Get Complete Project Material File(s) Now! »



The following section provides an overview of the traditional methods of dietary assessment and their associated strengths and limitations. This section also includes information on the common sources of error in dietary assessment, validity of the traditional methods of dietary assessment, and discusses the implications of measurement error. The sources for this information were obtained from MEDLINE (1966 to September 2014), Web of Science, Google Scholar, and PubMed (to September 2014). The limitations of the traditional methods of dietary assessment, and implications of measurement error, provide the rationale for investigating the use of wearable cameras to reduce the reporting bias for dietary EI.
This literature overview is structured as follows:
Section 3.2 provides an overview of the traditional methods of dietary assessment
Section 3.4 summarises the common sources of error in dietary assessment
Section 3.5 describes methods used to assess validity, common results from reviews of validation studies, and discusses the implications of measurement error
Section 3.6 summarises sections 3.2 to 3.5
85, 86, 91
41, 42
2, 84

Dietary assessment

The traditional methods of dietary assessment have been primarily designed to estimate the quantity and/or frequency of food consumption, and are essential to evaluate individuals’ or populations’ dietary intake and nutritional status. Four traditional methods of dietary assessment have been used extensively in nutrition research and in clinical and community settings: 24-hour dietary recall, food record (FR), food frequency questionnaire (FFQ), and the diet history interview (DH). 1, 2, 84 Each method has specific strengths and limitations, making them appropriate for use in different situations (discussed in sections 3.2.1 to 3.3.3). However, all methods rely on self-report without the ability to verify intake objectively (at the time of assessment). Therefore, both unintentional and intentional misreporting can contribute undetected to measurement error. 84
Although traditional methods of dietary assessment are primarily designed to record dietary intake, the 24-hour dietary recall and FR can also be used to collect additional limited information on the context of eating episodes (meals and snacks), such as time and location (e.g. food eaten at home vs. at work). 85, 86 87, 88 Context describes the interrelated conditions in which something exists or occurs, 89 but at present there is no standard terminology for defining context or classes of contextual variables for dietary intake. 42 Both lab-based and real-world studies investigating human eating and nutrition behaviours have shown that many contextual factors affect dietary intake; such as meal time, location, effort required (e.g. distance of food away from table), cost, décor, ambient music, plate size, television viewing, social interaction, positive and negative cues (e.g. suggestions provided regarding the popularity of a food), meal frequency, and meal duration. Equally, contextual information can also reveal where and when certain foods are likely to be obtained or consumed. 90 As such, information on contexts is now recorded in some large-scale dietary surveillance programmes in conjunction with details of the foods and beverages consumed.

