Advancing dietary assessment in nutrition and dietetics

Accurately measuring dietary intake remains one of the most persistent challenges in nutrition and dietetics research and practice. Dietary behaviours are complex and dynamic, influenced by people's food preferences, habits, hunger, emotion, cultural contexts and access.1 Being able to capture what people eat, how they eat, when they eat and why requires tools that are scientifically robust and tailored to the study's purpose. Choosing the right dietary assessment method or combination of methods—that are reliable, valid and appropriate for the question being addressed—is of fundamental importance. Without such methods, there is a risk of collecting poor-quality data that can mislead research findings and compromise clinical care. A variety of dietary assessment methods are available for use in the public health and clinical settings, each with its own strengths, limitations and appropriate applications.2 The diversity of methods featured in this issue of Nutrition and Dietetics reflects this breadth: Hammad et al.3 used 3 days of 24-h recalls in a haemodialysis population to compare nutrient intakes between dialysis and non-dialysis days; Yap et al.4 applied a food frequency questionnaire (FFQ) and short dietary questions to explore dietary behaviours in adults with inflammatory bowel disease; Hoch et al.5 extracted data from medical records to evaluate perioperative nutrition practices in patients undergoing a pancreaticoduodenectomy; and Fisher et al.6 leveraged clinical nutrition informatics to predict malnutrition risk in a hospital setting. In addition, nutrition risk screening tools were used in two studies. Popiolek-Kalisz et al.7 used the well validated Nutritional Risk Screening 2002 (NRS-2002) tool to examine the association between nutritional risk and length of stay in patients undergoing coronary angiography, while Teasdale et al.8 evaluated the feasibility of a newer tool, the NutriMental screener in Australian mental health settings. The traditional methods—food records, 24-h recalls, FFQs and diet histories—have formed the foundation of dietary assessment for decades. More recently, many of these tools have been digitised, making it easier to collect standardised dietary data with less burden on researchers. For example, automated self-administered 24-h recall systems such as ASA24-Australia9 and Intake2410 offer standardised multi-pass protocols. Shorter recall blocks (i.e., 2 vs. 24 h) spread across different days and times of day are another variation of the 24 h recall showing promise combined with ecological momentary assessment (also known by the acronym EMA).11 Mobile dietary tracking apps (e.g., EasyDietDiary, along with many others) enable users to record real-time intake with a user-friendly interface. These tools are generally preferred over paper-based alternatives by children, adolescents and adults comfortable with digital technology. However, for people with lower computer literacy, cognitive impairment or language barriers, these digital tools could limit feasibility. Self-reported dietary data has attracted considerable criticism over the years. Errors in dietary reporting, particularly underreporting of food intake, can substantially compromise data quality.12 Yet, as Subar et al.13 have argued, when used appropriately, self-report tools have considerable value and have played an essential role in identifying diet–disease relationships and in dietary guidance. However, it remains essential to recognise and transparently report the limitations of any method used, ensuring that data are analysed and interpreted appropriately. It is recommended that tools should undergo validation, ideally against objective biomarkers, and ongoing research is needed to refine and expand methodological approaches.13 Some limitations in self-reported dietary data collection are difficult to overcome. People forget meals or snacks, or have poor knowledge of ingredients in a composite meal, particularly when the food is purchased outside of home. Social desirability may be a factor when reporting intake, and changing eating behaviours when recording food intake is common. Portion size estimation is difficult for most people, particularly for amorphous foods (that take the shape of the plate or bowl), meat and spreads.14 Additionally, with over 20 000 distinct food and drink items in a typical Australian supermarket, being able to identify specific products and nutrient compositions remains a major barrier. Emerging technologies such as wearable sensors, image-based analysis, and advanced informatics platforms or big data may be able to overcome some of the inherent limitations in traditional self-report methods. For example, wearable devices—such as wrist-worn accelerometers, glasses, ear-worn microphones or neck sensors—are being tested to detect chewing, swallowing, or eating events using combinations of gyroscopes, accelerometers, microphones, pressure sensors and cameras.15 Algorithms are continually being refined to improve the detection of eating occasions and distinguish between different types of intake. Sensor-based data can offer a more granular understanding of eating behaviours; not only dietary intake but also the timing, frequency and duration of eating episodes. Continuous, real-time data collection reduces participant burden and may be particularly useful in understanding contextual factors that influence food intake, such as emotional state, social environment or time constraints. Image-based dietary