*2.3. Training of Field Workers*

Five female field workers were recruited to conduct the interviews, of which two spoke Somali and three Arabic. All of them spoke fluent Norwegian and one of the Arabic-speaking field workers also spoke Kurdish. Thus, the mothers could choose to speak either Norwegian or their own language during the interviews.

The field workers received 1–2 weeks of training on how to conduct 24-h recalls according to the protocol using the forms and tools. Practice took place in pairs and in plenary using different languages. The training particularly emphasized how to ask follow-up questions to make sure all food items were registered, to identify the correct food items and to estimate portion sizes as accurately as possible.

### *2.4. Pilot Testing of the Procedures for 24-H Recall*

The pilot study enabled a full appraisal of all aspects of the 24-h recall procedure. The field worker and the observer recorded data and answered questions regarding the method after each interview using an evaluation form (Table 2). The mothers were asked for their views on the method, including the visual tools, after the second interview.

The dietary data obtained from the 24-h recalls were manually coded and entered by C.M., M.B.A., and N.K.G. in a software system (KBS, database AE-10) developed at the Department of Nutrition, University of Oslo, Norway. The food database in KBS was mainly based on the official Norwegian food composition table. Breast milk intake among the breastfed children was calculated by multiplying the number of feeding events by an estimated breast milk intake per feed of 124 mL. This amount of breast milk per feed was derived from an estimated daily breast milk intake of 497 mL among 12-month old children in developed countries [17] divided by the average breastfeeding frequency in Norwegian 12-month old breastfed children of 4 times per day [9] (497 mL/4 feeds = 124 mL/feed).

As an objective measure of validity for this pilot test, each child's estimated energy intake (EEI) was compared with its estimated energy requirement (EER). According to the Nordic Nutrition Recommendations, estimated average daily energy requirement for 12-month-old boys is 337 kJ/kg and 333 kJ/kg for girls [18]. We did not measure the children's weight, but recorded the weight registered at clinic for the 12-month health check-up (Table 3). However, eight of the children were interviewed at ages between 13 and 21 months and two were interviewed younger than 12 months of age. One of the children younger than 12 months of age had his body weight measured at the 8 month consultation at the child health center. Thus, for 9 of the 11 children with registered body weight, at least one month had passed between the weighing and the 24-h recall. To adjust for this, an estimate of monthly weight gain was calculated using the World Health Organization's growth standards for children between 0 and 24 months [19]. The estimated weight gain for boys 8–21 months of age varies by month and is highest from 8 to 9 months (3.25%) and decreases gradually to 1.76% from 20 to 21 months. For girls, weight gain from 12 to 13 months was found to be 2.64% and from 13 to 14 months average weight gain is 2.46%. Each child's body weight was thus calculated by adding the monthly estimated weight gain to its weight. For the child with no records of body weight registered, the average weight of 11-month old girls was used. Using the estimated body weight at the time of interview, EER for each child was calculated and compared with the EEI calculated from the 24-h recalls, using the mean intake of the two recalls. Differences between EER and EEI were tested with paired samples *t*-test. Bland-Altman plot [20] was used to visualize the dispersion between EER and EEI. Linear regression analysis was applied to study whether there was any relationship between the mean of the estimates EER and EEI and the difference between the two estimates.


**Table 2.** Evaluation form for the pilot study.


**Table 3.** Estimated energy requirements and energy intake among children (10–21 months) with mothers from Iraq (ID 1–5) and Somalia (ID 6–12) living in Norway.


a Estimated body weight at time of interview calculated based on average growth rate from World Health Organization's growth standards [19] multiplied by number of months between time of weighing and time of interview; <sup>b</sup> Estimated body weight at time of interview multiplied with estimated average requirement per kilogram; <sup>c</sup> Mean estimated energy intake of the two recalls; <sup>d</sup> Calculated as percent difference of mean. Difference between EER and EEI tested with paired samples *t*-test: *p* = 0.58; <sup>e</sup> Body weight not registered. Average weight for girls at 11 months of age used as reference [19]; <sup>f</sup> Body weight at 8 months of age; <sup>g</sup> Calculated using absolute values of percentage differences

#### **3. Results**

#### *3.1. Subjects*

A total of 28 Somali-born mothers were asked to participate in the pilot study, and 13 consented. However, only seven of these showed up to the appointed interview. Of the fourteen 24-h recalls, eight were conducted in Norwegian and six in Somali. Likewise, 48 Iraqi mothers were contacted, seven consented, but only five showed up at the interview. Five of the 24-h recalls were conducted in Kurdish, four in Arabic and one in Norwegian. Among the 12 participating mothers, mean age was 31 (range 22–42) and the average number of years lived in Norway was 15 (range 3–24). Three mothers had no education from Norway. Two of them had, however, completed a Norwegian course. Seven mothers had completed high school education and two had completed higher education. Seven mothers had more than one child.

