Hostname: page-component-586b7cd67f-t8hqh Total loading time: 0 Render date: 2024-11-26T00:08:15.109Z Has data issue: false hasContentIssue false

Energy misreporting is more prevalent for those of lower socio-economic status and is associated with lower reported intake of discretionary foods

Published online by Cambridge University Press:  18 September 2020

Amanda Grech*
Affiliation:
The Charles Perkins Centre, The School of Life and Environmental Sciences, Faculty of Science, The University of Sydney, NSW2006, Australia
Megan Hasick
Affiliation:
The Charles Perkins Centre, The School of Life and Environmental Sciences, Faculty of Science, The University of Sydney, NSW2006, Australia
Luke Gemming
Affiliation:
The Charles Perkins Centre, The School of Life and Environmental Sciences, Faculty of Science, The University of Sydney, NSW2006, Australia
Anna Rangan
Affiliation:
The Charles Perkins Centre, The School of Life and Environmental Sciences, Faculty of Science, The University of Sydney, NSW2006, Australia
*
*Corresponding author: Amanda Grech, email [email protected]
Rights & Permissions [Opens in a new window]

Abstract

The role of socio-economic status (SES) on the misreporting of food and energy intakes is not well understood with disagreement in the literature. The aim of this study was to examine the associations between low energy reporting, dietary quality and SES in a representative sample of adults. Dietary data were collected using 2 d of 24-h recalls for 6114 adults aged 19 years and over, participating in the Australian National Nutrition and Physical Activity Survey 2011–2012. Low energy reporters (LER) and plausible reporters (PR) were identified. Discretionary food intake was used as a proxy indicator of diet quality. SES was determined using area-level SES and educational attainment. Regression analysis was applied to examine the effects of LER and SES on diet quality, adjusting for potential confounders. LER was more common in populations of lower SES than higher SES (area-level OR 1·46 (95 % CI 1·06, 2·00); education OR 1·64 (95 % CI 1·28, 2·09). LER and SES were independently associated with diet quality, with LER reporting lower percentage energy from discretionary foods compared with PR (27·4 v. 34·2, P < 0·001), and those of lower area-level SES and education reporting lower diet quality compared with those of higher SES (33·7 v. 31·2, P < 0·001; and 33·5 v. 29·6, P < 0·001, respectively). No interaction effect was found between LER and SES, indicating percentage energy in discretionary foods was not differentially misreported across the SES areas (0·3078) or education (P = 0·7078). In conclusion, LER and higher SES were associated with better diet quality.

Type
Full Papers
Copyright
© The Author(s), 2020. Published by Cambridge University Press on behalf of The Nutrition Society

Populations from lower socio-economic status (SES) backgrounds are in the unfortunate position of experiencing greater risk of non-communicable diseases such as depression, type 2 diabetes, CVD and obesity(Reference Everson, Maty and Lynch1Reference Rawshani, Svensson and Zethelius4). Poorer diet quality, along with other lifestyle factors such as physical inactivity, is key risk factors that cluster in lower SES groups(Reference James, Nelson and Ralph5). Measuring dietary patterns of populations from lower socio-economic backgrounds is therefore essential for examining diet–disease relationships and developing effective interventions to improve health. Assessment of diet predominantly relies on self-reported food intake, and the quality of the data collected relies on the accuracy of the reported intake by the participants. Misreporting of food and energy intake (EI), particularly low energy reporting, is a well-documented source of error in dietary assessment(Reference Lutomski, van den Broeck and Harrington6Reference Murakami and Livingstone9). If the degree of misreporting in large nutritional surveys is random and evenly distributed within the population and not specific to certain types of foods, it may reduce power but not prove too problematic if sample size is sufficiently large(Reference Rutishauser10). However, when systematic error is present, associations between nutrition and health can be distorted or misrepresented(Reference Rutishauser10).

The socio-demographic characteristics of misreporting have been frequently described, and greater rates of low energy reporting have recurrently been observed in overweight and obese individuals compared with individuals in the healthy weight range(Reference Freedman, Commins and Moler7,Reference Livingstone and Black8) . Other correlates of low energy reporting such as sex, smoking status, age, dietary restraint and physical activity have also been described, though less consistently across studies(Reference Livingstone and Black8,Reference Macdiarmid and Blundell11) . The direction of the effect of SES on reported EI has varied between studies(Reference Livingstone and Black8), and some studies report that populations of lower SES or education are more likely to misreport(Reference Lutomski, van den Broeck and Harrington6,Reference Murakami and Livingstone9,Reference Pryer, Vrijheid and Nichols12,Reference Gnardellis, Boulou and Trichopoulou13) , while other studies have reported that higher SES is associated with greater misreporting presumably due to greater health or diet consciousness and social desirability bias(Reference Pomerleau, Østbye and Bright-See14,Reference Lafay, Mennen and Basdevant15) .

