From coverage to capability: assessing equitable access to community-based senior meal programs in Shanghai

Abstract

Background

Shanghai is undergoing rapid population ageing, increasing demand for equitable community-based care. Government-led senior meal programs are a key local food-environment action to support healthy ageing in place, yet evidence on their effective service capability and spatial equity remains limited.

Methods

This study estimated older residents’ access to meal services using an improved Gaussian two-step floating catchment area (Ga2SFCA) model, integrating 2025 gridded population data (aged ≥ 60) with geocoded facility locations and district-level demographic calibration. Facility supply was parameterized as meals-per-session capacity under both policy-minimum and robustness scenarios. Overall inequality was assessed using the Gini coefficient, spatial clustering of supply–demand mismatch was examined using Local Moran’s I, and public–commercial coordination was evaluated using the Local Colocation Quotient (LCLQ).

Results

Citywide inequality in per-capita meal-service capability was substantial (Gini = 0.619), demonstrating a pronounced urban–rural contrast. Capability was lower in central urban areas and adjacent neighborhoods, whereas suburban town centers exhibited relatively higher capability; nevertheless, peripheral service gaps persisted. Supply–demand mismatch showed significant spatial clustering. LCLQ results indicated greater spatial feasibility for government-guided commercial participation in central districts, while peri-urban areas demonstrated weaker coordination.

Conclusions

Planning based solely on administrative coverage targets is insufficient. A capability-oriented framework is required, prioritizing joint public–commercial provision where spatial feasibility is high and targeted interventions in peripheral blind spots to improve equitable access for older residents.

Introduction

The aging of the global population is accelerating, and the daily care models are being greatly reshaped. By 2030, one-sixth of the world's population will reach or exceed the age of 60; by 2050, the global elderly population is expected to grow to approximately 2.1 billion [1]. China is also experiencing rapid population aging. By the end of 2024, the national population aged 60 and above had reached 310 million, accounting for 22% of the total population [2]. Many elderly people are in an "empty nest" state, living alone or living only with their spouses. Therefore, the ability to provide daily care and adequate and reliable nutrition has become a key issue for the elderly [3, 4]. Moreover, most elderly people tend to "stay at home", that is, they continue to live in existing housing and familiar community environments. For example, in the United States, 75% of adults aged 50 years and above want to stay in their current home, and 73% want to stay in their original community [5]. Various urban governance agendas—such as the "15-min city/life circle"—widely regard the community as the basic unit for providing proximate, accessible, and sustainable public services [6, 7]. Within this system of neighborhood-level facilities, community-based senior meal programs are directly related to whether older adults can "eat well" safely and affordably [8]. Therefore, this is not only a public health issue but also an action arena where issues of spatial equity and distributive justice are concretely manifested [9]. Evidence from the fields of geriatrics and public health indicates that community-dwelling older adults generally face a risk of malnutrition [10, 11]. Nutritional interventions, particularly when combined with community-embedded social interactions, can mitigate health decline, enhance subjective well-being, and strengthen social connections [12–14].

In terms of policy and in practice, many countries and cities are testing community meal programs that also create chances for social contact. Japan’s Long-Term Care Insurance System, introduced in 2000, built a universal, multi-level service system across home, community, and institutional care, and in doing so helped move the main care model from “family-based care” toward shared social responsibility [15]. In Singapore, the GoodLife! Makan program uses collective cooking and shared dining in community kitchens to build neighborhood support [16]. In the United States, Café Plus, launched by the non-profit group Mather Lifeways, brings together dining, learning, and leisure activities in community spaces [17]. Because nutritional status [11], healthy longevity [18], and the Sustainable Development Goals (SDGs) [19] are closely linked, the Chinese government treats community meal programs as a key part of the elderly care service system. The “14th Five-Year Plan for the Development of National Aging Undertakings and the Elderly Care Service System” (2021–2025) stresses the need to rely on home- and community-based elderly care service systems [20]. These programs may look different in terms of rules and management, but they share one goal: at the neighborhood scale, they offer nearby meal services in social settings, ease loneliness, and reduce nutrition risks. Therefore, whether older residents can reach timely, nutritionally adequate, and socially engaging meal services in their own communities has become an important way to judge how fair aging in place is.

Compared with the mature research paradigm for assessing the equity of public service spaces such as medical services and urban green spaces [21–23], the spatial assessment of the capacity of community meal programs lags behind. The existing literature has focused mainly on the following directions. First, attention is given to policy responses and governance pathways. For example, Wu et al. discussed the necessity and feasibility of "aging-friendly" service certification in Chinese community elderly canteens by systematically reviewing the service and certification standards related to meal assistance for the elderly [24]. In eastern England, Dickinson et al. examined the "door-to-door" service and found that the service not only provides food, but also provides the necessary care to support the elderly to maintain an independent life at home [25]. In view of those benefits, they urged local officials to adequately weigh the risks to the well-being of older persons before considering any service cuts. In addition to the basic food delivery services, the research has focused on the quality of the dining experience itself, emphasizing user satisfaction [24, 26, 27]. For example, Hung et al. investigated how renovating dining spaces impacts residents in long-term care. They found that improving the physical environment successfully reinforced person-centered care and encouraged more frequent social interaction among older adults [28]. The third category primarily concerns research on microscale site selection and the built environment, exploring how indoor and outdoor environmental characteristics influence dining volume and older adults' willingness to use meal-assistance points. Zou et al. evaluated the impact of indoor and outdoor built environment elements on dining volume, providing empirical references for the site selection of senior dining halls and the optimization of indoor and outdoor spaces [29]. The fourth category of studies evaluates the coverage of community meal programs starting from administrative units such as districts or communities, and on this basis, proposes layout optimizations within these basic units [30, 31]. Although these studies provide an important foundation for understanding community meal programs, much of the existing work remains at the case-study scale, leaving substantial room for further conceptual and empirical extension.

Shanghai, in the Yangtze River Delta (YRD), is one of China’s most economically developed megacities and one of the earliest to enter “deep aging” [32]. Rapid population aging has made eldercare a major public-welfare priority that requires sustained investment and institutional support [33]. Shanghai also includes a high-density urban core and an extensive suburban–rural hinterland, reflecting a typical urban–rural dual structure. This heterogeneity makes Shanghai a useful case for examining how public services are distributed across contrasting spatial contexts and for comparison with other large cities [34, 35]. To support aging in place, Shanghai has steadily expanded its community meal programs. In 2008, the program was incorporated into the municipal government’s key livelihood initiatives. Since 2019, a series of policies and regulations has strengthened service standards and promoted more systematic implementation [36]. In parallel, researchers have examined Shanghai’s community meal programs from multiple perspectives. For example, Li et al. reviewed the evolution of Shanghai’s community-based eldercare system through the lens of catering services and drew on international experience to discuss implications for strengthening eldercare in China [37]. Using community-level data and multiple time thresholds, Huang et al. estimated accessibility to community meal programs and identified low-access clusters and underserved areas using spatial autocorrelation [30]. Gong et al. conducted fine-grained assessments across Shanghai’s community units by integrating an enhanced Gaussian 2SFCA with interpretable random forests and geographically weighted regression [38]. They documented accessibility disparities and spatially heterogeneous determinants, providing more detailed evidence on the spatial organization of community meal programs.