#### *3.2. Results from the Evaluation Form*

The mean (minimum-maximum) time spent on the total 24-h recall interviews was 47 (20–75) min. There was a decreasing trend in the time spent on each interview conducted over time in both the Somali and Iraqi groups. The repeat interviews were conducted by the same field worker, except for three of the Somali mothers and one Iraqi mother where two different field workers conducted the interviews.

The field workers spent an average of four minutes during each interview showing pictures on the iPad. The visual tools were used during all interviews. Of the 19 folders in the picture library, eight were used by the Somali mothers to identify foods. The most frequently used folders were "baby cereal" (seven interviews) and "oils and butter" (four interviews). Among the Iraqi mothers, pictures from 13 of the 19 folders were used. The most frequently used folders in this group were "breads" (eight interviews) and "baby cereal" (four interviews). Although the mothers browsed through all folders to identify foods given to the child, none of the Somali or Iraqi mothers used or identified foods from the five folders "ready-made meals", "fruits and vegetables", "soda", "squash, lemonade, *etc.*" or "meat".

Eleven of the 17 colored photograph series were used by the Somali mothers to estimate portion sizes eaten by the child. The portion sizes of baby cereal and butter were the most frequently used total in 10 and 9 of the 14 interviews, respectively. The Iraqi mothers used 10 of the 17 photograph series of portion sizes to estimate foods consumed by the child. The most frequently used series were the portion sizes for milk and butter, which were referred to in 9 and 5 of the 10 interviews, respectively. The measuring equipment was used together with the photographic booklet in the first interviews, but over time the interviewers favored the photographic booklet over actual measurements. Reasons given for this shift were that measurements were time consuming and difficult to use when the interviews were conducted outside the informants' homes. When mothers were asked to identify amount with both the photographic booklet and by measurements of actual foods, these seemed to correspond well.

The protocol was mostly only used during the last pass, when the field workers were going through the checklist of foods and beverages often forgotten. When the interviews were conducted in Norwegian, the observers noted that the field workers consistently asked about added foods/ingredients, brands and amounts consumed. It was sometimes difficult for the field workers to write down recipes and cooking methods because of limited space on the forms. However, the amount of food eaten by the child was usually asked about and written down clearly.

The mothers expressed that the picture library was a good tool to be reminded of and to identify the type of foods given to the child. It was especially useful for remembering brand names. Among pictures missing in the picture library, some mothers mentioned different types of rice, fruit purees, bread spreads, breads, butter, baby cereals, and yoghurts, as well as Weetabix and prunes.

One of the topics that emerged repeatedly was how difficult it was to estimate portion sizes. Six mothers mentioned the difficulties in estimating the amount of bread eaten by the child, without pictures of bread in the booklet. Other pictures of portion sizes mentioned as missing were lasagna, spaghetti and pancakes. Five mothers expressed that pre-packed industrially produced foods were easier to estimate. All mothers found that the portion size options in the photographic booklet matched amounts the child usually ate. Ten mentioned that it was easier to show amounts of foods eaten using the booklet, whereas illustrating amounts of beverages was easier using the measuring equipment.

Most mothers said that the day of interview represented the typical foods given to the child, only three recall days were considered non-representative for the typical foods given. Two foods (bread and bulgur) were mentioned by two mothers as being typical foods given to the child, but not during the days in question. Pancakes (*anjera*), Weetabix and juice were mentioned by two Somali mothers as cultural relevant foods often given to children, while four Iraqi mothers mentioned different types of staple foods and vegetables, such as okra.

All field workers found the protocol easy to understand. Finding pictures in the picture library took time to begin with, but became easier after a few interviews. The photographic booklet was judged as a good tool for estimating portion sizes. However, similarly to the mothers, they missed portion size pictures of pasta and bread.

The registration form was described as clear and easy to understand, but the field workers missed more space to write down recipes.

