediate levels of restriction (both 50.17%), indicating that privacy-related dimensions significantly influence matching outcomes. Spatial accessibility and thematic activities are comparatively less restrictive (both 46.00%), as these dimensions are more evenly distributed across the territorial supply system (thematic activities) or patients do not have stringent needs about them (spatial accessibility). Finally, when all constraints are simultaneously applied, only 37.67% of patients achieve full compatibility, confirming the cumulative effect of strict multi-criteria intersection.

Overall, the sensitivity analysis demonstrates that organizational structure and spatial capacity represent the dominant limiting factors in demand–supply alignment, while thematic and accessibility-related variables are comparatively less constraining.

4. Discussion

4.1. Matching Therapeutic Needs and Territorial Supply: Key Mismatches and Implications

The multi-criteria spatial analysis integrating accessibility, privacy, spatial capacity, thematic preferences, organizational models, and environmental conditions enabled the identification of fully compatible facilities for 37.67% of participants. This value should not be interpreted as the proportion of therapeutically usable environments overall, but rather as the proportion of cases achieving complete compatibility under a strict non-compensatory matching logic. As such, the indicator reflects system coherence rather than partial suitability. At the same time, the relevance of demand–supply alignment in therapeutic landscapes must be situated within the broader context of mental health burden and recovery-oriented care. Eating disorders and related conditions represent a growing global health challenge with significant impacts on health, mortality, and quality of life [1,2,3,4]. Recovery frameworks increasingly emphasize the multidimensional nature of care, integrating clinical, psychosocial, and environmental determinants [5], including the role of nature-based and socio-ecological settings in supporting well-being [12,51,52].

The comparison between demand and supply highlights that the most significant mismatches concern organizational accessibility, privacy conditions, and spatial capacity. Organizational accessibility emerges as the most critical limiting factor. Half of the participants expressed a preference for freely accessible environments without temporal constraints, reflecting a desire for autonomy and self-directed engagement in therapeutic activities. However, the territorial supply is dominated by structured access models, including reservation-based or time-regulated systems. This discrepancy resonates with evidence from collaborative and community-based care systems, where flexibility and user agency are central to participation and adherence [17,18,19,34,53]. From a planning perspective, hybrid organizational models could enhance inclusivity while aligning with principles of community-based health promotion [33].

Spatial capacity represents a second major mismatch. The supply system is polarized toward large-scale facilities, whereas demand is more evenly distributed across medium and large spatial categories. This “missing middle” aligns with broader critiques in the spatial planning literature that emphasize the need for multi-scalar and adaptive green infrastructure [54,55]. Intermediate typologies are also increasingly recognized as critical for inclusive urban agriculture and community-based ecosystems [25,26,27].

Privacy conditions also contribute to misalignment. Although patient preferences are heterogeneous, the territorial supply is overwhelmingly characterized by highly attended environments. This structural imbalance limits opportunities for individuals requiring reduced social exposure due to anxiety or emotional vulnerability. Literature on restorative environments consistently highlights the importance of perceived safety, social buffering, and controlled exposure in supporting mental health recovery [40,51,52,56]. The absence of differentiated privacy gradients, therefore, reflects a structural gap in socio-ecological planning for health [32,33].

Environmental configuration further contributes to partial mismatches. Although both open and secluded environments are represented, the supply is predominantly oriented toward exposed settings. At the same time, a substantial proportion of users expressed a preference for quiet, protected environments, reinforcing the evidence that perceived landscape structure significantly influences restorative outcomes [12,40,51,52]. Micro-scale interventions in vegetation structure and spatial design have been shown to enhance psychological benefits and ecosystem service provision [14,15,57].

Interestingly, spatial accessibility appears to play a less decisive role than traditionally assumed in accessibility studies [13,47,58]. Although most facilities are located more than ten kilometers from the care center, a large majority of participants reported indifference toward distance. This challenges conventional accessibility models that prioritize proximity as a primary determinant of service utilization [13,47,58]. Recent GIS-based studies suggest that accessibility is increasingly multi-dimensional, involving both physical and perceived components [41]. At the same time, place engagement and experiential quality may outweigh distance effects in shaping well-being outcomes [40]. In therapeutic contexts, spatial transition itself may support psychological detachment and recovery-oriented meaning-making. Furthermore, the widespread indifference toward physical distance among participants (75.00%) suggests the possibility of a “psychological distance” hypothesis. As Higgs [47] notes, in FED recovery pathways, physical displacement may also support symbolic separation from the clinical environment and strengthen perceived autonomy. This interpretation is partially supported by Curzio [23], whose clinical findings indicate that horticultural engagement may reduce physiological stress by emphasizing the qualitative nature of the experience rather than purely logistical proximity.

Thematic activities represent one of the strongest areas of potential alignment. Agri-cultural and nature-based activities are widely available and strongly preferred by users. This is consistent with extensive evidence on gardening, horticultural therapy, and care farming as effective interventions for mental health and psychosocial recovery [16,18,19,23,55]. Expanding thematic diversity could further enhance therapeutic adaptability across heterogeneous patient profiles.