Given the increasing emphasis on foods and diet quality rather than the effect of single nutrients on health, understanding how diet quality is affected by low energy reporting has become increasingly relevant to nutrition research(Reference Schulze, Martínez-González and Fung16), but little research has been undertaken in this area(Reference Markussen, Veierød and Ursin17,Reference Garden, Clark and Whybrow18) . Some studies indicate that certain foods are disproportionately misreported and those that are perceived to be less healthy, that is energy-dense, nutrient-poor discretionary foods such as cakes or confectionery, are under-reported, while healthier foods such as fruits and vegetables are more accurately reported(Reference Pryer, Vrijheid and Nichols12,Reference Rosell, Hellénius and de Faire19) . However, other studies have found that all foods are misreported without discrimination between healthy and unhealthier foods(Reference Garden, Clark and Whybrow18,Reference Millen, Tooze and Subar20) . To our knowledge, the effects of misreporting of foods between different SES groups are yet to be investigated.

The aim of this study was to examine the associations between low energy reporting, dietary quality and SES in a representative sample of adults. Specifically, (a) whether low energy reporting is affected by SES, and (b) if diet quality is differentially reported according to low energy reporters (LER) status and SES.

Material and methods

2011–2012 National Nutrition and Physical Activity Survey

Data from the 2011–2012 National Nutrition and Physical Activity Survey were obtained with permission from the Australian Bureau of Statistics. This survey collected detailed self-reported information on the dietary intake and physical activity of over 12 000 adults and children (2 years of age and older) across Australia. The scope of the survey included usual residents of private dwellings in urban and rural areas of Australia, covering approximately 97 % of people living in Australia, using a multistage, probability sampling design(21). Trained Australian Bureau of Statistics interviewers collected information from one adult and one child (where applicable) of each sampled private dwelling, at least one of whom was 18 years or over, by a face-to-face, computer-assisted personal interview and by a computer-assisted telephone interview for the second interview(21). Interviews were conducted at least 8 d apart, 7 d/week over a 12-month period, and complied with the Census and Statistics Act of 1905(21). Ethics approval was obtained from the Australian Government Department of Health and Ageing’s Departmental Ethics Committee(21). The response rate for the computer-assisted personal interview was 77·0 % of eligible households, with 63·6 % of these respondents also participating in the subsequent computer-assisted telephone interview to provide a second day of nutrition data(21).

Two 24-h recalls were collected (at computer-assisted personal interview and computer-assisted telephone interview) for adults aged 19 years and over using the Automated Multiple-Pass Method(21). This is a five-pass method developed by the United States Department of Agriculture and adapted to the Australian food supply that records all food and beverages consumed on the day before the interview. It comprises several techniques to help participants recall their dietary intake, thus reducing the occurrence of under-reporting in nutrition surveys. A Food Model Booklet was provided to assist in the estimation of portion sizes. The AUSNUT 2011–2013 nutrient database, created by Food Standards Australia New Zealand specifically for the survey, was used to convert foods and beverages into energy and food groups(21,22) . Dietary data were averaged across the 2 d of recall to calculate mean estimates of energy and food intakes.

Energy reporting status

Participants were categorised as LER or plausible reporters (PR). The Goldberg method uses the BMR (using the Schofield equations for individuals based on weight, age and sex) and the ratio of reported EI:BMR to estimate the amount of energy available for activity. The EI:BMR ratio is then compared with a physical activity level. As no measure of physical activity level was available, a physical activity level of 1·55 was assumed to indicate the minimum energy requirement for a normally active but sedentary population. Goldberg et al. calculated the lower 95 % confidence limit based on 2 d of dietary intake, allowing for day-to-day variation in EI and errors in calculation of EI:BMR, as 0·96. Individuals with reported EI below this cut-off were classified as LER, and those at or above the cut-off were classified PR.

An alternative method of identifying LER was applied using energy prediction equations(Reference Huang, Roberts and Howarth23). This method and results are described in online Supplementary material and provide similar findings as reported with the application of the Goldberg method.

Selection of food types

Foods from the five food groups were included in the analysis: (1) fruit; (2) vegetables; (3) grains and cereals foods (total, refined grains and wholegrains); (4) milk, yogurt, cheese and non-dairy alternatives (total, high fat (>10 %), moderate fat (4–10 %) and low fat (<4 %)) and (5) meat and alternatives (total, red meat, poultry, fish, seafood, eggs, and nuts and seeds). The number of serves of each food group consumed by each participant was calculated using the Australian Health Survey–Australian Dietary Guidelines database, which contains the standard serve sizes of foods in each food group(24).

