What not to eat: inequalities in healthy eating behaviour, evidence from the 1998 Scottish Health Survey

Transcription

1 Journal of Public Health VoI. 27, No. 1, pp doi: /pubmed/fdh191 Advance Access Publication 7 February 2005 What not to eat: inequalities in healthy eating behaviour, evidence from the 1998 Scottish Health Survey Nicola J. Shelton Abstract Background The role that healthy eating plays in good health is well documented. Government policy in Scotland recommends reducing salt, refined sugar and saturated fat in the diet and increasing the consumption of fruit and vegetables, carbohydrates, fibre and oily fish. Methods Using data from the 1998 Scottish Health Survey a composite measure of healthy eating behaviour was derived by scoring low levels of consumption of salt, refined sugar and saturated fat and higher consumption of fruit and vegetables, carbohydrates, fibre and oily fish. This paper presents results from logistic regression analysis of the risk factors for this measure of healthy eating behaviour. Results Young people, men, those on low income, those without qualifications, those who take little exercise, who lack access to car, live in deprived areas and women who smoke were less likely to show healthy eating behaviour. Conclusion Healthy eating reflects dietary policy and guidelines for those with higher socio-economic status and who demonstrate other health seeking behaviour. Health education policy needs to be targeted at young people, men, and those of lower socio-economic status, and deprived areas, not just the population or country as a whole. Keywords: healthy eating, Scotland, food frequency, Scottish Health Survey Introduction People s health is affected greatly by what they do and do not eat. In Scotland, eating habits were the second major cause (after smoking) of poor health contributing to a range of serious illnesses, which include coronary heart disease, certain cancers, strokes, osteoporosis and diabetes. 1 Healthy eating was seen by the Scottish Executive as one of the key factors for improving Scotland s position as one of the most unhealthy nations in Europe. 2 January 2003 saw the launch of Scotland s national healthy living campaign. 3 The campaign aims to accelerate the implementation of the Scottish Diet Action Plan, which set targets for 2005, including doubling the consumption of fruit and vegetables, a reduction in the consumption of sugar, salt and fat, particularly saturated fats, and an increase in the consumption of oily fish and complex carbohydrates. 4 In 1995 the United Kingdom had the lowest per capita intake of fresh fruit and vegetables of the EU member states. 5 In the same year results from the Scottish Health Survey (SHS) showed that fruit consumption was even lower in Scotland than in England. 6 Though the 1998 SHS showed that between 1995 and 1998 there had been a 6 per cent increase in fruit consumption (at least once a day or more), an almost 10 per cent increase in potatoes, pasta or rice consumption (five times a week or more); and 2.5 per cent reduction in the number of people who added salt at the table, there were still national differences. 7 The Scottish diet remained higher in fat than the rest of the United Kingdom and consumption of meals containing vegetables was 23 per cent lower; a major improvement was required in the national diet overall, but particularly in lowincome communities. Evidence from the 1998 SHS showed that men and people in social classes IV and V were more likely to exhibit unhealthy eating habits, while regional differences were small. 7 Similarly the National Diet and Nutrition Survey for Great Britain reported that older men and women aged ate more healthy food and less convenience food than young people aged and that men and women on benefits were less likely to eat healthy foods. 8 Recent work on saturated fat consumption has argued that lower social class Europeans also chose more traditional foods, which were often unhealthier. 9 Work on the Dietary and Nutrition Survey, showed that women, older people, non-smokers and non-manual social classes were more likely to comply with national fruit and vegetable goals, with lower compliance in Scotland and northern England compared to the rest of Britain. 10 In Finland, unhealthy diet was found to be positively correlated with physical inactivity and smoking, but inversely correlated with alcohol consumption. 11 This study looks at the risk factors for composite healthy eating behaviour in Scotland using the 1998 SHS by aggregation of food frequencies. Four groups of risk factors are considered, Department of Epidemiology and Public Health, University College London, London WC1E 6BT Dr Nicola J. Shelton n.shelton@ucl.ac.uk The Author 2005, Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved.