 24-hour dietary recall

The 24-hour dietary recall is a retrospective method of dietary assessment usually conducted by a skilled interviewer (in person or over the telephone), but it can be completed solely by a participant using an automated website or software package. 2, 19-21 This flexibility makes the 24- hour dietary recall suitable for a range of participants including those with low literacy and people with mental and physical disabilities. 1
Typically, the interviewer follows a standardised format to prompt participants to recall all foods and beverages consumed over the previous 24-hours. The 24-hour period is usually reviewed multiple times to elicit further information from the participant, and this is referred to as the multiple pass 24-hour dietary recall (MP24). As described above, dietary recalls also allow some contextual information on each eating episode to be recorded, such as location (discussed in section 3.2). 11 The retrospective nature of dietary recalls places a reduced burden on participants compared to the FR, as there is no need to record any information prior to the assessment, and this therefore reduces the possibility of participants to change their dietary behaviour. 1
Research-specific software is not required to conduct the 24-hour dietary recall interview but is commonly used in large-scale research to help standardise data collection and reduce the burden of analysis. 23, 86, 92 The United States Department of Agriculture (USDA) developed the Automated Multiple Pass Method (AMPM) 24, 86 for use in The National Health and Nutrition Examination Survey (NHANES), which uses five distinct steps (see Figure 3). The first step asks participants to recall all the foods and beverages consumed for the previous day. The second step encourages additional information with a “forgotten foods” list to prompt recall of unreported foods. Successive passes gather additional details, such as time and occasion, food brand, portion size, method of cooking, and condiments added. A final review is then conducted to check all the information is correct.
Nutrition surveillance programmes in N.Z. and Australia incorporate the MP24 and have adopted the method using bespoke software but differ slightly in the number of steps and details collected at each successive step. 25 23, 93 In N.Z., the LINZ24 MP24 software developed by the University of Otago has been used in two previous adult national nutrition surveys (NNS97 and ANS 08/9).
The LINZ24 system was also adapted for use in the 1995 Australian National Nutrition Survey, 93 but Australia recently developed their own system based on the USDA AMPM for the most recent Australian Healthy Survey 2011/13. 25
Software has also been developed to assist 24-hour dietary recalls for a range of both large and small scale research. An early example is EPIC-SOFT developed for the European Prospective Investigation into Cancer and Nutrition, used to standardise the 24-hour dietary recall procedures across 23 different research centres in 12 European countries. 92 Similar to other software, EPICSOFT assists an interviewer to follow the structured method and consists of a quick list followed by detailed steps and probing questions to gather further details. 92 For small scale research some nutrient analysis software packages, used to convert foods into energy and nutrients (discussed in section 3.4.1), include a 24-hour recall feature to assist researchers. More recently, web-based systems have been developed (for both large and small scale research), such as the National Cancer Institute ASA24, which can be easily modified for any country. 20

Limitations of the 24-hour dietary recall

Due to the wide variation in an individual’s daily dietary intake, use of a single 24-hour recall is only appropriate to describe the average dietary intake of a group, not individuals. As a retrospective method, the portion size of foods and beverages must be estimated, and details regarding the brand and variety may be forgotten or misreported. Moreover, the accuracy of the recall can be affected by interviewer bias and social desirability bias (discussed in section 3.4). 2, 96 Overall, the data obtained from the 24-hour recall method usually under-estimates dietary intake, especially amongst the elderly, women and overweight/obese respondents (specific details regarding the validity of the 24-hour dietary recall are presented in section 3.5).

Food record

The FR is a self-reported prospective record of all food consumed by a person over a given time period (usually 3 to 7 days), and is used widely in community and clinical settings, and small and large-scale nutrition research. 1, 2 Typically, participants are required to record everything they eat or drink for 1 to 7 days. Quantities of foods and beverages recorded are either weighed or estimated (using portion size guides) and training is usually provided to ensure thorough records are kept. 1 Traditionally, FR’s were completed by participants using pen and paper in booklets. An example of the FR booklet used for the National Diet and Nutrition Survey (NDNS) in the United Kingdom (U.K.) is presented in Figure 5. 98 However, electronic food records have been developed for hand-held devices, personal digital assistants (PDAs), smartphones and tablet computers. More recently consumer friendly FR apps have been developed for smartphones, such as MyNetDiary, Easy Diet Diary, and My Fitness Pal shown in Figure 4.
As the FR is the only method of dietary assessment which attempts to record dietary intake prospectively (at the time food is consumed), the FR has been referred to as the gold standard measure (seven-day weighed FR) and is used as a reference method to validate other methods of dietary assessment. 1 However, the data obtained from the FR is far from optimal, as discussed in section
As described above in section 3.2, the prospective nature of the FR makes it suitable to record the context of eating episodes, and is the recommended choice in research of human eating and nutrition behaviours in free-living settings. Moreover, since 2008 the NDNS has included an extra column in the FR booklet for participants to record contextual information on the location (where?), whether there was social interaction (with whom?), whether television was viewed (TV on?), and seating position (at table?) of the eating episode (see Figure 5). 85

