Child temperament is associated with energy intake and its day‐to‐day variability among preschoolers—Results from a cross‐sectional DAGIS study

Although research has linked children's temperament with weight outcomes, the associations between temperament and dietary outcomes, particularly energy intake, remain understudied. Furthermore, little is known about how temperament is associated with diet in the early childhood education and care (ECEC) context, which is an important environment for many children.


| INTRODUCTION
Tackling childhood overweight and obesity is an important public health objective worldwide. In the World Health Organization European region, one-third of children aged between 6 and 9 years are overweight (including obesity). 1 Current overweight-prevention efforts, however, have yielded limited success. 2 One suggested reason for unsuccessful prevention attempts has been that interventions fail to acknowledge children's individual behavioural differences behind their eating behaviours. 3 A better understanding of how individual differences are linked to weight-related behaviours, such as diet, could aid the design and implementation of obesity-prevention efforts. 3 One major factor contributing to individual behavioural differences among children is temperament. 4 Temperament refers to a biologically based emotional and behavioural style that characterizes the child's reactions to environmental stimuli as well as the self-regulation of those reactions. 4 Rothbart and colleagues 5 have examined temperament through three broad dimensions: surgency, effortful control and negative affectivity. Surgency describes a child's reactivity in terms of its approach to potential rewards and inclination towards positive emotions. Surgency includes components, such as impulsivity, a high activity level and low shyness. Effortful control describes a child's level of self-regulatory capacity and includes components, such as inhibitory control, attentional focusing and perceptual sensitivity. Negative affectivity describes a child's general proneness to distress and tendency to experience uncomfortable emotions. Negative affectivity includes components, such as anger, frustration, fear and low soothability.
Different temperament dimensions have been linked with different weight outcomes among children. Specifically, low selfregulation, 6,7 facets of negative affectivity, [6][7][8] and higher surgency/extraversion 6,9 are positively linked with weight outcomes among children of different ages. Furthermore, impulsivity, a characteristic included in the surgency dimension of temperament, predicts a higher weight status and increases the risk of overweight in children. 10 Studies investigating the associations between temperament and different dietary outcomes among young children have found a negative/low regulation temperament profile to predict unhealthier food consumption, 11 whereas higher effortful control has been linked with healthier food consumption 12 and lower daily energy intake. 13 Results regarding surgency, however, are somewhat inconsistent. Surgency, as a broad temperament dimension, has been associated with higher odds for daily vegetable consumption among young children, 14 and lower consumption of sugar-sweetened beverages among adolescents. 15 Impulsivity, a part of the surgency dimension, in turn, has been linked among primary school children with higher consumption of palatable energy-dense foods, 16,17 and with higher energy intake specifically when palatable energy-dense foods were available. 17 However, two studies among preschoolers investigating links between temperament and energy intake did not find associations between surgency 13 or impulsivity 18 and energy intake assessed with a food frequency questionnaire.
In addition to mean energy intake, another potentially relevant perspective on energy intake is its individual day-to-day variability, which has been shown to vary between individuals. 19 Among adults, few dietary studies have examined individual day-to-day variability in energy intake [20][21][22] in relation to different healthrelated outcomes. Some of these studies linked greater variability with adverse outcomes, such as poorer diet quality 21 and metabolic risk factors, 22 including weight outcomes, 22 although one study did not find an association with weight. 20 Only one study has examined associations between health outcomes and individual variability in eating, specifically in eating frequency, among children. 23 The study found that greater variability in eating frequency was associated with higher total and LDL cholesterol but not with weight. In addition, few studies among children have examined day-to-day variability in energy intake as an indicator of energy intake regulation, showing large variation between children. 24,25 Previously, it has been noted that it is still unclear where these individual differences in eating self-regulation stem from. 3 To the best of our knowledge, studies to date have not examined temperament-diet associations in preschoolers separately at home and in early childhood education and care (ECEC) contexts. In addition to their home environment, preschool children spend a substantial part of their time at ECEC centres. 26 Home and ECEC centre differ as food environments; for example, the availability of foods is more restricted at ECEC centres compared with home, and the social mealtime context (number of peers and adults present, interactions between children and adults) is also different. Environment may affect children's behaviour differently, depending on a child's individual characteristics 27 and therefore temperament-diet associations in different contexts should be examined. The associations between a child's temperament and energy intake, as well as day-to-day variability in energy intake in different everyday life environments, remain largely under-examined. Furthermore, the correlates of day-to-day variability in energy intake among children are largely uncharted. Thus, this study expands the literature by (1) examining concurrent associations between a child's temperament dimensions and mean energy intake, (2) investigating associations between temperament dimensions and individual day-to-day variability in energy intake and (3) examining these dietary factors in two contexts: home and ECEC. Based on the previous literature, we hypothesize that surgency and negative affectivity are positively and effortful control is negatively associated with mean daily energy intake. The analyses regarding day-to-day variability in energy intake are exploratory, as is the examination of the associations in the ECEC context. This study only includes the children who had complete data available for the three temperament dimensions and a 3-day food record, including 2 days at the ECEC centre and one weekend day.
