Obesity is a major risk factor for non-communicable disease including cardiovascular disease, diabetes and some cancers [1
]. Obesity is at pandemic levels, affecting 10–15% of the global population [2
], and up to one-third of adults in developed countries such as Australia and the United States of America (USA) [3
]. Diet, specifically excess energy intake relative to energy expenditure, is a key modifiable cause of obesity [5
It has been a challenge to identify the dietary patterns clearly linked to excess energy intake [6
]. Research on dietary patterns aims to capture the behavioural complexity of food and beverage intake combinations that underpin the associations between diet and health [7
]. While significant associations between dietary patterns and weight status are observed, these findings are weak and inconsistent and warrant further investigation [6
Dietary indexes are a composite indicator of diet quality, where adherence to an a priori
set of components or recommendations is reflected in a single score [8
]. Indexes can incorporate dietary quality, diversity, adequacy, moderation and balance [10
]. Some indexes are nutrient-based, others food-based, or a combination of both [12
]. In the context of obesity, diet quality could be defined as the degree to which a dietary pattern reduces the risk of positive energy balance. In developed countries, diet quality tools measuring adherence to national dietary guidelines are consistently inversely associated with obesity [9
]. For example in the USA, the Healthy Eating Index is inversely associated with weight status in 10 of 13 studies examined in a recent systematic review [9
]. Similarly in Australian research, dietary guideline adherence measured using the Dietary Guideline Index (DGI) has shown similar associations [13
]. Conversely, diet quality conceptualised as dietary diversity has been positively associated with risk of obesity [9
]. A priori components selected for inclusion in a diet quality score is likely to influence the utility of diet quality measures in obesity research.
In 2016, Livingstone and McNaughton compared the association between two food-based diet quality scores, the DGI and the Recommended Food Score (RFS), and obesity [15
]. The DGI includes components reflecting adherence to recommendations for core food groups—‘healthy foods’—and discretionary choices—‘unhealthy food and beverages’. In contrast, the RFS conceptualises diet quality as variety of ‘healthy’ or core foods only. When the DGI and RFS scores were applied to nationally representative food intake data of Australian adults, only the DGI score was associated with lower risk of overall and central obesity [15
]. The authors concluded that inclusion of both healthy and unhealthy components appears to be important in conceptualising diet quality as a risk factor for obesity. Whether similar findings can be observed for other components of diet quality, for example within the core food groups, remains unexplored. Given the multifaceted nature of diet quality indexes, there remains unanswered questions around whether particular aspects of diet quality or compliance with guidelines differs between weight status groups. This warrants further investigation.
The DGI reflects compliance with the 2013 Australian Dietary Guidelines and has been applied to food intake data measured via 24 h recalls [15
], food frequency questionnaires [16
] and a short food survey [17
]. Regardless of dietary assessment method, the DGI is significantly associated with weight status in large population surveys [15
]. The Commonwealth Scientific and Industrial Research Organisation (CSIRO) Healthy Diet Score survey is a freely available online survey designed to assess diet quality using the DGI scoring approach. The survey allows individuals to enter their food intake and receive immediate feedback in the form of a numerical diet quality score as well as three brief statements on how to improve their score [17
]. The CSIRO Healthy Diet Score survey enables the examination of population diet quality in the context of the provision of individualised feedback to improve diet quality. In this paper, we use data from this online survey to explore whether compliance with particular dietary guidelines varies by weight status, and to identify key food groups to target to improve diet quality for different weight status groups.
This paper aimed to assess the relationship between the CSIRO Healthy Diet Score and weight status and determine whether it may have utility in framing dietary recommendations for the prevention and management of overweight and obesity. Access to this large dataset has allowed for detailed subgroup analysis, increasing our capacity to understand the specific components of diet quality which differ by individual characteristics, in this case gender and weight status. Given the statistical power of the sample we used cut offs in addition to statistical significance to moderate our interpretation of results, and focus the discussion on results with medium to large effect sizes.
In this large sample of Australian adults who have completed the online CSIRO Healthy Diet Score survey we found that low compliance with dietary guidelines was associated with an almost 3-fold higher likelihood of being obese. We also found that the total score per se may not be as important as the dietary pattern within the score. For example, there was a small difference in overall diet quality score between obese and normal weight individuals, however, moderate to large differences in component scores by weight status category were observed for discretionary foods, fruit and healthy fats. Discretionary foods are an important intervention target to improve diet quality, regardless of weight status or gender.
The relationship between diet quality and weight status has shown mixed results [6
], and the definition of diet quality appears to be important [9
]. In this study, an online diet quality score was used that reflected adherence to the 2013 Australian Dietary Guidelines which was applied to food intake data collection using a short food survey. A key finding was the apparent stepwise increase in likelihood of being classified as overweight or obese with decreasing compliance with dietary guidelines. The findings are consistent with other dietary pattern research that has used the same scoring approach, the Dietary Guideline Index, applied to food intake data derived from a nationally representative sample using 24-h recalls [15
] and a large sample of older adults using a Food Frequency Questionnaire [16
]. The present study reinforces that diet quality, conceptualized as compliance with guidelines, is a relevant intervention target for obesity prevention and management, and that specific elements of the score could be used to provide personalised feedback to individuals.
