Examination of Varying Statistical Methods to Determine Meal Patterns Using Data From the National Adult Nutrition Survey (NANS) Ireland.

The study described here applies a variety of statistical methods to develop strategies to determine meal based dietary patterns within nationally representative food consumption databases. The work will apply methods such as Principal component analysis, latent class analysis, basket set analysis, topic models and others to explore methods to accurately identify meal based dietary patterns for use in personalised nutrition.

The work will initially focus on National Food Consumption Data collected within Ireland, specifically using the National Adult Nutrition Survey (NANS), with details of data collection fully outlined below.

Within NANS dietary, demographic and phenotypic data was collected from a Nationally representative sample within Ireland from 2008-10. The fieldwork phase was carried out between October 2008 and April 2010. Adults aged 18-90 years (740 males, 760 females) were recruited in the Republic of Ireland. Eligible respondents were adults aged 18 years and over who were free-living and who were not pregnant or breast feeding. A sample of adults was randomly selected from a database of names and addresses held by Data Ireland (An Post). An introductory letter and information leaflet were posted to each person selected from the database. A researcher called to potential respondents" homes to introduce the survey and invite participation. If the individual agreed to participate, a consent form was signed and the survey commenced. If the person was not at home, the researcher called on three more occasions on different days and at different times, before deeming them ineligible. For groups that were not highly represented in the database, particularly 18-35 year olds, a second level of recruitment was introduced. A second database of names and addresses was compiled through referrals from respondents and participation was invited for those that were contactable.

Food intake data collection - A four-day food diary was used to collect food and beverage intake data. The researcher made three visits to the respondent during the four-day period: a training visit to demonstrate how to keep the food diary and how to use the weighing scales; a second visit 24-36 hours into the recording period to review the diary, check for completeness and clarify details regarding specific food descriptors and quantities; and a final visit one or two days after the recording period to check the last days and to collect the diary. Respondents were asked to record detailed information regarding the amount and types of all foods, beverages and nutritional supplements consumed over the recording period and where applicable, the cooking methods used, brand names of the foods consumed and details of recipes. Data were also collected on the time of each eating or drinking occasion, the respondent's definition of each eating or drinking occasion (e.g. morning snack, lunch) and the location of the preparation of the meal or snack consumed.

Food quantification and coding - A quantification protocol that had been established by the IUNA (Irish Universities Nutrition Alliance) for the North/South Ireland Food Consumption Survey (NSIFCS) (Harrington et al., 2001) was updated for the NANS. It is summarised as follows: (1) Weighed by respondent/manufacturer weights) - A portable food scales (Tanita, Japan) was given to each respondent. The researcher gave detailed instructions (including a demonstration) as to how to use the food scales during the training session. This method was use to quantify 46% of foods and drinks consumed. A further 10% of weights were derived from manufacturer's weights. To facilitate collection of such data, researchers asked respondents to collect all packaging of food and beverages consumed in a storage bag provided. (2) Food Atlas - A photographic food atlas (Nelson et al., 1997) was used to quantify 16% of foods and beverages consumed. (3) IUNA Weights - Average portion weights that had been ascertained for certain foods by the IUNA survey team were used. This method was used to quantify 4% of foods and beverages consumed. (4) Food Portion Sizes - "Food Portion Sizes" (Ministry of Agriculture, Fisheries and Food, 1997) was used to quantify 11% of foods and beverages consumed. (5) Household Measures - Measures such as teaspoon, tablespoon, pint etc. were used to quantify 11% of foods and beverages consumed. (6) Estimated - Food quantities were defined as estimated if the researcher made an estimate of the amount likely to have been consumed based on their knowledge of the respondent's general eating habits as observed during the recording period. This method was used to quantify 2% of foods and beverages consumed. Nutrient composition of foods and estimation of nutrient intake - Food intake data were analysed using WISP© (Tinuviel Software, Anglesey, UK). WISP© uses data from McCance and Widdowson's The Composition of Foods, sixth (Food Standards Agency, 2002) and fifth (Holland et al., 1995) editions plus all nine supplemental volumes (Holland et al., 1988; Holland et al., 1989; Holland et al., 1991; Holland et al., 1992; Holland et al., 1993; Chan et al., 1994; Chan et al., 1995; Chan et al., 1996; Holland et al., 1996) to generate nutrient intake data. During the NANS, modifications were made to the food composition database to include recipes of composite dishes, nutritional supplements, generic Irish foods that were commonly consumed and new foods on the market. All previous modifications to the food composition database were also checked and updated from current manufacturers" information as necessary. The above modifications and additions comprise "The Irish Food Composition Database" (Black et al., 2011).