Outcome Measures

Most data collection in MESA is structured around a series of clinical exams, scheduled at approximately two year intervals. The baseline exam occurred between July 2000 and July 2002, and the most recent exam, “Exam 5”, spanned April 2010 through February 2012. The analysis presented here primarily employs data collected at Exam 5. All participants attending this exam were asked to complete a food frequency questionnaire that inquired about eating habits over the previous year and included items about organic produce consumption. Specifically, participants were asked how often the fruit and vegetables they ate were “organically grown”, defined as “[having] a ‘USDA Organic’ label, purchased locally from an ‘organic farm’, or grown without pesticides in a home garden”. Options were “Seldom or Never”, “Sometimes”, and “Often or Always”. For the primary analysis, participants who reported that they sometimes, often or always ate either organic fruit or organic vegetables were categorized as “organic consumers”, and those who reported that they seldom or never ate organic fruit and organic vegetables were categorized as non-consumers. A separate, secondary analysis restricted the definition of organic consumers to just those who reported they “often or always” consumed organic fruit and vegetables.

Individual-Level Variables

We hypothesized organic consumption to be associated with individual-level factors, including sex, age, race/ethnicity, metropolitan area, marital status, per capita income [total household income divided by number of persons living in the household], education, and employment status. With the exception of age, all variables were evaluated categorically.

Neighborhood-Level Variables

We also hypothesized a relationship between organic consumption and a set of complementary measures of the local food environment, after control for individual-level variables. Specifically, we hypothesized organic produce to more likely be consumed by individuals living in areas with more supermarkets and where there is a perception of a larger selection of produce in general. These measures were developed by the MESA Neighborhood Study, an ancillary study to MESA that characterized the local food environments of MESA participants [17]–[18]. Each measure is briefly described here.

The first was a GIS-based measure representing the density of supermarkets within 1 mile of participants' homes. The density of supermarkets was determined using data obtained from the National Establishment Time Series (NETS) database from Walls and Associates [19]. Additional supermarket data was obtained from Nielsen/TDLinx to enhance the identification of supermarkets [20]. Supermarkets were defined as grocery stores (SIC code #5411) with at least $2 million in annual sales or at least 25 employees. Participant addresses were geocoded using TeleAtlas EZ-Locate web-based geocoding software [21], and simple densities per square mile were created for 1-mile buffers around each address using the point density command in ArcGIS 9.3.

The second measure was the participants' self-report of the selection of fruits and vegetables available in their neighborhoods, defined as the area within approximately 1 mile of their home (“MESA self-reports”). At Exam 5, participants were asked the extent to which they agreed with the statement “A large selection of fresh fruits and vegetables is available in my neighborhood”, and responses were coded on a five-point Likert scale (strongly agree; agree; neutral; disagree; and strongly disagree).

The third measure, the Aggregated Neighborhood Survey (ANS), was constructed by aggregating responses of multiple respondents residing in each participant's census tract (as a proxy for neighborhood). Survey respondents used in the calculation of the ANS included other MESA participants living within a given census tract as well as other residents in those census tracts who were included to increase the sample size in areas with few MESA respondents [22]. This supplementary survey was conducted on a random sample of residents in selected tracts, identified through address-based sampling methods. Availability of healthy food was ascertained based on responses to three survey items: “A large selection of fresh fruit and vegetables is available in my neighborhood”, “A large selection of low-fat food is available in my neighborhood”, and “The fresh fruits and vegetables in my neighborhood are of high quality”, with responses coded on five-point Likert scales. Conditional empirical Bayes estimates, which borrow information across all tracts in order to increase reliability, were derived from three level hierarchical linear models to account for the nested structure of the data [22].

Statistical Analyses

For both individual and neighborhood analyses, we first conducted bivariate comparisons exploring the relationship between each variable and organic consumption, using either chi-squared tests or log-binomial regression, as appropriate. We then included the full set of individual-level variables in a log-binomial regression to model the association with organic consumption. Individual-level variables found to be statistically significant in the full model were included in the analyses of organic consumption and the local food environment. Log-binomial models were used in the primary analysis due to the relatively high prevalence of sometimes, often or always consuming organic produce (40%, n = 1,644). In the secondary analysis, where the outcome was the smaller set of individuals who reported they “often or always” consumed organic food (5%, n = 204), we employed logistic regression models.