Limitations of the food record

Similar to the 24-hour dietary recall the FR is self-reported without any means to objectively verify dietary intake (at the time of assessment). Moreover, 1-day FRs like 24-hour dietary recalls are only appropriate to describe the average dietary intake of a group, not individuals, due to the wide variation in an individual’s daily dietary intake. A selection bias (non-response) is possible as participants must be literate, and a high burden is placed on participants due to the level of detail, concentration, and time required for completion of the FR correctly. Investigations have demonstrated that the quality of the data obtained from FRs diminishes over successive days of assessment. Early studies using electronic food records revealed the majority of participants record their meals between 3-6 hours after consumption, thus the FR can still be prone to similar limitations of retrospective methods i.e. memory lapses, and incorrect estimation of portion sizes. 103 The prospective design of the FR is also prone to respondent biases such as changes in dietary behaviour, under-recording of food intake, and forgetting to record foods. 104 Together, these limitations affect the quality of the data obtained when using FRs, and often result in under-estimation of dietary intake. 8, 9 Specific details on validity of the FR method are presented in section 3.5.

Food frequency questionnaires

Food frequency questionnaires assess the frequency with which certain foods are consumed over a standard period of time (e.g. one month), using questions such as “how often over the past month did you eat fish”? A limited number of possible responses are available. The FFQ was traditionally an interviewer-led or self-administered paper-based questionnaire, but can be completed by a parent or caregiver proxy, or conducted over the telephone, or the internet. 2 A FFQ designed to assess the full diet usually contains approximately 120-180 pre-specified food items (questions) but can be shortened to focus on specific nutrients and food groups. The frequency at which certain foods are consumed can be used to rank individuals according to nutrient intake, or monitor changes in usual intake, and semi-quantitative FFQs that incorporate portion size questions also allow estimates of energy and nutrient intakes to be derived. The strength of the FFQ is therefore its ability to assess a participant’s habitual intake of foods, and is also better suited to capture intake of episodically consumed foods/seasonal foods such as, strawberries or mandarins. Further, being a retrospective method, it does not influence dietary behaviours like the FR (though participants can still misreport intake). 1 Moreover, FFQ has a low participant burden compared to the FR, and the limited number of pre-specified responses make the FFQ easy to analyse compared to other methods. 1, 2 These attributes make the FFQ a popular choice in dietary surveillance programmes, diet-related intervention studies, and prospective and retrospective research investigating diet and disease relationships.

Limitations of food frequency questionnaires

Food frequency questionnaires lack accuracy compared to other methods, as many details regarding dietary intake are not captured by the limited possible responses. Respondent biases and memory lapses are problematic, as it can be challenging to estimate how frequently certain foods are consumed, especially with foods only eaten occasionally. 1 Further inaccuracies result from incomplete lists of foods in the questionnaire, and FFQ must be tailored and validated for specific population groups, otherwise it may not contain the most appropriate and commonly consumed foods for the population assessed. Therefore, data obtained from the FFQ method should only be considered an approximation. 1, 2 Additionally, information on context cannot be recorded as the FFQ captures information on habitual dietary intake, not specific eating episodes.

Diet history

The DH method is usually conducted by a skilled interviewer and comprises a range of questions, which can vary depending on the exact technique used, to assess an individual’s historical and habitual dietary intake. 108-110 However, computer-assisted systems for automated self-report have also been developed. 111 The DH typically obtains information on commonly consumed meals and snacks, cooking methods, and the frequency with which certain foods and food groups are consumed. 1, 2 The major strength of the DH is the assessment of habitual intake, rather than daily intake, as assessed by the 24-hour recall and FR. Therefore, the DH is a popular method for dietitians and nutritionists in clinical and community settings, 1 but in recent years has become less common in nutrition research, and is not commonly used in dietary surveillance programmes due to the limitations discussed below.

Limitations of a diet history

Substantial interviewer biases may arise due to the variability of skills and techniques used to obtain dietary intake information, and the lack of standardisation can make comparisons between studies challenging. 1, 96 Respondent biases are also problematic as many judgements are required regarding the frequency of consumption of different meals or food. 1 Social desirability biases may affect reported intake during the interview, and, as it is a retrospective method, lapses in memory may cause further errors. 2, 96 The validity of different DH techniques is also hard to assess as there is no independent reference measure for habitual intake. 1, 28 Additionally, information on context cannot be obtained as the DH assesses habitual dietary intake, not details of specific eating episodes.