Intake of discretionary foods, defined as foods and beverages high in added sugars, saturated fat, Na and/or alcohol, was also assessed, with one serve being equivalent to 600 kJ (total; cakes, muffins, cookies and pastries; pizza and burgers; fried potatoes; beer; savoury snack foods and crackers; confectionery and ice cream; sugar; sugar-sweetened beverages; beer; and wine). The proportion of energy derived from discretionary foods as a proportion of total energy (%E from discretionary food) was assessed for each participant and used as an indicator of diet quality.

Socio-economic status and co-variates

SES was defined by area-level and individual level. An area-level index of relative socio-economic disadvantage (SEIFA) was based on postcode for variables such as income, educational attainment, unemployment and dwellings without motor vehicles and represents an average of all people living in an area(25). Participants were categorised into three groups: low (quintile 1), middle (quintiles 2–4) and high (quintile 5). Individual level SES was represented by educational attainment(Reference Winkleby, Jatulis and Frank26) categorised into three groups: no tertiary education qualification; college or vocational qualification; and university qualification. Weight and height measures were taken to one decimal point, by trained Australian Bureau of Statistics staff during the household interview using digital scales and a portable stadiometer(21).

Statistical analysis

Collinearity between SEIFA and education was assessed using a correlation matrix in SAS, and no correlation was >0·3 and therefore no high correlations observed. Crude and multiple logistic regression was used to calculate the OR of being classified as a LER for different socio-demographic variables including age group (18–50 years; 51–70 years and 71+ years); sex (male and female); BMI category (underweight <18·5 kg/m2; normal ≥18·5 kg/m2 to <25·0 kg/m2; overweight ≥25·0 to <30·0 kg/m2; and obese ≥ 30·0 kg/m2); SEIFA; education; geographic area; whether on a low-energy or weight-loss diet and country of birth.

Multiple linear regression was used to determine the relationship between %E from discretionary food and energy reporting status (LER or PR), SEIFA and educational attainment and adjusted for age (continuous), BMI (continuous), country of birth and whether on a low-energy or weight-loss diet. To determine if there was effect modification between SES and energy reporting status, interaction terms were applied between SEIFA and energy reporting status (LER or PR) and educational attainment and energy reporting status.

Differences in proportions of LER and PR that reported consuming different food groups were determined with Pearson’s χ 2 test. Mean differences in per consumer intake and in %E from discretionary food were determined between LER and PR using ANCOVA, adjusted for BMI (continuous), age (continuous), sex, and country of birth, SEIFA and education and differences between groups were assessed with Bonferroni post hoc tests. All outcome distributions were checked for normality. All analyses were conducted in SAS® version 9.4: SAS Institute Inc.(27). To account for selection probability and the clustered survey design, person-specific weights were applied to compute point estimates and replicate weights (the Jackknife group delete one method) were used to compute standard errors(21). Significant differences were considered as those P < 0·05.

Results

Socio-demographics of plausible reporter and low energy reporter

The final sample consisted of those participants who had completed 2 d of dietary recalls and provided height and weight data (n 5421). The mean age was 45·9 (se 0·09) years. About 23·0 % of participants were classified as LER using the Goldberg cut-offs. The socio-demographic characteristics of participants classified as LER and PR and the OR for different socio-demographic variables and LER are shown in Table 1. The characteristics associated with LER were similar in the unadjusted and adjusted multivariate models. The odds of being classified as a LER differed by groups, with participants of lower education attainment being more likely to be classified as a LER than those with university education, in both the crude and adjusted model (Table 1). Being classed as overweight or obese, and being on a low-energy or weight-loss diet were also associated with higher odds of being classified as a LER compared with their counterparts (Table 1).

Table 1. Risk of being a low energy reporter (LER) (n 1289) compared with a plausible reporter (PR) (n 4132) for different socio-demographic groups in the Australian National Nutrition and Physical Activity Survey 2011–2012*

(Odds ratios and 95 % confidence intervals)

Ref, reference; SEIFA, socio-economic index for area.

* All estimates are weighted.

P values derived from logistic regression.

Multivariate model adjusted for all variables in column.

Differences in percentage energy from discretionary food by energy reporting status

Regression analysis showed that LER, SEIFA and educational attainment were all associated with reported %E from discretionary foods (Table 2). The %E from discretionary foods was lower for LER than PR (27·4 % (26·5–28·4) v. 34·2 % (33·7–34·7), P < 0·0001). As an indicator of SES, educational attainment showed a slight stronger effect than SEIFA but both indicators showed higher SES was associated with lower %E from discretionary foods. Testing for effect modification using interaction terms (energy reporting status and SEIFA, P = 0·3078 or energy reporting status and educational attainment, P = 0·7078) revealed no significant effect modification. Fig. 1 demonstrates the same gradient in %E for LER and PR according to educational attainment.