2 WHAT NOT TO EAT: INEQUALITIES IN HEALTHY EATING BEHAVIOUR 37 demographic factors, socio-economic factors including education and access, health seeking behaviour (smoking behaviour, drinking habits, and exercise), and health status, taking in account the interrelation of factors. Methods The 1998 SHS was conducted to monitor the health and diet of the Scottish population aged 2 74 living in private households. The sample is described in detail in the survey report. 12 In brief, a systematic selection of 312 postcode sectors by level of deprivation was made and a letter sent to random sample of addresses from those sectors. Interviewers then collected information on health-related topics and measured height and weight and nurses collected a range of anthropometric and physiological measurements. The analysis detailed here was restricted to adults aged adults were interviewed (response rate 76 per cent) and of these 8549 had their height measured and 8230 their weight. 63 per cent of adults and 82 per cent of those interviewed (n = 7455) saw a nurse, of these 7326 agreed to a waist and hip measurement. All the data were weighted to adjust for sampling design features; such as the selection of only one adult in the households leading to adults living in large households having a smaller chance of selection. Most responses were entered directly into a laptop. Smoking and drinking behaviour were collected by face to face interview for older adults and by self completion booklet for all young adults aged and some aged (>99 per cent) adults interviewed and 7434 (>99 per cent) of those who saw a nurse answered questions about their intake of key dietary components that reflected policy targets. Refusals and pregnant women (n = 66) who were not measured, were excluded from the analysis, giving an achieved unweighted sample size of Showcard options were: six or more times a day; four or five times a day; two or three times a day; once a day; five or six times a week; two to four times a week; once a week; one to three times a month; less often or never. For the purposes of this study each frequency was grouped into tertiles and then assigned a score. The highest frequency tertiles of saturated fat, sugar and salt and lowest tertiles of fruit, vegetables, starches, oily fish and fibre consumption were assigned zero scores. +1 scores were assigned to the mid-range tertiles. +2 scores were assigned to the best dietary tertiles. Dichotomous positive dietary factors e.g. eating high fibre breakfast cereal were assigned +1 if present. Don t knows were coded as zero. To establish the risk factors between those adopting a healthier composite diet and those not, the scores were cumulated for men and women separately. Exact numbers meeting all dietary targets could not be calculated with the data and given that only three in 10 adults ate cooked green vegetables once day or more, this group would have been too small for comparison with the rest of population. Therefore, the bottom and top half of the total scores were dichotomized into unhealthy and healthy eating respectively and analysed with logistic regression. Four groups of risk factors were considered: demographic, socio-economic, health behaviour and health status. Demographic variables used in the analysis included age, gender, marital status and living alone. Socio-economic status of the head of household was used in the analysis, own economic activity and area level deprivation using a grouped Carstairs index. 13,14 Region of residence was considered at the health board area. Education level was grouped into degree, A level pass or equivalent, O level grade C pass or equivalent, and other/no qualifications; current housing tenure was grouped into owner occupier, public sector tenant and other rented. Car ownership was dichotomized into does and does not have access to a car to represent wider access to food stores, as well as an indicator of socio-economic status. Physical activity was grouped into activities that were or were not part of daily routine. Housework, gardening and DIY were grouped into three activity levels: inactive, light (some non-heavy gardening or housework) and moderate (some heavy gardening and/or housework). Physical activity as part of employment was coded into inactive, light or vigorous depending on type of job. Walking was grouped and coded into three levels, inactive, light (15 min or more at slow/steady pace) and moderate (15 min or more at brisk/fast pace). Sporting activities were grouped and coded into three levels of activity, light (gentle exercise), moderate (not likely to be sweaty or out of breath) and vigorous (likely to be sweaty or out of breath). Smoking behaviour was grouped by own smoking status and attitude to smoking (never smoked, previously smoked, currently smokes and wants to give up, does not want to give up, not answered). Alcohol consumption was dichotomized into drank more than or up to the weekly unit limit (14 for women, 21 for men). (A unit of alcohol is 8 g of ethanol, and is the amount contained in half a pint of beer or lager, a small glass of wine, or a measure of spirits. Though weekly totals are used for analysis purposes, limits for alcohol consumption were changed in the late 1990s to take into account binge drinking, to daily limits of 2 3 units/ day for women and 3 4 units/day for men.) 15 Obesity was derived from grouped body mass index: <20 (underweight), (normal weight), (overweight), (obese) and 40+ (morbid obesity) following WHO guidelines. 16 The waist and hip ratio (WHR), a measure of abdominal obesity and a risk factor for cardiovascular disease, which is major cause of death in Scotland 17 was calculated and dichotomized into 0.95 and over and below 0.95 for men and 0.8 and over and below 0.8 for women. This was derived from a quintile measurement used for the Health Survey for England and the 1995 Scottish Health Survey; unlike BMI there is no consensus about appropriate cut of points to define high WHR 18 (The respondents were asked to remove shoes, jackets and heavy cardigans, heavy jewellery and heavy items from pockets before weight and height were measured. Waist and hips were measured after the removal of all outer layers of clothing such as cardigans, jumpers and waist-

