Demographic variation in nutrition knowledge in England

This paper describes a nutrition knowledge survey carried out on a cross-section of the adult population of England (n = 1040), looking at knowledge relating to current dietary recommendations, sources of nutrients, healthy food choices and diet–disease links. Serious gaps in knowledge about even the basic recommendations were discovered, and there was much confusion over the relationship between diet and disease. Significant differences in knowledge between socio-demographic groups were found, with men having poorer knowledge than women, and knowledge declining with lower educational level and socio-economic status. Possible reasons for these differences and implications for public education campaigns and socio-economic inequalities in health are discussed.

Introduction ents mentioned as many as three out of four of the core recommendations (to eat more fruit, veget-In the richer countries of the world at the end of the 20th century, consumers may be literally ables and salad; to cut down on fat; to eat more fibre; to eat more starchy carbohydrate). Although 'spoiled for choice' in the food domain. They must select their foods from hundreds or even thousands these were the most common answers, they were given by only 67, 45, 27 and 20% of respondents, of products, many of which are designed and marketed to maximize their appeal to the consumer.
respectively. Over two-thirds of respondents (68%) endorsed the statement that 'experts never agree To select a healthy diet, they must be able to ignore the advertisers' blandishments and the immediate about what foods are good for you', demonstrating that clear messages about healthy eating are not appeal to the palate, and draw on a complex being conveyed to the general public. The ignorance about nutritional advice was mirrored by little elsewhere, e.g. (de Graaf et al., 1997)], two-thirds education and socio-economic status (SES). These findings are in keeping with other research. In the either did not want to change or wanted to but did not think that they would. US, Levy et al. (Levy et al., 1993) found that levels of knowledge about fat and cholesterol were Studies in the UK looking at other aspects of nutrition knowledge have had equally concerning highest for more educated people, people of middle years and White people. In a study in Australia, findings. Buttriss (Buttriss, 1997) found that people were poor at identifying foods containing starch Crawford and Baghurst (Crawford and Baghurst, 1990) found that knowledge about the links and even worse at knowing which foods contained fibre, with 35% of the sample failing to correctly between disease and fat, sodium and sugar was generally better for women, people of higher SES categorize as many as half of the foods presented as high or low in fibre. Tate and Cade (Tate and and older people, and Tate and Cade (Tate and Cade, 1990) reported similar results from a UK Cade, 1990) reported positive findings regarding knowledge of dietary fat and coronary heart dis-sample. Buttriss (Buttriss, 1997), also in the UK, found significant gender and class differences in ease. However, there were areas of knowledge which were extremely poor (e.g. 84% of people ability to identify foods containing starch and fibre, with women performing better than men, and believed that sunflower margarine contains less fat than butter). Although the majority of items were middle and upper-middle class respondents better than working class respondents. answered correctly by at least 70% of people, this still means that over a quarter of respondents were To summarize, then, studies have tended to find women more knowledgeable than men (Crawford incorrect. Research in other countries has reported similar and Baghurst, 1990;Tate and Cade, 1990;Buttriss, 1997;Hansbro et al., 1997), knowledge to increase results. Analysing surveys conducted in the USA in 1983, 1986and 1988 with higher SES (Tate and Cade, 1990;Buttriss, 1997;Hansbro et al., 1997) and education (Cremer 1993) found that although nutrition knowledge relating to fat and cholesterol was improving, it and Kessler, 1992;Levy et al., 1993;Hansbro et al., 1997), and generally to be better in middle was still unacceptably low, with no more than 60% of respondents correctly answering any individual aged than older or young groups (Anderson et al., 1988;Tate and Cade, 1990;Levy et al., 1993;item. Patterson et al. (Patterson et al., 1995) found that 23% of Americans were not aware of any of Hansbro et al., 1997), although the patterns found across age groups have not always been the same. the National Cancer Institute's dietary recommendations, and 36% believed links between diet Existing studies of nutrition knowledge in the UK have tended to focus on one particular aspect and cancer to be weak or non-existent. Cotugna et al. (Cotugna et al., 1992), in a larger US survey, of nutrition, e.g. fat, fibre, etc., rather than taking a broad perspective. There has also been a tendency found that 52% of those asked about diet-disease links did not mention cancer unprompted, and 27% to use ad hoc measures with little attention to issues of reliability and validity. This has limited still did not acknowledge any link between diet and cancer when asked specifically about it. True our understanding of the state of nutrition knowledge in the UK. levels of knowledge may even be lower than those reported, as survey response rates tend to be around Most of the studies looking at demographic differences in nutrition knowledge and dietary 70% and it is probable that non-responders would have lower knowledge than responders.
