Goryakin, Y; Rocco, L; Suhrcke, M; Roberts, B; McKee, M (2015)Fruit and vegetable consumption in the former Soviet Union: the roleof individual- and community-level factors. Public health nutrition,18 (15). pp. 2825-35. ISSN 1368-9800 DOI: 10.1017/S1368980015000105

Downloaded from: http://researchonline.lshtm.ac.uk/2115596/

DOI: 10.1017/S1368980015000105

Usage Guidelines

Please refer to usage guidelines at http://researchonline.lshtm.ac.uk/policies.html or alterna-tively contact researchonline@lshtm.ac.uk.

Available under license: http://creativecommons.org/licenses/by/2.5/

Fruit and vegetable consumption in the former Soviet Union:the role of individual- and community-level factors

Yevgeniy Goryakin1,2,*, Lorenzo Rocco3, Marc Suhrcke2,4, Bayard Roberts5 andMartin McKee51Norwich Medical School, University of East Anglia, Norwich NR4 7TJ, UK: 2UKCRC Centre for Diet and ActivityResearch (CEDAR), Cambridge, UK: 3Department of Economics, University of Padua, Padua, Italy: 4Centre for HealthEconomics, University of York, York, UK: 5European Centre on Health of Societies in Transition, Department of HealthServices Research and Policy, London School of Hygiene and Tropical Medicine, London, UK

Submitted 19 August 2014: Final revision received 28 November 2014: Accepted 7 January 2015: First published online 17 February 2015

AbstractObjective: To explain patterns of fruit and vegetable consumption in nine formerSoviet Union countries by exploring the influence of a range of individual- andcommunity-level determinants.Design: Cross-sectional nationally representative surveys and area profiles wereundertaken in 2010 in nine countries of the former Soviet Union as part of theHealth in Times of Transition (HITT) study. Individual- and area-level determinantswere analysed, taking into account potential confounding at the individual andarea level.Setting: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova,Russia and Ukraine.Subjects: Adult survey respondents (n 17 998) aged 18–95 years.Results: Being male, increasing age, lack of education and lack of financialresources were associated with lower probability of consuming adequate amountsof fruit or vegetables. Daily fruit or vegetable consumption was positivelycorrelated with the number of shops selling fruit and vegetables (for women) andwith the number of convenience stores (for men). Billboard advertising of snacksand sweet drinks was negatively related to daily fruit or vegetable consumption,although the reverse was true for billboards advertising soft drinks. Men livingnear a fast-food outlet had a lower probability of fruit or vegetable consumption,while the opposite was true for the number of local food restaurants.Conclusions: Overall fruit and vegetable consumption in the former Soviet Unionis inadequate, particularly among lower socio-economic groups. Both individual-and community-level factors play a role in explaining inadequate nutrition andthus provide potential entry points for policy interventions, while the nuancedinfluence of community factors informs the agenda for future research.

KeywordsNutrition

Fruit and vegetable consumptionSocio-economic determinants

The publication of the 2010 Global Burden of DiseaseStudy reinforced the importance of adequate fruit andvegetable consumption(1), primarily via its impact on car-diovascular health(2) and some cancers(3,4). The WHO andthe FAO recommend a minimum fruit and vegetableconsumption of at least 400 g/d for adults(5).

The volume of research on determinants of fruit andvegetable consumption in high-income countries(6–8) isnot matched by its scarcity in the countries of the formerSoviet Union (FSU), even though global agricultural tradedata suggest that consumption there is especially low(4).One survey found fruit and vegetable consumption to beinadequate (defined as eating less than 400 g/d, or five

servings of 80 g/d) among 80 % of people in Russia, 92 %in Kazakhstan and 55 % in Ukraine(9). Another study foundthat 93 % of men living in Russian Karelia consumedinadequate vitamin C, compared with only 2 % in neigh-bouring Finnish Karelia(10), subsequently linked to lowfruit consumption(11).

Although studies of environmental determinants of diet-ary consumption have increased globally, the existing evi-dence remains insufficient to draw robust conclusions(12)

and what does exist is limited in scope. Brug(13) has con-trasted the relative lack of evidence on macro-levelenvironmental determinants of nutrition with that onmicro-level determinants, with most research worldwide

Public Health Nutrition: 18(15), 2825–2835 doi:10.1017/S1368980015000105

*Corresponding author: Email y.goryakin@uea.ac.uk

© The Authors 2015. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/3.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the originalwork is properly cited.

having focused on biological, psychological, behaviouraland social factors acting at the individual level(13).However, there is growing interest in the role of environ-mental determinants(14,15) as the explanatory power ofindividual factors alone has proved limited(16).

By assessing both individual- (e.g. age, gender, maritaland socio-economic status) and community-level (e.g.advertising for high-energy food and drinks, availability ofshops selling fruit and vegetables, ease of access to fast-food outlets) drivers of fruit and vegetable consumption innine FSU countries, the present study contributes to theglobal body of research on determinants of diet andobesity. Our aims are: (i) to present new estimates of theprevalence of fruit and vegetable consumption in nine FSUcountries; and (ii) to identify relevant individual- andcommunity-level determinants.

Methods

Study designData are from household surveys in nine countries of theFSU as part of the Health in Times of Transition (HITT)study(17). This used the same standardized questionnaire ineach country to capture a range of health outcomes, healthbehaviours and demographic, socio-economic and environ-mental characteristics. Surveys were nationally representa-tive and conducted among adult respondents (aged ≥18years) in Armenia, Azerbaijan, Belarus, Georgia, Kazakh-stan, Kyrgyzstan, Moldova, Russia and Ukraine.

Multistage random sampling with stratification by regionand rural/urban settlement type was applied. Sample sizefor the urban and rural population was determined pro-portionally to these populations in each study country.Primary sampling units were selected randomly using theprobability-proportional-to-size technique from routinedata. Within each primary sampling unit (about 100–200per country, except Russia and Ukraine with 329 and 435primary sampling units, respectively) households wereselected by random route procedures. Within each selec-ted household one person was chosen (based on nearestbirthday). If after three visits (on different days and times)there was no one at home, the next household on theroute was selected.

