The relationship between diet and health is now fully recognized(1); there is strong evidence for health benefits associated with the consumption of fruits and vegetables (F&V). The role of a supportive environment, i.e. one that promotes making the healthiest choice into the easiest choice at the physical, economic, sociocultural or political level, was highlighted at the Ottawa Conference for Health Promotion in 1986(Reference Milio2, 3). The evidence base on the role of the food environment in the food–health relationship is also growing(Reference Ball, Timperio and Crawford4, Reference Raine5). The food environment can be referred to as a ‘set of conditions in which one person (or a group of persons) has access to, chooses, prepares and eats foods’(Reference Ouellet, Hamelin, Auger and Galarneau6). As Raine noted: ‘individual determinants are necessary, but not sufficient, to explain eating behavior’(Reference Raine5). Food availability and food access, as conditions for food security, should not be overridden(Reference Hamelin, Habicht and Beaudry7–Reference Koc and Dahlberg11). Food availability is understood as the level of supply in the food market, and food accessibility as the physical and economic ability to access the food supply(Reference Clifford and David12).
Studies that focus on the physical access to food in American urban settings, and on clarifying the extent of social disparities, point to a consistent picture(Reference Cummins and Macintyre13). Supermarkets are more prevalent in wealthier neighbourhoods than in poorer ones, and this unequal spatial distribution is also associated with the racial make-up of communities(Reference Morland, Wing, Roux and Poole14–Reference Zenk, Schulz, Israel, James, Bao and Wilson19). Some findings also show that people living in high-income areas or those characterized by better physical access to a supermarket eat more F&V or have significantly lower odds of obesity and overweight than those living in conditions of deprivation(Reference Diez-Roux, Nieto, Caufield, Tyroler, Watson and Szklo20–Reference Morland, Roux and Wings22). Outside the USA, studies of food access draw more on the socio-economic status of the area than on wealth and racial composition when assessing deprivation. In Australia and the UK, studies show either no or small differences in the spatialization of food purchasing infrastructure and of food purchasing behaviour while considering the varying socio-economic factors across the target settings(Reference Turrell, Blakely, Patterson and Oldenburg23–Reference Cummins and Macintyre26).
In Canada, a few studies have found that unequal distribution of food stores has negatively impacted low-income neighbourhoods(Reference Latham and Moffat27, Reference Bertrand, Thérien and Cloutier28), while others conclude that urban households have generally good food access(Reference Apparicio, Cloutier and Shearmur29, Reference Smoyer-Tomic, Spence and Amrhein30). A recent research in Montréal found that 40 % of the population studied had to walk more than 500 m to reach a source of F&V, without regard to the district’s socio-economic levels for which no significant relationship was found(Reference Bertrand31). Generalizations are difficult to make due to the varying landscape and culture observed within and across provinces and territories. Even so, the link between food access and deprivation still needs to be clarified. In rural settings, however, studies have argued that supplying food might be more difficult with regard to the travel distance and its quality and freshness(Reference Burns, Gibbon, Boak, Baudinette and Dunbar32–Reference Cummins and Macintyre34). To our knowledge, food access has yet to be explored in rural Québec and merits closer examination since roughly 20 % of the total Québec population lives in a rural setting.
In order to fill this gap, the present study explores fresh F&V supplies by considering the setting and its level of deprivation. The main questions being addressed are:
1. Is there a difference in the spatial distribution of food stores found in rural and urban settings, the intra-urban setting and the intra-rural setting?
2. Is there a relationship between fresh F&V in-store availability, the type of food store, the specific setting and its level of deprivation?