From a policy perspective, Shanghai has recently clarified institutional targets and implementation requirements for community meal programs. Documents issued by the Shanghai Civil Affairs Bureau specify indicators for service coverage and quality and set a target of approximately 250,000 persons served per day by the end of 2025 [39]. By August 2025, Shanghai had established about 400 senior dining halls and more than 2,000 meal-assistance sites, achieving broad coverage at the subdistrict/township level [40]. In parallel, Shanghai has explored collaborative mechanisms to engage commercial restaurants in meal assistance. The 2024 implementation guidelines encourage commercial restaurants to establish “senior dining tables” and facilitate participation through government service procurement and co-development partnerships [39]. District-level implementation has varied across Shanghai. For example, constrained by limited siting options and long construction timelines in the urban core, Huangpu District has increasingly mobilized time-honored restaurants and other commercial catering resources [41]. In Xuhui District, the “Kanglehui” community canteen provides multi-cuisine options and co-locates several time-honored restaurant brands in a single venue, creating a hybrid model of a publicly built platform with commercial restaurant provision [42]. Together, these initiatives indicate that meal assistance involves not only government facility siting and coverage but also the coordinated development of a community food-service system that integrates public and commercial provision.

Despite progress in research and practice, several issues warrant further refinement. First, prior studies often represent supply using facility counts or threshold-based accessibility measures. However, they rarely clarify the institutional definition of “service capacity” or how it varies across scenarios, even though policy documents now specify explicit capacity targets [30, 38]. This gap requires transparent capacity parameterization and robustness analyses to operationalize “effectively obtainable supply.” Second, demand patterns in megacities change rapidly. Moreover, the urban core and suburban areas differ markedly in residential forms, population registration, and how populations are represented spatially. Using a single spatial scale or a single-source population proxy may miss “important but statistically inconspicuous” demand hotspots, including mixed residential–commercial corridors and peri-urban transition zones. Third, although policies encourage commercial participation, additional evidence and methods are needed to (i) identify where collaboration is spatially feasible and cost-effective; (ii) develop reproducible criteria for screening eligible restaurants; and (iii) translate these criteria into operational strategies for coordinated siting and service integration.

This paper is organized to guide readers through the analysis. Figure 1 provides an overview of the study design and methodological framework. Sect. "Materials and methods" details the study area, data, and methods. Sect. "Results" presents the empirical results. Sect. "Discussion" discusses the implications, policy recommendations, and limitations. Finally, Sect. "Conclusion" offers conclusions and research significance.

Fig. 1.

Methodological framework

Materials and methods

Data collection and processing

Senior meal program facilities

Data on senior meal program facilities were sourced from the Shanghai Elderly Care Service Platform (https://shyl.mzj.sh.gov.cn/). The existing meal program facilities in Shanghai can be categorized into two main types: the first type is the senior dining halls, which are mostly constructed under government guidance, with facility site selection referencing the "15-min life circle" standard. These dining halls usually have their own kitchens and larger dining areas, ensuring a strong capacity to serve the community. Some have been upgraded into “Canteen-plus” spaces that link dining with daily life functions such as health support, cultural and recreational activities, and basic convenience services [31]. In these spaces, older adults can enjoy a kind of “one-stop” community care. Another type is the small meal-assistance point that is usually located on the Residents’ Committee (RC), which is the office space for grassroots self-governance bodies that handle local public and welfare affairs. These points mainly provide simple meal services, have a single core function, and focus on meeting the immediate dining needs of nearby older residents.

The two facility types were parameterized by daily service capacity and incorporated into the Ga2SFCA model as supply-side inputs. Because current policy documents in Shanghai specify minimum service-capacity requirements for different facility types but do not define a unified upper limit for actual operating capacity, a two-scenario parameterization strategy was adopted. Under the policy-minimum scenario, the daily serving capacities of senior dining halls and meal-assistance sites were set at 150 and 10 persons served per day, respectively [39, 43]. Under the robustness scenario, the corresponding values were set at 300 and 50 persons served per day. These values were not taken directly from a single administrative standard. Instead, they were specified as scenario-based estimates that considered differences in venue size, kitchen configuration, meal provision mode, and service functions between the two facility types. Specifically, the value of 50 persons served per day for meal-assistance sites was informed by field visits to relevant facilities across multiple districts in Shanghai. The value of 300 persons served per day for senior dining halls was back-calculated from Shanghai’s 2025 target of approximately 250,000 persons served per day and the planned total number of facilities [43]. This scenario was included as a robustness check to assess the sensitivity of the findings to alternative supply-side assumptions.

Elderly population data

In accordance with Shanghai’s community meal program policies, which define the target population as adults aged 60 years and above, this study used the permanent resident population aged 60 years and above as the demand-side population. Population data were obtained from the age-structured gridded population dataset released in WorldPop 2025 A v1 at 100 m spatial resolution [44]. Because this dataset represents a modeled population surface rather than observed census counts, district-level calibration was performed to improve consistency with the policy evaluation period and official demographic statistics. Specifically, the total population aged 60 years and above in each district was estimated using the most recently published 2025 district statistical yearbooks, based on the permanent resident population size from late 2024 to early 2025 and the district-level aging rate. These estimates were then used to generate district-specific calibration coefficients, which were applied to proportionally rescale the corresponding WorldPop grid cells for the population aged 60 years and above in each district. This procedure preserved the fine-scale spatial distribution within districts captured by WorldPop while reducing potential discrepancies between gridded estimates and officially reported district-level totals.

For the subsequent accessibility analysis, the calibrated population grid was aggregated into a 500 m × 500 m fishnet, and the geometric center of each cell was used as a demand point (Fig. 2). The 500 m fishnet balances spatial detail with computational feasibility and helps identify potential demand areas that may be underrepresented in community population records, including mixed residential–commercial areas, roadside settlements, and peri-urban transition zones.

Fig. 2.

Elderly population aged 60 years and above in Shanghai (500 m × 500 m grid)

Commercial dining facilities

To assess the spatial feasibility of expanding community meal programs through collaboration with existing market actors, this study constructed a citywide database of potentially age-friendly commercial dining facilities in Shanghai. The database was derived from food-service merchant data collected from the Dianping platform in August 2025. As an open local lifestyle platform, Dianping provides structured information on merchant names, category hierarchy, average per-capita expenditure, review activity, delivery availability, and signature dishes. These platform variables allowed us to identify the collaborative potential of commercial dining facilities using transparent and reproducible rules rather than relying solely on subjective judgment.

The screening procedure comprised two stages. In the first stage, a set of structural filtering rules was applied to generate an initial pool of candidate commercial restaurants. After basic data cleaning, removal of invalid coordinates, and spatial deduplication, business types clearly not comparable to routine community meal provision were excluded, including dessert and bakery shops, beverage and coffee shops, bars and nightlife venues, buffets, hotpot restaurants, barbecue and skewer restaurants, and other food-service formats primarily oriented toward leisure consumption, social dining, or highly stimulating taste preferences. Only merchants with an average per-capita expenditure of RMB 10–40 were retained. This price range was intended to approximate the cost level of existing community senior dining halls in Shanghai while allowing moderate flexibility across neighborhoods and commercial settings. To reduce the influence of newly opened or low-activity establishments on subsequent spatial statistical analysis, each merchant was required to have at least 20 reviews. Finally, because community meal programs in Shanghai are increasingly operating through a combined dine-in and delivery model, only merchants offering takeout or delivery services were retained. After these structural filters were applied, the initial candidate pool comprised 60,464 merchant records. The structural filtering rules used in the first stage are summarized in Table 1.