The surveys were conducted between March and May2010, except in Kyrgyzstan where there was a delay untilMarch to May 2011 due to political violence. Face-to-faceinterviews were conducted by trained fieldworkers inrespondents’ homes. Response rates varied from 47·3 % inKazakhstan to 82·9 % in Georgia. Each country had 1800respondents, except Russia (n 3000) and Ukraine (n 2200)to reflect their larger and more regionally diverse popu-lations, and Georgia (n 2200) where a booster survey of400 additional interviews was undertaken in November2010 to ensure a more representative sample. The finalsample used in the individual-level regression analysis was

slightly smaller due to a small number of missing obser-vations. The sample that was used in the community-levelanalysis was considerably smaller due to the fact thatonly a sub-sample of communities was selected for datacollection. However, since the communities were randomlydrawn from the larger number of sampling units used in themain HITT household survey, there is unlikely to be anybias introduced by this drop in the sample size (as theindividual observations are also missing at random).

The draft questionnaire was forward- and backward-translated into each of the languages in which it wasadministered, and then piloted with approximately fifteenpeople in each country. Except in Russia and Belarus(where all interviews were conducted in Russian),respondents were given the choice of answering inRussian or a national language.

The research was approved by the ethics committee ofthe London School of Hygiene and Tropical Medicine andwas conducted in accordance with the ethical standardslaid down in the 1964 Declaration of Helsinki. All personsgave informed consent. Quality control procedures inclu-ded re-interviews to assess the work of both the inter-viewers and the interviewers’ supervisors.

VariablesOur main dependent variable was daily or almost dailyconsumption of fruit or vegetables (excluding potatoes). Thisranges from 1, referring to daily or almost daily consumption,to 4, referring to consuming fruit or vegetables less than onceweekly (with information only on the frequency ofconsumption being available, and not on the quantity).

Our independent variables included indicators forpeople who believe that good diet is unimportant, age,being female, having primary or secondary levels ofeducation as the highest attainment, reporting good eco-nomic status, number of people in the household, beingmarried, asset classes and living in the rural area. Moredetails are given in the online supplementary material,Annex 1 and Annex 2.

Additional community-level variables were recorded ina sub-sample of 333 primary sampling units randomlyselected from those in the main household surveys. Thecommunity-level variables were measured using a stan-dardized Community Observation Form, based upon thevalidated Prospective Urban and Rural EpidemiologyStudy’s Environmental Profile of a Community’s Health(EPOCH) instrument(18). Two trained data collectors percommunity systematically recorded aspects of the environ-ment relating to general social/economic situation (e.g.community-specific architecture such as conditions ofhomes and roads), nutrition and physical activity (e.g.walkability and food environment) and tobacco andalcohol (e.g. availability and advertising). Thirty commu-nity profiles were conducted in each country, exceptRussia (seventy-three profiles) and Ukraine (fifty profiles)to reflect their larger and more regionally diverse

2826 Y Goryakin et al.

populations. Additional information on the communityprofile instrument is available elsewhere(19).

We used data on the number of outdoor advertisements(e.g. billboards, adverts on shop windows, bus sheltersand other easily accessible locations) for fast foods,snacks, fizzy carbonated drinks, as well as for sweet drinks(including juices). These were collected independently ofthe interviews.

Community-level data also included the number ofshops/other outlets selling sweets, biscuits and crisps, aswell as fruit and vegetables, with kiosks being included inthis category. Incidentally, in the FSU kiosks rarely appearto sell fruit and vegetables and instead appear primarily asoutlets for alcohol and tobacco(20). Finally, informationwas obtained on whether people lived within an ‘easywalk’ to fast-food outlets within a community, as well ason the number of restaurants and cafés in the communityselling local food. Cafeterias were counted separately fromfast-food restaurants.

All variables used in the analysis were based on closed-ended questions, except the information on the number ofadvertisements/outlets, collected by two observers percommunity.

Statistical analysisWe examined the statistical association between fruit andvegetable consumption and potential individual- andcommunity-level determinants. Analyses were conductedusing the statistical software package Stata 11. We beganthe analysis with ordinary least squares (OLS) regression,where the outcome variable was daily or almost dailyconsumption of either fruit or vegetables, and performedthe regression of the outcome variable v. individual-levelcovariates and country dummies. We then controlled forpotential confounding with community fixed effects (CFE).

Next, as our interest shifted to estimating the associationbetween several community-level variables and fruit orvegetable consumption, we performed the regression of thisoutcome v. the set of community-level covariates of interest,as described below. To reduce the potential for confounding,we controlled for several potentially relevant community-level variables, as well as for regional fixed effects.

Finally, to take advantage of the ordered nature of theunderlying variables used to define consumption, orderedprobit results were estimated, separately for fruit andvegetable consumption. As the underlying outcomevariable in this specification ranges from 1 to 4, with thevalue of 1 referring to eating fruit or vegetables less thanonce weekly and 4 to eating daily or almost daily, positiveordered probit estimates indicate a greater probability ofeating fruit or vegetables and, conversely, negative esti-mates are associated with factors that reduce the like-lihood of consuming fruit and vegetables. Although, inprinciple, one may apply multinomial logit or probitmodels to estimate effect of covariates on these outcomes,ordered probit specification takes advantage of the natural

ordering of the data, also allowing a more parsimoniouspresentation of results(21).

The initial specification was as follows:

Yisc ¼ α0 + α1Zisc + ηc + eisc; (1)

where Y is the dummy variable for individual i living incommunity s located in country c, with the value of 1assigned to individuals reporting daily or almost dailyconsumption of either fruit or vegetables. In equation (1),the main interest is in the parameters contained in a vectorα1 obtained from regression of the outcome variable Yisc v.the vector of individual-level determinants Zisc and coun-try effects ηc, as described in the ‘Variables’ section.

To control for additional area-level confounders thataffect both the covariates of interest included in Zisc andthe outcome variable, a richer specification was con-sidered that replaced country with community fixedeffects. For example, community-level infrastructure andemployment opportunities may be a determinant of bothfruit and vegetable consumption, as well as of reportinggood health and of good economic status. Also, to controlfor any correlation of the error term eisc among individualsbelonging to the same community, we clustered standarderrors on the community level.

Next, the association of community-level determinantswith the same outcome of interest was estimated accord-ing to the following specification:

Yisc ¼ α0 + α1Xsc + α2Zisc + α3Ssc + μrc + eisc: (2)

The parameters contained in vector α1 are associatedwith a vector of community determinants (also used assimultaneous controls), Xsc ¼ X1

sc;X2sc;X

3sc

� �, that included

three sets of community determinants:

1. X1sc included variables measuring exposure to different

types of advertising for high-energy foods and drinks. Asan ad hoc hypothesis, it is expected that greaterexposure to these advertisements will negatively affectthe probability of daily fruit and vegetable consumption.

2. With X2sc, the focus was on availability of healthy and

unhealthy foods in stores. A priori, one expects fruitand vegetable outlets to increase availability of thoseproducts, as well as to positively affect preferences fortheir greater consumption, while the reverse will betrue for stores selling sweets and crisps.