READ  Current Inlet Pressure Distortion Practice

Sources of error in dietary assessment

There are many potential sources of error in dietary assessment that can be random or systematic.96 Random errors occur with all respondents/participants but can be minimised by increasing the number of observations and/or days diet is assessed. 8, 84 In contrast, systematic errors may be associated with particular respondent/participant characteristics (e.g. among people with obesity), with certain foods (e.g. socially undesirable snack foods), or with specific interviewers conducting the dietary assessment. 84
Extensive work has been conducted to help minimise the potential sources of error in dietary assessment. These include standardising data collection with software and websites, 19 and developing sophisticated portion size guides 22, 27 or food atlases (books that contain images of commonly eaten foods in various portion sizes). 112 However, the traditional methods still rely on self-report without the ability to verify intake, 112, 113 therefore systematic biases in dietary studies are problematic. 1, 84, 114 The common sources of measurement error are briefly described below.
Non-response bias
Non-response or selection bias may arise from use of a sample that does not reflect the population of interest. This can occur due to restraints of the research design, recruitment method, or resources. 96 Moreover, the associated burden of dietary recording for participants can cause a reduced response rate, 115, 116 and therefore people who do choose to participate may not reflect the population of interest.
Respondent bias
Respondent bias may occur if a participant provides socially desirable answers to avoid criticism (social desirability and approval biases), under or over-reports food intake, or changes dietary behaviours during the period of assessment, thereby altering dietary intake. The psychosocial and behavioural characteristics related to respondent bias have been extensively reviewed with a range of characteristics commonly associated with bias: lower leisure physical activity, increased social desirability, fear of a negative evaluation, body size dissatisfaction, recent weight loss, fluctuation of body weight, attempted weight loss in previous 12 months, eating restraint, and eating disinhibition (overeating/loss of self-control over hunger). 28
Interviewer bias
Interviewer bias may result if different interviewers use different techniques to probe for information to varying degrees, intentionally omit questions (to save time), or record responses incorrectly. 96 Thorough training regimens and audits of interviewer techniques can be used to ensure interviewer bias is minimised. 117 Furthermore the development of computer-assisted methods commonly used in large-scale research has helped to minimise interviewer bias.
Incorrect portion size
Errors in portion size can arise in all methods of dietary assessment as participants may fail to record estimate and/or record the portion size correctly. Other errors may arise due to misconceptions of what is considered a “normal” portion size, and differences in food characteristics and volume make some foods more difficult to estimate accurately than others, such as single foods items and complex mixed dishes. 96, 118 27 Reduced accuracy in portion size estimation has also been shown among obese people. 119
Supplement usage
Food and nutrient databases usually contain a limited number of dietary supplements, but due to the number of products on the market and the large differences in dosage and ingredients, supplement intake is inherently hard to assess. The errors associated with supplement intake affect both macronutrients (e.g. protein powders or oil-based supplements) and micronutrients (e.g. vitamin and mineral products).
Coding errors and mixed dishes Coding errors are errors that arise when converting foods into nutrients due to the limited number of foods present in nutrient databases (the coding errors are discussed below in section Mixed dishes are problematic to assess as the specific ingredients and portions may be unknown.