Table 1.

Structural screening rules for the preliminary commercial dining candidate pool

Step	Screening dimension	Operational rule	Purpose
1	Data source and scope	August 2025 Dianping food-category merchant records in Shanghai; fields used included merchant name, category hierarchy, average spending, review count, takeaway availability, and major dishes	To build a citywide and reproducible dataset of commercial dining venues
2	Data cleaning	Remove duplicate records, invalid coordinates, and abnormal or unidentifiable entries where possible	To reduce platform noise and avoid double-counting in spatial analysis
3	Hard business-format exclusion	Exclude non-ordinary-meal or non-comparable formats at the category level, including dessert/bakery shops, beverage-only outlets, bars and nightlife venues, buffet restaurants, hotpot restaurants, barbecue/skewer formats, and other leisure- or stimulation-oriented dining types	To retain only venues broadly comparable to routine neighborhood meal provision
4	Affordability filter	Retain merchants with average spending between RMB 10 and 40 per person	To approximate the price band of community senior dining halls and low-cost neighborhood meals
5	Operational stability filter	Retain merchants with review count ≥ 20	To reduce the influence of newly opened, weakly active, or noisy stores on spatial statistics
6	Delivery feasibility filter	Retain only merchants with takeaway or delivery services	To capture venues with practical potential for off-site meal support and collaboration
7	Candidate pool generation	Venues meeting all the above criteria were retained as the preliminary commercial dining candidate pool (n = 60,464)	To form the input dataset for the second-stage empirical dictionary screening

The second stage introduced an empirical lexicon-based screening process to identify establishments more likely to be compatible with government-guided senior meal-assistance collaboration. Rather than attempting to infer the actual nutritional composition of each restaurant, this step used observable business characteristics and dish information from secondary data as proxy indicators of compatibility with routine meal provision, price accessibility, and potential age-friendliness. The screening logic integrated four sources of evidence: (1) second- and third-level category labels, (2) merchant names, (3) signature dish names, and (4) publicly available collaboration cases. On this basis, the empirical lexicon was applied to the candidate pool retained after the first-stage screening.

The lexicon was developed using a bottom-up empirical replication approach. Rather than relying solely on an abstract concept of “healthy catering,” the keyword system was calibrated against publicly released collaboration cases from several districts in Shanghai, particularly the public list of “senior restaurants” in Huangpu District and age-friendly meal practices promoted in Xuhui District, such as smaller portion sizes and reduced oil and salt. These local practice samples were treated as evidence of actual policy implementation rather than as idealized normative assumptions. Based on these cases, repeatedly occurring brand names, service formats, and dish characteristics were extracted, and the identification rules were refined for application to citywide Dianping data. The structure of the empirical lexicon is presented in Table 2.

Table 2.

Empirical lexicon for identifying collaboration-compatible candidates within the retained commercial dining pool

Dictionary module	Field used	Positive signals	Cautionary signals within retained formats	Interpretation
Routine meal format	Level-2 and level-3 category labels	noodle shops, porridge shops, dumpling/wonton shops, simple Chinese fast meals, home-style cooking, vegetarian restaurants, Shanghai/Jiangsu–Zhejiang cuisine	highly processed snack-heavy subtypes or strongly flavored specialty subtypes	Used to distinguish ordinary neighborhood meal providers from less suitable subtypes that remained after first-stage screening
Merchant identity	Merchant name	terms indicating porridge, noodles, soup, rice meals, home-style cooking, or neighborhood dining	terms implying strongly stimulating, heavily seasoned, or snack-dominant positioning	Used as a supplementary cue when category labels were too broad
Dish structure	Major dishes/featured dishes	porridge, soup noodles, wontons, dumplings, rice-meal sets, steamed dishes, boiled dishes, braised dishes, and vegetable dishes	repeated cues of preserved meats, heavily oily toppings, strongly spicy dishes, or snack-dominant combinations	Used to infer whether a venue was more compatible with routine elderly dining under secondary-data conditions
Practice-based calibration	Public district cooperation rosters	brands and store types repeatedly appearing in publicly disclosed elderly-dining cooperation lists in Huangpu and Xuhui	stores without correspondence to observed cooperation practice and simultaneously showing multiple cautionary signals	Used to refine the dictionary through empirical replication of observed local practice rather than purely top-down assumptions

Within the retained candidate pool, positive signals mainly included routine neighborhood-oriented food-service formats such as noodle shops, congee restaurants, dumpling and wonton shops, simple Chinese fast-food restaurants, home-style restaurants, vegetarian restaurants, and Shanghai or Jiangsu–Zhejiang cuisine, especially when dish information indicated an emphasis on soup noodles, steamed or boiled dishes, braised foods, vegetables, or other conventional meal structures. By contrast, risk signals were primarily reflected in certain highly processed snack-type subcategories that remained in the candidate pool, recurring indications of pickled or heavily seasoned dishes, and menu structures dominated by strongly flavored, oily, or snack-based items. These signals were used only as operational proxy indicators of collaboration feasibility under secondary-data constraints rather than as direct measures of nutritional composition.

After application of the second-stage empirical lexicon, 7,494 commercial dining facilities were retained as the final sample for collaborative analysis between government meal program facilities and commercial dining resources. Among them, 1,433 establishments could be matched at the brand or store-name level with publicly available district-level collaboration lists and were therefore identified as a more conservative policy-validated subset. This two-tier sample structure has clear methodological significance: the full sample reflects the broader spatial potential for future government–market collaboration in senior meal assistance at the city scale, whereas the matched subset represents a more conservative group that has already received partial validation in local practice. Accordingly, the commercial dining facilities used in this study should not be interpreted as an officially certified list of senior meal-assistance merchants, but rather as a reproducibly screened candidate pool that is spatially accessible, operationally active, and policy-relevant for collaborative meal-assistance provision.

Methods

A time-weighted Gaussian Two-Step Floating Catchment Area method (Ga2SFCA)

The two-step floating catchment area (2SFCA) method was first proposed by John et al. in 2000 and is now widely used in studies of urban public resource accessibility [45, 46]. The conventional 2SFCA procedure consists of two steps. In the first step, each supply point is taken as the center, and all demand points located within a predefined threshold distance are identified to calculate the facility-level supply-to-demand ratio . In the second step, each demand point is taken as the center, all accessible facilities within the same threshold are identified, and the corresponding supply-to-demand ratios are summed to obtain the overall accessibility . The basic formulation is as follows:

1

2

Although the traditional 2SFCA method can identify facility accessibility within a fixed service radius, it has two important limitations. First, it does not incorporate a continuous distance-decay mechanism and therefore implicitly treats all locations within the catchment as equally accessible. This assumption is problematic in the context of senior meal programs, because older adults are particularly sensitive to travel time and walking burden. Second, the conventional structure may overestimate effective supply because the same demand can be repeatedly counted across multiple accessible facilities, which may lead to an inflation bias in accessibility estimates [47]. To address these limitations, this study adopted a time-weighted Gaussian 2SFCA (Ga2SFCA) framework and replaced the conventional distance threshold with a travel-time threshold . Walking travel times between demand point and facility were obtained from the Amap (Gaode) API and used as the impedance measure. A Gaussian distance-decay function was then introduced:

3

and , when.