Overall, the results indicate that mismatches are not primarily driven by spatial scarcity but rather by qualitative dimensions of design, organization, and experiential structure. This is consistent with socio-ecological perspectives, which emphasize that health outcomes emerge from the interaction between environmental configuration and social organization [13,33,52].

Despite its contributions, the study presents several limitations. First, the patient sample (n = 24) reflects the exploratory nature of this study and limits the ability to conduct robust inferential or clustering-based analyses. Second, the analysis is context-specific to a single care center and its surrounding territorial system, which may limit direct generalizability. Future developments will involve expanding the patient sample and the number of studied care centers.

A further weakness of this study is that the matching criteria were qualitatively selected and agreed upon with the medical staff solely to establish spatial (size, distance, environmental settings) and management (uses, opening hours, level of confidentiality, activities) links between demand and supply. Consequently, the current results do not account for potential hierarchies of relevance among these criteria. Future research will involve testing and validating a broader range of criteria to select the most significant ones through longitudinal clinical evidence and patient feedback.

4.2. Non-Compensatory GIS-Based MCDA Framework

The decision-support system was intentionally implemented as a non-compensatory GIS-based MCDA framework. Unlike weighted additive approaches, in which poor performance in one criterion can be offset by strong performance in another, the proposed model treated each criterion as a mandatory constraint rather than as a compensable preference factor [59,60]. Although several non-compensatory MCDA families, such as PROMETHEE, may incorporate differentiated weighting schemes [50], the present study deliberately avoided the use of weights. Criteria were therefore conceptualized as non-substitutable therapeutic requirements rather than preference scores. This methodological choice was motivated by the clinical nature of the application. In therapeutic contexts related to eating disorders, certain environmental characteristics cannot be considered mutually substitutable. For example, inadequate privacy conditions or excessive social exposure cannot be clinically compensated by greater accessibility, larger spaces, or broader thematic opportunities. In weighted compensatory models, facilities with clinically unsuitable characteristics could still achieve high overall suitability scores due to compensatory effects across criteria. The Boolean intersection approach adopted in this study ensured that each recommended site fully satisfied the therapeutic profile defined by clinicians and patients, thereby avoiding algorithmic trade-offs in which strong performance in one dimension could mask clinically relevant deficiencies in another. This approach also improved transparency and interpretability, as clinicians and patients could directly identify why a facility was excluded and which specific criterion prevented its selection. When no facility satisfied all criteria simultaneously, the model implemented a progressive criterion-relaxation procedure, enabling patients to identify which requirements could be waived according to their individual priorities. The resulting sensitivity analysis assessed the system’s robustness by evaluating how removing each constraint affected both the proportion of satisfied patients and the number of eligible facilities. The adopted methodology is consistent with non-compensatory MCDA approaches commonly discussed in healthcare decision-support literature, particularly in contexts where avoiding inappropriate compensatory effects among sensitive criteria is considered methodologically preferable [61].

The main limitations of this methodological choice are twofold. First, the strict non-compensatory logic, although clinically justified, may limit the identification of partially compatible facilities that could still provide therapeutic value under supervised or adaptive conditions. Second, the absence of differentiated weighting prevents the model from capturing potential variability in the relative clinical importance of criteria across different patient profiles or treatment stages.

Future developments should therefore explore hybrid decision-support approaches that integrate the current constraint-based framework with adaptive or patient-specific weighting strategies defined through participatory clinical protocols. Additional developments may include integrating clinician-driven prioritization schemes, longitudinal monitoring of patients’ responses to selected environments, and applying clustering or profiling methods to identify recurrent therapeutic-environment preference patterns among patients with eating disorders.

4.3. Innovation, Planning Implications, and Theoretical Contributions

The results of the paper advance understanding of eating disorder recovery by explicitly considering spatial environments as active components of therapeutic pathways. Recovery is increasingly conceptualized as a multidimensional process shaped by clinical, relational, and environmental factors [5], including exposure to restorative and nature-based settings that support emotional regulation and psychosocial reintegration [12,51,52,56].

Second, the study extends research on nature-based solutions by operationalizing their integration into a decision-support framework. While previous studies have demonstrated the health benefits of green environments [16,51,52], care farming and community gardening literature further highlight their role in fostering inclusion, resilience, and social cohesion [17,18,19,25,31,34]. However, fewer studies have translated these insights into spatially explicit tools for individualized care planning.

Third, the research contributes to GIS-science and health geography by reframing GIS as a clinical support instrument rather than solely a descriptive–analytical tool. GIS has been widely used for accessibility and spatial inequality analysis [13,58,62,63], but recent advances emphasize its role in modeling complex socio-spatial systems and health-related decision-making [64,65]. In this study, GIS becomes a mediating interface between healthcare systems and territorial ecosystems, aligning with emerging approaches in spatial health analytics [41].

A key conceptual contribution lies in reframing mobility as part of the therapeutic process. Movement from a clinical to a natural or agricultural environment represents not only spatial displacement but also a symbolic transition toward autonomy and recovery. This aligns with research on place engagement and psychological restoration in everyday mobility contexts [40]. The study also introduces a more nuanced understanding of accessibility by incorporating psychological and experiential dimensions. This extends traditional GIS-based accessibility frameworks [13,47,58] by integrating evidence that perceived environment and subjective experience can outweigh distance in determining well-being outcomes [40,41].