3 38 JOURNAL OF PUBLIC HEALTH coats, or any tight garments such as belts or corsets. Hip circumference was taken at the widest circumference over the buttocks and below the iliac crest and waist at the midpoint of the costal margin and the iliac crest, unreliable measurements were not included in this analysis.) The relationship between eating behaviour and current health was explored using self reported mental and physical health. Mental health was measured using the General Health Questionnaire which asked twelve questions about recent happiness, anxiety, depression and sleep disturbance to derive a GHQ12 score. The score was dichotomized into four and above (higher risk of psychiatric disorder), 1 3 (some unhappiness) and 0 (for details of the scoring method see Goldberg and Williams 19 ). Self-assessed general health was grouped into very good/good, fair, bad/very bad. Limiting longstanding illness was defined by self-reported conditions that limited activities that would, or had, affected respondents, over a period of time. Results Table 1 The classification and scoring of variables used in calculating the healthy eating score Table 1 shows the classification of the dietary data by their contribution to healthy eating in Scotland. Most fruit and vegetable consumption came from eating fruit and cooked green vegetables. There was less frequent consumption of fresh fruit juice, root vegetables, pulses and raw vegetables or salad. Around a third of the population met oily fish consumption targets, eating it more than once a week. The highest tertile of the population were eating starches other than bread at least daily. The lowest frequency tertile consumed saturated fat and sugar in the form of fried food, cakes, scones, sweet pies, pastries or puddings, and soft drinks (not diet) less than once a week. Consumption of chocolates, crisps or biscuits, was high in all groups: up to four times a week in the group with the lowest intake tertile. Table 2 shows the weighted, mean healthy eating score (mean of scores for all components) for men and women in Zero 1 2 Usual breakfast cereal Not high fibre, Does not eat High fibre breakfast cereal Usual type of bread White and other breads Brown breads including wholemeal Does not eat bread Usual spread on bread High fat spread on bread Does not eat bread/no usual type Low fat spread/no spread used Usual fat for frying High-fat cooking fats Low-fat fried food/no usual type Not eating fried food Usual milk Whole, evaporated, condensed Semi-skimmed, skimmed, soy milk milk, does not drink milk Frequency of eating cooked Less than 5 times a week Five to six times a week Once a day or more green vegetables Frequency of eating raw Less than once a week Once a week Twice a week or more vegetables or salad Frequency of drinking fresh fruit Less than once a week One to five times a week Six times a week or more juice Frequency of eating cooked Less than twice a week Two to four times a week Five times a week or more root vegetables Frequency of eating fresh fruit Less than 5 times a week Five to seven times a week More than once a day Frequency of eating pulses Once a week or less 2 4 times a week 5 times a week or more Frequency of eating potatoes Less than 5 times a week 5 6 times a week Once a day or more pasta, or rice Frequency of eating slices of Less than 2 a day 2 3 per day 4 or more per day bread or rolls Frequency of eating oily fish Less than once a week Once a week More than once a week Frequency of eating red meat Once a day or more Less than once a day Frequency of eating fried food 5 times a week or more One to four times a week Less than once a week Frequency of eating cheese 5 times a week or more 2 4 times a week Less than twice a week Adds salt at the table Usually without tasting Generally, occasionally after tasting Rarely, never Frequency of eating sweets or 5 times or more a week 2 4 times a week Once a week or less ice cream Frequency of drinking soft Twice or more a day 1 7 times a week Less than once a week drinks (not diet) Frequency of eating Twice or more a day Five to seven times a week Less than five times a week chocolates, crisps or biscuits Frequency of eating cakes, 5 times or more a week One to four times a week Less than once a week scones, sweet pies, pastries or puddings Use of sugar in hot drinks Takes sugar in tea/coffee Does not take sugar in tea/coffee, does not drink tea or coffee Vitamin supplementation Does not take vitamins Takes vitamins