In this specification, the Gaussian decay function does not introduce an additional free coefficient. Instead, the decay intensity is determined by the normalized ratio , with serving as the bandwidth parameter of the decay curve. Under this formulation, accessibility weights decrease smoothly as travel time increases and approach zero at the threshold boundary. Following recent methodological refinements [48], normalized demand-side and supply-side weights were introduced to reduce inflation bias in the allocation of supply and demand. The normalized demand-side weight from facility to demand point is defined as:

4

The normalized supply-side weight from demand point to facility is defined as:

5

Based on these normalized weights, the modified facility-level supply-to-demand ratio and demand-point accessibility are calculated as follows:

6

7

where denotes the daily service capacity of facility (persons served per day), denotes the elderly population at demand point , and represents the per-capita reachable effective service supply available to demand point under a given travel-time threshold.

Considering Shanghai’s “15-min life circle” planning framework and the greater mobility constraints of older adults, this study evaluated accessibility under 15-, 20-, and 25-min travel-time thresholds [49]. The 20-min threshold was used as the primary specification, while the 15-min threshold served as the planning benchmark and the 25-min threshold was used for sensitivity analysis. This design allowed the robustness of the accessibility estimates under alternative bandwidth settings to be directly examined. Compared with simple per-capita facility measures, the improved Ga2SFCA framework simultaneously incorporates service capacity, travel-time impedance, and competition for demand, thereby providing a more realistic estimate of the effective meal-service supply available to older residents across different urban and rural contexts [50].

Bivariate Local Moran’s I

To examine localized spatial associations between meal-service supply and population demand, this study employed bivariate local Moran’s I. The analysis used the Ga2SFCA output for each populated grid cell as the service variable, representing the potential meal-service supply accessible to grid under the assumed facility capacities of 300 persons served per day for senior dining halls and 50 persons served per day for meal-assistance points. The elderly population in neighboring grid cells was used as the demand-side contextual variable to evaluate whether each grid’s service level was spatially associated with surrounding demand intensity. Analyses were conducted separately under 15-, 20-, and 25-min travel-time thresholds. Prior to the local spatial analysis, both variables were transformed using log(x + 1) to reduce right-skewness and the influence of zero values and were subsequently standardized as z-scores. Spatial weights were constructed using an 8-nearest-neighbor matrix (KNN, k = 8) with row standardization among populated grid cells. Statistical significance was assessed using 999 random permutations, and grid cells with p < 0.05 were considered to exhibit significant local spatial association. The bivariate local Moran’s I statistic was calculated as:

8

where denotes the standardized value of the service variable in grid , denotes the standardized elderly population in neighboring grid , and represents the spatial weight between grids and . In this study, refers to the Ga2SFCA-derived potential supply index, and refers to the elderly population. According to the signs of and the spatial lag term, four types of local spatial association can be identified: high supply-high neighboring demand (H–H), high supply-low neighboring demand (H–L), low supply-low neighboring demand (L-L), and low supply-high neighboring demand (L–H). Among these, the L–H type indicates grid cells with relatively low service supply situated within high-demand surroundings, thereby highlighting localized resource-allocation mismatch.

Gini coefficient

The Gini coefficient is a widely used measure of inequality and is commonly applied to assess distributive equity in resource allocation [50]. In this study, it was used to evaluate the degree of inequality in the spatial distribution of senior meal program resources across Shanghai at multiple spatial scales. From a people–service matching perspective, the analysis examined disparities in the allocation of meal-assistance resources from the subdistrict level to the district level. The Gini coefficient was calculated based on the Lorenz curve as follows:

9

where is the cumulative proportion of the elderly population in spatial unit k, is the cumulative proportion of service resources in the same unit, and represents the ordered spatial units.

Local Colocation Quotient analysis

To identify the local spatial relationship between government-led meal program facilities and commercial dining facilities, this study employed the Local Colocation Quotient (LCLQ) [51]. The LCLQ measures whether points of type A tend to be collocated with points of type B within a local neighborhood. In this study, government-led meal program facilities were defined as type-, and the commercial dining facilities retained after the screening procedure were defined as type B. For each type- facility i, the LCLQ is defined as:

10

where denotes the tendency of a type-A facility at location to be surrounded by type-B facilities; is the number of neighboring facilities included in the local neighborhood of the type- facility; is the number of type- facilities within that neighborhood; is the total number of type- facilities in the study area; and is the total number of all facilities. When > 1, type-A facilities are more likely than expected under spatial randomness to be collocated with type-B facilities, indicating positive local spatial association. When < 1, the two facility types are relatively separated at the local scale. Local neighborhoods were defined for each senior meal program facility using a k-nearest neighbors (KNN) approach based on projected planar distance. LCLQ values were calculated under k = 10, 20, and 30, with k = 20 used as the primary specification and the other two settings used for robustness checks. Statistical significance was assessed using 999 random permutations, and p < 0.05 was used as the threshold for significance. Points with insufficient neighboring observations for valid classification were labeled as undefined. Because LCLQ is a local statistic involving multiple simultaneous tests, significant results were interpreted primarily as exploratory spatial diagnostics rather than strict confirmatory inference.

Results

Spatial accessibility of senior meal program facilities

Based on Point of Interest (POI) data from Amap, the spatial distribution of senior meal program facilities in Shanghai is highly uneven (Fig. 3). The number of facilities is significantly greater in the northern part of the city than in the southern part; facilities are highly concentrated in the central urban area, with rapid decay toward the urban periphery. The distribution in the outer suburbs is generally sparse, with local agglomerations forming around urban satellite towns—such as Songjiang New City, Jiading New City, Chuansha Town in Pudong, and the Wusong area in Baoshan—and the administrative centers of suburban districts. In terms of the type of facilities, the dining halls for the elderly are mainly concentrated in the urban area of the city center. The elderly population in this area is dense, and the urban construction environment is crowded. The operation modes of community meal programs are different, and their distributions range are wider. In the suburbs, their distributions are relatively scattered, and they rely mainly on township offices, village committees and other local administrative units.

Fig. 3.

Spatial distribution of senior meal program facilities in Shanghai

Figure 4 presents the spatial accessibility of senior meal program facilities in Shanghai under two service-capacity scenarios and three travel-time thresholds (15, 20, and 25 min). The Ga2SFCA results were classified into five levels. In this study, a value of 1 indicates that, for a given demand point, the older population can access an average daily service capacity of approximately one person served per day, that is, within the specified catchment, the accessible facility supply is broadly sufficient to provide each older adult with one service opportunity per day. Accordingly, values below 1 indicate a relatively insufficient level of per-capita reachable effective service supply, whereas values above 1 indicate a relatively sufficient level of supply.

Fig. 4.