Converting food to nutrients using a food composition database

The dietary intake data (e.g. foods and beverages) recorded during dietary assessments is converted into energy and nutrient values using food composition databases. Food composition databases are usually country specific and provide the macronutrient (energy protein, carbohydrate, and fat) and micronutrient (e.g. vitamins and minerals) values of local foods. The values for every food and nutrient are derived by chemical analysis or can be estimated from other databases. 120 The first nutrition composition tables (paper-based) were published in 1896 by Atwater et al. 121 More recently, electronic formats are available for many countries and are often freely available to view online, or can be accessed and manipulated using nutrition-specific software for nutrient analysis.
There are a range of software packages available to analyse dietary intake data, but these are usually region-specific or only have access to a select number of food composition databases. The software accesses the appropriate database and codes foods and beverages weight/volume with the corresponding macronutrient and micronutrient values. In N.Z., the most common nutrient analysis software is FoodWorks (Xyris software, Queensland, Australia) and was used for the PICTURE study reported in Chapter 7. FoodWorks accesses the electronic version of the N.Z. Food Composition Database (NZFCD) called FoodFiles. 122 The NZFCD is N.Z.’s most comprehensive food composition database and contains information on nutrient data for 59 nutrient components of 2710 commonly consumed foods. The NZFCD is maintained by the N.Z. Institute for Plant and Food Research Limited in partnership with the Ministry of Health. In the U.K., WISP (Tinuviel Software, Warrington, United Kingdom) is a common nutrient analysis software and was used for the UK-based feasibility study reported in Chapter 6. WISP accesses the Composition of Foods Integrated Data Set, which is maintained by the Institute of Food Research in partnership with the Biotechnology and Biotechnology Sciences Research Council. The dataset contains nutrient information for 3423 foods and was last updated in 2002. 123 An updated version is scheduled for

Sources of error associated with food composition databases

Although food composition databases are the only method researchers can use to derive nutrient intakes from dietary data, there are several limitations of food composition databases that need to be acknowledged. Firstly, food databases only contain a limited number of foods, and thus may not contain all foods consumed by respondents. In such circumstances a similar food must be chosen. Secondly, there is a constant change in the food supply as manufacturers produce new products and reformulate existing products. Therefore, it is a challenge to keep the databases current and many manufactured foods are not included. Thirdly, seasonal variations, changes to agricultural practices and natural variability in natural and manufactured products can change the nutrient composition of foods. Fourth, coding errors may result when mixed dishes are reported (or recorded) due to the ingredients in the mixed dishes being different than the ingredients used in the food composition database. Lastly, different food composition databases vary in how foods are described, grouped and analysed. Therefore, it is difficult to make comparisons across databases and/or countries. 125.

Validity in dietary assessment

Validity describes the degree to which a method of dietary assessment measures what it is intended to measure. Errors that affect the validity of dietary assessment methods are usually systematic. The relative or concurrent validity of a method can be assessed by comparing the method of interest with another method of dietary assessment, usually FR and 24h-hour dietary recall. However, caution must be taken when interpreting results from concurrent validation studies as no method of dietary assessment can accurately assess dietary intake in free-living settings, therefore correlated errors are almost certainly present. 8-10, 18


Biomarkers provide an independent and objective measure to assess the validity of specific nutrients as assessed by dietary assessment. At present, the DLW method to assess total energy expenditure (discussed further in section 3.5.2 below), 24-hour urinary nitrogen to assess protein intake, and 24-hour urinary potassium to assess potassium intake are routinely used to validate the results of dietary assessment methods. 128 24-hour urinary nitrogen is the most well-known biomarker, and is used to validate protein intake at the group level. 129 Metabolic studies have demonstrated a moderate correlation between protein intake, and urinary nitrogen excretion when dietary intake is controlled. 130 However, its use relies on the assumption that participants are in nitrogen balance. In other words, nitrogen is not accumulated in the body due to the growth or repair of tissue, or lost due to starvation. Due to the daily variation in nitrogen excretion, multiple 24-hour periods of urine collection are usually conducted to validate protein intake estimated from dietary assessment. 24-hour urinary potassium is a suitable biomarker to assess dietary data as it is abundant in a wide range of foods and 24-hour urinary potassium is highly correlated with dietary intake. 128 Other urinary biomarkers (sodium and iodine), plasma biomarkers (Vitamin C, β-carotene, Vitamin E, and Vitamin D), serum selenium, and folacin are also used. However, these biomarkers are only weak to moderate correlates of intakes and some of the plasma biomarkers can also be affected by smoking and alcohol intake, particularly for those which are prone to oxidation (e.g. Vitamin C, tocopherols, β-carotene, folate), and are therefore better used as a measure of nutrition status. 132

