Online focus groups

Key informant interviews

Desktop analysis

Desktop analysis involved examination of the format of six FCDB. Aside from the major Australian databases, NUTTAB 2010 and AUSNUT 2011–13 [13], databases from the US [19], UK [20], Canada [21] and NZ [22] were included as they are current, publically available, published in English and likely to be used by Australian FCDB users due to similarities in their food supply systems [11, 19]. The formal website of each country’s database was used to assist in gathering information surrounding format, related to the themes extracted from schematic analysis. The webpage and supporting documents were examined to extract details. Relevant components of each database were identified and tabulated.

Database use

Database use and its capacity to meet the needs of the user were reported to be influenced by the type, purpose, choice, and usability of a database. This study identified that due to the wide variety of uses and users of FCDB, finding or developing a database to meet the needs of all professionals is challenging.

“We have so many different users that have such different needs; it’s really difficult to develop… a product that really meets all of them. So there’s always … compromises and frustrations for users” (Interview 2)

This is supported by Egan et. al. [3] who states that as the users and applications of FCDB are largely varied; this presents a significant issue when identifying standards for format. Conversely, from a user’s perspective this makes database choice and usability equally challenging.

An Australian example identified in the focus groups and interviews is that the primary purpose of FSANZ is in food regulation. Hence, the development of food composition data is driven by regulatory needs. This creates constraints for users of this data for purposes other than regulation of food standards or the Australian National Health Survey [11].

Furthermore, this reiterates the importance of understanding a databases purpose, and hence it’s likely limitations [5]. Additionally, this study has highlighted that many database users are reportedly not conversant in making informed choices, as to which database to use, as well as understanding the differences between a database and its supporting software. Whilst many studies [12, 23–25]. provide supporting information on how to make an appropriate database choice, it is apparent from this study that further training is likely still required. This is further supported by McCabe-Sellers and Chenard [10] who came to a similar conclusion when examining the FCDB needs of US dietitians.

Framework

Data presentation, database layout and software components and functions were all closely linked with formatting. Few concerns were raised about the database layout from a users’ perspective, in this study. Participants were generally comfortable with the table (columns and rows) design and the associated file formats e.g. Excel. It is also likely that user’s associate software design with FCDB and do not differentiate the two tools. However, in contrast from a FCDB development perspective:

“Making decisions about whether to combine data sets, replace data sets ….. about how to actually present the information to the public…are huge challenges” (Interview 2)

This is supported by Stumbo [25] who highlights that how a software program summarises data entry, food components and meals were an important consideration. Secondly, ensuring the output generated is presented in a way that is meaningful to users is critical [25, 26]. It must be emphasised, that users need to understand the difference between software and the FCDB, which the software draws on.

One suggestion for data presentation that arose from this study was to provide a range for each data value rather than one number. The purpose being to better represent and emphasise the variability of nutrients in food, especially phytochemicals. However, a range could present difficulties when calculating and applying data for use. Pennington et. al. [5] suggests that by providing the standard deviation and information on the number of samples analysed, users may obtain a better understanding of the data. Additionally, from the literature, the search functions of a software package were identified as a key framework component, which is also tightly linked with food classification [23, 25, 27].

Food classification

How foods are grouped and named in a food composition database was highlighted as a major theme in this study. Interpretation of food descriptors presented a challenge to many participants, as identified by the quote below.

“actually understanding how users, use the data, is important I think in looking at the descriptions, often I think maybe there isn’t enough, consultation with the users about the descriptors and maybe that’s something that could be used to make it better and …simpler for people to use than it has been in the past” (Focus group 2)

Similarly, Charrondiere and Burlingame [28] identified the accurate and unambiguous description of food as an important component of FCDB, noting that descriptors commonly include a unique abbreviation, definition, common and scientific names of a food item [28]. Stumbo [7] also identified food description as an area of difficulty in database structure. Stumbo [7]explains that food nomenclature differs depending if a database is a reference or survey database. This was further highlighted in the desktop analysis (Table 3), where food nomenclature systems were found to differ between reference and survey databases, as well as between countries. Additionally, alternative names, language and spelling can be a limitation [2, 25, 29]. Many food nomenclature systems are currently available, such as LanguaL, [29, 30]. These systems aim to make database searching and food classification more user friendly. However, these systems have yet to be fully incorporated into an Australian database [9, 30]. Though it should be noted that FoodWorks software (Xyris Software Australia PTY. LTD.), which draws upon NUTTAB and AUSNUT databases, utilises INFOODS tagnames. Additionally, a recent reflection on the new Czech FCDB identified that visual descriptors may be an effective and simple means to make a databases written food descriptions more user friendly [31].

Accessibility and Availability

Accessibility and availability of food composition data and databases can be influenced by location and proprietorship. The idea of a central repository, a place where all food composition data, from all sources could be stored and accessed produced mixed responses from participants. One common viewpoint was the capability of a central repository to improve the amount of readily available data and the associated advantages.