Spatial accessibility of senior meal program facilities in Shanghai. Panels (a–c) show the robustness scenario, with senior dining halls serving 300 persons per day and meal assistance points serving 50 persons per day, under 15-, 20-, and 25-min travel-time thresholds, respectively. Panels (d–f) show the policy-minimum scenario, with senior dining halls serving 150 persons per day and meal assistance points serving 10 persons per day, under 15-, 20-, and 25-min travel-time thresholds, respectively

Across the six maps, the spatial accessibility of senior meal program facilities in Shanghai exhibits a broadly stable pattern, characterized by relatively low accessibility in the central urban area and its adjacent neighborhoods, and relatively high accessibility in suburban town centers and some peripheral nodes. Specifically, the inner-city districts, including Huangpu, Jing’an, Xuhui, and Yangpu, are predominantly composed of low-accessibility grid cells, with extensive areas remaining below 0.25 under all three travel-time thresholds. Similar low-value clusters are also observed in southern Baoshan and parts of Minhang, reflecting intense competition between high concentrations of older residents and limited facility service capacity. By contrast, relatively high accessibility values are more likely to occur in suburban districts such as Jiading, Songjiang, Jinshan, Fengxian, and Chongming, particularly around town centers and some village centers. In many of these areas, values exceed 1, suggesting that facility service capacity and local demand are more favorably matched within the specified travel-time range. However, this advantage is not continuously distributed; rather, it shows a marked pattern of local concentration. High-value clusters are typically centered on towns or settlement centers, while their surrounding areas still contain numerous low-accessibility or zero-accessibility grid cells.

Comparison across the three travel-time thresholds shows that, although the overall spatial structure remains stable, both the extent and continuity of accessible areas increase as the threshold expands. Under the 15-min threshold, low-accessibility and zero-accessibility grid cells are more widely distributed, particularly in peripheral areas outside established town centers. Under the 20-min threshold, some moderate- and high-accessibility areas begin to show greater continuity. Under the 25-min threshold, the spatial extent of high-accessibility areas expands further, although substantial service gaps remain evident in the peripheral areas of many districts.

Comparison between the two service-capacity scenarios further indicates that increases in facility service capacity have a substantial effect on accessibility outcomes. Relative to the policy-minimum scenario, the robustness scenario not only expands the spatial extent of moderate- and high-accessibility areas, but also increases accessibility levels in many existing high-value and intermediate-value clusters. Nevertheless, this improvement does not fundamentally alter the overall citywide pattern: insufficient supply in the central urban area persists, while the relative advantage of suburban areas remains concentrated mainly in town and village centers rather than being evenly distributed across the wider peripheral hinterland.

Overall, the spatial accessibility of senior meal program facilities in Shanghai is shaped not simply by the presence of facilities, but by the combined effects of facility service capacity, demand concentration, travel-time constraints, and the urban–rural settlement structure.

Analysis of supply–demand patterns of urban–rural senior meal program facilities

To further clarify the accessibility patterns described above, this study used GeoDa to examine the supply–demand relationship of senior meal programs in Shanghai by analyzing the spatial association between the effective service supply of senior meal program facilities and the older population. The LISA clustering results (Fig. 5) show that, under all three travel-time thresholds, the supply–demand relationship of senior meal program facilities in Shanghai exhibits clear spatial differentiation.

Fig. 5.

Bivariate LISA cluster maps of effective service supply and elderly population for senior meal programs in Shanghai

Overall, high supply–high demand (H–H) clusters are concentrated primarily in the central urban area and several suburban town centers. These areas represent locations where relatively strong service capacity coincides with relatively high concentrations of older residents, indicating comparatively favorable local matching between supply and demand. By contrast, low supply–high demand (L–H) clusters are distributed mainly around the edge of the urban core and in parts of the urban–rural transition zone, forming discontinuous belts or patches surrounding the main H–H areas. These clusters indicate areas where demand pressure is relatively high but effective reachable supply remains insufficient, suggesting persistent local service shortfalls.

In the more peripheral suburban and rural areas, two additional patterns are evident. Low supply–low demand (L–L) clusters are widely distributed in outer suburban areas, particularly in places with relatively dispersed settlement patterns and comparatively low concentrations of older residents. In these locations, both service supply and demand remain limited. High supply–low demand (H–L) clusters are fewer in number and tend to appear as scattered nodes embedded within outer suburban towns or village-center areas. These clusters suggest that service capacity is relatively concentrated at specific local nodes, despite a comparatively weak surrounding demand base. A substantial number of units are classified as not significant, especially in the outer urban periphery, parts of the suburban interior, and transitional zones between urban and rural settlements. These results suggest that, in many areas, the local spatial association between service supply and the older population remains weak or spatially heterogeneous, rather than forming a clearly clustered pattern.

Comparison across the 15-, 20-, and 25-min thresholds shows that the overall spatial structure of supply–demand association remains broadly stable, but the extent of clustering changes with the travel-time threshold. Under the 15-min threshold, L–H and not-significant units are more extensive, indicating stronger local mismatches under a more restrictive accessibility assumption. Under the 20-min threshold, some H–H clusters become more continuous, particularly in the central urban area and selected suburban nodes, suggesting improved local matching as the accessible service range expands. Under the 25-min threshold, the overall pattern remains similar, but some previously mismatched areas shift toward more balanced cluster types, while peripheral L–L areas remain widespread. This indicates that extending the travel-time threshold can alleviate some local supply–demand tension, but does not fundamentally alter the broader urban–rural structure of uneven provision.

Taken together, the bivariate LISA results indicate that the supply–demand relationship of senior meal program facilities in Shanghai is characterized by a central concentration of relatively well-matched areas, surrounded by belts of local mismatch, and by more fragmented patterns in suburban and rural areas. This further suggests that improving spatial equity requires not only increasing facility capacity, but also addressing the uneven alignment between facility locations and the distribution of older residents.

Spatial equity analysis of senior meal program facilities

To assess the equity of service distribution within administrative units, this study calculated Gini coefficients at three spatial scales: municipality, district, and subdistrict/township. The results indicate that the spatial distribution of senior meal program facilities in Shanghai remains markedly uneven. At the municipal scale, the Gini coefficient calculated based on township-level units was 0.619, indicating substantial inequality in the citywide distribution of service capacity.

At the district scale, the Lorenz curves (Fig. 6) reveal pronounced variation in internal equity across districts. Overall, inequality is most pronounced in the outer suburban districts. Chongming had the highest Gini coefficient (0.883), followed by Fengxian (0.839), Jinshan (0.810), Qingpu (0.786), and Songjiang (0.742). These values indicate that service capacity in these districts was concentrated in a limited number of local units rather than being evenly distributed across each district as a whole. Among the suburban districts, Jiading (0.511) had the lowest level of internal inequality, suggesting a comparatively more balanced distribution. Transitional districts adjacent to the urban core, including Pudong (0.674), Baoshan (0.668), and Minhang (0.633), also had relatively high Gini coefficients, indicating persistent internal disparities between better-served nodes and surrounding areas.

Fig. 6.

District-level Lorenz curves and Gini coefficients of senior meal program in Shanghai

The central urban districts exhibited a different pattern. Huangpu (0.232), Hongkou (0.267), Jing’an (0.285), Xuhui (0.322), Changning (0.357), and Yangpu (0.380) all had comparatively low Gini coefficients, indicating a relatively balanced internal distribution of service capacity. Putuo (0.415) remained more unequal than the other central districts but still performed substantially better than most suburban districts. Taken together, the district-level results indicate a clear urban–rural gradient: the central city was characterized by a relatively even intra-district allocation of service capacity, whereas the outer suburbs exhibited a much stronger internal concentration of service resources.