3. Finally, in X3sc, the focus was on outside eating

establishments, such as ease of access to fast-foodoutlets and general service restaurants. Our a prioriexpectation is that easier access to fast-food outlets willbe associated with worse dietary attitudes and lower fruitand vegetable consumption; at the same time it is notclear what association to expect between our outcomesof interest and the number of local restaurants.

The main problem with estimating equation (2) is poten-tial area-level confounding. For example, some previousresearch found fruit and vegetable consumption to be

Fruit and vegetable consumption in the former Soviet Union 2827

positively correlated with neighbourhood average income(7),but wealthier neighbourhoods may also have better accessto supermarkets and a wider variety of foods(22). Takentogether, this evidence suggests that area-level socio-economic status may drive the observed associationbetween dietary outcomes of interest and environmentaldeterminants, by affecting both simultaneously.

To control for potential area-level socio-economic con-founders, we included a vector of neighbourhood controlvariables Ssc in equation (2), such as dummy variables for:living in the capital city; living in communities where garbageis collected by authorities from all homes; living in com-munities where all homes have a cold water supply; living incommunities where all homes have a central heating system;and living in communities where there are no derelict homespresent. In addition, regional fixed effects μrc were includedin equation (2) which account for potential confounders thatvary at that geographic level. Differently from equation (1),community fixed effects cannot be included as they will beperfectly collinear with the vector Xsc.

Finally, a vector of individual-level determinants Ziscaccounted for the remaining variation at the individual level.

Results

Descriptive statisticsTable 1 presents the main descriptive statistics. In all coun-tries, the proportion of people consuming fruit or vegetables

daily or several times weekly exceeded the proportion ofpeople eating them once weekly or less than once weekly.This was also confirmed in formal tests (results availableupon request), as the P value is in all cases less than 0·001.

However, in only one country (Azerbaijan) did morethan half of the surveyed population eat fruit or vegetablesdaily/almost daily; in four other countries this proportionwas about one-third. Table 1 shows that the gender dif-ference was relatively small except in Belarus (34·8 % forwomen v. 28·8 % for men) and Russia (45·1 % for womenv. 39·1 % for men). Weighted average values for commu-nity variables used in the analysis are also presented inTable 1 (the weights are numbers of respondents living inrespective communities).

Individual determinantsOur main ordinary least squares regression results arepresented in Table 2, for the whole sample and separatelyfor men and women (columns 1–3). Each year of agereduced the probability of daily/almost daily fruit orvegetable consumption by about 0·1 %.

It has already been shown in Table 1 that women ten-ded to consume more fruit and vegetables in all countries,a finding confirmed in the multivariate analysis, withwomen having about 4 % greater probability of eating fruitor vegetables daily compared with men. Education waspositively correlated with daily consumption of fruit orvegetables, with people with tertiary education being5·4 % more likely to eat them daily, compared with those

Table 1 Selected descriptive statistics, by country

Armenia Azerbaijan Belarus Georgia Kazakhstan Kyrgyzstan Moldova Russia Ukraine

Response rate (%) 60·1 56·8 48·1 82·9 47·3 78·4 74·8 59·2 60·1Fruit or vegetable consumptionMen and womenDaily (%) 49·8 55·5 32·2 32·6 47·4 36·9 30·4 42·7 48·9Several times weekly (%) 36·2 29·5 42·9 41·7 30·8 33·4 34·2 42·3 34·3Once weekly (%) 12·1 12·8 15·4 15·9 13·7 16·6 16·8 12·5 11·2Less than once weekly (%) 1·9 2·2 9·6 9·8 8·1 13·0 18·6 2·6 5·6

Women onlyDaily (%) 50·4 55·8 34·8 34·0 49·5 37·6 30·7 45·1 49·7Several times weekly (%) 35·1 30·0 43·1 40·4 29·3 31·6 34·2 42·2 33·7Once weekly (%) 12·3 11·7 13·7 15·2 13·4 17·4 16·9 10·6 10·7Less than once weekly (%) 2·2 2·4 8·4 10·4 7·9 13·3 18·2 2·1 6·0

Men onlyDaily (%) 49·1 55·0 28·8 30·1 45·0 36·2 30·0 39·1 47·8Several times weekly (%) 37·5 29·0 42·6 44·1 32·4 35·4 34·2 42·3 35·2Once weekly (%) 11·8 14·2 17·4 17·0 14·2 15·7 16·7 15·3 11·8Less than once weekly (%) 1·6 1·8 11·1 8·9 8·4 12·6 19·1 3·3 5·2

Community-level determinantsNo. of fast-food adverts 3·0 3·5 0·0 0·5 0·1 0·5 3·5 0·6 1·3No. of snack adverts 2·4 4·5 1·1 0·5 0·5 0·3 3·0 2·0 1·8No. of soft drink adverts 2·4 2·3 2·3 2·0 0·7 1·1 3·7 2·0 4·5No. of sweet drink/juice adverts 4·0 2·0 3·3 2·5 1·8 2·3 9·1 3·3 4·5No. of shops selling crisps and sweets 6·5 6·5 8·8 6·9 6·7 6·7 9·7 7·9 8·2No. of shops selling fruit and vegetables 4·9 5·0 8·2 3·4 4·8 4·1 6·4 6·1 5·0No. of local restaurants 3·4 2·4 2·7 2·2 2·6 0·8 4·9 1·9 2·8

Source: Health in Times of Transition (HITT) data set, 2010.In all columns, mean values are presented.Summary of individual-level data represents average proportion of people eating fruit or vegetables daily, several times weekly, once weekly and less than once weekly.Community-level data represent mean values per community, weighted by community size. Each community represents a separate primary sampling unit, equivalent toa ‘rayon’ or a small administrative region.

2828 Y Goryakin et al.

with secondary education only. Reporting a good financialsituation (even controlling for wealth) was associated withabout 8 % greater probability of eating fruit or vegetablesdaily. Similarly, people whose combined wealth placedthem in the top 25% of the asset score in their countrieswere about 11% more likely to report daily fruit or vege-table consumption, compared with the bottom 25%. Livingin the capital was associated with about 5·6% higherprobability of reporting daily fruit or vegetable consumption.Being married was significantly positively related to theprobability of reporting daily fruit or vegetable consumption.Finally, the perception that diet is not important to goodhealth was unrelated to fruit or vegetable consumption.