The relationship between each measure of the local food environment and organic consumption was evaluated separately with and without control for individual-level variables. For each of these measures, we also examined the impact of including a random intercept for census tract. In sensitivity analyses, we examined the effect of stratification by education and income category in the individual-level analyses. All analyses were conducted in SAS v9.3 [Cary, NC].

Organic Produce Consumption and Individual-Level Variables

Table 1 shows descriptive individual-level statistics by reported organic consumption habits. In bivariate analyses, organic consumption was significantly more common among women, younger individuals, and those currently employed. Metropolitan area was also significantly associated with organic consumption, as were higher per capita household income and education.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographic and socioeconomic characteristics of the Multi-Ethnic Study of Atherosclerosis cohort at Exam 5 (2010–2012), by organic produce consumption habits.

https://doi.org/10.1371/journal.pone.0069778.t001

Table 2 shows the associations between individual-level variables and organic consumption in a multivariate log-binomial regression model including all variables with statistically significant bivariate associations. Race/ethnicity was also included because of the importance of this variable in this cohort and to diet in general. After accounting for other individual-level factors, women were more likely to be organic consumers than men (prevalence ratio [PR]: 1.21, confidence interval [CI]: 1.12–1.30, p<0.0001). Chinese participants were less likely than other participants to be organic consumers, though this difference was not large and overall, race/ethnicity did not show a statistically significant effect. Age was highly associated with organic consumption; for every +10-year increment in age, there was a 13% reduction in the likelihood of being an organic consumer.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Prevalence ratios and 95% confidence intervals for the association between organic food consumption and individual-level demographic and socioeconomic characteristics in adjusted models.

https://doi.org/10.1371/journal.pone.0069778.t002

Metropolitan area was also significantly associated with organic consumption in this cohort: participants living in more the populated cities (Chicago, LA and New York) were more likely to be organic consumers compared to those living in Winston-Salem, Baltimore and St. Paul. Education was found to be an important predictor: comparing the highest and lowest education categories (less than high school compared to graduate school) resulted in a 68% greater likelihood of organic consumption.

Being in the highest income category compared to the lowest (per capital household income of <$14,999 versus >$45,000) was associated with a 10% greater likelihood of organic consumption, but the overall relationship between organic food consumption and per capita household income was not statistically significant (p = 0.06), and higher income was not always associated with greater consumption. For example, participants with per capita household income between $15,000 and $25,000 were less likely to consume organic produce than those in the <$15,000 category. Employment status was not associated with organic consumption in multivariable analyses. Results were not sensitive to stratification of the sample; prevalence ratio point estimates were similar when restricted to just those participants in the lower and higher education and income brackets (data not shown).

Organic Produce Consumption and the Local Food Environment

Of those participants included in the individual-level analyses, 84% (n = 3,428) consented to MESA Neighborhood and had complete data for the neighborhood-level analyses (see Figure 1). The distribution of demographic and socioeconomic characteristics between this group and those shown in Table 1 is nearly identical. Table 3 shows the frequency of organic consumption among these participants by each measure of the local food environment. In bivariate analyses, whether measured by self-report, supermarket density, or ANS score, participants for whom accessibility was greater were more likely to be organic consumers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Frequency of organic food consumption in relationship to measures of the local food environment.

https://doi.org/10.1371/journal.pone.0069778.t003

This association remained in fully adjusted models as well (Figure 2). After adjusting for individual-level variables, self-reported produce availability within a participant's neighborhood was positively associated with organic consumption; each unit increase on the Likert scale, was associated with a 7% greater likelihood of eating organic food (PR: 1.07, CI: 1.02–1.11, p = 0.002). The ANS score analysis also suggested an effect of local food environment on organic consumption; the likelihood of being an organic consumer was 8% higher per interquartile change in score (0.5 units) (PR: 1.08, CI: 1.00–1.17, p = 0.05). Inclusion of a random intercept for each census tract did not substantially modify estimates or standard errors in any of the three models. The GIS-based supermarket density measure was not significantly associated with organic consumption after control for individual-level variables, though the direction of the effect was unchanged (PR: 1.02, CI: 0.99–1.05, p = 0.16). All individual factors associated with organic consumption remained significant with the inclusion of the measures of the local food environment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Associations of organic food consumption with neighborhood food accessibility.