Methods
Sample
Household food security and social inequalities in the greater Québec City area are already under investigation. Field work was thus conducted in the same area to benefit from an increasing pool of relevant data that could provided us with a basis for analysis(Reference Pampalon, Hamel, DeKoninck and Disant35–Reference Hamelin, Mercier and Bédard38). Urban and rural settings were differentiated by their population: the cities with more than 10 000 inhabitants being defined as ‘urban’ and those with less as ‘rural’(Reference Martinez, Pampalon, Hamel and Raymond39). In addition, the influence exerted by the metropolitan area, dubbed ‘metropolitan areas’ and ‘census agglomeration-influenced zones’, was also taken into account to identify rural areas(40). Data were collected in a sample of six contiguous urban districts of the subdivision La Cité and of seven rural communities in the regional county municipalities (RCM) of Portneuf, Charlevoix and of St-Anne-de-Beaupré.Footnote * The subdivision districts and communities were chosen because of their contrasting socio-economic characteristics as assessed using the Material and Social Deprivation Index. This index combines three material deprivation parameters (education level; activity rate; mean income) and three social deprivation parameters (proportion of separated, divorced or widowed persons; proportion of single parents; proportion of persons living alone) at the dissemination area scale, creating a nine-category typology of deprivation(Reference Pampalon and Raymond41).
Data collection procedure
Food stores census
Research (e.g. references 8, 42 and 43) and one study outlining a typology(Reference Pouliot and Hamelin44) show that food may be obtained from different sources. In Québec in 2006, 71 % of total food demand was met by the food retail sector against 29 % by the food service industry(Reference Brouillette45). As we focused on the most current way of acquiring fresh F&V, the food stores census was limited to the retail sector and precisely to those stores that are likely to carry them. Table 1 provides a definition for each type of fresh fruit and vegetable store (FVS) identified in the present study.
Table 1 Definition of the types of fresh fruit and vegetable store (FVS) under study
*The definitions are extracted and translated from Marquis and SimardReference Marquis and Simard(66).
†In the present study, specialty stores were further divided into ‘natural food stores’ and ‘delicatessen’.
To create a list of FVS, various databases were utilized, including the software Business 411 (2005; Tame Inc., Flintstone, GA, USA), accessible with the agreement of the Centre for Research in Regional Planning and Development (Laval University), and a registry from the Québec Ministry of Agriculture (2005). During data collection, the local telephone directory allowed us to verify addresses of stores that could not readily be located. Local visits also helped to complete and validate the list.
In-store availability
With reference to current Canadian studies, two main indicators – quantity and diversity – were defined to measure F&V in-store availability(Reference Bertrand, Thérien and Cloutier28, Reference Bertrand31, Reference Winson46, Reference Miedema47). First, a proxy measure of F&V quantity was developed, by estimating the length and width of stalls to calculate total shelf-surface (ft2). Second, to measure diversity, the different varieties of F&V available were counted, with variety being understood here as belonging to a similar taxonomic group. For example, all types of apples (e.g. McIntosh, Delicious, Granny Smith) were considered as belonging to one variety of fruit: the apple. A predefined list of the varieties of fruits (thirty-eight) and vegetables (thirty-seven) was used as a base by the observer.
Processes
Data were collected in September 2006, when F&V supply is bountiful. In order to minimize possible bias, only one observer (N.P.) visited the FVS(Reference Miedema47). The process of observation, described on the answer sheet (prepared and pre-tested in order to facilitate data collection), was carefully applied at each food store included in the study. Information was gathered by using a digital recorder. A second visit was made to a random sample (10 %) of stores located in the La Cité subdivision to assess the reliability of the measure of F&V quantity and diversity(Reference Wechsler, Basch, Zybert, Lantigua and Shea48).
Analyses
Statistical analyses were performed with the SPSS software package version 11·5 for Windows (SPSS Inc., Chicago, IL, USA), with a significance level of P < 0·05. For each community and district, the FVS spatial distribution (frequency of stores per 10 000 population and per km2) was estimated using population and territory data from Statistics Canada’s 2001 census results in order to compare their frequency rates(Reference Barbolet, Cuddeford, Jeffries, Korstad, Kurbis, Mark, Miewals and Moreland42, Reference Kaufman49, Reference Morris, Neuhauser and Campbell50). One-way ANOVA were then performed to evaluate the significance of differences in distribution between rural and urban settings. Due to sample sizes, no statistical analyses were performed at the intra-urban and intra-rural levels. General linear model (GLM) univariate procedures were employed to identify any interaction between the quantity and the diversity (dependent variables) and the type of FVS, the setting and the Material and Social Deprivation Index (as fixed factor). Dunnet T3 post hoc procedures were used when a significant association was perceived(Reference Dunnet51). Moreover, test–retest reliability was confirmed with paired-sample t tests.