In addition, on the days spent in the ECEC centre, the child must have consumed both lunch and an afternoon snack there. As a result, 505 children were included (58% of participants, 14% of invited children) in the present study.

| Food record data
Parents reported the child's food consumption with a 3-day food record (including two weekdays and one weekend day) that they received by mail. The days were predefined by research personnel and were not always consecutive to capture data from all days of the week. Some parents wished to change the given dates because of reasons such as illness or a holiday trip, and in these cases, more suitable days were agreed on.
The food record included written instructions and an example page. Instructions guided to report all the foods and beverages the child had consumed during the day, excluding foods and beverages consumed in the ECEC centre. The time and place of consumption were also requested. Parents were advised to estimate the amounts by weighing, using household measures, or using a food picture book developed for the project. 29 For composite dishes, parents were instructed to list all the ingredients. For prepacked foods, the exact brand and product name were to be recorded.
In the ECEC centres, early educators recorded the foods children consumed there during the same food record days that were given to the parents. They were given a separate food record, and to aid completion, the food record was pre-coded with separate sections for meal type (breakfast, lunch, afternoon snack or other snack) and predetermined rows for different parts of the meal (e.g., main course, side dishes, salads, breads and toppings). Early educators could report the amounts either by household measures or by using a food picture book. 29 Researchers or research assistants thoroughly scrutinized all food records for possible omissions and made follow-up phone calls if important information was missing. Special attention was paid to fruits and vegetables and sugary products, as these were the focus of the study. Food record data were entered into Aivo-Diet dietary software, which utilized the food composition database established and maintained by the Finnish Institute of Health and Welfare (release 16,2013). When entering the food record data, the research assistant used existing recipes/items from the database, modified existing ones, or created new ones according to each parent's report. As for foods served in the ECEC centres, five of the municipalities gave detailed recipes of all the foods, one municipality gave some of the recipes, and two declined to give any. For those municipalities that did not give recipes for the research, estimates based on the recipes from other municipalities were used. After data were entered, outlying values of food consumption in grammes and outlying energy and nutrient intakes were checked. A detailed description of the food record data processing is described elsewhere. 30

| Confounding variables
The child's age (continuous) and gender were selected as confounders, as they correlated both with temperament and mean energy intake in the present data, and previous research has demonstrated their relevance for both mean energy intake 31 and dayto-day variability in energy intake. 32 Based on found correlations in the data and existing literature, 31,33 the mother's educational level and the child's moderate-to-vigorous physical activity (MVPA) were selected as additional confounders for the analyses, with mean energy intake as an outcome. Furthermore, analyses conducted separately for ECEC and home contexts were controlled for the place where a child had their breakfast, as it varied among participants.
The child's age and gender and the mother's education were inquired by a questionnaire. The mother's education was categorized into three levels: (1) low (comprehensive, high school or vocational school); (2) middle (bachelor's degree or equivalent); and (3) 34 The place for breakfast consumption was coded into four categories based on the food record reports: (1) varying place of breakfast, (2) breakfast at home on both the days, (3) breakfast at ECEC centre on both the days and (4) breakfast both at ECEC centre and at home on both the days.

