"Food deserts," areas characterized by relatively poor access to healthy and affordable food, may contribute to social disparities in diet and diet-related health outcomes, such as cardiovascular disease and obesity (1-3). The term "food desert" reportedly originated in Scotland in the early 1990s and was used to describe poor access to an affordable and healthy diet (4). Although the term "food desert" can mean a literal absence of retail food in a defined area, studies of food deserts more commonly assess differential accessibility to healthy and affordable food between socioeconomically advantaged and disadvantaged areas.

In the last decade, research on food deserts has become increasingly common, but this research is also a source of debate (4). Many primary studies have been conducted on food deserts, and some attempts have been made to compile the existing evidence (2,5-7). Despite the burgeoning interest in food deserts, we could find no systematic review of them. We address this gap by systematically reviewing the evidence for food deserts, specifically in socioeconomically disadvantaged areas. Food deserts in disadvantaged areas are arguably of more concern because a process of "deprivation amplification" may contribute to social disparities, whereby area-level deprivation compounds individual disadvantage (8,9). The objective of this review was to systematically and critically review the literature to determine whether access to healthy, affordable food in retail stores varies by area socioeconomic status to the disadvantage of socioeconomically deprived areas. In addition, we critique the methodologic rigor of existing evidence.

We included quantitative studies from high-income countries (10) if they compared 2 or more geographic areas that differed by socioeconomic indicator of disadvantage and, for market-basket comparisons, involved more than 1 type of food. We considered market-basket studies that compared a selection of food items across areas and stores in terms of availability, variety, price, or quality; geographic studies that compared the accessibility of different types of food stores; and studies that combined geographic and market-basket approaches. We allowed both random and nonrandom sampling methods; however, we excluded studies that compared preidentified food deserts to other areas. We also excluded studies that investigated change as a function of time or intervention and studies that did not involve food retail stores (eg, restaurants) (11,12). Eligible outcomes included average distance to nearest food stores, store density by area or population, average selling space, availability and variety of items, price of food, and food quality. Variety was defined as number of different types or brands of the same food item.

We searched the following electronic databases for all years since inception of the database through September 2007: CINAHL, EMBASE, MEDLINE, PsycINFO, Social Services Abstracts, Sociological Abstracts, and ABI INFORM/Global. We searched these databases for original research articles written in English or French. The OVID search strategy used free-text terms and subject headings (Appendix A). We searched other databases and the Journal of Business online by using relevant free-text terms and conducted a grey literature search by using Google and Google Scholar. We hand-searched reference lists of included studies and contacted key researchers in the field (Diez-Roux, Giskes, Raine, Turrell, and Zenk) to identify other relevant studies.

Two authors independently reviewed search results, reviewed retrieved studies to determine whether they should be included, and abstracted outcomes data. We independently assessed the methodologic quality of studies that met our inclusion criteria. Instead of rating or scoring each study, we rated the quality to describe how methodologically robust each included study was. Six criteria were common to all types of studies, 3 were unique to market-basket studies, and 1 was unique to geographic studies (Appendix B). When findings were reported in several articles, we used the most recent published version to rate quality and outcomes, unless different outcomes were published in different articles (13,14).

The literature search identified 2,826 studies; of these, 106 were retrieved. After detailed inspection of retrieved studies, 57 were excluded and 49 were included. Reasons for exclusion were irrelevant outcomes or comparisons (n = 21), not a study (n = 6), inappropriate method of gathering or using data (n = 2), methods not relevant (n = 16), inappropriate or no comparison area (n = 15), only 1 food item studied (n = 1), inadequate information (n = 10), and preselected food deserts (n = 3). Some studies were excluded for multiple reasons.

The 49 included studies consisted of 22 market-basket studies, 17 geographic studies, and 10 mixed-methods studies (Appendix C). Thirty-seven were from published sources, and 12 were from unpublished sources. Thirty-four were conducted in the United States, 5 in the United Kingdom, 6 in Canada, 3 in Australia, and 1 in New Zealand. They varied in scope from a comparison of 2 areas in 1 city (15) to a comparison of food stores across the United States (16). Dates ranged from 1966 to 2007; 3 studies were published in the 1960s, 2 in the 1970s, 2 in the 1980s, 12 in the 1990s, and 29 after 2000; 1 study had no date. Some geographic studies considered all types of food retail stores, whereas others only considered supermarkets.

Among market-basket studies, the number of food items in baskets varied; the largest basket included more than 80 items (17) and the smallest included 18 items (18). Some studies focused on 1 type of basket (eg, typical market basket or a healthy food basket), and others compared a healthy food basket to an unhealthy food basket.

The 17 geographic survey studies were of moderately high quality. Reporting of raw data and statistics and precision of data were problematic in some studies. The 10 mixed-methods studies were higher quality; however, nearly all failed to report interrater reliability. The quality of the 22 market-basket studies was variable and generally poorer. Interrater reliability was rarely reported; many also had problems with store selection, reporting raw data, statistics, and participation rate. In addition, many of these studies did not control for item quality and only compared outcomes within type of store and not across neighborhood.

Food deserts exist, at least in the United States. Evidence is both abundant and robust enough for us to conclude that Americans living in low-income and minority areas tend to have poor access to healthy food. However, studies on the price of food were generally of low quality, and their findings were mixed. Evidence from other countries is sparse and equivocal. The evidence that is available is much less compelling than evidence from the United States. On this basis, evidence from other countries does not warrant firm conclusions at this time on whether access to healthy, affordable food systematically varies to the disadvantage of socioeconomically deprived areas.

The environment in which we live, work, and play contributes to health and socioeconomic differences in health over and above the influence of individual characteristics (62). We found evidence for structural inequalities in the food retail environment and believe that these inequalities may contribute to inequalities in diet and diet-related outcomes. For the United States, our findings suggest a process of deprivation amplification (9), since structural problems related to food retail appear to further disadvantage low-income and minority Americans, who are already limited in their ability to purchase healthy food.

Less access to supermarkets and chain stores in low-income neighborhoods may create barriers to accessing healthy, low-cost food for those who lack access to transportation; in 2001, 26.5% of Americans with incomes below $20,000 did not own a car (63). This barrier, combined with the increased presence of small, independent stores and convenience stores in low-income neighborhoods, may limit shopping to stores that charge higher prices (24,42). These stores have a poor selection of healthy foods and a wide selection of unhealthy foods, which can contribute to poor diet. In the United States, increased access to supermarkets is associated with lower prevalence of overweight and obesity (64), improved fruit and vegetable consumption, and better diet quality among African Americans (13,65), low-income households (66), and pregnant women (67). In contrast, increased access to convenience stores is associated with increased risk of obesity (64).

In considering the influence of the environment on health, the quality of existing amenities must be taken into account (9), in addition to their availability and accessibility. This review shows that low-income areas are also disadvantaged in this regard. Availability of healthy food is associated with better diet (68); perception of better selection and quality of fresh fruit and vegetables is associated with increased fruit and vegetable consumption (65).

Evidence on neighborhood food price from the United States was inconsistent. However, studies showed higher prices among convenience stores and small, independent stores, which are more prevalent in low-income and African American communities. Furthermore, the finding of higher prices for healthier food baskets (45) is supported by a recent economic analysis (69) that demonstrated that energy-dense diets cost less than healthier diets. These observations suggest that people with limited food budgets may not be able to purchase healthy food. Indeed, people on low incomes cite the high cost of healthy foods as a barrier to eating healthily (45,70). Lower fruit and vegetable prices are associated with lower body mass index among elementary school children; effects are particularly strong for those children who are disadvantaged by poverty or race (71).