Table 2. Linear regression for discretionary food (DF) intake (% energy (%E)) by energy-reporting status for different groups in the Australian National Nutrition and Physical Activity Survey*

LER, low energy reporters; PR, plausible reporters; SEIFA, socio-economic index for area.

* Model 1: univariate model. Model 2a adjusted for age, sex, BMI, country of birth, low-energy or weight-loss diet, energy reporting status, SEIFA and educational attainment. Model 2b adjusted for age, sex, BMI, country of birth, low-energy or weight-loss diet, energy reporting status and educational attainment. Model 2c adjusted for age, sex, BMI, country of birth, low-energy or weight-loss diet, energy reporting status and SEIFA. There was no significant effect modification between energy reporting status and SEIFA (P = 0·3078) or educational attainment (P = 0·7078), and the interaction terms were removed from the models. All estimates are weighted.

Fig. 1. Percentage energy (%E) from discretionary food by highest tertiary education attainment for plausible reporters (PR) and low energy reporters (LER). Mean differences determined with ANCOVA and Bonferroni post hoc tests; means were adjusted for BMI, age, sex, country of birth, low-energy or weight-loss diets and socio-economic index for area (SEIFA). *P < 0·05, **P < 0·01, ***P < 0·0001. All estimates are weighted. , No tertiary education; , vocational education; , university education.

Differences in food intake by energy reporting status

The differences between LER and PR for the mean intake reported by consumers of different foods and the proportion of consumers of different foods are shown in Table 3. LER reported smaller portions of five food group foods and discretionary foods with the exception of fish and seafood, and eggs. Similarly, the proportion of participants reporting consuming foods from the five food groups and discretionary foods was lower for LER compared with PR except for lean meat and alternatives, fish and seafood, legumes and beans, poultry, eggs and low-fat dairy products which were not significantly different (Table 3).

Table 3. Mean intake of foods and food groups per consumer for plausible reporters (PR) and low energy reporters (LER)

(Mean values and 95 % confidence intervals; proportions and standard errors)

Df, difference; N/A, not applicable.

* Pearson’s χ 2 test. Significant differences were considered at P < 0·05.

One serve of vegetables = 75 g; one serve of fruit = 150 g; one serve of milk and alternatives = 550 kJ; grains and cereals = 500 kJ; meat and alternatives = 500 kJ; one serve of discretionary food = 600 kJ.

ANCOVA for means adjusted for BMI, age, sex, country of birth, low-energy or weight-loss diets and socio-economic index for area (SEIFA).

§ %Df: Percentage difference in number of serves for all foods. All estimates are weighted.

LER tended to report lower intake of five food group foods that may have been perceived to be less healthy; for example, LER reported fewer serves of medium and high fat dairy products than PR and were 38·1 and 29·9 % lower than PR, respectively, compared with low-fat dairy products, which was only 17·4 % lower; red meat serves were reported 25·9 % less by LER compared with PR, whereas the differences in the serves of fish (−1·9 %) and eggs (−8·8 %) were negligible and/or not significantly different (Table 3).

Discussion

In our analysis of a nationally representative nutrition survey, low energy reporting was more prevalent for groups living in lower-SES areas and in those without a university education. Compared with PR, adults classified as LER reported less frequent consumption of all foods (five food group foods and discretionary foods). However, LER reported better diet quality shown by a lower %E from discretionary foods which contributed 26·6 % of energy for LER, compared with 35·4 % for PR. This relationship was not modified by area-level SES or educational attainment, with populations from the lowest-SES backgrounds having poorer diet quality.

Our findings that low energy reporting is more common in groups living in lower-SES areas or with lower educational attainment are largely consistent with other studies that used large, representative samples of the population and collected dietary data with 24-h recalls(Reference Freedman, Commins and Moler7,Reference Mattisson, Wirfält and Aronsson28Reference Briefel, Sempos and McDowell30) . This has been attributed to having fewer skills needed to complete dietary assessments accurately and/or less time invested into diet or health(Reference Livingstone and Black8). A smaller study that used biomarkers to validate EI found no association between education and misreporting, but cautioned that this may have been due to the limited education range in the sample(Reference Tooze, Subar and Thompson31). Other studies have found that higher SES is associated with greater low energy reporting; in a Canadian study, higher educational attainment was associated with lower EI:BMR ratios(Reference Pomerleau, Østbye and Bright-See14). These conflicting results may be due to differences in the dietary assessment method, for example, a FFQ(Reference Pomerleau, Østbye and Bright-See14) can be prone to over-estimation of the amounts of foods consumed(Reference Bogers, Dagnelie and Westerterp32). Greater rates of over-reporting for participants with lower educational attainment have been found in several, but not all, studies that assessed diet with FFQ and it may depend on the specific FFQ used(Reference Freedman, Commins and Moler7,Reference Braam, Ocké and Bueno-de-Mesquita33) . A further study in northern France found a greater prevalence of low energy reporting in 3-d diet diaries in groups of higher socio-professional class(Reference Lafay, Mennen and Basdevant15). However, a more recent study using 7-d food diaries in a representative sample of the French population found those with the lowest educational attainment misreported to a greater extent than those of higher educational attainment(Reference Berta Vanrullen, Volatier and Bertaut29). Overall, these results seem to suggest that lower SES is associated with higher rates of low energy reporting, especially in 24-h recalls, when representative population samples are used.