At the subdistrict/township scale, the spatial pattern of inequality became more explicit (Fig. 7). Most suburban towns and rural townships had Gini coefficients above 0.50, indicating high internal inequality in the distribution of service capacity. In these areas, facilities were concentrated mainly in administrative centers or major settlement nodes, while peripheral villages and dispersed residential areas remained comparatively underserved. By contrast, subdistricts in the central urban area generally had lower Gini coefficients, with many units falling below 0.30 or within the 0.20–0.40 range, indicating a relatively more even internal distribution. Intermediate values were concentrated mainly in the urban–rural transition belt and selected suburban nodes.

Fig. 7.

Subdistrict-level Gini coefficients of senior meal program facilities in Shanghai

Overall, the multi-scale Gini results consistently show that the spatial equity of senior meal program facilities in Shanghai follows a distinct urban–rural gradient. Inequality was relatively limited in the central city but became substantially more pronounced in suburban districts, rural townships, and urban–rural transition zones. This pattern suggests that improving spatial equity requires not only expanding overall service capacity but also reducing the strong internal concentration of facilities within suburban administrative units.

Colocation analysis of senior meal program facilities and commercial dining facilities

Given the growing demand for senior meal programs and the predominantly government-led nature of Shanghai’s current meal-support system, this study employed the Local Colocation Quotient (LCLQ) to examine the spatial relationship between senior meal program facilities and screened commercial dining facilities. The aim was to determine whether government-led facilities are embedded in urban environments more conducive to public–private collaboration. Figure 8 presents the results under three KNN neighborhood specifications (k = 10, 20, and 30). Throughout this section, “significant” and “not significant” refer to permutation-based local colocation results evaluated at p < 0.05.

Fig. 8.

Local colocation patterns between senior meal program facilities and commercial dining outlets under KNN-10, KNN-20, and KNN-30 neighborhood specifications. Significant and non-significant types were classified based on permutation tests at p < 0.05

Overall, the results reveal a broadly stable dual pattern. Local colocation is observed in the central urban area and in some suburban nodes, whereas local isolation is more widespread across suburban and peripheral areas. Across all three neighborhood specifications, “Isolated” types outnumber “Colocated” types. This indicates that, in much of Shanghai, commercial dining resources that could potentially support collaboration around existing government-led meal program facilities remain relatively limited, thereby constraining the formation of a stable spatial foundation for public–private collaborative meal provision.

From the perspective of neighborhood sensitivity, the overall spatial structure remains stable, whereas significant patterns become more pronounced ask increases. Under KNN-10, the results are dominated by non-significant types, particularly Isolated – Not Significant (n = 1035) and Colocated – Not Significant (n = 758), whereas significant colocated points remain comparatively few (Colocated – Significant, n = 107). At the same scale, Isolated – Significant already reaches n = 555, indicating that even within smaller neighborhoods, local isolation is more common than local colocation. Under KNN-20, significant patterns become clearer: Colocated – Significant increases to n = 147, and Isolated – Significant rises to n = 677, while both non-significant categories decline. Under KNN-30, this trend becomes more pronounced: Colocated – Significant increases further to n = 160, whereas Isolated – Significant rises to n = 737. Overall, as neighborhood scale expands, local spatial structure becomes more readily detectable, but the dominant pattern remains one of relative isolation rather than strong colocation.

In spatial terms, Colocated – Significant points are concentrated mainly in the central urban area and in several suburban nodes with relatively mature residential and commercial environments. Within the urban core, these points are found mainly in parts of Huangpu, Xuhui, and central-western Pudong, indicating that existing senior meal program facilities in these areas are more likely to be located near screened commercial dining facilities. Outside the urban core, significant colocated points also appear in some large suburban residential areas and town centers, suggesting that these places may provide favorable spatial conditions for future government-guided participation of social catering in senior meal programs. By contrast, Isolated – Significant points are more widely distributed and constitute the dominant significant type across much of the city. These points occur both along the edge of the urban core and, more extensively, in suburban and peripheral areas, with particularly notable concentrations in Jiading and in eastern and southern Pudong. In these locations, existing senior meal program facilities and nearby commercial dining facilities are often clearly separated in space, indicating a relatively weak local foundation for public–private collaboration. This pattern is especially evident in many outer suburban towns and peripheral settlements, where either public facilities or screened commercial dining facilities tend to exist in isolation rather than forming a close colocated relationship.

The two non-significant categories also warrant attention. Colocated – Not Significant points are widely distributed across mixed-function urban neighborhoods and some suburban built-up areas, indicating that nearby commercial dining resources are present but have not formed statistically significant local clustering. Isolated – Not Significant points are widely but discontinuously distributed across the suburban interior and urban–rural transition zones, suggesting that although some degree of local separation exists, it is not strong enough to differ significantly from the overall citywide background distribution.

Overall, the LCLQ results indicate that the spatial relationship between senior meal program facilities and commercial dining facilities in Shanghai is characterized by limited but identifiable local synergy in the central urban area and some suburban nodes, whereas a much larger share of suburban and peripheral areas is dominated by local isolation. This suggests that government-guided public–private collaborative meal provision is not uniformly spatially feasible across the city; rather, its feasibility is strongly constrained by local settlement form and the existing commercial service environment.

Discussion

Factors shaping the spatial pattern of senior meal programs in Shanghai

Consistent with earlier Shanghai studies, the present analysis also identifies a clear central–peripheral contrast in senior meal programs, with relatively low accessibility in the urban core and comparatively higher service levels in some suburban and peripheral nodes. Huang et al. similarly reported relatively low meal-service accessibility in central Shanghai and comparatively better conditions in some peripheral residential communities [30]. Building on this observation, the present study further shows that suburban advantages are concentrated mainly in town centers and a limited number of settlement nodes, whereas extensive peripheral hinterlands still contain many low- or even zero-accessibility units. By using a 500 m × 500 m fishnet and a district-calibrated gridded elderly population surface, this study is better able to detect internal suburban differentiation, urban–rural transition zones, and peripheral service gaps.

An important reason for the relatively low accessibility observed in the central urban area is the combined effect of high demand density and limited effective service capacity, which is broadly consistent with the findings of Gong et al. [38]. In central Shanghai, the problem is not simply the absence of facilities. Rather, although meal facilities are not scarce in terms of administrative coverage or facility counts, they must serve a highly concentrated older population within severely constrained urban space. High land values, limited room for expansion, and the cost of adding new welfare facilities all constrain the ability of the meal-support system to respond to rapidly growing demand. As a result, the central city may appear to have relatively complete formal coverage, while per-capita reachable effective service supply remains low in practice.

A different mechanism can be observed in some old industrial residential areas and workers’ new villages in Jiading and Minhang. These places often show relatively favorable local supply–demand matching not because resources are exceptionally abundant, but because of their historically formed community spatial structure. In these areas, work, residence, welfare, and daily services were originally organized in a relatively compact and integrated spatial form [52]. Existing residents’ committees, activity centers, and neighborhood service nodes continue to provide a favorable built environment for the organization and delivery of senior meal programs. Long-standing neighborhood ties may also facilitate older adults’ access to service information and actual service use [53].