behaviour have used occupation as a measure of SES and have not taken into account differences People with poorer nutrition knowledge may fall into particular socio-demographic groups. In in education. Given that SES is a complex construct including economic status (income), social status the HEMS (Hansbro et al., 1997) mentioned above, women had better knowledge than men, and know-(education) and work status (occupation) (Adler et al., 1994), it is important to try to establish the ledge tended to be better at higher levels of contribution of different socio-economic variables with eating too much or too little of various types of food. to the variation in knowledge.
Demographic questions covered age, gender, The aim of the present study was to examine ethnic origin, work status, occupation and partner's nutrition knowledge and demographic variations occupation, level of education, marital status, numin knowledge, in a wide cross-section of the adult ber of children, and children under 18 living at population of England. Knowledge was assessed home. Questions were also asked about specialist using the recently developed and validated Nutrinutritional training and specific dietary requiretion Knowledge Questionnaire (Parmenter and ments. Wardle, 1999). This instrument covers knowledge relating to current dietary recommendations, Procedure sources of different nutrients, everyday food cho-Each participant was written to personally in a ices and diet-disease links. Demographic characletter from their GP asking them to help with a teristics were assessed in the same instrument with study which was described as looking at 'people's simple questions.
understanding of nutrition advice'. The questionnaire was enclosed, together with a postage paid

Subjects and methods
envelope for returning it. A second letter and another questionnaire was sent to anyone who had Subjects not returned the questionnaire after 2 weeks, in an As the majority of the people in England are attempt to maximize the response rate (Dillman, registered with a general practitioner (GP), GP 1991). practices were recruited to participate in the study in the hope of reaching a wide cross-section of the Results population. Three practices were used, one each in Essex, Lancashire and Oxfordshire, thereby Sample giving cluster samples from very different areas A response rate of 73.6% (n ϭ 1040) was achieved, of the country. Five hundred patients aged 18-75 comprising 43.8% men and 56.2% women. were selected at random from each of the patient The mean age of respondents was 51.5 years lists (750 men and 750 women in total) and and the majority of respondents were White. Sociocontacted by post with a request to take part in economic status followed a normal distribution, the survey.
although a large number of people could not be classified on the basis of the occupational Materials information which they supplied. The level of Nutrition knowledge was assessed using the Nutrieducation followed a straight line, with numbers tion Knowledge Questionnaire, the development decreasing as educational level increased. The of which is described elsewhere (Parmenter and demographic characteristics of the sample are Wardle, 1999). The questionnaire has four sections shown in Table I. covering (1) experts' recommendations regarding Comparison with the demographic characterincreasing and decreasing intake of different food istics of the UK population from the 1991 Census groups, (2) nutrient knowledge, (3) food choice (Office of Population Censuses and Surveys, 1991) (which asks people to choose between different showed that the sample was biased in favour options, e.g. to pick the snack which is low in fat of women, people of high SES and educational and high in fibre), and (4) the relationships between qualifications, White people, and older age-groups. diet and disease. This last section looks at beliefs These differences, typical of postal surveys of this about which foods can cause particular diseases, sort, obviously limit the degree to which results can be generalized to the English population as a as well as knowledge of any diseases associated groups, the mean score was 45.6 (SD 11.63). When asked to categorize various foods as either high or low in sugar, fat, starch, salt, protein, fibre or whole and the implications of this will be discussed saturates, mistakes were generally made by just in more detail later. However, the sample was under a third of respondents. There were, however, sufficiently large to allow trends in knowledge a few items which stood out as being particularly across the demographic characteristics to be clearly poorly answered. Eighty-five percent of people identified.
failed to realize that low-fat spread is actually high

Nutrition knowledge
in fat, with the majority of respondents believing it to be a low-fat food. Only just over half of Figure 1 shows the mean percentages of correct responses for all sections and the whole ques-respondents knew that nuts are low in starch and fewer than half realized that cheese is high in salt. tionnaire.