4 WHAT NOT TO EAT: INEQUALITIES IN HEALTHY EATING BEHAVIOUR 39 Table 2 Mean score for healthy eating, adults aged 16 74* Score N Min Max/42 Mean SD Women Men All Unweighted base The results were weighted for non-response and survey design. Scotland. Women had higher mean healthy eating scores than men, scores ranged from 7 to 41 (the maximum score was 42) for women, and from 7 to 39 for men. The distribution (not shown) was approximately normal. Table 3 shows the weighted, age standardized mean healthy eating scores for men and women. In all sub-groups men were more likely to have lower healthy eating scores than women. Younger men and women were less likely to be eating healthily than older adults. Unmarried men were less likely to be eating healthily than married and widowed men. Household composition showed little differences, but varied in opposite directions for men and women. There was a gradient in mean healthy eating score for men and women for social class, area level deprivation, tenure, education and access to a car. The sub-group with the lowest mean healthy eating score was unemployed men. There were small regional differences in healthy eating at the health board level. Some health behaviours were related with healthy eating: men reporting themselves as inactive in walking had lower mean scores than those with moderate walking activity levels. Current smokers (both those who wanted to give up and those who did not) had lower mean healthy eating scores than nonsmokers (both never and previously smoked). Sporting activity was less closely related to healthy eating than walking, with the lowest mean healthy eating scores for men and women in those who engaged in moderate levels of sporting activity. Alcohol consumption above the recommended weekly limit was associated with unhealthy eating in men and healthy eating in women, but the differences were small. Routine domestic activity (such as gardening and housework) was not associated with healthy eating. Physical activity in the workplace also showed no obvious pattern. Low BMI (<20) for men and women and high BMI (>30) for men only was related to lower mean healthy eating score; high WHRs were also associated with healthy eating for men and women. Self reported mental and general health for men and women and limiting long term illness for women were positively related to healthy eating. As many of these variables were associated with each other, multivariate analyses were needed to identify associations between particular variables and healthy eating, taking into account the effect of the other variables included in the analysis. Initially models using all the covariates were fitted. In a later iteration, variables were retained that that had a statistically significant association (at the 5 per cent level) with the outcome contrasts. The results in Tables 4 and 5 show that once confounding variables have been controlled for, significant differences remained for several covariates. Age remained important with men aged and women aged having significantly higher odds of healthy eating than younger men and women (aged 16 24) with odds ratios increasing with age. Marital status and household composition were no longer significant. Low socio-economic status class was strongly associated with lower levels of healthy eating in the multivariate analysis. Both men and women in social classes III manual (but not III non-manual) and social classes IV and V had significantly lower odds of healthy eating than social classes I and II. There was a strong gradient in education level, men and women had increasingly higher odds of healthy eating with higher qualifications. The odds of healthy eating were significantly lower in the areas with highest deprivation for women and in the worst two groups of areas for men compared to the least deprived areas. No access to a car also was associated with significantly lower odds of healthy eating for women. Housing tenure, region and individual economic status were no longer significant. Current women smokers and those who did not answer the smoking questions had significantly higher odds of unhealthy eating than those who had never smoked. Female ex-smokers did not have significantly different odds of unhealthy eating than the never smoked. Men who engaged in moderate walking activity and vigorous sporting activity were significantly more likely to have high healthy eating scores than those who engaged in no walking activity or light sporting activity respectively. Men who engaged in moderate levels of sporting activity were significantly less likely to have healthy eating scores. Women who were engaged in both light or moderate walking and vigorous sporting activity were significantly more likely to have higher odds of healthy eating than those who engaged in no walking or light sporting activity respectively. Domestic physical activity was no longer significant. A low BMI (<20) was associated with lower odds of healthy eating for women, but high BMIs (30 39) were associated with higher odds of healthy eating for men. Self reported mental and physical health and WHR were no longer significant once other factors had been controlled for. Discussion The aim of the paper was to see which population sub-groups were most likely to have diets that better reflected dietary policy targets. The results shown here are similar to that in other European countries, healthy eating in Scotland is associated with age, gender, social class, and health seeking behaviour. Low healthy eating scores in young people, may be due to domestic circumstances, but could be evidence of the emergence of unhealthier diets, so should be monitored. Poverty, especially unemployment is another key area to address for health education policy; campaigns may need to be specifically