Being older reduced the probability of eating fruit orvegetables daily by women, but not by men (Table 2,columns 2–3). On the other hand, the role of economicsituation appeared stronger among men than amongwomen. Being married was related to greater likelihood ofeating fruit or vegetables daily, but only by men. Otherassociations were similar for men and women.

As a robustness check, community fixed effects wereincluded in equation (1), as their inclusion should controlfor local heterogeneity more precisely. The estimates,reported in Table 2, columns 4–6, show that there waslittle difference compared with the baseline estimates(columns 1–3), although parameters tended to be some-what smaller (but still significant).

Community determinantsTable 3 presents the main community-level parameters. Thenumber of snack and sweet drink advertisementsin the community was significantly negatively relatedto the probability of eating fruit or vegetables daily (Table 3,column 1). An additional advertisement for snacks reducedthis outcome by about 3 %, while another advertisement forsweet drinks reduced it by about 1·6%. This association wassignificant (and of much larger in size) only for women,compared with men (compare columns 2 and 3). Livingwithin an easy walk to a fast-food outlet was associated witha very large (16·1%) reduction in the probability of eatingfruit or vegetables daily for men (although the associationwas insignificant for women). Similarly, more shops sellingfruit and vegetables was positively correlated with theprobability of eating fruit or vegetables daily, although onlysignificantly so for women. Furthermore, more restaurants inthe community was positively correlated with daily con-sumption of either fruit or vegetables (although only sig-nificantly so for women). Finally, a greater number of softdrink adverts, as well as a greater number of shops sellingcrisps and sweets, was positively related to daily fruit orvegetable consumption (in the former case, the effect wassignificant among women; in the latter, among men).

Additional checksTable 4 presents the ordered probit results. In columns 1and 2, the focus is on individual-level determinants; while in Ta

ble

2Individu

al-le

veld

eterminan

tsof

daily/alm

ostda

ilyfruitor

vege

tableco

nsum

ptionin

nine

coun

triesof

theform

erSov

ietUnion

Colum

n1

Colum

n2

Colum

n3

Colum

n4

Colum

n5

Colum

n6

All

Wom

enMen

All

Wom

enMen

Coe

fficien

t†SE

Coe

fficien

t†SE

Coe

fficien

t†SE

Coe

fficien

t‡SE

Coe

fficien

t‡SE

Coe

fficien

t‡SE

Goo

ddiet

notim

portan

t0·03

20·04

50·02

50·06

30·04

10·05

7−0·02

40·03

80·04

30·05

4−0·07

70·05

5Age

−0·00

1***

0·00

0−0·00

1***

0·00

00·00

00·00

0−0·00

1***

0·00

0−0·00

2***

0·00

0−0·00

1**

0·00

0Fem

ale

0·04

1***

0·00

8–

––

–0·03

9***

0·00

7–

––

–Prim

aryed

ucation

−0·07

2***

0·01

5−0·07

3***

0·02

0−0·06

6***

0·02

0−0·06

3***

0·01

3−0·07

8***

0·01

8−0·03

7*0·02

1Sec

onda

ryed

ucation

−0·05

4***

0·01

0−0·05

3***

0·01

3−0·05

0***

0·01

4−0·03

5***

0·00

9−0·05

1***

0·01

2−0·01

70·01

4Goo

dec

onom

icsituation

0·08

1***

0·01

20·06

5***

0·01

50·09

8***

0·01

60·07

9***

0·01

00·07

3***

0·01

40·08

0***

0·01

5Hou

seho

ldsize

0·00

00·00

3−0·00

20·00

30·00

20·00

4−0·00

20·00

2−0·00

5*0·00

30 ·00

10·00

4Marrie

d0·02

3***

0·00

80·01

50·01

10·02

6*0·01

30·03

1***

0·00

70·01

40·01

00·04

6***

0·01

3Villag

e−0·01

80·01

6−0·00

60·01

8−0·03

3*0·01

9–

––

––

–Cap

ital

0·05

6***

0·02

00·05

3**

0·02

30·06

2**

0·02

4–

––

––

–Asset2

0·02

00·01

20·01

90·01

50·01

90·01

70·02

5**

0·01

10·03

2**

0·01

50·01

30·01

7Asset3

0·05

6***

0·01

30·05

8***

0·01

70·05

0***

0·01

90 ·05

3***

0·01

10·06

3***

0·01

60·03

2*0·01

8Asset4

0·11

4***

0·01

40·13

5***

0·01

80·09

1***

0·01

90·10

6***

0·01

10·13

2***

0·01

60·07

0***

0·01

8No.

ofob

servations

1730

597

7875

2717

305

9778

7527

Sou

rce:

Hea

lthin

Tim

esof

Tran

sitio

n(H

ITT)da

tase

t,20

10.

Cluster-rob

uststan

dard

errors

arepres

ented.

Allsp

ecifica

tions

also

includ

eco

untrydu

mmies.

*Significan

tat

10%

leve

l;**sign

ifica

ntat

5%

leve

l;***significan

tat

1%

leve

l.†Ordinaryleas

tsq

uares(O

LS)mod

el.

‡Com

mun

ityfixed

effects(C

FE)mod

el.

Fruit and vegetable consumption in the former Soviet Union 2829

columns 3 and 4 community variables are added. Age wasnegatively related to frequent fruit or vegetable consump-tion. Conversely, women were more likely to eat fruit orvegetables more frequently. These parameters were alsopositive for education, good economic status, being married,greater wealth and negative for living in a village. Finally, allcommunity-level parameters were now insignificant (butrecall that the outcome is defined differently from that in

Table 3 and that results are now presented separately forfruit and vegetable consumption).

Discussion

Prevalence of fruit and vegetable consumptionOverall, fruit or vegetable consumption in the FSU appearsinadequate, consistent with other evidence from this

Table 3 Community-level determinants of daily/almost daily fruit or vegetable consumption in nine countries of the former Soviet Union

All Women Men

Coefficient† SE Coefficient† SE Coefficient† SE

Food adverts 0·020 −0·018 0·028 −0·020 0·013 −0·025Snack adverts −0·031*** −0·011 −0·040*** −0·011 −0·019 −0·015Soft drink adverts 0·023* −0·012 0·032*** −0·011 0·027 −0·017Sweet drink/juice adverts −0·016* −0·008 −0·025** −0·011 −0·015 −0·009Shops selling crisps and sweets 0·005 −0·008 −0·011 −0·009 0·022** −0·009Shops selling fruit and vegetables 0·009 −0·010 0·023** −0·012 −0·004 −0·012No. of local food restaurants 0·015 −0·009 0·029** −0·012 −0·005 −0·010Easy walk to fast-food outlet −0·069 −0·046 −0·012 −0·064 −0·161** −0·062No. of observations 1680 946 734

Source: Health in Times of Transition (HITT) data set, 2010.Cluster-robust standard errors are presented.*Significant at 10% level; **significant at 5% level; ***significant at 1% level.†Ordinary least squares (OLS) parameter estimates are presented. All specifications include the same control variables as in Table 2. In addition, allspecifications include the following community control variables: dummy indicators for living in communities where no homes have garbage collected byauthorities from all homes; for living in communities where no homes have a cold water supply; for living in communities where no homes have a central heatingsystem; and for living in communities where there are no derelict homes present. Finally, all specifications include regional fixed effects.