There was evidence that all foods (healthy and unhealthy) were misreported and both were reported less frequently and smaller amounts by LER. Reasons for misreporting are multi-factorial and can be due to the deliberate misrepresentation of food intake but also due to other errors such as memory lapse and underestimation of portion sizes(Reference Livingstone and Black8). Some foods that are recalled during the interview and snacks are more frequently forgotten, which may explain why fruit and nuts were reported by a smaller proportion of LER(Reference Gemming, Rush and Maddison34). Eggs and fish and seafood were not differentially reported between PR and LER, which is consistent with other studies(Reference Gemming and Ni Mhurchu35) as these foods may be recalled more easily or may be less prone to omission due to social desirability bias.

Whilst all food groups were misreported, the significantly higher degree of under-reporting of discretionary foods suggests some bias towards selective under-reporting. Social desirability is likely to have played a significant role in both the types of foods from the five food groups that were reported, as well as the amounts of discretionary foods(Reference Macdiarmid and Blundell11,Reference Maurer, Taren and Teixeira36) . For example, there were only small differences in the proportion of LER that reported intake of low-fat dairy products but there were more substantial differences in moderate- and high-fat dairy products between LER and PR. Similarly, the number of serves of wholegrains reported by LER was closer to the number of serves reported by PR, whereas there was a larger difference in the number of serves of refined grains reported. This form of under-reporting is likely an attempt by the participant to avoid disapproval by adhering to what society perceives as ‘acceptable dietary behaviour’(Reference Maurer, Taren and Teixeira36). These findings are in agreement with previous studies which have found that LER tends to report healthier dietary patterns or lower consumption of discretionary (unhealthy) foods, namely products that are high in added sugar and/or saturated fat, such as biscuits, chips, cakes and confectionery(Reference Macdiarmid and Blundell11,Reference Gemming and Ni Mhurchu35Reference Scagliusi, Ferriolli and Pfrimer38) . Further studies have found that when estimates were adjusted for energy, healthier foods were over-estimated by LER(Reference Pryer, Vrijheid and Nichols12,Reference Rosell, Hellénius and de Faire19) . This is consistent with the present results that healthier foods represent a higher proportion of foods reported by LER. However, some studies have found that reported intake of both healthy and unhealthy foods were lower for LER compared with PR, with no apparent bias towards unhealthy foods(Reference Garden, Clark and Whybrow18,Reference Millen, Tooze and Subar20,Reference Krebs-Smith, Graubard and Kahle39) .

Exclusion of low energy reporter in data analysis

Understanding self-reported dietary intake errors can help to improve the collection of data, and analysis of relationships between nutrition and health. For example, there has been much debate over whether to exclude participants identified as LER when analysing dietary intake data(Reference Thompson, Kirkpatrick and Subar40,Reference Poslusna, Ruprich and de Vries41) . A few studies revealed that the inclusion of LER in their analyses led to weak and/or misleading relationships between obesity and dietary intake, as obese individuals more commonly overeat and under-report foods high in sugar and fat(Reference Gnardellis, Boulou and Trichopoulou13,Reference Poslusna, Ruprich and de Vries41) . However, if the aim of the study is to obtain information on the food and EI of a nationally representative sample, the exclusion of LER would introduce an alternative source of error – selection bias(Reference Thompson, Kirkpatrick and Subar40). This is particularly problematic as low energy reporting is associated with socio-demographic characteristics, including BMI and SES(Reference Livingstone and Black8). Future research should include SES as part of their models/analysis when examining dietary associations. Additionally, the excluded LER would include participants who genuinely under-eat for any number of reasons, thus distorting the results. Some participants classified as LER may have genuinely reduced EI and improved their diet quality due to attempted weight loss, and in the present study, 12·3 % of LER reported they were on a low-energy or weight-loss diet compared with 5·1 % of PR. Excluding LER from analysis is therefore likely to over-estimate intake of the population, and some low EI will represent actual intake. It has been recommended to use sensitivity analysis to determine the effect of misreporting and use alternatives to removing participants including adjusting the data analyses for factors linked to low energy reporting or stratifying by LER status, in order to avoid introducing selection bias(Reference Thompson, Kirkpatrick and Subar40,Reference Tooze, Freedman and Carroll42) . Adjusting for EI has been shown to improve estimates for macronutrients(Reference Thompson, Kirkpatrick and Subar40); however, further research on how this improves estimates for dietary patterns is needed(Reference Poslusna, Ruprich and de Vries41).