| Data processing and statistical analysis
For the analyses, a score describing each temperament dimension (mean of the items in each dimension) as well as the mean energy intake over the three measured days were calculated for each participant. In addition, the mean daily energy intake was separately calculated for foods eaten at the ECEC centre and at home over the two weekdays. The day-to-day variability in total daily energy intake was quantified as the coefficient of variation (CV), which was calculated by dividing the individual standard deviation of daily energy intake with the individual mean energy intake and multiplying it by 100. The result describes the extent of day-to-day variability in daily energy intake during the 3 days in relation to the mean energy intake in percentages. The CV was also calculated separately for the daily energy obtained at the ECEC centre as well as the energy obtained at home during the two weekdays. CV at the ECEC centre represents the daily energy intake variability during the two weekdays at the ECEC centre, and CV at home represents the daily energy intake variability at home over the two weekdays.
First, the distributions of the study variables were examined. As the distributions of all the outcome variables were skewed to the right, square root transformations were made and transformed variables were used in the analyses. The preliminary data examination and all variable transformations were made using IBM SPSS (version 28.0).
Linear regression analyses were used to examine crude and adjusted associations separately between each of the temperament dimensions and outcomes of mean energy intake in three different models: (1) no adjustments; (2) adjusted for the child's age, gender, and mother's educational level (and place for breakfast when home and ECEC contexts are separately analysed); (3) adjusted for Model 2 variables and the amount of the child's MVPA. The relevance of the child's MVPA was examined in a separate analysis, as it is a potential mediating factor between temperament and mean energy intake. Analyses with CV as an outcome were examined with two models: (1) no adjustments and (2) adjusted for the child's age and gender (and place for breakfast when home and ECEC contexts were separately analysed). We also examined all models simultaneously, including all three temperament dimensions. However, the results did not change significantly and, therefore, are not further presented in this article. The regression analyses were run with R version 4.1.2 and RStudio 35,36 and all were adjusted for data clustering on the preschool level using cluster-robust standard errors (package "sandwich" 37 ).

| RESULTS
The characteristics of children with the required food record and temperament data (n = 505) are presented in Table 1. The mean age of the children in the sample was 4.7 (2.75-6.83) years. On average, children's energy intake was 5900 kilojoules (kJ) per day, which aligns with the estimated energy requirements stated in the Finnish nutrition recommendations of 5000 kJ per day for 2-5-year-olds and 6900 kJ per day for 6-9-year-olds. 38 The CV in energy intake, that is, the extent of the individual standard deviation of daily energy intake relative to the individual mean energy intake as percentages, was approximately 17%. The number describes the magnitude of day-to-day variability of a child's energy intake; a higher percentage means greater variability. The energy intake variability of 17% is in between estimates reported in previous studies that range between 10% and 29%. 25 Children included in the analyses did not differ from the children not included (n = 359) in terms of age or gender; however, more children had mothers with a high educational level in the analysed sample compared with children not included in the analyses (32% vs. 25%, p = 0.04, respectively, data not shown).
No differences were found between children included in the analyses and those not included in temperament dimension scores (compared with 246 children with temperament data but missing other data) or energy intake variables (compared with 63 children with required food record data but missing other data).   A child's negative affectivity was positively associated with the total CV in a crude model (F(1, 503) = 6.46, p = 0.02, R 2 adjusted = 0.01). When adjusted for confounders, the association remained statistically significant; however, the overall adjusted model had a poor fit to the data (F(3, 501) = 2.23, p = 0.08, R 2 adjusted = 0.01). The variance explained by negative affectivity was 1%. When examined separately in the home and ECEC contexts, the association between negative affectivity and energy intake variability was non-significant in both contexts. Neither T A B L E 2 Unstandardized regression estimates and their 95% confidence intervals (CI) for associations between a child's temperament dimensions and different outcomes of energy intake. Note: Model 1: no adjustments. Model 2: a Models with mean energy as an outcome adjusted for child's age and gender and mother's highest education (and for place of breakfast when separately analysed at ECEC and home contexts). b Models with CV as an outcome adjusted for child's age and gender (and for place of breakfast when separately analysed at ECEC and home contexts). Model 3: Model 2 variables and moderate-to-vigorous physical activity. Statistically significant results indicated in bold ( p < 0.05).
effortful control nor surgency were associated with the overall energy intake variability or separately in either context.