Table 4 Ordered probit results for fruit or vegetable consumption† in nine countries of the former Soviet Union, without and with communitydeterminants

Column 1 Column 2 Column 3 Column 4

Vegetables Fruit Vegetables‡ Fruit‡

Coefficient SE Coefficient SE Coefficient SE Coefficient SE

Good diet not important −0·071 0·103 0·032 0·106 0·265 0·363 0·072 0·293Age −0·002*** 0·001 −0·004*** 0·001 −0·002 0·002 −0·004** 0·002Female 0·073*** 0·017 0·111*** 0·017 −0·003 0·053 0·091 0·060Primary education −0·226*** 0·035 −0·226*** 0·036 −0·079 0·099 −0·166 0·112Secondary education −0·147*** 0·022 −0·185*** 0·022 −0·155** 0·065 −0·180*** 0·069Good economic situation 0·186*** 0·029 0·298*** 0·028 0·219*** 0·078 0·370*** 0·070Household size 0·002 0·006 0·005 0·007 0·004 0·023 0·010 0·017Married 0·060*** 0·019 0·066*** 0·019 0·057 0·062 0·043 0·060Village −0·077** 0·038 −0·209*** 0·035 0·066 0·120 0·022 0·115Capital 0·074 0·049 0·108** 0·047 0·318** 0·161 0·250* 0·144Asset2 0·084*** 0·028 0·173*** 0·029 0·099 0·084 0·216** 0·085Asset3 0·170*** 0·03 0·279*** 0·031 0·149 0·092 0·202** 0·093Asset4 0·334*** 0·032 0·390*** 0·033 0·347*** 0·094 0·418*** 0·102No. of food adverts – – – – 0·049 0·039 −0·030 0·032No. of snack adverts – – – – −0·037 0·026 0·007 0·023No. of soft drink adverts – – – – 0·033 0·028 0·007 0·021No. of juice adverts – – – – −0·007 0·017 0·021 0·014No. of shops selling sweets and crisps – – – – −0·003 0·016 −0·012 0·011No. of shops selling fruit and vegetables – – – – 0·009 0·02 0·022 0·014Number of local food restaurants – – – – 0·009 0·016 −0·017 0·013Easy walk to fast food outlet – – – – −0·097 0·103 −0·011 0·099Observations 17 395 17 372 1692 1685

Source: Health in Times of Transition HITT data set, 2010.Original parameters are presented. Cluster-robust standard errors are presented.*Significant at 10% level; **significant at 5% level; ***significant at 1% level.†The outcome variable range is from 1= less than once weekly to 4= daily/almost daily.‡Specifications include the following community control variables: dummy variables for living in communities where garbage is collected by authorities from allhomes; for living in communities where all homes have a cold water supply; for living in communities where all homes have a central heating system; and forliving in communities where there are no derelict homes present.

2830 Y Goryakin et al.

region(9,23). However, it should be noted that existing stu-dies in the FSU region only cover some countries and do notexamine determinants of dietary patterns. One exception isa recent study(17) which found that fruit and vegetableconsumption in eight former Soviet countries has worsenedin the past decade, especially among the poor and those inrural areas. However, it is much more descriptive than ours.While we consider community-level determinants in addi-tion to individual ones, that study did not take advantage ofthe community-level data set. It also focused mainly on thedeterminants of inadequate fruit and vegetable consumption(i.e. fruit once weekly or less often, or vegetables onceweekly or less often), while we consider determinants ofgood (i.e. daily/almost daily fruit or vegetable consump-tion). Another (less important) difference is that the previousstudy considered prevalence and determinants of fruit andvegetable consumption separately by fruit and vege-tables, while we aggregated consumption. Our approach ismore relevant in our view, because international guidelinesprescribe combined fruit and vegetable consumption of400 g/d, rather than separately for fruit or vegetables.

Some findings are unexpected: despite its large agri-cultural sector and warm climate, Moldova has the smallestproportion of people reporting fruit or vegetable con-sumption more than once weekly. This may be because itis one of the poorest countries in the FSU, with a rapidlygrowing share of its agricultural output now beingexported(24). Interestingly, only 0·1 % of respondents livingin Moldova agreed with the statement that good diet is notimportant for health (see online supplementary material,Annex 3). At the same time, in Russia, where a relativelylarge proportion of people reported daily fruit or vegetableconsumption, 1·85% (the highest number) agreed with thisstatement. This gap between the perceived importance ofgood diet and actual fruit or vegetable consumption meritsfurther study, although research elsewhere has found asimilar disconnect between knowledge and practice(25). Itshould also be noted that these proportions are estimatedfor a very small number of respondents; fifty-four out of2922 in the case of Russia, for example.

Socio-economic and demographic determinantsOur findings are consistent with other evidence on the socialpatterning of fruit and vegetable consumption (i.e. accordingto socio-economic status)(7,26–29). Thus, variables such aseducation, household economic situation and householdsize, as well as wealth, are all independently associated withdaily fruit or vegetable consumption in the FSU.

Similarly, the lower probability of daily fruit or vegetableconsumption with increasing age is consistent with someprevious studies(30) but not all(28). While older people mayhave less disposable income to spend on nutritious food, itis also likely that age has an independent effect, as a rangeof socio-economic variables are controlled for in allregression results reported in Table 1. One potentialexplanation is that older people living in the FSU may

prefer to eat more traditional diets, which in many coun-tries in that region are based on meat and carbohydrate-rich foods such as potatoes and grains.

Like us, some previous studies also found that women,and those who are married, are more likely to eat enoughfruit and vegetables(7,28), including in Russia and severalCentral and Eastern European countries(31,32). This is inline with findings that in Russia, for example, women aremuch less likely to engage in dangerous health behaviourssuch as smoking and excessive alcohol consumption(33),which suggests that women living in the FSU may be morehealth-conscious then men. Since in that region, womentraditionally spend more time cooking than their hus-bands, this may also explain why married people are morelikely to eat healthily.