Limitations and strengths

Two days of 24-h recall were used to assess diet and reported EI which may not reflect usual intake for individuals. As with many self-reported diet measures, measurement errors including social desirability and portion size estimation were not able to be objectively quantified. The Goldberg cut-off used to identify LER in this study is not only able to determine extreme LER(Reference Black43,Reference Black44) but has also been validated against doubly-labelled water, the gold standard for total energy expenditure, and shown to be a reasonable approach to characterising low energy reporting for 2 d of recalls(Reference Tooze, Krebs-Smith and Troiano45). The use of two separate indicators of SES, area-level SEIFA and individual level education, resulting in same results, strengthen the evidence of the association. The findings of the present study may be generalisable to Western populations given the large, nationally representative sample of Australian adults.

Future research would benefit from objectively measuring dietary intake within a variety of free-living settings in order to truly capture what food types and at what times low energy reporting is likely to occur. Recent studies have shown wearable cameras to be a useful, passive measure of dietary assessment, with findings that low energy reporting was most common for snack foods, condiments and beverages(Reference Gemming and Ni Mhurchu35). Further evaluation of such technology with larger study populations has the potential to greatly improve the accuracy of dietary assessment.

Conclusions

This study indicates that lower SES is associated with a greater prevalence of low energy reporting. However, there did not appear to be a differential effect of SES on the types of foods that were misreported. Differences in reported food intake were observed between PR and LER and while the amounts and proportions of almost all foods were reported by fewer LER, discretionary foods were disproportionately affected and made a smaller contribution to total energy. These results indicate that low energy reporting is likely to over-estimate the diet quality and obscure diet–disease relationships. This will disproportionality effect socio-economically disadvantaged groups and need to be considered when interpreting studies on diet–disease relationships.

Acknowledgements

The authors would like to thank the Australian Bureau of Statistics for providing the data set for analysis.

This research received no specific grant from any funding agency, commercial or not-for-profit sectors.

M. H. and A. G. analysed the data; all authors interpreted the data; M. H. and A. G. wrote the draft and A. G., L. G.; A. R. revised it critically for important intellectual content; all authors had final approval of the version to be published.

The authors declare that there are no conflicts of interest.