The FVS spatial distribution was mapped using a specific symbol for each type of store. The F&V in-store availability (quantity and diversity) was also illustrated with four tones of colour ranging from white (nil), light grey (moderate), dark grey (good) to black (excellent). These categories were defined according to the natural clusters: no variety and no shelf-surface (nil); one to nineteen varieties and 1–99 ft2 of shelf-surface (moderate); twenty to thirty-nine varieties and exactly 100 ft2 of shelf-surface (good); and forty or more varieties and 101 ft2 or more shelf-surface (excellent). All maps were created using Mapinfo version 8·5 (Pitney Bowes MapInfo, Troy, NY, USA).
Results
Characteristics of the studied area
The area under study (2230·1 km2) is home to a population of 84 874 inhabitants. The majority lives in an urban setting (62 110 inhabitants in an area of 11·8 km2v. 22 764 inhabitants in a rural area of 2218·3 km2). There are significant differences in the socio-economic and demographic characteristics between urban and rural settings. Globally, the rural setting is more deprived materially than the urban setting, as already observed for the same region by DeKoninck et al.(Reference DeKoninck, Clément, Hamelin, Pampalon, Paquet, Trudel, Vézina and Giguère36).
Frequency and spatial distribution of fresh fruit and vegetable stores
A total of 144 FVS were visited. Among these, 74 % were located in the urban subdivision. In all stores found in the studied area, half were convenience stores (51 %) and few of them were superstores and greengrocers (11 %). Within the urban setting, one district (Vieux-Québec/Cap-Blanc) was served mostly by convenience stores and had no supermarkets, superstores or greengrocers. Also, two districts (St-Jean-Baptiste and St-Sauveur) were relatively underserved by supermarkets; otherwise they enjoyed many more grocery stores than the other districts. The number of specialty stores was similar in all districts. Within the rural setting, three communities (Les Éboulements, St-Hilarion and Rivière-à-Pierre) relied only on convenience or grocery stores and one (Ste-Christine-d’Auvergne) had no FVS at all. At least one community visited per RCM was home to a supermarket (St-Raymond, Baie St-Paul and Ste-Anne-de-Beaupré), but only one of the communities had a greengrocer (St-Raymond).
Table 2 shows the mean distribution of each type of FVS by setting. With regard to the FVS/10 000 population measure, one-way ANOVA analyses indicated no differences, except for a higher number of delicatessen and natural food stores per 10 000 population (P = 0·018 and P = 0·033, respectively) in the urban setting. This implies that city dwellers have better access to more specialty stores than rural inhabitants. Likewise, an examination of the distribution of FVS/km2 showed that with the exception of greengrocers and superstores, there were relatively more FVS in the urban setting (convenience store P = 0·000; grocery store P = 0·003; supermarket P = 0·001; delicatessen P = 0·000; natural food store P = 0·037).
Table 2 Mean distribution of fresh fruit and vegetable stores by type of setting: greater Québec City area, Canada, September 2006
*Significant difference compared with the urban setting (P < 0·05).
In-store availability
Thirty-four per cent of all food stores surveyed did not carry F&V; almost half of them (48 %) being convenience stores (data not shown). The total shelf-surface area means for F&V were much higher in supermarkets and greengrocers (>200 ft2) than in any other type of FVS (<100 ft2; Table 3). The same observation can be made regarding the total variety means, where supermarkets and greengrocers carried more than fifty varieties while other FVS had less than twenty (Table 4).
Table 3 Fresh fruit and vegetable mean total shelf-surface area (ft2) per type of fresh fruit and vegetable store (FVS) in urban and rural settings: greater Québec City area, Canada, September 2006
1 ft2≈0·1 m2.