Few studies have examined how fruit and vegetableconsumption varies among those living in rural and urbanareas. One study from the USA found people in rural areasmore likely to consume fruit and vegetables(34); in con-trast, a European study found living in rural areas to beassociated with lower consumption(7). There is no sig-nificant association in the ordinary least squares modelsreported in Table 2, but living in rural areas is negativelyrelated to fruit or vegetable consumption in the orderedprobit regression model (Table 4). This finding may looksomewhat counterintuitive but again one possible expla-nation is the preference for the traditional diet rich ingrains, potato and meat (recall that potatoes are excludedfrom the definition of vegetable consumption).

Food stores and supermarketsTheoretically, greater availability of food stores andsupermarkets may increase access to fruits and vegetables,thus contributing to increased consumption, for reasonssuch as lower travel and time costs of obtaining suchfoods; stimulation of consumption by visual cues; and theeffect of exposure on food preference(35). However, betteraccess to supermarkets and food stores may also providegreater exposure to unhealthy foods and therefore,a priori, the overall effect is far from clear.

The available evidence does not clearly support theassertion that better access to food stores improves fruitand vegetable consumption(36). However, most of theexisting evidence is derived from cross-sectional stu-dies(37) conducted in high-income countries, so theirfindings may not be transferable to poorer countries in theFSU. Adding to the complexity, several studies foundconsistent positive associations between healthy dietarypatterns and supermarket access in the USA(38), but not inEurope(39). One potential explanation is the greater loca-tional segregation in the USA(38), with supermarkets dis-tributed more evenly among poor and wealthy districts inEurope, or because of better access to retail food outlets inEurope due to better public transport.

Although our data do not capture the number ofsupermarkets in the neighbouring area, they show that

Fruit and vegetable consumption in the former Soviet Union 2831

access to shops selling fruit and vegetables is positivelyand significantly correlated with daily fruit or vegetableconsumption for women. One potential explanation is thatit is not really the proximity of additional stores selling fruitand vegetables that influences fruit and vegetable con-sumption, but rather the fact that they are situated inwealthier areas, where people may be better educatedabout the importance of nutritious food and have higherincomes to purchase them(22). In addition, access toremotely located stores in the FSU may be limited due tothe lack of convenient and affordable public transport andscarcity of cars. Nevertheless, one can be more optimisticabout a causative interpretation of our findings becauseregional fixed effects are also included in the analysis,which should account for interregional variations insocio-economic indicators. In another middle-incomecountry – Brazil – a study that also controlled for areasocio-economic status found a similar positive correlationbetween regular fruit and vegetable intake and density offood markets specializing in fruit and vegetables(40).

It should also be mentioned that the consequences ofbetter access to supermarkets can differ from those ofbetter access to convenience stores, and that our data setdoes not make a clear distinction between these two kindsof stores. Thus, some studies have found either no ornegative associations between availability of conveniencestores and fruit and vegetable consumption(38,39). This canbe because such stores may provide less choice of freshfruit and vegetables, and thus encourage people to buymore unhealthy food items. Alternatively, such stores maybe located in more economically disadvantaged areas andthus the observed association between dietary patternsand convenience store access may be partly driven byvariations in neighbourhood socio-economic status.Although there is no proxy for convenience store avail-ability, there is a variable measuring the number of storesselling sweets and crisps in the neighbourhood. Whilethere is no significant association between this variableand daily fruit or vegetable consumption for the wholesample and women only, surprisingly it is significant andpositive for men. Nevertheless, it is important to empha-size that these stores may not necessarily be limited intheir supply of fruit or vegetables (and thus not properlyfall in the category of convenience stores) and thereforeone should not over-interpret this finding.

Nutrition and advertisingThe relationship between food advertising and dietarybehaviours is also of interest, as sums spent on advertisingare very large, most promoting unhealthy foods(41).However, the existing literature on the effect of foodadvertising on either fruit or vegetable consumption islimited and tends to focus on adolescents, as well as ontelevision advertising only(42). A considerable part of thisliterature is based on small-scale experimental studies ofquestionable generalizability.

We find billboard advertising of snacks and sweetdrinks (including juices) to be significantly negativelyrelated to daily fruit or vegetable consumption. While thisdoes not prove that billboard advertising for unhealthyfoods causes less consumption of fruit or vegetables (as itmay well be that such advertisements are deliberatelyplaced in communities where unhealthy eating is moreprevalent), the fact that the effect is significant even withthe inclusion of regional effects, as well as of a range ofboth community and individual controls, does increaseconfidence in our findings. Also the fact that this associa-tion is much stronger for women in both cases suggeststhat local confounders are unlikely to be the main expla-nation. Surprisingly, billboard adverts for soft drinks arepositively related to daily fruit or vegetable consumption(although not for men). One can speculate that the posi-tive sign found for soft drinks advertising might be due to acomplementarity or substitutability between fruit andvegetables, and other goods. For instance, while juicedrinks could be perceived as substitutes for fruit and soconsumed as an alternative, soft drinks or sweets couldinstead be more often consumed with fruit. We could notfind any other studies that measured this association.

Fast food and restaurantsThe role played by availability and access to fast-food outletsis also unclear. Thus, although, theoretically, one can expecteasier access to fast-food stores to be associated with worsedietary patterns(43), such findings may be due tocommunity-level confounding by neighbourhood socio-economic status, with less well-off communities morelikely to provide access to fast-food establishments.

Empirical evidence on this topic has so far been incon-clusive. One New Zealand study found neighbourhoodaccess to fast-food establishments unrelated to fruit andvegetable consumption(38). The previously cited study fromBrazil found no association between fast-food outlet densityand fruit and vegetable intake(40). Several US studies foundeasier access to fast-food outlets to be negatively related todiet-related outcomes(38). Our finding of a significant negativeassociation between the ease of access to fast-food outletsand the probability of daily fruit or vegetable consumption formen is thus more consistent with the US studies.

As for easier access to full-service restaurants, theoreti-cally it is unclear how they may affect dietary attitudes andbehaviours. On one hand, the effect may depend on thefood choice on offer (traditional menus are quite heavilymeat- and potato-based in many FSU countries). Con-versely, any empirical finding of an association betweenthese variables should be tempered by the risk of con-founding by neighbourhood-level characteristics. As it is,the existing empirical evidence is more limited than forfast-food outlets. One study found, for example, that betteraccess to a full-service restaurant was related to lowerintake of saturated fat among black Americans(41). Anotherstudy reached a different conclusion after finding that

2832 Y Goryakin et al.

away-from-home eating (with both restaurant and fast-food consumption) was related to worse quality of diet(22).Our finding of a small positive association between thenumber of local food restaurants and greater fruit orvegetable consumption in the FSU countries (especiallyfor women) adds to this growing literature.