From a planning perspective, improving alignment does not necessarily require increasing infrastructure supply. Rather, it requires a qualitative redesign of existing systems. Evidence from urban agriculture and green infrastructure research supports diversification of spatial typologies, hybrid access systems, and multi-functional landscapes [25,29,54,55]. Similarly, health geography studies highlight the importance of integrated service networks combining accessibility, quality, and social inclusion [33,34,53].

The DSS developed in this paper demonstrates transferability across care contexts, consistent with broader calls for integrated and place-based health planning frameworks [66,67]. The minimum requirements for implementation align with established recommendations for spatial health data infrastructures and participatory planning systems [64,65]. Implementation requirements include: (i) the availability of georeferenced socio-therapeutic facilities; (ii) structured collection of user preference data; (iii) institutional collaboration between healthcare providers and territorial authorities; and (iv) locally adapted definitions of therapeutic environmental quality.

Finally, the policy implications are consistent with international evidence on urban green space as a public health resource [32,33,51,52]. Integrating healthcare systems with care farming and green infrastructure planning can support more equitable and sustainable health strategies, positioning spatial planning as an active component of public health governance. Public administrations may use similar tools to identify service gaps, optimize resource allocation, and design integrated therapeutic landscapes. In this sense, spatial planning becomes an active component of public health policy rather than parallel discipline.

From this perspective, GIS is not merely a technical infrastructure for spatial analysis, but also a cognitive framework capable of translating subjective therapeutic needs into spatial decision logic. It operates as a mediating layer between lived experience and territorial organization, enabling the materialization of otherwise abstract recovery pathways.

5. Conclusions

This study presents a novel integrative framework connecting food systems, health, and spatial planning through a GIS-based decision-support approach applied to FED. By operationalizing therapeutic needs as spatially explicit variables, the proposed DSS reframes agricultural and green environments not as passive contexts of care, but as active components of recovery-oriented pathways. In this sense, the study contributes to a broader conceptual shift in which food, health, and landscape are understood as interdependent dimensions of well-being rather than as separate domains.

The results highlight a clear and structurally consistent mismatch between therapeutic demand and territorial supply. This mismatch is not primarily quantitative but qualitative, and mainly concerns organizational accessibility, privacy conditions, and intermediate spatial scales. These findings indicate that current socio-agricultural systems are not fully aligned with the differentiated and heterogeneous needs of individuals undergoing treatment for FED. Importantly, this does not imply a lack of available resources, but rather a lack of functional differentiation in the way these resources are designed, organized, and accessed. From a planning and policy perspective, the findings suggest that improving alignment between health needs and territorial resources does not necessarily require the creation of new infrastructures. Rather, it calls for the reorganization and diversification of existing socio-agricultural spaces. Three suggestions for care farmers emerge clearly: (i) introducing hybrid organizational models that combine open access with structured activities; (ii) designing differentiated privacy environments within farms and gardens; and (iii) developing intermediate spatial typologies between small urban gardens and large rural farms. These interventions are realistic, scalable, and potentially high-impact in terms of improving therapeutic accessibility.

From a methodological perspective, the study demonstrates the feasibility of translating qualitative therapeutic preferences into structured spatial decision rules. At the same time, it highlights some limitations of strict Boolean approaches, particularly in contexts where clinical and psychosocial variables are not mutually substitutable. Future developments could incorporate weighted or hybrid decision models that reflect the differentiated therapeutic relevance of individual criteria, as well as participatory approaches to improve the co-definition of variables with patients and clinicians.

The study also supports a broader reinterpretation of accessibility. While traditional spatial planning frameworks prioritize physical proximity, the results suggest that psychological and experiential dimensions of distance may be equally relevant in therapeutic contexts. In particular, spatial separation from clinical environments may contribute to perceived autonomy, emotional regulation, and symbolic transition toward recovery. This perspective encourages the integration of psychological accessibility into future GIS-based health-planning frameworks.

A further contribution of this study lies in its transferability. The proposed DSS is not limited to FED, but may also be adapted to other vulnerable populations requiring personalized environmental support. Its applicability depends primarily on the availability of georeferenced facilities and the capacity to collect structured preference data. However, institutional coordination and local interpretations of “care environments” remain important contextual factors that may influence implementation outcomes.

Finally, the study reinforces the strategic role of spatial planning in public health governance. By linking healthcare systems with territorial resources, GIS-based decision-support tools can support more equitable, place-based, and adaptive models of care. Within this framework, planning is not only concerned with the distribution of spaces, but also with shaping the conditions under which environments become therapeutically meaningful.

Overall, the proposed approach demonstrates that integrating food-related environments, social agriculture, and spatial analysis can expand the scope of recovery-oriented care. It provides a replicable methodological foundation for future research and offers practical guidance to planners, healthcare providers, and policymakers seeking to design inclusive and health-supportive landscapes.