Participants and procedure
Families with children born in 2003 and 2004 in Trondheim, Norway were recruited at routine health check-ups for 4 year olds (97.2% attendance). Of those invited to participate (n = 3,016), 82.2% consented; a subsample (n = 1,250) was drawn to participate in the Trondheim Early Secure Study (TESS). Details of the TESS study are described in Wichstrøm et al. . The present inquiry used data from the second wave of collection only, including 752 children (376 boys, 376 girls) with CEBQ FF data (mean age = 6.7 years, SD = .18). Attrition analyses revealed drop-out was not predicted by gender or pickiness assessed by clinical interview. Parents were mainly of Norwegian origin (92.3%) and were comparable to the Norwegian parent population for the parents’ level of education . The majority of informants were mothers (81.1%).
Picky eating was assessed at 6 years by: 1) the 6-item FF scale from the CEBQ (5 point Likert response scale) , a parent-reported measure of pickiness assessing both neophobia (e.g. ‘My child refuses new foods at first’) and more general ‘fussy’ eating (e.g. ‘My child is difficult to please with meals’) (α = .89); and 2) the PAPA, a semi-structured psychiatric interview . The CEBQ was developed to capture the range of eating styles seen in children, and the inclusion of eating style constructs was based on an 1) evaluation of the existing literature; and 2) interviews with parents about their children’s eating. Fussiness was a common observation and the FF scale constitutes one of eight eating style dimensions captured by the CEBQ . The questionnaire has been validated against behavioral measures of eating , and has shown good test-retest reliability .
The PAPA interview was administrated by trained personnel holding at least a bachelor’s degree in a relevant field and with substantial practice in working with children and families. Parents were interviewed about their child’s food preferences, appetite, restricted consumption of foods, and resulting impaired functioning. The interviewer decided whether pickiness was present and probed until s/he could categorize the children: 0 = no restricted intake; 1 = moderately restricted intake (child only eats food s/he likes); 2 = severe pickiness (substantial pickiness; separate meals must be made for the child). Nine percent of videotaped recordings were recorded by blinded interviewers, with high interrater reliability (ICC = .92).
Children’s height and weight were measured using digital scales (Heightronic digital stadiometer: QuickMedical,Model 235A and Tanita BC420MA). Weight was measured to the closest .1 kg and correction for light indoor clothing (0.5) was applied. BMI SDS (Body Mass Index Standard Deviation Score) was estimated [14–16]. Parental occupation coded according to the International Classification of Occupations  (6-point scale: 1 = Manual workers, 6 = Leaders) was used to measure socioeconomic status.
Descriptive analyses were conducted in SPSS 22, using the Complex Samples option. General Linear Modeling (GLM) was applied to examine the association between the mean FF scale score and BMI SDS.
The screening efficiency of the FF scale was evaluated using non-parametric receiver operating characteristic (ROC) curve analysis, which determines the area under the curve (AUC) for the FF scale against the PAPA. The AUC indicates the probability that a randomly selected subject with PAPA-defined pickiness and a randomly selected subject not categorized as a picky eater according to the PAPA would be correctly distinguished based on their FF scale scores. The sensitivity/specificity pairs generated through the ROC analysis were used to select a threshold for identification of clinical cases (PAPA-defined pickiness). The analysis was performed twice; first using both moderate and severe pickiness as the main outcome (0 = ‘normal’; 1 = ‘moderate/severe’); secondly, using severe cases only (0 = ‘normal/moderate’; 1 = ‘severe’). At a given cut-point, sensitivity shows the proportion of children positively screened with the FF scale among children with PAPA-defined pickiness; specificity denotes the proportion of children with a negative FF scale screen among non-picky eaters. The positive predictive value (PPV) and negative predictive value (NPV) were also calculated; the PPV provides the probability that a child who screens positive on the FF scale does have PAPA-defined pickiness, the NPV provides the probability that a child who screens negative does not have the condition (i.e. is a noncase). ROC analyses were performed in STATA 13. Analyses were performed using probability weights because we oversampled for emotional and behavioral problems at baseline to increase variability and statistical power (for details, see ).