It is valuable to understand if specific components of dietary guideline adherence are of particular importance in the context of the overweight and obesity [23
]. In this study, components of diet quality that contrasted most by weight status were fruit, discretionary foods, and healthy fats. For men, vegetables, grains and variety component scores also showed moderate differences by weights status groups—all favouring greater compliance with guidelines in normal weight individuals. For women, there were small differences in compliance with the guidelines for dairy but further analysis showed this was due to the fat type of dairy products consumed and not the amount. Interestingly, dairy and meat were the only two food groups for which obese individuals were more compliant with guidelines. While these differences were considered to be small they are worth further investigation. For example, these may reflect true differences in intake but may also reflect the construction of the index. The scoring algorithm does not penalize for overconsumption of healthy foods. Scoring the idea of eating beyond one’s needs, could be explored by applying a bell-shaped scoring system whereby at a threshold of overconsumption scores start to reduce. However, this threshold may need to be food group specific and not follow a general rule, given there is evidence that overconsumption of some food groups, such as meat, may be more detrimental to health [24
] than overconsumption of other food groups such as vegetables.
The finding that discretionary foods—also termed unhealthy foods—are an important component of diet quality in the context of obesity are consistent with recent analysis comparing diet quality indexes that are comprised of only healthy food based components compared with those comprised of healthy and unhealthy food based components [15
]. The Recommended Food Score is based on consumption of five healthy food groups, and shows no association with obesity risk [15
]; likewise dietary diversity indexes are also not associated with obesity [25
]. The Australian Dietary Guidelines make population level recommendations for appropriate types and portions of foods to consume for health and wellbeing [19
]. In contrast to the US Dietary Guidelines, the Australian guidelines avoid explicitly linking food intake recommendations to set energy requirements. Rather there is a food-based recommendation to limit discretionary foods that are higher in saturated fat, added sugar, alcohol and/or sodium to “sometimes and in small amounts” [19
]. This focus on nutrient-poor foods appears to sufficiently address consumption of energy dense foods and prevention of positive energy balance [23
It is generally accepted that reduction in energy intake is a necessary focus in nutrition interventions targeting obesity. However there is less consensus around specific food group-based strategies required to achieve a moderation in energy intake, and whether these targets are consistent across all subgroups within the population. To improve diet quality, this study suggests that obese individuals need to increase their consumption of fruit, and choose behaviours that improve the quality of grains and dietary fats, that is choosing wholegrains, unsaturated spreads and trimming meat. For men, increasing vegetable consumption and including a wide variety of core foods were also identified as key differentiating factors in diet quality between normal weight and obese adults. Fruit and vegetables initiatives are common targets for obesity prevention and important targets given their low energy density, and association with reduced disease risk [26
]. However, other less explored areas of diet quality such as healthy fats, wholegrains and variety may be additional targets to consider and provide a more nuanced approach to population nutrition and obesity prevention programs. However, in the context of obesity prevention, dietary factors should be considered together with non-dietary factors such as physical activity, as their effects alone on body weight may be small over the longer term. It has proven difficult to separate out the effects of diet from physical activity in weight gain because few nutrition epidemiology studies adequately control for physical activity [29
]. Therefore to better understand relationships between dietary components and weight status a broader range of covariates need to be consider such as age and physical activity. Delivering the survey online has resulted in large volumes of data being collected in a time and cost effective way. Exploration of this dataset has identified elements of diet quality associated with increased risk of obesity, which can now be used to inform interventions that address its management and prevention at the population level. The survey platform currently provides brief and immediate feedback to individuals, and the results of this analysis can be used to further refine this feedback to improve the overall diet quality of the population. The opportunity “big data” provides in terms of tailoring feedback is an emerging area of public health research, and moves towards the concept of quantified population health [30
]. The online environment also helps to accelerate the progress in intervention development, as it speed up the temporal lag associated with traditional data collection and dissemination cycles [30
Limitations of our research are that due to the nature of recruitment, being an online food intake survey, males, older adults and obese individuals were underrepresented relative to the Australian population. However, these were partially accounted for in our adjustment for age and gender. While extreme values were removed, self-reported height and weight could have led to an inaccurate weight status classification. The ability to self-report anthropometric data may vary systematically by demographic characteristics such as age, gender and socioeconomic status [31
]. Misreporting of food intake can also vary by individual characteristics such as weight status. Underreporting of intake is more frequent and of a greater magnitude in obese individuals [32
], and more likely to be due to underreporting of discretionary foods, which could have led to inaccurate diet quality scores as well.
Another limitation of this study was its cross sectional design, which limits any inference of causal relationships. Therefore longitudinal data are needed to determine if diet quality predicts risk of obesity and/or energy imbalance, and whether changing intake of particular foods groups to be more consistent with guidelines reduces risk of future obesity. In addition, this study did not assess levels of physical activity in relation to diet quality and obesity status. Therefore, future research should also consider energy expenditure, as the other important element of energy balance, as well as other potential confounding factors, including individual and environmental level factors such as socio-economic status, physical activity and accessibility to healthy foods. Regardless, this study adds to the evidence base that diet quality is associated with health outcomes, including weight status.