Supplementary material

For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114520003621

References

Everson, SA, Maty, SC, Lynch, JW, et al. (2002) Epidemiologic evidence for the relation between socioeconomic status and depression, obesity, and diabetes. J Psychosom Res 53, 891895.CrossRefGoogle Scholar
Kivimäki, M, Virtanen, M, Kawachi, I, et al. (2015) Long working hours, socioeconomic status, and the risk of incident type 2 diabetes: a meta-analysis of published and unpublished data from 222,120 individuals. Lancet Diabetes Endocrinol 3, 2734.CrossRefGoogle Scholar
Ding, D, Chong, S, Jalaludin, B, et al. (2015) Risk factors of incident type 2-diabetes mellitus over a 3-year follow-up: results from a large Australian sample. Diabetes Res Clin Pract 108, 306315.CrossRefGoogle Scholar
Rawshani, A, Svensson, A-M, Zethelius, B, et al. (2016) Association between socioeconomic status and mortality, cardiovascular disease, and cancer in patients with type 2 diabetes. JAMA Intern Med 176, 11461154.CrossRefGoogle ScholarPubMed
James, WPT, Nelson, M, Ralph, A, et al. (1997) Socioeconomic determinants of health: the contribution of nutrition to inequalities in health. BMJ Case Rep 314, 1545.Google ScholarPubMed
Lutomski, JE, van den Broeck, J, Harrington, J, et al. (2011) Sociodemographic, lifestyle, mental health and dietary factors associated with direction of misreporting of energy intake. Public Health Nutr 14, 532541.CrossRefGoogle ScholarPubMed
Freedman, LS, Commins, JM, Moler, JE, et al. (2014) Pooled results from 5 validation studies of dietary self-report instruments using recovery biomarkers for energy and protein intake. Am J Epidemiol 180, 172188.CrossRefGoogle ScholarPubMed
Livingstone, MB & Black, AE (2003) Markers of the validity of reported energy intake. J Nutr 3, 895s920s.CrossRefGoogle Scholar
Murakami, K & Livingstone, MBE (2016) Prevalence and characteristics of misreporting of energy intake in US children and adolescents: National Health and Nutrition Examination Survey (NHANES) 2003–2012. Br J Nutr 115, 294304.CrossRefGoogle ScholarPubMed
Rutishauser, IH (2005) Dietary intake measurements. Public Health Nutr 8, 11001107.CrossRefGoogle ScholarPubMed
Macdiarmid, J & Blundell, J (1998) Assessing dietary intake: who, what and why of under-reporting. Nutr Res Rev 11, 231253.CrossRefGoogle ScholarPubMed
Pryer, JA, Vrijheid, M, Nichols, R, et al. (1997) Who are the ‘low energy reporters’ in the dietary and nutritional survey of British adults? Int J Epidemiol 26, 146154.CrossRefGoogle ScholarPubMed
Gnardellis, C, Boulou, C & Trichopoulou, A (1998) Magnitude, determinants and impact of under-reporting of energy intake in a cohort study in Greece. Public Health Nutr 1, 131137.CrossRefGoogle Scholar
Pomerleau, J, Østbye, T & Bright-See, E (1999) Potential underreporting of energy intake in the Ontario Health Survey and its relationship with nutrient and food intakes. Eur J Epidemiol 15, 553557.CrossRefGoogle ScholarPubMed
Lafay, L, Mennen, L, Basdevant, A, et al. (2000) Does energy intake underreporting involve all kinds of food or only specific food items? Results from the Fleurbaix Laventie Ville Sante (FLVS) study. Int J Obes Relat Metab Disord 24, 15001506.CrossRefGoogle ScholarPubMed
Schulze, MB, Martínez-González, MA, Fung, TT, et al. (2018) Food based dietary patterns and chronic disease prevention. BMJ Case Rep 361, k2396.Google ScholarPubMed
Markussen, MS, Veierød, MB, Ursin, G, et al. (2016) The effect of under-reporting of energy intake on dietary patterns and on the associations between dietary patterns and self-reported chronic disease in women aged 50–69 years. Br J Nutr 116, 547558.CrossRefGoogle ScholarPubMed
Garden, L, Clark, H, Whybrow, S, et al. (2018) Is misreporting of dietary intake by weighed food records or 24-hour recalls food specific? Eur J Clin Nutr 72, 1026.CrossRefGoogle ScholarPubMed
Rosell, MS, Hellénius, M-LB, de Faire, UH, et al. (2003) Associations between diet and the metabolic syndrome vary with the validity of dietary intake data. Am J Clin Nutr 78, 8490.CrossRefGoogle ScholarPubMed
Millen, AE, Tooze, JA, Subar, AF, et al. (2009) Differences between food group reports of low-energy reporters and non–low-energy reporters on a food frequency questionnaire. J Am Diet Assoc 109, 11941203.CrossRefGoogle ScholarPubMed
Australian Bureau of Statistics (2013). Australian Health Survey: Users’ Guide 2011–12. Canberra: ABS.Google Scholar
Food Standards Australia New Zealand (2014). AUSNUT 2011–13 Food Nutrient Database. http://www.foodstandards.gov.au/science/monitoringnutrients/ausnut/foodnutrient/Pages/default.aspx (accessed December 2016).Google Scholar
Huang, TTK, Roberts, SB, Howarth, NC, et al. (2005) Effect of screening out implausible energy intake reports on relationships between diet and BMI. Obesity Res 13, 12051217.CrossRefGoogle ScholarPubMed
Food Standards Australia New Zealand (2017) Australian Health Survey – Australian Dietary Guidelines Database. Canberra: FSANZ.Google Scholar
Australian Bureau of Statistics (2013) Census of Population and Housing: Socio-Economic Indexes for Areas (SEIFA), Australia, 2011. Catalogue no. 2033.0.55.001. Canberra: ABS.Google Scholar
Winkleby, MA, Jatulis, DE, Frank, E, et al. (1992) Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. Am J Public Health 82, 816820.CrossRefGoogle ScholarPubMed
SAS Software [Computer Program]. Version 9.4 for Windows. Cary, NC2002–2012.Google Scholar
Mattisson, I, Wirfält, E, Aronsson, CA, et al. (2005) Misreporting of energy: prevalence, characteristics of misreporters and influence on observed risk estimates in the Malmö Diet and Cancer cohort. Br J Nutr 94, 832842.CrossRefGoogle Scholar
Berta Vanrullen, I, Volatier, JL, Bertaut, A, et al. (2014) Characteristics of energy intake under-reporting in French adults. Br J Nutr 111, 12921302.CrossRefGoogle ScholarPubMed
Briefel, RR, Sempos, CT, McDowell, MA, et al. (1997) Dietary methods research in the third National Health and Nutrition Examination Survey: underreporting of energy intake. Am J Clin Nutr 65, 1203S1209S.CrossRefGoogle ScholarPubMed
Tooze, JA, Subar, AF, Thompson, FE, et al. (2004) Psychosocial predictors of energy underreporting in a large doubly labeled water study. Am J Clin Nutr 79, 795804.CrossRefGoogle Scholar
Bogers, RP, Dagnelie, PC, Westerterp, KR, et al. (2003) Using a correction factor to correct for overreporting in a food-frequency questionnaire does not improve biomarker-assessed validity of estimates for fruit and vegetable consumption. J Nutr 133, 12131219.CrossRefGoogle Scholar
Braam, LAJLM, Ocké, MC, Bueno-de-Mesquita, HB, et al. (1998) Determinants of obesity-related underreporting of energy intake. Am J Epidemiol 147, 10811086.CrossRefGoogle ScholarPubMed
Gemming, L, Rush, E, Maddison, R, et al. (2015) Wearable cameras can reduce dietary under-reporting: doubly labelled water validation of a camera-assisted 24 h recall. Br J Nutr 113, 284291.CrossRefGoogle ScholarPubMed
Gemming, L & Ni Mhurchu, C (2016) Dietary under-reporting: what foods and which meals are typically under-reported? Eur J Clin Nutr 70, 640641.CrossRefGoogle ScholarPubMed
Maurer, J, Taren, DL, Teixeira, PJ, et al. (2006) The psychosocial and behavioral characteristics related to energy misreporting. Nutr Rev 64, 5366.CrossRefGoogle ScholarPubMed
Rangan, A, Allman-Farinelli, M, Donohoe, E, et al. (2014) Misreporting of energy intake in the 2007 Australian Children’s Survey: differences in the reporting of food types between plausible, under- and over-reporters of energy intake. J Hum Nutr Diet 27, 450458.CrossRefGoogle ScholarPubMed
Scagliusi, FB, Ferriolli, E, Pfrimer, K, et al. (2008) Under-reporting of energy intake is more prevalent in a healthy dietary pattern cluster. Br J Nutr 100, 10601068.CrossRefGoogle Scholar
Krebs-Smith, SM, Graubard, BI, Kahle, LL, et al. (2000) Low energy reporters vs others: a comparison of reported food intakes. Eur J Clin Nutr 54, 281287.CrossRefGoogle ScholarPubMed
Thompson, FE, Kirkpatrick, SI, Subar, AF, et al. (2015) The National Cancer Institute’s dietary assessment primer: a resource for diet research. J Acad Nutr Diet 115, 19861995.CrossRefGoogle ScholarPubMed
Poslusna, K, Ruprich, J, de Vries, JHM, et al. (2009) Misreporting of energy and micronutrient intake estimated by food records and 24 hour recalls, control and adjustment methods in practice. Br J Nutr 101, S73S85.CrossRefGoogle ScholarPubMed
Tooze, JA, Freedman, LS, Carroll, RJ, et al. (2016) The impact of stratification by implausible energy reporting status on estimates of diet-health relationships. Biom J Biometrische Zeitschrift 58, 15381551.CrossRefGoogle ScholarPubMed
Black, AE (2000) The sensitivity and specificity of the Goldberg cut-off for EI:BMR for identifying diet reports of poor validity. Eur J Clin Nutr 54, 395404.CrossRefGoogle ScholarPubMed
Black, AE (2000) Critical evaluation of energy intake using the Goldberg cut-off for energy intake:basal metabolic rate. A practical guide to its calculation, use and limitations. Int J Obes Relat Metab Disord 24, 11191130.CrossRefGoogle ScholarPubMed
Tooze, JA, Krebs-Smith, SM, Troiano, RP, et al. (2012) The accuracy of the Goldberg method for classifying misreporters of energy intake on a food frequency questionnaire and 24-hour recalls: comparison with doubly labeled water. Eur J Clin Nutr 66, 569576.CrossRefGoogle Scholar
Figure 0

