Overall, parental response accuracy of fourth-grade children's usual school-meal participation was moderately high, and better concerning lunch participation than breakfast participation. With regards to breakfast participation, responses on consent forms from approximately three in four parents on average were in agreement with nametag records compiled by researchers during direct meal observations. Parental responses concerning breakfast participation were more accurate for older children and did not differ significantly by BMI, sex, race, or study. These results for breakfast participation are similar to those in a 2002 article  that showed that responses from three in four parents were in agreement with nametag records and that parental response accuracy did not differ by race or sex . (For that 2002 article, BMI was not included in analyses, and data were from a single study.) With regards to lunch participation, parental response accuracy was even higher at 92% agreement with nametag records. Parental responses concerning lunch participation were also more accurate for older children and differed by race; specifically, parental responses were more accurate for Black children than for White children. These disparate effects related to age and race warrant caution when relying on parental responses of children's participation in school-provided meals. In a 2009 publication, Moore and colleagues  also expressed caution in relying on parental responses of children's participation in school-provided meals.
Analyses indicated relationships in opposite directions between fourth-grade children's usual breakfast participation and usual lunch participation (when based on parental responses) and BMI. Specifically, when breakfast participation and lunch participation were included in the model (Model 2A), results showed an inverse relationship between school lunch participation and BMI but a (marginally significant) positive relationship between breakfast participation and BMI. When lunch participation was not in the model (Model 2B), breakfast participation remained marginally significant (and positive). These results differ from a recent article  that used data from the same four studies and found no significant relationship between school-meal participation (based on nametag records compiled by research staff for direct meal observations) and BMI. These results also differ from studies published in 1998  and 2010  that showed no significant relationship between school-meal participation (based on parental responses and daily administrative records, respectively) and BMI. In addition, these results conflict with a study published in 2009 that showed an inverse relationship between breakfast participation (based on student or parental responses) and BMI among White students  and with a study published in 1994 that showed a positive relationship between children who participated in school lunch (based on parental responses) and BMI . These conflicting results emphasize the need to consider the source of information concerning children's school-meal participation when investigating the relationship between school-meal participation and childhood obesity.
It is possible that the conflicting results may be explained by children's food preferences, as children in the studies for this article's analyses had a choice each day of breakfast options and a choice each day of lunch entrées. Food preference data were collected for the subset of children who were interviewed for three of the four dietary-reporting validation studies which provided data for this article's analyses. However, preference data were collected only for foods observed or reported eaten during the time frame covered in the dietary recalls; thus, food preference data were not collected for all foods available on school-provided meals, no food preference data were collected for children not interviewed, and no food preference data were collected for one study (D).
It may be possible to improve parental response accuracy by rewording questions that ask parents about children's participation in school meals. For past studies, questions (and response options) have been worded as "Does your child eat school lunch" (yes, no)
, "During the school year, about how many times a week (does the child) usually get a complete school lunch (/breakfast at school)" (0-5)
, "Does (CHILD) usually eat breakfast at school under the School Breakfast Program" (yes, no)
, "Does (CHILD) usually eat a complete hot lunch offered at school? Usually is defined as about 3 or more days a week" (yes, no)
, and "How many days a week does (CHILD) usually eat a school lunch/breakfast" (0-5)
. Some publications have not specified the wording of the question(s) , or have indicated that parents were asked whether their children obtained a school breakfast and/or lunch on the most recent school day, and on each day during the most recently completed week . For the four studies that provided data for the current article's analyses, questions did not define "usually", so it is possible that parents interpreted the questions differently. The authors are not aware of any research that has investigated the relationship between the wording of the questions and accuracy of parental responses. Research on this topic is needed.
The current analyses have limitations. First, the four studies from which the data arose were not designed statistically to address these research questions. Second, consent forms for only two of the four studies asked parents about children's usual participation in school lunch. Third, weight and height measurements were available for only one point in time per child rather than at the start and at the end of the fourth-grade school year. Fourth, the time of year at which parental responses regarding their child's school-meal participation was collected on consent forms (September and November of each study's school year) did not correspond with the time at which nametag records of school-meal participation were compiled (which occurred throughout each study's school year). In addition, because of the wording of questions on consent forms, it was possible to dichotomize but not to quantify parental responses of children's school-meal participation. Fifth, although childhood obesity may be related to physical activity, socioeconomic status (SES), and parental BMI, data concerning physical activity, SES, and parental BMI were not collected for the four studies. Finally, analyses for this article did not differentiate between free, reduced-price, and full-price school meals (i.e., definitions for breakfast participation and lunch participation did not change as a result of meal price); however, this was appropriate because children's participation in meals provided at school, rather than meal price, was the independent variable [7, 8]. Although it is possible that the accuracy of parental responses concerning their children's participation in school-provided meals differs according to whether meals were free, reduced-price, or full price, information concerning SES at the individual child level was not collected for the four studies. However, based on the percentages of children at the schools in the four studies who were eligible to receive free or reduced-price school meals, most children in the sample were likely from families of low to moderately-low SES.
The current analyses also have several strengths. The sample for each of the four studies had approximately equal numbers of children by race (Black, White) and by sex. Results regarding the relationship between BMI and age, and between BMI and race, were similar to previous studies and national studies [31–34]. The objective measure of participation in school meals was collected during direct meal observations on nametag records compiled by research staff. Rigorous and consistent quality control procedures were implemented in all four studies for weight and height measurements. The statistical analyses specifically acknowledged the nested structure of children within school.