Table 4 Fresh fruit and vegetable mean total variety (counted number of varieties) per type of fresh fruit and vegetable store (FVS) in urban and rural settings: greater Québec City area, Canada, September 2006
The results of the GLM univariate procedures are presented in Table 5. No significant association was observed in the interaction model. When looking at the separated effects of the fixed factors on the indicators of quantity and diversity, the type of FVS was the only factor that was significantly associated with shelf-surface area and number of varieties. The total shelf-surface area means and the total variety means were higher in greengrocers and supermarkets than in the other FVS types regardless of the setting or level of deprivation. This suggests that the FVS type is a good predictor of the F&V supply.
Table 5 Variables influencing fresh fruit and vegetable in-store availability: greater Québec City area, Canada, September 2006
FVS, fruit and vegetable store.
All results shown before post hoc procedure.
Finally, the paired-sample t test demonstrated no difference for shelf-surface area and number of varieties between the collected data and the data from the 10 % random sample.
Mapping
Among the FVS located in the six districts comprising the urban setting (Fig. 1), those rated excellent for their in-store availability of F&V were distributed evenly across the subdivision, with the exception of the St-Sacrement and Vieux-Québec/Cap-Blanc districts. However, people living in St-Sacrement could access FVS located just outside the study area, thus addressing part of the issue of food access. Vieux-Québec/Cap-Blanc, situated by the St-Lawrence River, clearly provides limited access to F&V. The three rural settings revealed a very different picture, which corroborates the results obtained from the one-way ANOVA. Figure 2 presents the map of the RCM of Charlevoix (observations for the other two RCM are identical, therefore no map is included). It illustrates that F&V supply in two of the three rural communities is nil or moderate (Les Éboulements and St-Hilarion), while one community’s supply is excellent (Baie St-Paul).
Fig. 1 Map of the urban setting (La Cité subdivision) showing location of fruit and vegetable store types (○, convenience store; ◊, grocery store; □, supermarket, ⋆, greengrocer; ▵, natural food store; ▿, delicatessen) and their in-store availability of fresh fruits and vegetables (␣, nil; ░, moderate; ▒, good; █, excellent): greater Québec City area, Canada, September 2006
Fig. 2 Map of one regional county municipality of the rural setting (Charlevoix) showing location of fruit and vegetable store types (○, convenience store; ◊, grocery store; □, supermarket, ⋆, greengrocer; ▵, natural food store; ▿, delicatessen) and their in-store availability of fresh fruits and vegetables (␣, nil; ░, moderate; ▒, good; █, excellent): greater Québec City area, Canada, September 2006
Discussion
Two major findings emerge from the present study. First, differences exist between rural and urban settings in the greater Québec City area in terms of FVS spatial distribution. Second, observations made of the quantity and diversity of fresh F&V in-store availability suggest that the type of food retailer is the best indicator to predict fresh F&V supply. We believe the study also yields new insights into methodology and goes beyond the mere identification of the location of FVS on a map. Indeed, by qualifying F&V variety and quantity, we have provided a better assessment of true population access to F&V.
Trends in the distribution of fresh fruit and vegetable stores
Whether living in an urban or rural setting, people seem to have access, in the same proportion, to almost all types of FVS studied (except for specialty stores) even though their spatial distribution is less scattered in the urban setting. However, the distance to FVS is considerably higher for people in rural settings. This unequal access to F&V, which could have an even larger impact on non-motorized households, can lead to health disparities(Reference Apparicio, Cloutier and Shearmur29). The present study argues that, in general, spatial access to FVS is significantly higher for those living in urban settings. Still when comparing FVS intra-urban spatial distribution, some inconsistencies appear between the districts as suggested in other Canadian studies that found disparities in food store distribution(Reference Latham and Moffat27, Reference Bertrand, Thérien and Cloutier28). Therefore, the scope of future research on the intra-urban distribution of food stores in the area should be broadened to include other relevant aspects such as transportation.