Food accessibility is estimated by a) density of supermarkets (per increase in one supermarket per mile); b) self-report of fruit and vegetable selection in a participant's neighborhood (per one point increase on the Likert scale); and c) Aggregated Neighborhood Survey (per interquartile difference, represented by a 0.5 increase on the Likert scale). Models are adjusted for sex, age, education, income, metropolitan area, and race/ethnicity.

https://doi.org/10.1371/journal.pone.0069778.g002

Frequent Organic Consumers

While the primary analysis aimed to understand the factors associated with the decision to consume organic produce at least occasionally, this secondary analysis explored the question of whether individual and local food environment factors were associated with more frequent organic produce consumption. Here, the definition of organic consumers was restricted to those who reported that the fruit and vegetables they ate were “often or always” organic. In general, the relationships between organic produce consumption and individual-level factors were similar to those reported in the primary analyses. In fully adjusted models of individual factors, female gender, younger age, more urban metropolitan areas, and higher levels of education were all significantly associated with “often or always” consuming organic produce, as was the case in the primary analysis. Race/ethnicity and marital status were not significantly associated with organic produce consumption. Per capita household income was significantly associated with “often or always” consuming organic produce (p = 0.04), but the relationship was not linear. Instead, the lowest and highest income groups were more likely to report that they “often or always” consumed organic produce, and the middle income groups were significantly less likely to be frequent organic produce consumers.

However, in contrast to the results of the primary analysis, the local food environment was not associated with the decision to “often or always” consume organic produce. Though density of supermarkets within 1 mile of the residence remained strongly associated with organic produce consumption in bivariate analyses (p = 0.0002), aggregated neighborhood survey and self-report of accessibility were no longer significantly associated (p = 0.09 and p = 0.17, respectively). Further, once individual-level variables were accounted for in a fully adjusted model, no significant relationship was found between “often or always” consuming organic produce and any of the measures of the local food environment (supermarket density: OR = 1.05, CI = 0.95–1.17; self-report: OR = 1.00, CI = 0.85–1.17; ANS Score: OR = 1.01, CI = 0.52–1.98).

Individual-Level Findings

Organic food consumption is increasing; consistent with our findings, several studies over the past decade have reported that 40 to 50% of individuals and households purchase organic food at least occasionally [23]–[27]. However, the specific factors associated with organic food consumption have not been well understood, as early studies painted contradictory pictures of the socioeconomic status and demographics of organic food consumers.

Organic food consumption has been found to be associated, variously, with higher education [28]–[29], lower education [30], or not associated with education at all [31]. Results were also mixed for the relationship between income and organic food; some studies observed consumers with high incomes to have less tolerance for food with blemishes and to be less likely to purchase organic food [29], while others found people with higher income to be more likely to make organic purchases [28], [30], and others found no association [31]–[32]. Findings with respect to age and ethnicity were also inconsistent; in fact, the only demographic attribute to be reliably associated was gender, with women purchasing more organic food than men [28], [30]–[31]. However, all of these studies employed convenience samples, and typically included people who were already shopping at either food cooperatives or at expensive specialty grocers, missing substantial segments of the general population. More recent research has capitalized on Nielsen Consumer Panel studies, in which thousands of American households are provided handheld scanners to scan each item they purchase [24], [27], [33]. These studies have consistently found higher income and education to be associated with purchasing organic food, but age and ethnicity have continued to show inconsistent effects.

Our results are consistent with previous research showing that women purchase organic food more frequently than men, and with the Nielsen Consumer Panel studies' observation that higher education is associated with more organic food purchasing. In MESA, older participants were less likely than younger participants to consume organic food. However, since all participants in this study were aged 45 and older, this result could also be consistent with a “U-shaped” relationship between age and organic consumption, in which middle aged people are more likely to consume organic food than both younger and older individuals, as other researchers have suggested [34].

The relationship between per capita household income and organic produce consumption was sensitive to adjustment for other individual level variables, and to the categorization of organic consumers. In bivariate analyses, self-report of either “sometimes” or “often or always” consuming organic produce was strongly associated with income category, but the strength of this relationship was attenuated by adjustment for other individual level factors. When restricted to individuals who “often or always” consumed organic produce, the relationship with income was decidedly non-linear; individuals with the lowest and highest per capita household incomes were more likely to report frequent consumption of organic produce.