A partly similar but institutionally distinct pattern can also be observed in some large residential areas and population-receiving communities in Pudong, Songjiang, Fengxian, and Jiading. As part of the strategy of guiding population redistribution outward from the urban core, public service facilities are often incorporated into planning at an early stage of development [54]. This has enabled some large residential projects to establish a basic public service foundation from the outset, thereby reducing the risk of the severe service shortfalls commonly seen at the urban fringe. At the same time, the present results suggest that such advantages are often localized rather than continuous. In many outer suburban districts, relatively high service levels remain concentrated in town centers or major settlement nodes, while the surrounding hinterland continues to exhibit marked fragmentation and weak service provision. This helps explain why suburban Shanghai can simultaneously display localized areas of relatively favorable service conditions and extensive peripheral service gaps.

Challenges facing senior meal programs in Shanghai

Although Shanghai has made substantial progress in advancing the 15-min community life circle, the current organization of senior meal programs still faces several structural challenges. The key issue is not simply whether facilities exist, but whether their service capacity is sufficient, whether resource allocation matches the actual spatial distribution of demand, and whether formal coverage can be translated into access that is genuinely usable for older adults [55].

The most prominent challenge in the central urban area lies in the tension between relatively balanced internal distribution and insufficient effective service capacity. The Gini results show that several central districts perform relatively well in terms of intra-district equity, yet the accessibility analysis indicates that per-capita reachable service supply remains low in these areas. This points to a density paradox: areas with relatively dense facility networks do not necessarily provide more adequate per-capita service supply for older residents [56]. In practice, relative balance within the central city does not automatically imply sufficient provision. Under conditions of high elderly population concentration, even a relatively even facility network may still lack the capacity needed to meet everyday demand. This contradiction resembles findings from studies of public sports facilities in Shanghai, where high density coexists with low per-capita resources [57]. It also reflects the spatial constraints faced by megacities in the provision of public services.

By contrast, the main challenge in suburban and rural areas lies in internal fragmentation beneath broad administrative coverage. The Gini analysis shows that inequality is more pronounced within outer suburban districts and townships, where service capacity is concentrated in town centers, administrative nodes, and a limited number of settlement centers, thereby producing a typical pattern of nodal concentration and peripheral weakness. This top-down approach has been important for rapidly expanding the service network, but it also shapes facility siting and service organization in ways that remain strongly tied to existing administrative nodes while paying insufficient attention to the actual residential locations and mobility patterns of older adults. As a result, there remains a gap between institutional coverage and effective accessibility in practice. For older adults living at village edges or in dispersed settlements, administrative coverage does not necessarily imply a meaningful improvement in service convenience.

A third challenge is the uneven spatial foundation for public–private collaboration in meal provision. The LCLQ results show that Isolated types consistently outnumber Colocated types, indicating that in much of Shanghai, government-led senior meal program facilities are not embedded in sufficiently developed commercial dining environments. This issue is especially evident in many suburban and peripheral areas, where public facilities and screened commercial dining resources are often either sparse or spatially separated from one another. By contrast, the central urban area and a limited number of suburban nodes show clearer local potential for collaboration, suggesting that the spatial feasibility of public–private coordination varies markedly across the city and that future policy should therefore be more differentiated and fine-grained.

In addition, dynamic demographic change and changing conditions of service use continue to increase the organizational complexity of senior meal programs. In some large residential developments, urban fringe areas, and population-receiving zones, population growth may outpace service adjustment, creating a clear lag between settlement formation and the maturation of the service network [58]. Meanwhile, digitalized and platform-based modes of service organization may improve efficiency for some groups while creating new barriers for older adults with limited digital skills or lower educational attainment [59]. The challenge facing Shanghai is therefore not only to improve the spatial distribution of facilities, but also to narrow the gap between formal provision, practical usability, and long-term adaptability.

Recommendations for optimizing the spatial layout of senior meal program facilities in Shanghai

Considering the findings above, future optimization of senior meal programs in Shanghai should move beyond the single administrative objective of full subdistrict-level coverage and adopt a configuration logic that integrates service capacity, demand identification, and conditions for collaboration. The policy implications can be organized into two major spatial policy zones and one dynamic adjustment context, thereby clarifying the practical scale of the recommended strategies.

First, in the central urban area, the area within Shanghai’s Inner Ring Road contains approximately 15.8% of all senior meal program facilities citywide but accommodates about 19.39% of the city’s older population. Although formal facility coverage appears relatively dense in this area, per-capita reachable effective service supply remains insufficient. Optimization should therefore not rely primarily on constructing additional stand-alone facilities, but instead prioritize strengthening the service capacity and collaborative potential of existing nodes [60]. On the one hand, service capacity could be enhanced through embedded micro-space retrofitting, the mixed use of community public space, and the expansion of existing canteens [61, 62]. On the other hand, drawing on the local collaboration zones identified by the LCLQ analysis and the pool of potentially suitable commercial dining establishments screened in this study, areas such as Huangpu, Xuhui, and central-western Pudong appear particularly suitable for piloting institutionalized pathways for introducing external market resources [41]. These areas account for approximately 8.32% of the city’s elderly population demand, whereas their share of existing senior meal program facilities is only 5.9%. Their common characteristic is that commercial dining resources are relatively abundant, whereas the expansion of new government-led facilities is constrained by both space and cost. The government could therefore select reputable and compliant food businesses, such as long-established local restaurants and chain fast-food outlets, and incorporate them into a senior-friendly meal-support network. Mechanisms such as senior dining tables, off-peak meal provision, customized set meals, and targeted subsidies could then be used to encourage eligible commercial establishments to participate, while price support could help older adults make use of these services. Such approaches are more likely to provide a practical way to relieve supply pressure in the central city.

Second, in suburban and rural areas, the priority should be to address fragmentation beyond town centers and service gaps in the surrounding hinterland. The results show that relatively high accessibility in suburban Shanghai is concentrated mainly in town centers and a limited number of settlement nodes, while extensive peripheral areas still contain many low- or zero-accessibility units. These areas are therefore not well suited to simply replicating the central-city model of filling gaps through collaboration with commercial dining establishments. Instead, they require a graded service network built around central nodes with flexible outward extension. In practical terms, town centers, nursing institutions, or central kitchens could serve as core nodes, while smaller pick-up points, reheating points, and flexible delivery nodes could be arranged in peripheral villages and dispersed residential locations. Through centralized meal preparation, third-party delivery, and multi-node coordination, service continuity could be improved. Rural areas alone account for approximately 16.6% of all senior meal program facilities citywide, representing a substantial share of Shanghai’s overall meal-support system. This indicates that they should not be regarded as marginal service spaces, but rather as an important component of the city’s senior meal-support system. Compared with public–private collaboration, the more urgent issue in suburban areas is how to improve spatial reach and delivery continuity under low-density settlement conditions [63].

Third, some areas should be treated as a dynamic adjustment context rather than as a fixed administrative zone. These areas are identified by recurring local low-supply–high-demand patterns and weak or non-significant public–commercial colocation, and are more likely to occur along the outer edge of the urban core, in urban expansion areas, and in parts of the suburban interior. This suggests that they are neither fully served core areas nor stable low-pressure zones. In such places, a one-off static allocation is unlikely to be effective. A more appropriate approach would be to establish a rolling monitoring mechanism based on fine-scale demographic change, walking accessibility, and the local commercial environment, so that areas where demand is growing faster than facility provision can be identified in time. This would reduce the risk that new population-receiving areas again develop into places where coverage exists, but service remains inadequate.

