Healthy, 3 year old Swedish children (N=30) were recruited in 2010-2011 [10-12]. Their mean TEE was measured for 14 days using the DLW method [11,12]. During this period, the children’s intakes of foods and drinks were assessed using TECH, and parents completed the KidMeal-Q. A complete data collection was obtained from 30 children. The mean change in body weight between day 1 and 14 was -0.008 kg (SD 0.317). All children originated from a well-educated, middle-income population. The study was approved by the Research Ethics Committee, Linköping, Sweden. Written informed consent was obtained from all parents.

Parents and other caretakers were instructed to take pre- and post-meal photographs of all the food items and beverages consumed by their child during one 24-hour period using a mobile phone provided for the study. At each meal, they also answered 6-7 questions regarding the type of milk, butter/margarine/oil, meat, bread, cereal, and sauce using a JAVA-based questionnaire installed on the mobile phone (see Figure 1). The parents were instructed to photograph the meals from three angles, to use tableware provided specifically for the study, and to place a matchbox in each image. Volumes of foods were assessed from images using known sizes of the tableware and matchbox identifications by means of the software Paint (Microsoft, version 6.1), and converted into weight by being multiplied by the appropriate weight per volume [13]. The energy intake was calculated from intakes of foods and drinks through linkage to the Swedish Food Database [14].

KidMeal-Q was developed in 2008 for the LifeGene study [15]. It is an online, meal-based food frequency questionnaire designed for children aged 3-5 years for assessing dietary intake during previous months, and covers 42-86 food items, drinks, and dishes, depending on the number of follow-up questions (see Figure 2). For each child, we converted the reported frequency for vegetables into daily intakes by multiplying them by the reported portion sizes (using six pictures). KidMeal-Q does not provide any portion sizes for juice, fruits, berries, or sweetened beverages, and thus standard portion sizes [14] were used to convert reported frequencies into daily intakes.

KidMeal-Q was validated against a 7-day weighed food record in 23 healthy Swedish children with a mean age of 4.6 years (SD 1.5), weight of 18.4 kg (SD 3.7), and height of 1.09 m (SD 0.11) (data to be published). In that study, correlation coefficients between intakes of vegetables, fruits, and juice to sweetened beverages assessed using KidMeal-Q and food record estimates were .45 (P=.03), .59 (P=.003), and .53 (P<.001), respectively. These correlation coefficients are similar to those reported for adults when comparing food-frequency questionnaires with food records [16,17]. Therefore, although not an established reference method, we consider KidMeal-Q to be an appropriate reference in this first evaluation of TECH.

Values are given as means (SD). Significant differences between mean values were identified using the Wilcoxon matched pairs test. Correlation analyses were performed using Pearson or Spearman rank order correlations. The Bland-Altman procedure [18] was used to assess the agreement between methods. Analyses were performed using Statistica Software, version 10 (STAT SOFT, Scandinavia AB, Uppsala, Sweden).

This is the first study that has evaluated energy intake assessed using mobile phones versus the DLW method in preschoolers. The average energy intake estimated using TECH was not significantly different from the average TEE (mean difference +7%). However, the limits of agreement in the Bland-Altman plot were wide, indicating low accuracy for TECH in estimating energy intake for individuals. Furthermore, we observed a bias where high energy intakes were overestimated while low energy intakes were underestimated. Although we only used TECH for one day, the mean difference and our limits of agreement for energy are comparable to previous validation studies for established dietary methods in children aged 3-6 years [20-25]. In our study, part of the inaccuracy for individuals may be the result of TECH being applied for only one day.

There is no reference method for intakes of foods and drinks. Hence, as commonly applied, we compared TECH with another dietary method to evaluate its relative accuracy for foods. Despite that TECH was only applied for one day, which may not represent habitual intakes, we obtained correlations for fruits and berries, vegetables, and juice with KidMeal-Q (which assesses habitual intakes in past months) that were of similar magnitude as those commonly reported when comparing two dietary methods [16,17]. Although the accuracy in individual children was low, the average intakes of fruits and berries, vegetables, and juice assessed using TECH were comparable to corresponding figures with KidMeal-Q. However, our results for foods need confirmation due to the different assessment periods.

The major strength of our study is that we compared energy intake to TEE using the DLW method, which is the gold standard when validating reports of energy intake [3,26]. This is superior to using another dietary method as a reference since all such methods are well known to be associated with systematic errors [3,27]. The DLW method can be used as a reference since energy intake and TEE should be equal for subjects in energy balance during the measurement period. This criteria is valid both in normal-weight and overweight subjects, and was fulfilled since our children were weight stable through the 14-day-period, and since the energy content of retained tissue corresponds to only approximately 1% of energy intake at this age [28,29]. A limitation is that we compared one-day TECH data with mean TEE data from 14 days. However, this has unlikely influenced our results since the day-to-day variation in TEE is small [30,31].

The major limitation of this pilot study is that we applied TECH during only one day. The reason for this is that this first evaluation of TECH was conducted within an on-going study and we did not want to affect the parents’ participation in the original study by adding more assessment days. Furthermore, our sample size was small and participating families represented a selected group, which may limit generalizability. However, since the accuracy of using TECH for one day was comparable to established dietary methods, future research should evaluate if the accuracy for TECH can be improved with more days. In addition, identification of the underlying reasons for the observed bias in TECH could be the topic for future studies.

In conclusion, one day of recordings using TECH is not able to accurately estimate intakes of energy or certain foods in 3 year old children.