The present results show that spatial distribution patterns are complex in the intra-rural setting. Rural Québec has been witnessing a steady economic and population decline(Reference Martinez, Pampalon, Hamel and Raymond39), which is reflected in the food supply services of some rural areas. The picture seems standard in rural settings, where some communities act as central hubs providing the main services to the periphery(Reference Dugas52). For example, Baie St-Paul and St-Raymond are hubs servicing the surrounding communities like Ste-Christine-d’Auvergne. In fact, Ste-Christine-d’Auvergne residents need to drive about 30 km to and from the nearest food store (located in St-Raymond). Other communities like Neuville, outside the study area, could also provide them with services, but distances remain a problem, emphasizing the need to address disparities in access to food.
Outlying communities located in rural settings provide nothing but convenience or grocery stores. One can expect that these communities would have tried to improve their supply and better meet the needs of local residents. In Fig. 2, it is clear that in-store availability of fresh F&V is poor or even non-existent (moderate and nil). Thus, these communities appear ideal places to launch a community-based nutrition programme in partnership with food stores(Reference Connell, Yadrick, Simpson, Gossett, McGee and Bogle53). Some authors point out as well that local stores have the capacity to provide quality services to residents(Reference Clifford and David12, Reference Eisenhauer54, Reference Llewelyn55) and help support the local economy while respecting social and environmental values(Reference Burns, Gibbon, Boak, Baudinette and Dunbar32).
FVS spatial distribution may vary according to type of setting and population characteristics. For example, in the Montréal metropolitan area, peripheral neighbourhoods are less dense and more motorized and presumably have fewer but larger food stores, whereas central neighbourhoods are denser and less motorized and presumably have a greater number of smaller food stores(Reference Apparicio, Cloutier and Shearmur29). As mentioned before, the focus of the present study was only on the retail sector. Further information is thus needed to help interpret results from our study and others. Alternative or informal ways of acquiring food, such as directly from farms, public or farmers’ markets, or from home gardening, were not taken into account. It would be relevant to include the relative impact of these other sources on food provisioning. Food supply differences between the two settings may be a factor in determining food access. Therefore, in a public health context, it must be made clear from the start that urban and rural settings are two distinct food environments.
The type of fresh fruit and vegetable store: an indicator of availability
The type of FVS serves as a better indicator of the availability of fresh F&V than the setting and the level of deprivation. This finding corroborates results of studies conducted by Cummins and Macintyre in the UK(Reference Cummins and Macintyre34) and Winkler et al. in Australia(Reference Winkler, Turrell and Patterson24). This finding also aligns with the conclusion reached in Bertrand’s Montréal study, arguing that poor access to the F&V supply is not associated with population median income (a socio-economic measure of deprivation) in a given dissemination area(Reference Bertrand31). However, a study in a New York City neighbourhood reveals some disparity in in-store availability of fresh fruit and vegetables between a poor community and an affluent community living side by side(Reference Horowitz, Colson, Hebert and Lancaster56). Comparisons are difficult to make given that the indices of deprivation used in these studies are drawn on social and economic variables that differed from ours. Furthermore, as suggested by Ball et al., the FVS spatial distribution and the nature of supply chains could be country-specific, thus reflecting differences in institutional and legal frameworks and urban planning policies(Reference Ball, Timperio and Crawford4).
Our research shows that the variety of fresh F&V available in supermarkets is superior to that in smaller food stores. This finding corroborates those of Horowitz’s and Bodor’s groups(Reference Horowitz, Colson, Hebert and Lancaster56, Reference Bodor, Rose, Farley, Swalm and Scott57). It also reaches the conclusion that convenience stores do not represent a viable source of fresh F&V, even in rural settings. We suggest, moreover, that the significant relationship observed between the type of FVS and in-store availability could be explained in part by the increasingly oligopolistic food retail sector, which some refer to as the North American Agro-Food Complex(58, Reference Winson59). Since the 1980s, independent chains have been unable to maintain their market share(Reference Brouillette45, 58) and the food store types are being standardized(Reference Sone, Aguda, Malaviya and Mercer60). These changes seem to dictate how the retail sector is evolving and the availability of products. From a population health perspective, it is necessary to ensure that the changing nature of the retail sector leads to an increase in the fresh F&V supply and their access.