### *3.3. Results from the 24-H Recalls*

Six mothers were still breastfeeding their children (four Iraqi and two Somali mothers). The average breastfeeding frequency was two times per day (range one to four times per day). Foods given to the children included bread, porridge, different fruit and vegetables, snacks and supplements. The porridge was either industrial produced or home-made using oatmeal or bulgur. A couple of mothers also added margarine, olive oil, salt and/or sugar when preparing the porridge. Some of the foods registered that are not commonly given to Norwegian children were nan bread, feta cheese, Turkish delight, bulgur porridge, pancakes (*anjera*), and seeds. None of the mothers reported using ready-made dinners, as most of them mentioned that they did not trust the contents and the ready-made dinners were not considered to be fresh. The home-made dinners often constituted of different staple foods, vegetables, meat, and fish.

#### *3.4. Results from Energy Intake Estimation*

Table 3 presents the comparison of each child's EER with its EEI calculated from the 24-h recall. The percent difference between EER and EEI was in the range of ±0%–10% for five of the children and in the range of ±11%–20% for four of the children, whereas for three of the children the percentage of difference between the two estimates was ±38%, 41% or 45%, respectively. The mean (SD) for EER and EEI was 3407 (527) kJ/day and 3543 (773) kJ/day, respectively, and the difference was not significant, as tested with paired samples *t*-test. The Bland-Altman plot (Figure 1) showed large individual variations in the differences between EER and EEI but no clear pattern. A linear regression analysis testing the relationship between the mean of the estimates EER and EEI and the difference between the two estimates was not significant, *p* = 0.24. This indicates that the difference between the two estimates is not related to the magnitude of the estimates. Excluding child number 4, for whom there was no registered weight, did not change the results.

#### **4. Discussion**

For the InnBaKost study, we developed a protocol for a 24-h recall procedure, including a picture library to assist in identifying the correct foods eaten. In addition, a photographic booklet was used for portion size estimation. Although the latter approach has become a common method for portion size estimation [21,22], including dietary assessment in children [23,24], the use of a picture library is a rather novel approach. The hypothesis was that the picture library would be a useful tool to identify the correct food and brand, particularly for dietary assessment among immigrant mothers with varying levels of language and literacy skills.

A review conducted by Burrows *et al.* (2010), indicates that weighed food records provide the best dietary estimates for younger children aged 0.5 to 4 years, while 24-h multiple-pass recall that uses parents as reporters is the most accurate method to estimate total energy intake in children aged 4 to 11 years [25]. The weighed food record method requires both motivation and good literacy skills and is often time-consuming. Thus, the method has been considered to be less suitable for dietary assessment in immigrants, as the method has led to misreporting and dropout in immigrant groups due to the burden and time consumption the method carries [26]. The face-toface FFQs and multiple-pass 24-h recalls are reported to be the two most frequently used methods with immigrant populations in Europe [6]. The 24-h recall is more flexible because it can capture all foods and beverages consumed the preceding day, with no assumptions about the food culture or dependency on literacy levels. In addition, as seen from the few recalls in this pilot, some mothers gave selected atypical foods to their children and mostly made home-made dinners, which may vary from the general Norwegian population in regards to composition and preparation method. The 24-h recall has therefore been recommended as the most optimal method for many immigrant groups and is considered to provide valid information among children [26,27]. In addition, the interactive nature and the personal contact of the method may contribute to more reliable data collection, although social desirability bias may cause some misreporting [28]. The multiple-pass technique is considered to give the most exact estimates, and limit misreporting, because the probing questions encourage the respondent to remember more of the foods consumed [16]. The respondent burden is usually small compared to weighed records [29].

The protocol was used sparingly during the interviews, because the field workers expressed that they already knew the content in the protocol and that it was difficult to focus on the protocol while registering the child's food consumption. Thus, it was recommended that important guidelines from the protocol could be included in the 24-h recall registration form instead. The decreasing time spent on the second interview with each mother was mostly due to the mother being more prepared and that the background information was already collected. Another reason for the decline in time spent may have been that the field workers became more familiar with the method and navigated the picture library and photographic booklet more easily. The measuring equipment was initially used together with the photographic booklet to see how well both measurements corresponded, but both the field workers and mothers expressed that it was too time-consuming.