Table 1. Risk of being a low energy reporter (LER) (n 1289) compared with a plausible reporter (PR) (n 4132) for different socio-demographic groups in the Australian National Nutrition and Physical Activity Survey 2011–2012*(Odds ratios and 95 % confidence intervals)

Figure 1

Table 2. Linear regression for discretionary food (DF) intake (% energy (%E)) by energy-reporting status for different groups in the Australian National Nutrition and Physical Activity Survey*

Figure 2

Fig. 1. Percentage energy (%E) from discretionary food by highest tertiary education attainment for plausible reporters (PR) and low energy reporters (LER). Mean differences determined with ANCOVA and Bonferroni post hoc tests; means were adjusted for BMI, age, sex, country of birth, low-energy or weight-loss diets and socio-economic index for area (SEIFA). *P < 0·05, **P < 0·01, ***P < 0·0001. All estimates are weighted. , No tertiary education; , vocational education; , university education.

Figure 3

Table 3. Mean intake of foods and food groups per consumer for plausible reporters (PR) and low energy reporters (LER)(Mean values and 95 % confidence intervals; proportions and standard errors)

Supplementary material: File

Grech et al. Supplementary Materials

Grech et al. Supplementary Materials

Download Grech et al. Supplementary Materials(File)
File 48.4 KB