Data limitationsAlthough our rich data set helps alleviate potential endo-geneity concerns, there are certain limitations. The ques-tionnaires were not primarily designed to assess diet andonly recorded whether respondents had eaten any fruit orvegetables during the past week and not how often.Although eating fruit or vegetables daily or almost daily maystill not guarantee that adult people eat their recommendedamount of 400 g/d, at least this group is more likely to meetfruit and vegetable targets. The need for data collected withfood frequency or dietary recall questions is clear(44).

In addition, the fruit and vegetable consumption vari-ables have not been validated for the HITT study. Havingsaid that, very similar variables have been used in anotherpublished article(45). A possible concern regarding theexternal validity of our results comes from the fact thatdata were collected in the spring, between March andMay, a period when fruit and vegetables will be in rela-tively poor supply compared with June to September. Thistiming may lead us to underestimate the effect of proxi-mity to stores on fruit and vegetable consumption.

Also, observed associations may not be causal. Forexample, community-level exposure to advertising may bedetermined by the perceived attractiveness of the neigh-bourhood demographics to marketing organizations andplacement of stores may also depend on the perceivedwealth of the community. Having said that, this issue isaddressed in two main ways: (i) as all the variables ofinterest are included in equation (1) simultaneously, par-tial regression coefficients obtained for each covariatedemonstrate the association adjusted for any potentialconfounding by observable variables; and (ii) by includingcommunity fixed effects in equation (1), and regional fixedeffects in equation (2), any additional area-level con-founding affecting both the covariates of interest and theoutcome variable is controlled for.

Some of the community-level indicators may also beimperfect measures of the variables of interest. Thus, thedata set lacks information on size of outlets. Moreover, it ispossible that the same outlet may sell both healthy (e.g.fruit and vegetables) and unhealthy (e.g. biscuits) items.Nevertheless, given these limitations, it is encouraging thatour results are largely consistent with prior expectations.

Conclusions

The present study is the first one to examine both theindividual- and community-level determinants of fruit and

vegetable consumption in nine FSU countries. It confirmsthe inadequacy of consumption in this region and shedslight on which groups are most vulnerable: namely men,those at older ages, with less education and fewer financialresources. However, beyond these individual attributes,the local food economy also plays a role. Taken together,these findings provide potential entry points for policyinterventions.

Acknowledgements

Acknowledgements: The authors are grateful to all mem-bers of the Health in Times of Transition (HITT) Projectstudy teams who participated in the coordination andorganization of data collection for this paper. Financialsupport: The HITT Project was funded by the EuropeanUnion’s 7th Framework Programme (project HEALTH-F2-2009-223344). The European Commission cannot acceptany responsibility for any information provided or viewsexpressed. The contributions of Y.G. and M.S. were partlyfunded by the Centre for Diet and Activity Research(CEDAR), a UKCRC Public Health Research Centre ofExcellence. Funding from the British Heart Foundation,the Economic and Social Research Council, the MedicalResearch Council, the National Institute for HealthResearch and the Wellcome Trust, under the auspices ofthe UK Clinical Research Collaboration, is gratefullyacknowledged. The funders had no role in study design,data collection and analysis, decision to publish or pre-paration of the manuscript. Conflicts of interest: None.Authors’ contributions: Y.G. conducted the analyses anddrafted the initial manuscript, as well as subsequent revi-sions. B.R., M.S., L.R. and M.M helped interpret the results,revised the drafts and approved the final manuscript.Ethics of human subject participation: The research wasapproved by the ethics committee of the London School ofHygiene and Tropical Medicine and was conducted inaccordance with the ethical standards laid down in the1964 Declaration of Helsinki. All persons gave informedconsent.

Supplementary material

To view supplementary material for this article, please visithttp://dx.doi.org/10.1017/S1368980015000105

References

1. Lim SS, Vos T, Flaxman AD et al. (2012) A comparative riskassessment of burden of disease and injury attributable to 67risk factors and risk factor clusters in 21 regions, 1990–2010:a systematic analysis for the Global Burden of DiseaseStudy 2010. Lancet 380, 2224–2260.

2. Ness AR & Powles JW (1997) Fruit and vegetables, andcardiovascular disease: a review. Int J Epidemiol 26, 1–13.

Fruit and vegetable consumption in the former Soviet Union 2833

3. Danaei G, Vander Hoorn S, Lopez AD et al. (2005) Causesof cancer in the world: comparative risk assessment of ninebehavioural and environmental risk factors. Lancet 366,1784–1793.

4. Lock K, Pomerleau J, Causer L et al. (2005) The globalburden of disease due to low fruit and vegetable con-sumption: implications for the global strategy on diet. BullWorld Health Organ 83, 100–108.

5. World Health Organization & Food and Agriculture Organi-zation of the United Nations (2005) Fruit and Vegetables forHealth. Report of a Joint FAO/WHOWorkshop, 1–3 September2004, Kobe, Japan. Geneva: WHO; available at https://extranet.who.int/iris/restricted/bitstream/10665/43143/1/9241592818_eng.pdf

6. Shaikh AR, Yaroch AL, Nebeling L et al. (2008) Psychosocialpredictors of fruit and vegetable consumption in adults: areview of the literature. Am J Prev Med 34, 535–543.

7. Kamphuis CBM, Giskes K, De Bruijn GJ et al. (2006)Environmental determinants of fruit and vegetable con-sumption among adults: a systematic review. Br J Nutr 96,609–619.

8. Krølner R, Rasmussen M, Brug J et al. (2011) Determinantsof fruit and vegetable consumption among children andadolescents: a review of the literature. Part II: qualitativestudies. Int J Behav Nutr Phys Act 8, 112.

9. Hall JN, Moore S, Harper SB et al. (2009) Global variabilityin fruit and vegetable consumption. Am J Prev Med 36,402–409.e5.

10. Matilainen T, Vartiainen E, Puska P et al. (1996) Plasmaascorbic acid concentrations in the Republic of Karelia,Russia and in North Karelia, Finland. Eur J Clin Nutr 50,115–120.

11. Paalanen L, Prättälä R, Alfthan G et al. (2013) Vegetable andfruit consumption, education and plasma vitamin C con-centration in Russian and Finnish Karelia, 1992–2002. PublicHealth Nutr (Epublication ahead of print version).