assessment is also progressing, allowing users to capture photos of meals that are then analysed using machine learning to estimate food type and portion size. Advances in artificial intelligence have improved the classification and quantification of foods in digital images, with implications for both research and clinical practice.16 The growing availability of large datasets or 'big data' presents new opportunities in dietary assessment and monitoring.17, 18 Big data is typically characterised by volume, velocity and variety,17 and can include data sources such as supermarket purchases, geographic information, mobile health apps, social media activity as well as electronic health records and biological data. For example, metabolomics (metabolite profiling) could be used to complement well-validated dietary assessment methods,19 and data related to the microbiome, genome, lifestyle, environment and social determinants of health could be used to improve statistical analysis in nutrition research.17 Nutrition informatics dashboards are another example of big data that can be used in dietary assessment and monitoring. In this issue, Fisher et al.6 used a hospital food service monitoring system as an innovative opportunity to screen and identify patients at risk of malnutrition.6 The informatics dashboard, with food ordering and food intake data (collected by food service staff), in combination with demographic and admission details, enabled the authors to model and predict the risk of malnutrition based on patients' energy and protein intakes. Thresholds of up to 6000 kJ and 65 g protein were associated with an increased risk of malnutrition. This study is a great example of how nutrition informatics can be used to enhance care delivery. Choosing the appropriate dietary assessment method to capture dietary intake depends on the purpose of the study (what is being measured, why, and in whom) and the resources available. Is the purpose to evaluate intakes of food types, food groups, nutrients or eating occasions? Is habitual or usual intake required, or is a short-term snapshot sufficient? Is the target population young, old, in good or poor health, from a diverse cultural background or digitally literate? These questions can guide the selection of the method and help balance scientific rigor with the more practical constraints, including participant burden and cost. A recent scoping review on dietary assessment methods in a hospital setting involving 155 studies identified estimated plate waste as the most frequently used method, followed by food records and 24-h recalls. The use of novel technologies was notably limited, with data collection mostly paper-based.20 Of concern, only 15% of the studies had used validated tools. These findings underscore the need to strengthen the methodological standards for assessing dietary intake in clinical nutrition research. Validation of dietary assessment tools is essential before use, and best practice in validation is outlined by Kirkpatrick et al.12 and Cade et al.9 Validation refers to the process of evaluating how accurately a tool measures dietary intake by comparing it to a reference standard. Ideal reference methods are objective and based on recovery biomarkers such as doubly labelled water for energy expenditure or 24-h urinary nitrogen for protein intake. While these biomarkers reflect absolute (or 'true') intake, they are expensive and logistically complex. Concentration biomarkers (e.g., serum carotenoids and serum folate) are more common but do not reflect absolute intake and can also add significantly to study costs. Therefore, many validation studies rely on comparison to another self-report tool (e.g., multiple 24-h recalls), which provides a measure of relative validity. This approach, although less robust, can help determine whether a tool can correctly rank individuals by intake levels and detect between-group differences. It is important that tools are validated for the intended population. A method validated in healthy adults may not necessarily perform well in elderly people with cognitive impairment, or in culturally diverse communities. Similarly, applying a tool developed in the United States to an Australian population without any modification can result in misreporting due to differences in food supply, portion sizes and language variation. Despite the use of validated dietary assessment methods, data should be evaluated for plausibility prior to analysis. Misreporting of energy intake is prevalent and typically identified via the Goldberg cut-off or by comparing reported intake to estimated energy requirements.21 Current recommendations advocate for sensitivity analyses based on reporting status rather than excluding implausible reporters, enhancing transparency and enabling assessment of result robustness.21 To conclude, good quality dietary data comes down to careful planning, method selection and transparent reporting. Reporting guidelines such as STROBE-nut provide a useful framework for describing how dietary data were collected, processed and analysed.22 This includes naming the assessment tool used, describing portion size estimation, validation, the food composition database used and explaining how implausible data were handled. Advancing nutrition research requires dietary assessment methods that are rigorous, transparent and fit for purpose. Maintaining high methodological standards will ensure nutrition research continues to inform and improve health outcomes. The author declares no conflict of interest.