| DISCUSSION
This study examined associations between the child's three temperament dimensions (surgency, effortful control and negative affectivity) and mean energy intake, as well as day-to-day variability in daily energy intake overall and separately at home and at the ECEC centre. A child's higher surgency was associated with higher mean daily energy intake, and higher negative affectivity with greater day-today variability in total energy intake. When examined separately in the contexts of home and ECEC centre, the association between surgency and mean energy intake was only found in the ECEC centre. Furthermore, in separate analyses, the association between negative affectivity and energy intake variability was non-significant in both contexts. Effortful control was not associated with mean energy intake or energy intake variability in any model. A noteworthy caveat related to the results is that the associations found were weak, and the proportion of variance explained by temperament dimensions was rather low (1%-2%). On the other hand, this was expected, as general temperament is a distal factor to actual food consumption. Children's diets are affected by several factors on individual, family, and societal levels. 39 Thus, it was anticipated that individual temperamental dispositions would be modestly associated with energy intake. Also, other previous studies have reported modest numbers for explained variance. 17,40 The positive association between a child's higher surgency (i.e., disposition to experience positive emotions, be active, act impulsively, and approach potential rewards) and mean energy intake differs from two previous studies finding no association between surgency, 13 or impulsivity 18  Studies on children's eating behaviours, in turn, suggest that a link between surgency and higher energy intake is plausible, as surgency has been associated with food approach behaviours, such as food enjoyment and food responsiveness 41,42 and less satiety responsiveness. 41 Although these studies did not investigate whether surgency or food-approach behaviours also are associated with higher energy intake, in the study by Leung et al., 42 food-approach behaviours mediated the association between surgency and the child's body mass index (BMI) z-score. Furthermore, Hankey et al. 9 found a longitudinal association between surgency measured at the age of 5.5 and BMI zscore measured at second grade and in early adolescence. In our analyses with daily mean energy intake as an outcome, we also considered the amount of the child's MVPA as a potential confounder, as temperament is suggested to be linked with the child's physical activity level. 43,44 Our analyses showed that, when the amount of MVPA is accounted for, the association between surgency and mean energy intake is attenuated to some degree. This result implies that part of the association between surgency and higher energy intake is potentially explained by higher levels of physical activity.
The positive association between surgency and mean energy intake, particularly in the ECEC centre implies that children with higher surgency may be more sensitive to some aspect of the food environment there. A previous study has suggested that children higher in sociability (an aspect included in the surgency dimension of temperament) could be more receptive to social influences of the food environment compared with less sociable children. 45 Another study showed that, on average, young children tend to eat larger portions in ECEC centre compared with home, 46 although the study did not investigate the role of children's temperament in the association.
The positive association between a child's negative affectivity (i.e., disposition to experience distress and uncomfortable emotions, such as anger, fear and sadness) and higher day-to-day variability of energy intake is also in line with existing research on eating behaviours. Children with higher negative affectivity are more likely to have food-avoidant behaviours, such as fussy or picky eating and emotional undereating. 41 These food-avoidant behaviours could lead to more variable eating and energy intake depending on whether the foods offered are to the child's taste on a given day. On the other hand, the same study linking negative affectivity with emotional undereating have also linked negative affectivity with emotional overeating. 41 Furthermore, emotional over-and undereating tend to correlate, which means that the same children exhibit both the behaviours. 47 Thus, another potential explanation for more variable energy intake in children with higher negative affectivity could be that these children tend to undereat in some situations and overeat in others, resulting in higher variability of energy intake between days. Steinsbekk et al. 41 have suggested that children reacting with higher negative affectivity might be more susceptible to different environmental stimuli, and, depending on the nature of the stimuli (e.g., positive vs. negative), the reaction could be either over-or undereating.