To our knowledge, this is the first study to examine geographical differences in organic food consumption; we found participants in more populated cities (Los Angeles, Chicago, and New York) to consume more organic food than those in less densely populated regions (St. Paul, Winston-Salem and Baltimore). This difference may also be related to the local food environment and food access; more research is needed to fully understand the relationship between organic food consumption and urbanicity.

Local Food Environment Findings

Over the past decade, the US obesity epidemic – and in particular, disparities in obesity prevalence – has led the public health community to think much more broadly about factors influencing diet. No longer are dietary motivations understood only in the framework of individual lifestyle choices. Instead, the food environment has been increasingly recognized as important to diet [15], [18], [35]–[39], and the results of this study are consistent with the idea that this environment influences a variety of dietary choices.

While this is the first study to explicitly investigate the relationship between the local food environment and organic consumption, it is not the first study to look at factors beyond demographics and socioeconomics on this dietary choice. Zepeda and colleagues explored the motivations behind organic food consumption in a national survey of nearly 1,000 US adults [23], [40]. When variables related to food beliefs and shopping habits were considered, organic food consumption was not found to be associated with direct economic variables, such as household income or weekly food expenditures. Instead, the important factors in choosing organic food included both personal beliefs and opportunity, where opportunity was defined as shopping at food venues where organic food was more likely to be available. This represents a different approach to exploring the influence of food accessibility on organic consumption, but the results are consistent with our finding that access may play an important role in the decision to at least occasionally consume organic foods. The results from our secondary analysis support the notion that personal beliefs may matter, perhaps particularly for those who often or always consume organic food. This relatively small group of people may be willing to go out of their way to make this dietary choice, even if produce and supermarkets are not readily available in their neighborhoods.

Limitations

A primary limitation of this study was the lack of a direct measure of organic food availability. Instead, we employed supermarket density and both self-report and community perception of availability of produce and healthy food as proxies for organic food availability. The USDA's 2009 report, “Marketing US Organic Foods: Recent Trends from Farms to Consumers” shows that sales of organic food from conventional supermarkets and groceries now account for 46% of the total organic market share, with natural-products retailers and direct markets each accounting for another 44% and 10%, respectively [41]. This report also states that organic food is now available in more than 80 percent of all supermarkets. Given this high proportion of supermarkets in which organic food is available, we believe that it is reasonable to assume that areas with more supermarkets are also more likely to provide greater access to organic food. In addition, organic produce is more likely to be available in areas with a greater selection of produce in general. However, further research including more specific measures of organic food availability is warranted.

Recent literature has shown that subjective and objective measures of the local food environment do not always agree [42]–[43], and so we investigated three complementary measures of the local food environment, each with strengths and weaknesses. Supermarket densities are the most objective of the three but rely on the assumption that supermarkets offer organic foods. Further, the use of supermarket density within a straight-line distance neglects actual travel patterns along road networks and further assumes that people reliably shop at supermarkets near their homes. Recent research by Drewnowski et al. suggests that this assumption may not accurately reflect actual shopping patterns [44]. Self-reports reflect each individual's perceptions but their interpretation is affected by the possibility of same source bias, which may arise when using self-reports of the food environment if a person's behavior affects their reported perceptions of access to healthy foods. The strength of aggregate survey measure is that averages of multiple respondents are likely to eliminate the influence of individual subjectivities and eliminate the possibility of same–source bias. However, it may not accurately reflect access for a particular participant. The consistency of associations across the three types of measures increases confidence that the local food environment plays a role in organic consumption habits.

We chose to focus on organic produce rather than other food types. This was intended to be consistent with previous studies evaluating the relationship between consumption of organic food and organophosphate pesticide exposure [4]. Over 33 million pounds of organophosphates are applied annually in the US – more than any other class of insecticides – and their metabolites are found in the urine of 94% of the US population [45]. These compounds are registered for use on a wide variety of fruits and vegetables, but are not widely used in the production of meat or dairy [46]. We do not expect that this decision had a large impact on our results, as recent USDA research shows that US retail sales of organic fruits and vegetables are larger than all other organic food categories combined [41]. However, it is worth noting that there are several other categories of organic food not considered here.