5 40 JOURNAL OF PUBLIC HEALTH Table 3 Age-standardized mean score for healthy eating, adults aged 16 74* Men... X Women... X Age Marital status Single Married Divorced/separated Widowed Household composition Lives alone Lives with others Social class of chief income earner I and II IIINM IIIM IV and V Housing tenure Owner occupier Public sector rented Other rented Individual economic status In employment ILO unemployed Retired Other economically inactive Highest educational/vocational qualification Degree or higher A level/other below degree O level or equivalent Lower than O level Grade C Region Highlands and Islands Grampian and Tayside Lothian and Fife Borders, Dumfries and Galloway Glasgow Lanarkshire, Ayreshire and Arran Forth Valley, Argyll and Clyde Carstairs index Lowest Second lowest Second highest Highest Car ownership Has access to a car No access to a car Smoking behaviour Never smoked Ex-smoker Wants to give up Does not want to give up Not answered Walking activity level Inactive Light (15 min + at slow/steady pace) Moderate (15 min + at brisk/fast pace) Sport activity level Light activity Moderate Vigorous (Continued)

6 WHAT NOT TO EAT: INEQUALITIES IN HEALTHY EATING BEHAVIOUR 41 Table 3 (Continued) Men... X Women... X Housework/gardening Inactive Light Moderate Alcohol consumption Up to weekly unit limit Over weekly unit limit Occupational Activity Level Inactive Light Moderate Vigorous Body mass index Underweight (<20) Normal weight (20 24) Overweight (25 29) Obese (30 39) Morbid obesity (40+) Waist hip ratio 0.8/0.95 or more <0.8/ Mental health GHQ score GHQ score GHQ score Self-assessed general health Very good/good Fair Bad/very bad Limiting long-term illness Limiting Non-limiting None All *The means were weighted for non-response and age-standardized against the 1998 mid year estimate for the Scottish population. For more details see sections 15.8 and 15.9 of The Scottish Health Survey Report 1998, Volume 2. targeted at young people in lower social classes and in deprived areas. Healthy food availability may also be inhibited by lack of access to a car. The Data Food Networking initiative for Europe showed that mean availability of fresh vegetables in the UK had fallen for manual social classes between 1985 and In the late 1990s food deserts were assumed to be a public health concern, 21,22 though the link between access to food stores and diet had not been established. 23 Recent work in east Leeds, England found that more than two thirds of the local population with poor diets increased their consumption of fruit and vegetables, following the opening of a proximate large supermarket. 24 Distances to large towns are greater, especially in northern Scotland, than in England and the rural population is much higher, lack of access to a car may be a particular concern in more remote areas or areas with mainly out of town shopping. The launch of the Healthy Eating Education Campaign in Scotland in 2003 with advice phone lines will hopefully address some of the issues raised above. Health education messages must be targeted appropriately higher educationally qualified men and women had already significantly higher odds of healthy eating. Also poor diet in the manual social classes may not just be a matter of income inequalities; occupational dietary differences may be associated with food availability in and around the workplace and traditional food choices. Female ex-smokers achieved similar odds of healthy eating to the never smoked, whereas those who wanted to give up had odds that were similar to other smokers. Walking and sporting activity were both correlated with healthy eating, though this was reversed in the case of moderate sporting activity which may be an age effect or the effect of higher calorific consumption for sport. Being underweight was a significant risk factor for women. This suggests that the overcontrol of calorific input could have significant negative impact on other aspects of diet. Obesity in men was associated with higher odds of healthy eating; this seemed a surprising result. Possibly the most overweight were controlling their diet as a response to medical advice to improve health; if so, dietary advice could be routed through a health care setting.

7 42 JOURNAL OF PUBLIC HEALTH Table 4 Results from logistic regression: unhealthy eating among men aged 16 74, 1998 p Exp(B) 95.0% CI for Exp(B)... Lower Upper Age group Social class of chief income earner I and I IIINM IIIM IV and V Missing Highest education qualification No qualifications O Grade C or equivalent A Grade or equivalent Degree Missing Carstairs Lowest Second lowest Second highest Highest Missing Smoking behaviour Never smoked Ex-smoker Wants to give up Does not want to give up Not Answered Walking activity Inactive Light (15 mins + at slow/steady pace) Moderate (15 mins + at brisk/fast pace) Sporting activity Light activity Moderate Vigorous Missing x 10 9 Obesity Normal Overweight Obese Morbid obesity Underweight < Missing Constant Log-likelihood = ; Cox and Snell r 2 = 0.160; Nagelkerke r 2 = The principal limitation of this study is that it used crosssectional food frequency data which has well-documented problems including recall error; the score was not validated, but the results seem to suggest the measure is appropriate. Other, more detailed, studies of UK diet such as the National Diet and Nutrition Survey do not provide the sample size to examine risk factors for food intake in Scotland in such detail. This study suggests people in Scotland adopt healthy eating as one component of a healthy lifestyle and that health education campaigns may need to have composite foci as well as specific elements.