12. Giskes K, Kamphuis CB, Van Lenthe FJ et al. (2007) Asystematic review of associations between environmentalfactors, energy and fat intakes among adults: is there evi-dence for environments that encourage obesogenic dietaryintakes? Public Health Nutr 10, 1005–1017.

13. Brug J (2008) Determinants of healthy eating: motivation,abilities and environmental opportunities. Fam Pract 25,i50–i55.

14. Reidpath DD, Burns C, Garrard J et al. (2002) An ecologicalstudy of the relationship between social and environmentaldeterminants of obesity. Health Place 8, 141–145.

15. Chow CK, Lock K, Teo K et al. (2009) Environmental andsocietal influences acting on cardiovascular risk factors anddisease at a population level: a review. Int J Epidemiol 38,1580–1594.

16. Booth KM, Pinkston MM & Poston WSC (2005) Obesity andthe built environment. J Am Diet Assoc 105, 110–117.

17. Abe SK, Stickley A, Roberts B et al. (2013) Changing pat-terns of fruit and vegetable intake in countries of the formerSoviet Union. Public Health Nutr 16, 1924–1932.

18. Chow CK, Lock K, Madhavan M et al. (2010) EnvironmentalProfile of a Community’s Health (EPOCH): an instrument tomeasure environmental determinants of cardiovascularhealth in five countries. PLoS One 5, e14294.

19. Watson K, Roberts B, Chow C et al. (2013) Micro- andmeso-level influences on obesity in the former SovietUnion: a multi-level analysis. Eur J Public Health 23,291–298.

20. Pärna K, Lang K, Raju K et al. (2007) A rapid situationassessment of the market for surrogate and illegal alcoholsin Tallinn, Estonia. Int J Public Health 52, 402–410.

21. Cameron AC & Trivedi PK (2005) Microeconometrics:Methods and Applications. New York: Cambridge Uni-versity Press.

22. Popkin BM, Duffey K & Gordon-Larsen P (2005) Environ-mental influences on food choice, physical activity andenergy balance. Physiol Behav 86, 603–613.

23. Brainerd E & Cutler DM (2004) Autopsy on an Empire:Understanding Mortality in Russia and the Former SovietUnion. NBER Working Paper no. 10868. Cambridge, MA:National Bureau of Economic Research.

24. Certain S & Certain I (2012) Agriculture in the Republic ofMoldova: present and future. Manage Econ Eng Agric RuralDev 12, 13–22.

25. Sharma SV, Gernand AD & Day RS (2008) Nutritionknowledge predicts eating behavior of all food groupsexcept fruits and vegetables among adults in the Paso delNorte Region: Qu Sabrosa Vida. J Nutr Educ Behav 40,361–368.

26. Ball K & Crawford D (2010) Socioeconomic inequalities infruit and vegetable intakes. In Bioactive Foods in PromotingHealth: Fruits and Vegetables, pp. 195–205 [RR Watson andVR Preedy, editors]. Amsterdam: Elsevier.

27. Vlismas K, Stavrinos V & Panagiotakos DB (2009) Socio-economic status, dietary habits and health-related outcomesin various parts of the world: a review. Cent Eur J PublicHealth 17, 55–63.

28. Pollard J, Kirk S & Cade J (2002) Factors affecting foodchoice in relation to fruit and vegetable intake: a review.Nutr Res Rev 15, 373–388.

29. De Irala-Estevez J, Groth M, Johansson L et al. (2000) Asystematic review of socio-economic differences in foodhabits in Europe: consumption of fruit and vegetables. Eur JClin Nutr 54, 706–714.

30. De Vet E, De Ridder D & De Wit J (2011) Environmentalcorrelates of physical activity and dietary behaviours amongyoung people: a systematic review of reviews. Obes Rev 12,e130–e142.

31. Prättälä R, Paalanen L, Grinberga D et al. (2007) Genderdifferences in the consumption of meat, fruit and vegetablesare similar in Finland and the Baltic countries. Eur J PublicHealth 17, 520–525.

32. Boylan S, Welch A, Pikhart H et al. (2009) Dietary habits inthree Central and Eastern European countries: theHAPIEE study. BMC Public Health 9, 439.

33. Bobak M, Pikhart H, Hertzman C et al. (1998) Socioeconomicfactors, perceived control and self-reported health in Russia. Across-sectional survey. Soc Sci Med 47, 269–279.

34. Watters JL, Satia JA & Galanko JA (2007) Associations ofpsychosocial factors with fruit and vegetable intake amongAfrican-Americans. Public Health Nutr 10, 701–711.

35. Jago R, Baranowski T & Baranowski JC (2007) Fruit andvegetable availability: a micro environmental mediatingvariable? Public Health Nutr 10, 681–689.

36. Giskes K, van Lenthe FJ, Kamphuis CBM et al. (2009)Household and food shopping environments: do they playa role in socioeconomic inequalities in fruit and vegetableconsumption? A multilevel study among Dutch adults.J Epidemiol Community Health 63, 113–120.

37. Rose D, Bodor JN, Hutchinson PL et al. (2010) The importanceof a multi-dimensional approach for studying the linksbetween food access and consumption. J Nutr 140,1170–1174.

38. Pearce J, Hiscock R, Blakely T et al. (2009) A national studyof the association between neighbourhood access to fast-food outlets and the diet and weight of local residents.Health Place 15, 193–197.

39. Macdonald L, Ellaway A, Ball K et al. (2011) Is proximity to afood retail store associated with diet and BMI in Glasgow,Scotland? BMC Public Health 11, 464.

40. Jaime PC, Duran AC, Sarti FM et al. (2011) Investigatingenvironmental determinants of diet, physical activity, andoverweight among adults in Sao Paulo, Brazil. J UrbanHealth 88, 567–581.

2834 Y Goryakin et al.

41. Larson N & Story M (2009) A review of environmentalinfluences on food choices. Ann Behav Med 38, 56–73.

42. Lake A & Townshend T (2006) Obesogenic environments:exploring the built and food environments. J R Soc PromotHealth 126, 262–267.

43. Canoy D & Buchan I (2007) Challenges in obesity epide-miology. Obes Rev 8, 1–11.

44. Vaask S, Pomerleau J, Pudule I et al. (2004) Comparison ofthe Micro-Nutrica Nutritional Analysis program and theRussian Food Composition Database using data from theBaltic Nutrition Surveys. Eur J Clin Nutr 58, 573–579.

45. Hinote BP, Cockerham WC & Abbott P (2009) Psychologicaldistress and dietary patterns in eight post-Soviet republics.Appetite 53, 24–33.

Fruit and vegetable consumption in the former Soviet Union 2835