The section on fibre was generally well who were aware of the fat-disease link, over 90% also knew about the link between saturated fat and answered. This represents a substantial improvement on the findings of Buttriss (Buttriss, 1997), heart disease. As regards fruit and vegetables, well over a mentioned above, and Cremer and Kessler (Cremer and Kessler, 1992) who found people's knowledge third of respondents (41%) were unaware of a link between low intake and health problems. Only about which foods contained fibre to be significantly lower than equivalent knowledge about fat.
42% correctly thought that eating more fruit and vegetables can help reduce the risk of cancer, and People were generally better at identifying foods which are high in saturated fat than those which 47% knew that it could also reduce the chances of heart disease. are low in it. There was much confusion about whether foods could be high in fat but contain no Approximately two-thirds of respondents (62.1%) knew of health risks associated with a cholesterol (agree ϭ 27%, disagree ϭ 29%, not sure ϭ 44%). Over 70% of respondents either low fibre intake, with the majority of these people being aware of the specific risk of cancer. incorrectly believed that margarine contained less fat than butter or were unsure. Knowledge about Over 60% of people were also aware of links between sugar and salt intake and disease; 84% monounsaturated fat was also poor, with fewer than a quarter of people knowing that olive oil knew of the link between a high salt diet and heart disease. contains mostly this type of fat. Finally, people were confused about which food types contain When asked to specify diseases which were linked with different food types, respondents' most energy. Almost equal numbers believed it was fatty and sugary foods (33 and 35%, respect-answers varied enormously. For fruit and vegetables, the most commonly mentioned disorders ively), with 22% being unsure.
were scurvy and bowel problems, with answers Section 3-everyday food choices ranging from varicose veins to Beriberi. Bowel problems were also associated with insufficient Out of a maximum of 10 points on this section, the mean score was 7.4 (SD 1.83). Responses to the fibre by many people, with the most commonly mentioned disorders being bowel problems/cancer different items varied widely, seeming to depend largely on the distracter items. The most mistakes and constipation. Most people thought that sugar could cause diabetes and obesity, but only about a were made on the question which asked people to pick a low-fat, high-fibre snack. Only 36% chose quarter mentioned tooth decay. High blood pressure and heart disease were associated with excessive the correct answer (raisins), with almost half (47%) endorsing the muesli bar distracter option. As many salt intake by 57 and 43% of the respondents who answered the item. Finally, 81% of people as a third of respondents were unable to select the best choice for a low-fat cheese and approximately mentioned heart problems as being associated with high fat intake, with overweight/obesity being the 30% of people did not know that thick cut chips are 'healthier' than thin or crinkle cut ones.
second most popular response. The poorest scores in this section concerned Section 4-diet-disease relationships antioxidant vitamins with only 22% of respondents having heard of them. When these people were In this section participants were asked first whether they knew of any links between eating more or asked to say whether a particular vitamin was an antioxidant, less than half of them gave the correct less of particular foods and major health problems. The mean score was 7.35 (SD 3.41) out of a answer on any item. possible 20 points. The highest proportion of

Demographic differences in knowledge
people were aware of a relationship between high fat intake and disease, but almost 15% of people For the purposes of analysis, age was divided into five groups and educational level into four. Marital still did not know about this link. Of the people status was categorized as single, married or living group scored lower than people in middle years, with those aged over 65 obtaining the lowest as married, and separated, divorced or widowed. Ethnic origin was not used in the analyses as the scores. One way analysis of variance showed these differences to be significant (F[4,1038] ϭ 16.2, vast majority of respondents were White. Fewer than 6% of people had nutrition-related qualifica-P Ͻ 0.001 for total score). Respondents who were married or living as tions and fewer than 10% were on special diets, so these factors were also omitted from analyses. married achieved slightly higher scores (mean percentage correct ϭ 63.1%) than those who were Univariate analyses either single (58.8%) or separated, divorced or As predicted, women scored slightly, but significwidowed (58.9%). One-way analysis of variance antly higher than men on the knowledge questionshowed the differences between these groups to naire as a whole (t[d.f. ϭ 1037] ϭ 4.86, P Ͻ be significant (F[2, 1036] ϭ 7.5, P Ͻ 0.001). 0.001) and on each of the sections individually.
Respondents with children living at home scored The mean percentage of correct responses for men significantly higher (64.2%) than those without and women for each section are shown in Figure 1.
(61.3%) (t[d.f. ϭ 1038] ϭ 2.85, P Ͻ 0.01). Figure 2 shows that there was a linear relation-Multivariate analyses ship between knowledge and education level, with scores being lowest for people with no formal In order to establish whether education, gender, SES, marital status and the presence of children at qualifications while those with degrees scored highest (F[4,1009] ϭ 49.1, P Ͻ 0.001 for total home were all having separate effects on knowledge score, these variables were entered into score).
A similar pattern was found for social class a multiple regression model. Marital status was recoded to married/cohabiting versus not married. defined on the basis of occupation (see Figure 3) with total scores being lowest for those in class V, The results showed that gender, level of education and occupational social class all had significant rising progressively to class I (F[5,810] ϭ 16.2, P Ͻ 0.001). It should be noted that there was a independent effects at the 0.01 level, and the effect of marital status was significant at the 0.05 level large number of missing values for SES (where occupational information was insufficient for clas-(see Table II). Together these variables accounted for 22% of the variance in knowledge score. sification), so only about 77% of cases could be used for analyses including this variable.
The curvilinear relationship between age and knowledge score meant that it was not appropriate As illustrated by Figure 4, the youngest age   to enter age into the linear regression model. The There were, however, different patterns for educational level and SES across age, with level of distribution of respondents across age groups was very similar for men and women, indicating that education tending to decrease with increasing age and SES being higher for people of middle years its effect was not confounded with that of gender. than for the oldest or youngest groups. In order our results over-estimate the level of knowledge in England as a whole. to try to establish whether age was having an independent effect on knowledge score, the sample The results are partly encouraging, with many respondents being aware of most of the major was stratified by education and SES. The largest educational subgroup was those people with no guidelines on healthy eating. The results look better than those obtained in the HEMS (1997), which educational qualifications (n ϭ 431). If variation of score with age in this subgroup mirrored that could be attributable to the format of the questionnaire. People seem to be poor at spontaneously of the whole sample, it was likely that age was having an independent effect. This was indeed generating guidelines for healthy eating, as they were required to do in the HEMS (1997), but the case, with significant differences between age groups for total score (F[4,431] ϭ 7.48, P Ͻ .001) when asked about specific recommendations, they generally appear to know whether they should be and scores on the individual sections (all P Ͻ 0.05). The pattern was the same for the group with eating more or less of particular types of food.
The main exception to this was carbohydrate, O levels (n ϭ 281) (F[4,280] ϭ 8.49, P Ͻ 0.001), but the effect disappeared for the A level and indicating that more effort is needed to raise awareness of the importance of increasing intake degree groups, indicating that age might only have an effect on knowledge among people with lower of starchy foods. The question remains whether spontaneous recall or recognition would be most levels of education. It should, however, be noted that the numbers of respondents in these two influential in everyday life.
On the second section of the questionnaire, the groups were smaller (n ϭ 166 and 133, respectively), reducing the power of the ANOVA.
greatest confusion in categorizing foods as high or low in different nutrients concerned low-fat The same analysis was carried out stratifying by SES and the age pattern persisted in each SES spreads. Many people made the mistake of classifying these as low in fat. This could be because group. The effect was significant for SES group II, which was the largest (F[4, 235] ϭ 3.74, people genuinely believe these spreads to be low in fat, but it also seems possible that when quickly P ϭ 0.006).
looking down a list of foods, and ticking boxes, it might be almost automatic when faced with 'low-

Discussion
fat spreads' to tick the box marked 'low fat'. Confusion may arise because people know that The results of this survey give a clear and detailed picture of a broad range of nutrition knowledge in these foods are lower in fat than, for example, butter. This could be clarified by asking people to a large sample of the English general public. The response rate (over 70%) was excellent for a mailed classify the spreads into a particular food group rather than just saying whether they were high or survey of this kind. Nevertheless, we have to be cautious in extrapolating beyond this respondent low in fat. Generally people performed fairly well on the sample. As mentioned earlier, our sample, though large, was not wholly representative of the general food choice section indicating that they can translate their knowledge into actual choices. The most population. It was biased in favour of women, older people, high SES and education, and White common mistake was choosing a muesli bar as a low-fat, high-fibre snack. This might be attributable people, probably reflecting differential response rates by different demographic groups. Previous to marketing, presenting an image of muesli bars as a 'healthy' alternative to more fattening snacks. research (e.g. Anderson et al., 1988;Tate and Cade, 1990;Cremer and Kessler, 1992;Buttriss, The pervasiveness of the error could indicate that advertising is used by many people as a source of Hansbro et al., 1997) has shown that these groups tend to have better nutrition knowledge nutrient information and that people do not actually read the nutritional information labels on foods to than the general population, so it is probable that find out their fat or fibre content. This would tie more men are cooking for themselves and fewer are relying on women to make decisions about in with the HEMS (1997) research which found that large numbers of people never look at the their diets, this is not accompanied by an increase in nutrition knowledge, highlighting the need to ingredients of foods when shopping. More research is needed to clarify the effects of marketing on target men in education campaigns. Articles relating to diet are still very much the domain of beliefs about nutrition.
Knowledge about diet and disease was poor. women's magazines, something which must change if men are to learn more about eating healthily. Most people were inclined to believe in links between all sorts of foods and a huge variety of More research would be needed to identify effective media for conveying messages about healthy diseases. The only really well-known relationships seemed to be between high fat and salt intake and eating to men. The second major demographic trend involved cardiovascular disease, although approximately one in five people were still unaware of these level of education, with more educated people demonstrating significantly better nutrition know-links. Other than that, it seems that people are very confused about the effects of different foods on ledge. This may be because education incorporates the very information that is included in this survey, their health. This is in keeping with other research. For example, Anderson et al. (Anderson et al., but as many nutritional recommendations are relatively recent (e.g. eating five portions of fruit and 1998) found people in England and Scotland had poor knowledge about links between fruit and vegetables every day), yet are still better known by more educated respondents, this cannot be the vegetables and cancer. Similarly, Krebs-Smith et al. (Krebs-Smith et al., 1995) found that only 40% of whole explanation. People who are better educated may also be better able to make use of written adults in the US agreed with the statement that eating fruit and vegetables prevents cancer. The material like newspaper articles and leaflets, to gain information and implement it in their lifestyles. It worst-answered items in this section were those relating to antioxidant vitamins. Perhaps not sur-also seems likely that more educated people would be better able to understand the sometimes complex prisingly, given the recency of scientific findings relating to them, only about one in five people had information about diet-disease links. SES was also shown to have an effect independent of educational heard of them and very few knew which vitamins were classified as antioxidant.
qualifications, suggesting that aspects of the social and cultural milieu might modify exposure to This study had the advantage of a substantial sample and so although it was not entirely repres-nutritional advice. Much nutrition knowledge (e.g. the need to entative of the general population, groups were large enough so that multivariate analyses could reduce fat intake) is now assumed to be very widely held. This has the effect that many news-be used to establish whether the demographic variables exerted independent effects. Women paper articles, television programmes and other forms of popular journalism assume a certain level demonstrated superior knowledge of all areas of nutrition, as has been found in most studies looking of knowledge which, though common to most people, appears not to be reached by everyone. at nutrition knowledge. The gender difference does not seem to have diminished significantly over the This might be a particular problem for less educated or low SES people, making some sources of last 10 years. With the decline in the number of traditional family units where the husband earns nutrition information completely inaccessible due to lack of background knowledge and inability to and the wife is responsible for shopping and cooking, and the rise in the number of people deal with the new information. If this is the case, ways need to be found of targeting basic living alone (Bridgwood and Savage, 1993), it is increasingly important for men to know how to information at certain groups of people to bring their level of knowledge up to that of the rest of the eat healthily. So far, it appears that even though population. Once again, more research is needed to although they also ate more buns, cakes and chocolate, suggesting that knowledge differences find out which methods are effective for communicating to different groups.