Finally, the commercial dining screening framework developed in this study also has practical policy value. It should not be regarded as an official certification list, but it provides a relatively transparent and reproducible pathway for identifying candidate establishments on the basis of menu attributes, business characteristics, and documented collaboration cases drawn from public platforms. This helps translate scattered local pilot experience into a more operational pre-screening tool. In Shanghai, such a framework is particularly relevant to the central urban area, where commercial resources are abundant, but the expansion of public facilities is spatially constrained. More broadly, it may also serve as an initial spatial diagnostic tool for evaluating the collaborative potential between government-led facilities and commercial dining resources in other megacities.

Research contributions, limitations, and future directions

Compared with earlier Shanghai studies, the contribution of this paper is better understood as a refinement and extension of existing analytical frameworks rather than a replacement of previous conclusions. First, the study shifts the analytical focus from facility presence and administrative coverage toward an integrated evaluation of effective service capacity, fine-grained demand identification, and the spatial feasibility of public–private collaboration. In other words, the paper considers not only where facilities are located, but also how many people they can serve, which demand locations remain underserved under practical walking constraints, and where collaboration with commercial dining outlets is more likely to relieve supply pressure.

Second, in terms of demand identification, the study uses a district-calibrated 2025 WorldPop elderly population surface and a 500 m × 500 m fishnet as the analytical unit, allowing a finer-scale assessment of spatial patterns during the policy implementation period. On this basis, the analysis is more sensitive to internal suburban differentiation, mixed residential corridors along roads, and service gaps in urban–rural transition areas. In particular, beyond town centers, the fishnet perspective reveals more clearly the fragmented pattern of relatively stronger nodes and weaker peripheries.

Third, the study introduces service-capacity parameterization in terms of persons served and compares a policy-minimum scenario with a robustness scenario, while also examining 15-, 20-, and 25-min travel-time thresholds. This makes it possible to discuss more transparently the difference between facility coverage and effective supply and to interpret the results more directly in relation to the policy context. Finally, by incorporating screened commercial dining outlets into the analysis and applying the LCLQ, the study identifies local patterns of colocation and isolation between government-led facilities and social catering, thereby providing more direct spatial evidence on whether and where social catering may realistically participate in senior meal programs.

Several limitations should nevertheless be acknowledged. First, although the demand-side population data were calibrated using official district-level statistics, they remain a modeled gridded population surface and cannot substitute for observed fine-scale residential data. Second, supply-side service capacities were defined through scenario-based parameterization. While this improves transparency, it cannot fully represent variation across facilities in operating schedules, delivery functions, and fluctuations in actual throughput. Third, the commercial dining screening framework identifies collaborative potential rather than actual willingness to participate, nutritional quality, or realized service performance. Fourth, the study is cross-sectional and does not incorporate individual-level information on affordability, health status, dietary preferences, digital capability, or actual service utilization. The equity discussed here therefore refers primarily to spatial opportunity rather than providing a complete account of equity in realized use.

Future research may proceed in at least three directions. First, operational records, subdistrict data, and residents’ committee information could be used to validate actual numbers of persons served, delivery coverage, and peak-hour capacity, thereby strengthening the empirical basis of supply measurement. Second, older adults’ income, living arrangements, health conditions, digital capability, and actual meal-use behavior could be incorporated to assess more systematically the gap between spatial accessibility and practical usability. Third, Shanghai’s ongoing experiments in collaboration with commercial restaurants could be followed over time, allowing more direct comparison of the suitability and performance of different collaboration models in the central urban area, suburban towns, and peripheral areas. Such work would provide a stronger empirical basis for differentiated governance of senior meal programs in megacities.

Conclusion

This study examined senior meal program facilities in Shanghai using a 500 m × 500 m fishnet as the basic analytical unit. By integrating a district-calibrated 2025 gridded elderly population surface, facility point data, and screened commercial dining outlets, the study developed an analytical framework to evaluate the accessibility, supply–demand relationship, spatial equity, and collaborative potential of senior meal programs. Using Ga2SFCA, bivariate LISA, the Gini coefficient, and LCLQ, the study systematically assessed the spatial pattern of senior meal programs in Shanghai.

The results indicate pronounced spatial differentiation in senior meal programs across Shanghai. Per-capita reachable effective service supply is relatively low in the central urban area and its adjacent neighborhoods, whereas suburban town centers and some settlement nodes show comparatively higher service levels. However, extensive peripheral hinterlands still contain many low-accessibility or even zero-accessibility units, indicating that suburban advantage is expressed mainly through nodal concentration rather than overall balance. The supply–demand analysis further shows that relatively stable high supply–high demand clusters are concentrated in the central urban area and in some suburban nodes. By contrast, the outer edge of the urban core, urban–rural transition zones, and some newly expanded residential areas are more likely to exhibit low supply–high demand patterns. The spatial equity analysis indicates substantial inequality in the distribution of service capacity, with a municipal Gini coefficient of 0.619. Inequality is more pronounced in suburban, rural, and urban–rural transition areas. The colocation analysis shows that Isolated types outnumber Colocated types overall. This suggests that stable local collaboration between government-led meal facilities and commercial dining resources has not yet formed in much of Shanghai, whereas the central urban area and a limited number of suburban nodes show greater collaborative potential.

Overall, the main challenge facing senior meal programs in Shanghai is not simply an insufficient number of facilities, but a spatial mismatch among service capacity, demand distribution, and conditions for collaboration. Building on earlier Shanghai research, this study advances the spatial analysis of senior meal programs in three respects: fine-grained demand identification, service-capacity parameterization, and the assessment of collaborative potential with commercial dining resources. These findings provide more detailed spatial evidence for optimizing Shanghai’s senior meal program system and offer a useful reference for the differentiated governance of community-based meal provision in other megacities experiencing rapid population ageing.

Abbreviations

YRD

Yangtze River Delta

2SFCA

Two-Step Floating Catchment Area

Ga2SFCA

Gaussian Two-Step Floating Catchment Area

LISA

Local Indicators of Spatial Association

LCLQ

Local Colocation Quotient

POI

Point of Interest

SDGs

Sustainable Development Goals

LSCs

Large-scale Residential Communities

RC

Residents’ Committee

H-H

High-High

H-L

High-Low

L-L

Low-Low

L-H

Low-High

Authors’ contributions

Conceptualization, H.W.D. and M.Y.J.; methodology, S.J.W; software, M.Y.J; validation, H.W.D, M.Y.J. and S.J.W; formal analysis, M.Y.J..; investigation, H.W.D.; resources, S.J.W.; data curation, M.Y.J..; writing—original draft preparation, H.W.D.; writing—review and editing, S.J.W.; visualization, S.J.W; supervision. M.Y.J.; project administration, S.J.W.; funding acquisition, S.J.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by National Natural Science Foundation of China, grant number 42401247; Humanities and Social Science Foundation of Ministry of Education of China, grant number 24YJCZH262.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.