“I think it is ….. an important thing to have it compiled in the one spot and easy to access … for all people using… that information” (Interview 4)

However, in contrast, concern was raised surrounding resources, data access, standards and accuracy, including possible misinterpretation, of pooled data. Internationally, an example of a central repository is the USDA Nutrient Databank System [32]. This system methodically captures and aggregates nutrient information from various sources, such as lab analyses, food industry and scientific literature. This information is then used to provide data to many US databases, including the Standard Reference database and the My Pyramids Equivalent database [32]. Other examples include; EuroFIRs FoodEXplorer online tool allows users to simultaneously search 28 national FCDB’s [2]. Whilst, this multicentered system provides a more cost-effective strategy for international data distribution, than a single multinational system, it relies on all participating databases proprietors maintaining the quality of their individual FCDB’s [33]. Additionally, it may be beyond the needs of the average FCDB user.

Two index systems are also available internationally, the International Nutrient Databank Directory [23], managed by the US National Nutrient Databank conference and the International Food Composition Tables Directory [23], upheld by INFOODS. These two systems list available FCDB and software packages of various companies and governments, with the aim to assist users in choosing the most appropriate FCDB for their needs [23].

Data derivation

Data derivation methodology and associated data quality was an area of importance highlighted in this study. This included consideration of sampling and analytical techniques, data imputation as well as clearly documented analyses.

“It’s like a gold standard … its very time consuming and very expensive …. all the different places you have to sample and the different food outlets … to get the data for the database” (Focus group 1)

Stumbo[7] also noted that a key component of database structure is data derivation. Similarly, a study by Egan et al. [3] found analytical methods and documentation of methods to be important components of standard database criterion.

The quality and accuracy of data was found to be largely associated with methodology in this study. Concern was raised around the types of methods chosen and the use of imputed data.

“borrowing data from other countries… particularly borrowing only some of the data… I don’t think that’s… an accurate way of doing things” (Interview 5)

It was also expressed that careful consideration of the variability of a foods composition should be applied when choosing analytical techniques and imputing data, as well as understanding any associated database limitations. This further reiterates that the analytical methods used to obtain food composition data greatly influence the intended use and application of a FCDB, especially as data quality and accuracy are inherent to the reliability of dietary interventions, nutrition policies and research into diet-disease relationships [12]. Various methods of analysis, imputed data and recipe calculations are commonly used to provide food and nutrient data amongst all the current national databases [34]. Transparent documentation of assumptions and methods applied is therefore crucial to database users, and hence, should be incorporated as part of any database development and publishing, especially in relation to food fortification [34]. Finglas et al. [2] also notes that the desire for complete datasets has led to increased borrowing of data, and hence information on the data’s origin is crucial to ensure the borrowed data is appropriate. Currently, the national databases all provide supporting documentation. It is unclear from this study if these documents are regularly sought and used, highlighting an area for further research. Reasons for not using supporting documents may be: time poor or difficulty locating, especially if using dietary software, which does not highlight these limitations. Additionally, user understanding of the data and databases and awareness of supporting documentation may also limit use. EuroFIR’s FoodEXplorer online tool returns information on both the food item and its supporting documentation, for example; method of analysis. This is an ideal method to ensure all users are aware of supporting documentation and data derivation [2].

Another common viewpoint expressed in relation to methodology was the cost and time required to produce accurate data, as well as the resources required to use and develop the gold standard techniques.

“It also depends on … resources, time …because some of the…ideal methodologies are just not accessible for small group research” (Focus group 1)

Resources, particularly expenses are identified as a limitation by Harrison [34]. Funding is essential to maintaining, updating, developing and expanding FCDB in line with evolving knowledge in nutritional epidemiology, and diet disease relationships [34].

Guest et. al. [35] suggests at least six interviews are required for theoretical saturation of results. This was achieved in this study with six key informant interviews, complemented by three focus groups. However, despite this, we acknowledge that participation numbers were still small, and participants were mostly from New South Wales and were dietitians. Therefore, caution should be taken when extrapolating these results to users other than Australian dietitians. The low number of respondents may have been due to placement of advertisements in newsletters, which was largely outside the researcher’s control. [35] A longer recruitment period or conducting the focus groups in the evening or on the weekend may be useful for increasing participation rates in the future. Consideration should also be given to patient gifts for participation.

The use of online focus groups was beneficial in its potential to allow nation-wide participation and to include a variety of practices areas. However, online focus group methodology also presented difficulties in observing non-verbal cues, e.g. hand gestures and participant-participant interactions, key in face-to-face focus groups [36]. Technological issues such as time delays, compatibility of software programs and internet broadband also limited participant interactions and in some cases involvement in the study. Technical support (YP) typed suggestions to participants to assist in improving audio visual quality. Transcription also proved difficult when quality of Internet connection or participants computer microphones produced poor sound quality. Therefore, there is a small possibility that important data may have been missed. Literature available for online focus group methods is limited, and often refers to typed responses as part of an asynchronous discussion forum [37, 38]. Further research into improving data collection via online synchronous focus groups should be considered.

Clarification of the results from this study could be obtained by a short survey of a larger group to ensure data reliability and empirical generalisability. Focus should also be given to user’s knowledge of databases, as opposed to the software used, and their understanding of database limitations, including the use of supporting documentation. Additionally, a prototype database with data values presented as a range could be trialled with users.