Methodology issues
The present exploratory study examined FVS spatial distribution by comparing the frequency of each type of store in relation to land area population density. Its focus on fresh F&V in-store availability contributed to our understanding of the food environment.
Drawing on other Canadian studies(Reference Bertrand, Thérien and Cloutier28, Reference Winson46, Reference Miedema47), the data collection method used in the current study was first pre-tested before the research process began. High consistency in the results confirms the reliability of the process and adds some certainty to the method. Shelf-surface area appears also as a relatively precise measure of what is available even though the actual height or depth of the shelf used to hold F&V is not considered. It goes further than the simple ‘yes or no’ method used by Cummins and Macintyre(Reference Cummins and Macintyre34) or the F&V total surface area in food stores measure developed by Bertrand(Reference Bertrand, Thérien and Cloutier28). Indeed, our study allows for a precise assessment of in-store availability and, in so doing, it seeks to improve our understanding of methodological issues in research on food supply.
The relatively low number of FVS sampled in the rural setting could reflect the small sample size or rather the reality of this setting’s food environment, which is characterized by less FVS. An extended study area would further enhance our understanding of FVS spatial distribution in a rural or urban setting and at the intra-rural/intra-urban levels.
Measuring all aspects of food access in one study design represents a great challenge for researchers. Aspects to be addressed include where people shop and in which context, the convenience of opening hours, delivery services, the means of accessing stores, rates of stock turnover, the time of the year, food price and discounts, selection of food products that meet consumer preferences and the social perception of food stores. Also, creating a reliable measure of food accessibility is difficult because the conceptual construct could be better operationalized. The sociocultural acceptance of the food supply sources, or the supply itself, is also another key aspect of food access(Reference Sone, Aguda, Malaviya and Mercer60), but is rarely assessed in studies.
Future research: consumer perceptions and links with health
According to our results, supermarkets and greengrocers represent the main sources of fresh F&V. Knowing that these constitute a food group for which low consumption is associated with non-transmissible diseases, the limited spatial access to supermarkets in the rural setting is a source of health concern(1). With the intention of creating supportive environments for healthy eating, the results warrant further investigation on food consumption and its determinants. Thinking about solutions to health disparities raises the importance of considering how people interact with their food environment. Both individual and environmental determinants influence food behaviours, but their relationships have yet to be clarified(Reference Raine5, Reference Cheadle, Psaty, Curry, Wagner, Diehr, Koepsell and Kristal61). In health promotion, it is essential to consider the needs of the population and to involve them when trying to solve a problem(Reference Green and Kreuter62–Reference Sloane, Diamant, Lewis, Yancey, Flynn, Nascimento, McCarthy, Guinyard and Cousineau64). According to Carey, the population should be empowered to talk with their retailers about issues that are a concern to them(Reference Carey65). Being an economic sector, food retailing implies that merchants may not change or adapt their food supply unless pressure comes from consumers or a law forces them to do so. Moreover, according to a recent study carried out in the greater Québec City area, people’s perception of problems and social cohesion varies between neighbourhoods and/or localities, and perceptions of place appear to be significant predictors of people’s health(Reference Pampalon, Hamel, DeKoninck and Disant35). This observation reinforces the significance of getting people to address their issues of food access and then adapt the solutions to their local reality. Their involvement could also help highlight distinctions between urban and rural settings and develop an alternative framework for public health planning in rural settings.
Acknowledgements
We are grateful for the financial support provided to the first author by the Agence de la santé et des services sociaux de la Capitale-Nationale and the Fondation de l’Université Laval. We thank the Centre for Research in Regional Planning and Development (CRAD, Université Laval, Québec City) for their insight and expertise. We also acknowledge the contribution of the advisory committee members who helped us carry out this research. There are no conflicts of interest in this study. N.P., the first author, developed the study design, collected and analysed the data, and wrote the paper as part of her master’s degree. A.-M.H. provided guidance and assistance all along the research and writing process.