Use of doubly labelled water to assess total energy expenditure

Doubly labelled water contains two (doubly labelled) isotopes 2H and 18O. The isotopes occur naturally in the environment 133 but are concentrated for use in the DLW method, which allows them to be measured as they are turned over. Special attention to DLW water is given in this section as it is the biomarker used to validate EI in the PICTURE study (Chapter Seven). Total energy expenditure is usually assessed for a period of 7 to 15 days using the DLW method, though longer durations are possible. 134 A baseline urine sample is collected before a weight-specific loading dose of DLW (2H20 and H2180) is ingested orally by participants (approximately 60 to 130ml). The baseline urine sample is used to control for background levels of the isotopes 2H and 18O that occur naturally in the environment. 133 After the loading dose, urine samples are collected periodically (approximately 3 to 7 samples collected) over the testing period to measure the isotopes as they are eliminated. The isotopes are measured using ratio isotope mass spectrometry.
The labelled hydrogen deuterium (2H) exists in the body as water (2H2O), and the labelled oxygen (18O) exists in the body as water (H218O) and carbon dioxide (C18O2). 135 133 The 2H is eliminated through water losses in urine, respiration and sweat, and the 180 is eliminated through the same water losses (urine, respiration and sweat) and also as carbon dioxide in expired air  133 Total energy expenditure can be derived as the isotopes (2H and 18O) are turned over at different rates. The rate of turnover is used in calculations to determine carbon dioxide production, which in turn is used to derive total energy expenditure. Total energy expenditure can be used to validate EI data at a group level as the first law of thermodynamics states that energy is conserved. Therefore the energy put into a system (EI) is equal to the energy used (energy expenditure) and/or stored by the system (change in body mass). 137 More specific details of the methods and calculations used in this thesis are presented in the methods section of the PICTURE validation study (see section 7.5).
The DLW method was first developed in the early 1950’s 138 but its application as a method to objectively assess total energy expenditure in free-living individuals was not demonstrated until 1982 by Schoeller and Van Santeen. 139 The DLW method was later validated, with comparisons to direct (whole-room calorimetry) and indirect calorimetry demonstrating non-significant differences (<1%) between methods. 140 It is therefore considered an accurate method to assess total energy expenditure in free-living individuals. The DLW method has been used extensively as the gold standard method to evaluate EI data at a group level in nutrition research.
However, DLW is very costly (approximately $1000 NZD per participant excluding analysis) so is predominantly only used in studies with small of sample sizes between 5 – 50 participants; 9 although some larger DLW studies (N = 450+) have also been conducted

2.1. Overview
2.2. Epidemiology of nutrition-related disease
2.3. Trends in dietary intake
2.4. Cost of nutrition-related disease
2.5. Summary
3.1. Overview
3.2. Dietary assessment
3.3. Food record
3.4. Sources of error in dietary assessment
3.5. Validity in dietary assessment
3.6. Summary
4.1. Introduction to publication
4.2. Author contribution
4.3. Abstract
4.4. Introduction
4.5. Methods
4.6. Results
4.7. Discussion
4.8. Conclusions
5.1. Introduction to publication
5.2. Author contribution
5.3. Abstract
5.4. Introduction
5.5. Methods
5.6. Results
5.7. Discussion
5.8. Conclusions
6.1. Introduction to publication
6.2. Author contribution
6.3. Abstract
6.4. Introduction
6.5. Methods
6.6. Results
6.7. Discussion
6.8. Conclusion
7.1. Introduction to publication
7.2. Author contribution
7.3. Abstract
7.4. Introduction
7.5. Subjects and methods
7.6. Results
7.7. Discussion
7.8. Conclusions
8.1. Introduction to publication
8.2. Author contribution
8.3. Abstract
8.4. Introduction
8.5. Methods
8.6. Results
8.8. Discussion
8.9. Conclusion
9.1.Contributions of this thesis
9.2. Future implications
9.3. Future research recommendations
9.4. Conclusion

Related Posts