8 WHAT NOT TO EAT: INEQUALITIES IN HEALTHY EATING BEHAVIOUR 43 Table 5 Results from logistic regression: healthy eating among women aged 16 74, 1998 p Exp(B) 95.0% CI for Exp(B)... Lower Upper Age group Social class of the chief income earner I and I IIINM IIIM IV and V Missing Highest education qualification No qualifications O Grade C or equivalent A Grade pass or equivalent Degree Missing Carstairs Lowest Second lowest Second highest Highest Missing Car ownership No access to a car Smoking behaviour Never smoked Ex-smoker Wants to give up Does not want to give up Not Answered Walking activity Inactive Light (15 min + at slow/ steady pace) Moderate (15 min + at brisk/fast pace) Sporting activity Light activity Moderate Vigorous Missing Obesity Normal Overweight Obese Morbid obesity Underweight < Missing Constant Log-likelihood = ; Cox and Snell r 2 = 0.160; Nagelkerke r 2 = Acknowledgements Thanks to Dr Paola Primatesta, Senior Lecturer in Epidemiology, Department of Epidemiology and Public Health, University College London for comments on earlier drafts of this paper. The Health Survey for Scotland is funded by the Scottish Executive. References 1 The Scottish Office. Towards a healthier Scotland: a White Paper on health. Edinburgh: The Stationery Office, Department of Health, The Scottish Office. Coronary heart disease in Scotland: Report of policy review. Edinburgh: Public Health Policy Unit, The Scottish Office, Department of Health, 1996.

9 44 JOURNAL OF PUBLIC HEALTH 3 (last accessed 21 October 2003). 4 The Scottish Office. Eating for health: a diet action plan for Scotland. Edinburgh: The Stationery Office, Joffee M, Robertson A. The potential contribution of increased vegetable and fruit consumption to health gain in the European Union. Publ Hlth Nutr 2001; 4: Dong W, Erens B, eds. The Scottish health survey Volume 1. Edinburgh: Stationery Office, Shaw A, McMunn A, Field J, eds. The Scottish health survey Volume 2. Edinburgh: The Stationery Office, Henderson L, Gregory J, Swann G, eds. National diet and nutrition survey: adults aged 19 to 64. Volume 1. London: The Stationery Office, Prätälä RS, Groth MV, Olderstorf US, Roos GR, Sekula W, Tuomainen, HM. Use of butter and cheese in 10 European Countries, a case of contrasting educational differences. Eur J Publ Hlth 2003; 13: Hunt CJ, Nichols RN, Pryer JA. Who complied with national fruit and vegetable population goals? Findings from the dietary and nutritional survey of British adults. Eur J Publ Hlth 2000; 10: Laaksonen M, Prättälä R, Karisto A. Patterns of unhealthy behaviour in Finland Eur J Publ Hlth 2001; 11: Shaw A, McMunn A, Field J, eds. The Scottish health survey Volume 1. Edinburgh: The Stationery Office, Carstairs V, Morris R. Deprivation and health in Scotland. Aberdeen: Aberdeen University Press, Morris R, Carstairs V. Which deprivation? A comparison of selected deprivation indexes. J Publ Hlth Med 1991; 13: Marmot M. A not-so-sensible drinks policy. Lancet 1995; 346: World Health Organisation. Measuring obesity. Classification and description of anthropometric data. Report on a WHO consultation of the epidemiology of obesity. Copenhagen, Denmark: Nutrition Unit, WHO Regional Office for Europe, Kannel W, Cupples L, Ramaswami R, Stokes J, Kreger B, Higgins M. Regional obesity and risk of cardiovascular disease. The Framingham Study. J Clin Epidemiol 1991; 44: Molarius A, Seidell JC. Selection of anthropometric indicators for classification of abdominal fatness. A critical review. Int J Obesity 1998; 22: Goldberg D, Williams PA. Users guide to the General Health Questionnaire. London: NFER-Nelson, Trichopoulou A, Naska A. DAFNE III. Group patterns of unhealthy behaviour in Finland. Eur J Publ Hlth 2003; 13 (3 Suppl): Whitehead M. Food deserts: what s in a name? Hlth Educ J 1998; 57: Acheson D. Independent inquiry into inequalities in health. London: The Stationary Office, Cummins S, MacIntyre S. Food deserts evidence and assumption in health policy making. Br Med J 2002; 325: Wrigley N, Warm D, Margetts B. Deprivation, diet, and food-retail access: findings from the Leeds food deserts study. Environ Plan 2003; 35: