Correspondence to: Mr James C. Doidge, Health Economics and Social Policy Group, University of South Australia City East Campus, GPO Box 2471, Adelaide, SA 5001; e-mail: James.Doidge@UniSA.edu.au
Objective: To describe the pattern of dairy consumption in Australians aged 12 years and over, and assess the extent to which the population meets national recommendations.
Methods: We developed a new method of combining quantitative data from a 24-hour dietary recall questionnaire with semiquantitative data from a food frequency questionnaire (FFQ), to investigate the usual patterns of dairy consumption. We applied this technique to data from the 9,096 Australians aged 12 and over who completed the FFQ part of the most recent nationally representative nutrition study − the 1995 National Nutrition Survey.
Results: When weighted according to the characteristics of the Australian population, 58% of male and 73% of female FFQ respondents failed to regularly meet recommendations for consumption of dairy products. While mean dairy consumption was higher in adolescents, 62% of boys and 83% of girls failed to meet their higher recommendation of three serves per day. Breastfeeding women appeared to consume more dairy but 60% consumed less than two serves per day.
Conclusions and Implications: Given accumulating evidence of protective effects of dairy foods for a range of metabolic and cardiovascular disorders, our observations warrant a focus on the development of cost-effective public health interventions to increase dairy consumption.
In Australia, as in much of the developed world, there is concern about the quality of the diet. Diet-related diseases are now responsible for considerable ill health, resulting in premature death, morbidity and increased costs of healthcare as well as affecting the overall level of economic production.1,2 To reduce both the individual and societal impacts of unhealthy diet, many countries support a range of nutrition interventions.3–5 The information required to identify and develop well-targeted public health interventions are: (i) evidence-based nutrition guidelines (an understanding of optimal diets), (ii) knowledge of dietary patterns of the population (deviations from the optimum), (iii) a means of relating dietary patterns to recommended dietary guidelines (common units), and (iv) knowledge of effective intervention strategies for modifying diet. The Australian Government, through the National Health and Medical Research Council (NHMRC), publishes both broadly phrased food-based dietary guidelines (FBDGs) such as “eat plenty of vegetables, legumes and fruits” as well as a food selection guide (FSG) that provides quantitative recommendations for daily consumption, in standard ‘serve’ units defined for each of five core food groups. These are combined and summarised in the consumer-targeted booklet, Food for Health: Dietary guidelines for Australians − A guide to healthy eating.10
This paper focuses on one food group − dairy products. With their complex nutritional composition6,7 and biologically active components,8 dairy foods can be an integral component of a balanced diet.9,10 There is an increasing body of evidence that connects consumption of dairy foods with improved health outcomes, especially through improved weight control11 and reduction in the risk of metabolic and cardiovascular disorders.12,13 Given the increasing burden of these diseases, it is important to understand the patterns of dairy consumption in the community relative to recommendations. Here, the term ‘dairy’ primarily refers to milk, cheese and yoghurt; the specific dairy foods highlighted in NHMRC publications.
Australian recommendations are that people “include milk, yoghurts, cheeses and/or alternatives” and that those aged four years and older should aim for a minimum of 2–3 ‘serves’ of these per day, depending on age and gender.10 For example, the recommended minimum is three serves for people aged 12–18 years due to their increased requirements for calcium.10 Some publications based on the FSG also suggest that breastfeeding women aim for three serves due to their increased requirements for zinc.9
There is a paucity of information on the existing dietary patterns of Australians,14 and much of this is focused on the adequacy of intake of specific nutrients, rather than whole foods or food groups.15 The 2007 Australian National Children's Nutrition and Physical Activity Survey (ANCNPAS) was the first national survey to attempt to relate patterns of dairy consumption to recommended numbers of daily serves. This was done using calcium intake as a proxy; calculating the proportion of children, who met the estimated average requirement (EAR) for dietary calcium in one 24-hour period, adjusted for within-person variance. Based on dietary calcium intake, dairy consumption was estimated to be adequate in 99% of 2–3 year olds, but declined rapidly with age, to as little as 18% in girls aged 14–16 years.15
The most recent national survey of adult Australians’ diet is the 1995 National Nutrition Survey (NNS).16 When the NNS was conducted, there were no quantitative recommendations for dairy foods, and to date no publication has examined the dairy consumption data from the NNS in standard serve units or related dairy consumption to the more recent FSG. Published descriptions of this data provide only limited statistics (mean, median, etc)17 that are unsuitable for interpretation of skewed distributions with respect to any threshold such as a recommended minimum.
This article presents the findings of an analysis of dairy food consumption in the NNS,18 translating the data into standard serve units in order to relate findings to published recommendations. We demonstrate a new approach to estimating distribution parameters, providing healthcare researchers with quantitative data for use in modelling or for comparison with other studies such as the forthcoming Australian Health Survey. It also provides clinical healthcare workers and policy makers with an insight into the epidemiology of low dairy consumption in Australia.
A full description of the methods for data collection in the NNS has been reported by the Australian Bureau of Statistics (ABS).16 Data was collected from a subsample of 13,858 Australians aged two and over who were respondents to the 1995 National Health Survey (NHS). The NHS surveyed an area-based stratified random sample of 57,633 Australians, with a 97% response rate. From this group, 22,562 were selected to participate in a subsequent face-to-face interview for the NNS, of whom 61% participated. Differences between respondents and non-respondents (based on individual characteristics reported in the NHS) have been noted and discussed by the ABS, with corresponding adjustments made to sample weights.16
The NNS included three components: (i) a 24-hour dietary recall questionnaire which provided quantitative estimates of intake for a wide range of foods and nutrients (grams, milligrams, etc); (ii) a food frequency questionnaire (FFQ) completed by 9,096 (76% of) participants aged 12 or over, providing qualitative and semiquantitative information on the frequency of consumption for 107 foods and 11 dietary supplements, as well as questions about the frequency of several ‘food habits’ over the past 12 months; and (iii) a second 24-hour recall that was completed by approximately 1,500 participants that was used by the ABS to remove the statistical effect of within-person variance from their estimates of the standard error for specific nutrient values, but not for foods or food groups such as dairy. The present analysis uses a different method − combining data from the FFQ with data from the first 24-hour recall − to estimate the actual distribution of dairy consumption by age and gender.
The FFQ included questions on nine types of dairy food, which together account for nearly all the common foods for which milk is a main ingredient. The only apparent exceptions were butter and custard (use of butter as opposed to other spreads or cooking oils was considered separately in the FFQ, as a food habit). For each of the nine types of dairy, participants selected the most appropriate of either ‘not applicable’ or one of nine semiquantitative responses:“never/less than once a month”, “1–3 times per month”, “once per week”, “2–4 times per week”, “5–6 times per week”, “once per day”, “2–3 times per day”, “4–5 times per day”, or “6+ times per day”.
The ABS converted the 24-hour recall data into a set of values detailing the total intake for a range of micro and macronutrients and foods, including ‘milk products and dishes’. More detailed quantitative data on the individual types of dairy food were not provided in the available dataset. When weighted according to the characteristics of the 1995 Australian population, this data provides an estimate of the mean daily dairy consumption across the population as a whole that is theoretically unbiased; however, the distribution would have been affected by the within-person variance. The FFQ, on the other hand, provides less-precise measures of quantity but a relatively precise measure of frequency (because of the narrow categories for response) that is unaffected by daily variance in foods consumed. If the quantity consumed varies proportionally with frequency of consumption (an assumption explored further below), then the distribution of quantity can be approximated by adjusting individual measures of frequency proportional to the ratio of mean quantity to mean frequency (the mean portion size). To do this, a measure of ‘dairy frequency’ was constructed by converting the nine dairy items from the FFQ into quantitative measures of daily frequency and then summing the results. “Not applicable” and “never/less than once a month” were considered equal to zero times per day; for any other range of values, the response was considered equal to the midpoint of the range.
Dairy Australia maintains a database of annual estimates of the per capita consumption of milk, cheese and yoghurt from all sources, including retail sales, food service and industrial or manufacturing channels (thus encompassing some wastage and more processed dairy products).19 Adopting guideline-consistent standard serve sizes of 250 mL, 40 g, and 200 g for milk, cheese and yoghurt, respectively,9 the mean mass per serve for total dairy consumption was estimated for the years 1985/86 to 2009/10 (Figure 1; average mass per serve = total mass of dairy / total number of serves). The per capita consumption of cheese and yoghurt have increased in absolute and relative terms over the past few decades, while consumption of milk has remained relatively constant. This has resulted in a downward trend in the mean mass of a standard serve of dairy foods, from approximately 180 g in 1985 to 160 g in 2009/10. We divided the total mass of dairy foods consumed by each participant in the NNS by the mean mass of a standard serve (170 g in 1995/96) to estimate the number of standard serves of dairy consumed by each participant in the 24-hour recall questionnaire.
Measures of individual dairy frequency in standard serves were then multiplied by the age- and gender-specific ratio of mean dairy quantity to mean dairy frequency. In this way, we produced a distribution of daily dairy intake that had the same shape as the age- and gender-specific distribution of dairy frequency, but with a mean that was equal to the age- and gender-specific mean quantity. By allowing the ratio of mean quantity to mean frequency (i.e. mean portion size) to vary according to both age and gender, the underlying assumption was less restrictive, and simply assumed a proportional relationship within age and gender subgroups of the population. This assumption was supported by linear regression analyses of quantity upon frequency that returned highly significant coefficients (p<0.001) in each stratum. The statistical formula underlying the transformation is presented in the following equation (means are age- and gender-specific):
The relationship between breastfeeding status (“currently breastfeeding a child”, “has breastfed a child within the last three years”, and “not stated”, “not applicable”, “has not breastfed a child within the last three years”) and dairy consumption was also explored in female participants aged between 18 and 49 years; this group included all women who reported either currently or recently breastfeeding.
All estimates were weighted according to characteristics of the 1995 Australian population (weights provided by the ABS), and all analyses were performed using PASW Statistics 18 software (SPSS / IBM Corporation, Somers USA).
Both quantity and frequency of consumption were found to vary according to age and gender. The associations with quantity have previously been reported by the ABS.20 Frequency of dairy consumption was found to increase with both age and male gender up to about age 30, after which it remained fairly constant and similar between genders. The distribution of daily dairy consumption in Australia, by age and gender, is presented in Figure 2.
The results show that when the NNS FFQ respondents are weighted according to the characteristics of the 1995 Australian population, 24% of male and 32% of females consumed less than 1 standard serve of dairy per day, while 55% of male and 71% of females consumed less than two serves per day. Although mean consumption was higher amongst adolescents, most were still not meeting recommendations, especially adolescent girls of whom 83% consumed less than three serves per day (c.f. 62% for adolescent boys). The spike seen in Figure 3 at age 19 is an artefact of the change in recommendation from three to two serves per day.
The relationship between breastfeeding and dairy consumption was obscured by the fact that 88% of women aged 18–49 responded “not applicable”, with only a small sample of breastfeeding women. Among the 2% of the subgroup who were currently breastfeeding, 40% consumed at least two serves per day. Among women who had breastfed within the past three years, 32% consumed at least two serves per day, suggesting that consumption dairy falls after breastfeeding. Of the women who responded “not applicable”, 29% consumed at least two serves per day.
Our analysis indicates that most Australians consume less than the recommended minimum intake of dairy foods. While dairy consumption was somewhat higher in both adolescents and breastfeeding women, most still do not meet recommendations, which is of particular concern given their nutritional requirements.
While the dataset upon which this analysis was based is not recent, the results are consistent with the findings of the more recent 2007 ANCNPAS that 44% of boys aged 14–16 and 82% of girls aged 14–16 did not meet the EAR for calcium intake. Until more pertinent dietary data is collected, the actual situation will remain uncertain.
International comparisons of dairy intake are limited by the nature of published data, with only population means usually published. Mean Australian consumption of dairy products was 289 g NNS20 which, according to the industry-reported milk:cheese:yoghurt ratio for that year, equates to 1.7 standard serves. This is similar to the mean dairy intake for the US population of 1.8 ‘cups’ reported by the United States Department of Agriculture,15 and a bit less than the 366 g (women) and 404 g (men) reported in the United Kingdom.16
As far as we are aware, this analysis provides the only description of the pattern of dairy consumption in Australian adults that allows comparison with recommended daily consumption. The findings should enable an analysis of trends in dairy consumption in recent decades, by comparison with data from the forthcoming national dietary surveys. The methods could also be used in the interpretation of dietary data collected in various cohorts, such as the Women's Health Survey.15
Our approach of combining semiquantitative data from a food frequency questionnaire with quantitative data, such as provided by recall questionnaires or food diaries, provides information that enhances understanding of dietary patterns. With non-normal distributions, such as those found in consumption of dairy and other food groups, estimates of mean, median or variance cannot be used to accurately determine the extent to which the population is meeting recommendations. The data required to implement this analysis can be collected in a single encounter with the participant, which has the potential to reduce costs and increase participation rates.
The main limitation in our application of the approach was the grouping of all dairy foods together that, in effect, assumes that individuals consumed the same relative proportions of milk-, cheese- and yoghurt-based products. This assumption was adopted because of limitations in our quantitative data. While it would have been technically possible to adjust individual quantity estimates for proportional differences in frequency by dairy food, this would have involved a stronger assumption about the proportional relationship between quantity and frequency, specifically that it applies at the individual level, reintroducing an effect of within-person variance, which is undesirable. Future applications of the technique could more precisely estimate dairy consumption by analysing quantitative responses by type of dairy food.
Both in nutrition research and in the associated media, much attention is devoted to Australians’ low consumption of fruit and vegetables, and to the overconsumption of junk foods. Given the growing evidence connecting dairy consumption to prevention of chronic disease,11–13 and the gross departure from guidelines indicated by this and other analyses, there is strong justification for promoting increased dairy consumption in public health campaigns to improve diet in Australia, especially among adolescent populations. This would ideally be supported by the regular collection of data on dietary patterns, including consumption by type of dairy food and population subgroup to monitor the impact of public health nutrition interventions.
This study formed part of a larger analysis that was funded by a grant from Dairy Australia. Dairy Australia provided industry estimates for per capita consumption of milk, cheese and yoghurt from all sources, but otherwise had no role in the development, analysis, or drafting of this article, or in the decision to publish. The authors declare that there were no other personal or financial conflicts of interest. Findings based on use of ABS CURF data.