Both the mothers and field workers reported the picture library to be a good tool to identify foods given to the child. It was mostly used when the interviews were conducted outside the respondents' homes because the mothers could show foods available in the home. The use of a picture library similar to this has not been described by many; however, the use of photo images has been reported to be useful as a memory aid for respondents during 24-h recalls [30,31]. The picture library seemed to strengthen the mothers' ability to report the correct food and reduce misunderstandings. However, the pilot study revealed many desired additions to both the picture library and booklet.

Portion size estimation is one of the main challenges in dietary assessment studies. Estimating amounts eaten other than direct weighing may contribute to a source of error, both among children and adults [7,22,24,32]. The photographic booklet was considered to be a good tool for estimating portion sizes among the field workers and the mothers, as has also been reported in several other studies [21,24,30,33]. A study by Lillegaard *et al.* (2005) showed that children and adolescents could accurately estimate portion sizes of pre-weighed foods by viewing photographs, approximately 60% of the comparisons were made correctly [24]. The estimations were more accurate when the served portions had the exact appearance as the food portrayed in the photographic booklet [24]. Thus, the arguments can be made that more picture series in our photographic booklet may be favorable rather than using pictures of similar foods. The studies further emphasize the importance of validation studies to test the applicability of photographs for estimating current portions and actual consumption [21,22], especially among immigrant groups [13]. This was not done in this pilot study, but should be considered in the future.

Assessing children's food intake accurately can be difficult for a number of reasons. Infants and toddlers cannot account for their food intake, but parents are seen as reliable sources when affirming their children's consumption of food [25,34]. Efforts should be made to assess foods eaten outside the home or with other caretakers; for instance, at the kindergarten or with family members. A possible challenge may be that the level of reporting and motivation may vary for each caretaker [7]. In the pilot, one of the Somali fathers was on paternity leave and was in charge of the child's diet at the time; therefore, he was interviewed together with the mother. Among the Iraqi mothers, only one mother reported that her child had spent much of the day with a nanny. Although, this did not apply for many of the mothers in the pilot, it should be taken into consideration for larger studies and dietary assessment of somewhat older children. Potential solutions may be to ask the mothers prior to the interview if the child has other caretakers and if it may be possible to include them to obtain information about their child's food consumption during their supervision.

In regard to EEI, it seemed to correspond well with the EER for most of the children (within ± 2 SD of the average of the two estimates) except for three. The comparison of EEI and EER has some weaknesses and can only give an indication of whether the method is suitable for capturing habitual energy intake on a group basis. First, each child's EER might not reflect the true energy requirement of the child because an energy requirement is highly variable between children of the same age and weight [35]. There is also intraindividual variation in energy requirement for children, depending on their physical activity level and growth rate [35]. Second, the energy intake measure was simply averaged over the two days without adjustment for intraindividual variation over time. Thus, it may not be representative of habitual energy intake [36]. Although the sample size was small, it was encouraging that there was no consistent over- or underreporting of EEI compared to EER.

Recruitment of study participants in itself was challenging and time consuming in this pilot study, as it was difficult to come in contact with the target group. This was mostly due to wrong contact numbers registered on several mothers when tried to reach by phone. Some reasons for refusals were that they were not interested, skeptical, or had to consult their partner. It was necessary to seek the mothers through several methods and many did not show up to appointed interviews. The use of bilingual field workers was an advantage and enabled the recruitment of mothers who did not speak Norwegian. Challenges related to recruitment when conducting dietary studies with immigrants have previously been reported [6,26]. Most studies conducted with a European immigrant population group have also used nonprobability sampling methods, such as the convenience sampling method [6]. The need for extra effort in recruiting participants has been described, such as using bilingual field workers, involving key leaders and including places of worship and media, to overcome cultural barriers and ensure representativeness [6]. Although the convenience sampling method may lead to the inclusion of highly motivated participants, there seemed to be variations in the background characteristics of the mothers included in the pilot.

Based on the pilot study presented, some suggestions were made for improving the 24-h multiple-pass recall method. Observations of the interviews showed that the field workers were not actively using the protocol, and a possible solution is to incorporate the protocol into the registration form. Other important suggestion were to include more pictures in the library and supply the photographic booklet with portion sizes of bread in particular, but also of foods such as lasagna, pancakes and other portion sizes of meat, fish, fruits and vegetables. Furthermore, a more thorough training and follow-up of the field workers would be required to increase the quality of the data collection.