The association found between a child's higher negative affectivity and greater day-to-day variability in daily energy intake could also be of interest for research on energy intake self-regulation. Current research indicates that the ability to compensate for the energy from a preload in a subsequent meal differs markedly among children. 3 Also, studies that have investigated energy intake self-regulation over multiple days in more naturalistic settings have shown that day-to-day variability in energy intake differs substantially among children. 24,25 The determinants of these individual differences, however, are not well known. According to the review by Kral et al., 3 food approach and food avoidance behaviours are to some extent directed by genes. Furthermore, some but not all studies suggest that factors, such as age, ethnicity, weight status, and the parents' feeding practices could play a role in the child's ability to self-regulate energy intake according to energy requirements. 3 Our results imply that one potential correlate of energy intake self-regulation could be temperamental negative affectivity; however, its role should be further investigated in the future.
As noted earlier, the found associations were weak in this study.
Few previous longitudinal studies investigating associations between temperament and eating behaviours have shown that the relevance of temperament may increase over time. 40,41 For example, Kidwell et al. 40 showed that negative affectivity concurrently explained 3% of the variance in selective eating, while longitudinally, the explained variance was 6%. Also, Steinsbekk et al. 41 found that the association between negative affectivity and emotional overeating and between surgency and food responsiveness became stronger with time. Thus, future longitudinal studies should investigate whether this also applies to the association between temperament and dietary outcomes.
Moreover, future studies should investigate whether the relevance of day-to-day variability in energy intake for health outcomes is additional to that of mean energy intake among children.
We did not find associations between effortful control and mean energy intake or between effortful control and day-to-day variability in energy intake. In previous studies, effortful control measured as a temperament dimension in accordance with Rothbart et al.'s theory 4 has been associated with lower energy intake, 13 and healthier overall food consumption. 12 Furthermore, in some studies, a specific facet of temperamental effortful control, namely inhibitory control, has been inversely linked with BMI z-score and food approach behaviours. 48,49 Perhaps effortful control did not emerge as a significant factor in the present study because it is a broad measure of self-regulation covering aspects such as attentional focusing, perceptual sensitivity, and enjoyment of low-intensity stimuli, in addition to inhibitory control.
The strengths of this study include a widely used, validated measure of temperament. Furthermore, we were able to examine children's energy intake with a food record covering the child's whole day-both the time at the ECEC centre and the time outside daycare hours. We were also able to adjust the analysis with an objectively measured MVPA.
The study also has limitations that should be acknowledged. First, we only had three food record days available, whereas most of the prior studies focusing on day-to-day variability in eating have used 5-7 days of food record data. 22,23 Particularly, the results regarding home and ECEC contexts separately, are limited due to only 2 days of food record data available for the analyses. Examining day-to-day variability using only a few food record days increases the possibility that the estimates are biassed by a high degree of random error. 19 In a Finnish study among young children utilizing a 3-day food record, correlation coefficients between observed and true energy intakes were above 0.7 among children aged three and six. 32 According to the same study, between four and five food record days would have been required to achieve correlation coefficients above 0.8, depending on the children's age.
Another limitation of the study is that the DAGIS study was originally not designed to examine between-person differences in individual day-to-day variability in dietary outcomes. Factors that could potentially better explain these differences, such as a child's eating behaviours, a broad spectrum of food-related parenting practices, or information on situational factors of the eating occasions, were not included as part of the study. Thus, we failed to find a well-fitting regression model predicting energy intake variability; after including potential confounders in the model with day-to-day variability of energy intake as an outcome, the model fit was reduced. In these data, the common background factors explaining differences between children in mean dietary behaviour (e.g., age and gender) were irrelevant for day-to-day variability in energy intake. Only negative affectivity was a significant predictor of energy intake variability, but its relevance was small.
As children's food consumption and eating behaviours develop throughout the early years, laying grounds for later eating, it is important to understand how individual characteristics can potentially shape the forming path. By examining the associations between temperament dimensions and mean energy intake as well as day-to-day variability in energy intake, this study established that temperament dimensions are associated with mean energy intake and its day-to-day variability differently: higher surgency with higher mean daily energy intake, and higher negative affectivity with greater day-to-day variability in energy intake.