between SES groups are also mirrored by dietary variations. The variation in knowledge with age was less clear-cut. Different studies have found different Other studies have looked more generally at lifestyle patterns, including diet, and have found patterns, but it is often true that the middle-aged group perform best, as was found in the present similar demographic variations. Whichelow and Prevost (1996) grouped dietary behaviour into four study. The poor scores of the oldest group probably reflect the fact that the current dietary recommenda-components, one of which correlated with high intakes of fruit and vegetables, high-fibre foods tions are relatively recent and older people probably have more established views on food. In the first and low-fat spreads and milk. This component (the one which most closely meets current dietary half of the century when these people were growing up, the government was encouraging people to eat recommendations) was favoured by those of middle years rather than the very old or very young. It high fat, high sugar diets (Cannon, 1992), so it seems understandable that older people are less was also linearly associated with SES, being most popular with the professional group. When the receptive to new guidelines which are directly contrary to this. Low scores in the youngest group sample was divided into manual and non-manual groups, women from each group scored higher on might be indicative of a lack of interest in health care issues. One would expect that as people reach this component than men of the same SES. This patterning has striking similarities with the vari-middle age they become increasingly aware of diseases related to diet as they or their peers are ations in knowledge found in the present study. In a follow-up study of the same sample 7 years affected by them and information about nutrition becomes more salient. This increased knowledge later (Prevost and Whichelow, 1996), there was a significant increase in scores for the healthy diet might also be associated with having children and seeking out dietary information to ensure that the component. However, non-manual responders showed significantly greater health-related children are eating healthily. However, to gain a true understanding of the relationship between age improvements than those with manual occupations, indicating that differences in diet across socio-and knowledge, it would be necessary to conduct longitudinal research to differentiate between economic groups are persisting and even widening. Knowledge is not, of course, the only factor cohort effects and changes in knowledge related to life stages. underlying dietary behaviour. Other important influences include the quality and freshness of the These demographic patterns have some parallels in dietary behaviour. The Dietary and Nutritional food, taste, price and family preferences (Lennernäs et al., 1997). The relative importance of Survey of British Adults (Gregory et al., 1990) found that women were more likely to eat healthy these factors seems to vary with demographic characteristics. In their pan-European survey, Len-foods like wholemeal bread, fruit, vegetables and reduced-fat milk, although they also ate more nernäs et al. (Lennernäs et al., 1997) found that men tend to rate taste as more important than confectionery. Men, on the other hand, reported eating more sausages, meat pies and chips. This eating healthily. Eating healthily is regarded as a priority by people with tertiary education, whereas indicates that poorer nutrition knowledge in men could be being translated into less healthy eating price is more salient for those with only primary education. Price, taste and habit were identified habits.
In further analysis of the survey data (MAFF, as important barriers to change. However, these influences may to some degree be underpinned by 1994), men and women of higher SES reported eating more fruit, fruit juice, vegetables, salad, knowledge. Dowler and Calvert (1995) found that among lone-parents on very low incomes, those polyunsaturated margarine, oily fish and shell fish, who looked for fresh food and aimed to give their privileged groups seem to have more access. By targeting education at those who particularly need children a healthy diet consistently achieved a healthier diet (albeit within their financial con-it we can help to reduce the divides which lead to the perpetuation of inequalities in health. Know-straints) than those who did not. This indicates that nutrition knowledge could be important even ledge was identified by Link and Phelan (Link and Phelan, 1996) as one of the 'fundamental social when other barriers and constraints are present.
The implications of this research are two-fold. causes' of differences in health, so by empowering people with the knowledge to make appropriate Firstly, the indications are that an unacceptably large number of people are still unaware of the dietary decisions, we can take a step towards ending this pattern of social inequality. main dietary recommendations, as well as having poor knowledge about the sources of nutrients