Validation of a Questionnaire to Measure Fruits and Vegetables Selected and Consumed at School Lunch among Second- and Third-Grade Students.

BACKGROUND
Interventions designed to encourage fruit and vegetable (F/V) consumption within schools are increasingly common. Thus, there is a need for valid, practical dietary assessment instruments to evaluate their effectiveness.

OBJECTIVE
The aim of this study was to examine the validity of a group-administered, paper-and-pencil questionnaire to assess F/V selection and consumption at school lunch relative to digital photography.

DESIGN
This was a five-phase, method-comparison study in which the questionnaire was iteratively modified between each phase.

PARTICIPANTS/SETTING
The study examined sets of questionnaires and photographs of lunch trays (n=1,213) collected on 44 days between May 2015 and June 2016 among second-grade students from three New York City schools (phases 1 to 4) and second- and third-grade students from 20 schools across eight states (phase 5).

MAIN OUTCOME MEASURES
Outcomes assessed were selection, amount eaten, preference, and intention to consume F/V.

STATISTICAL ANALYSES PERFORMED
Validity was assessed by percent agreement (categorized as "match, omission, or intrusion" for items on or off tray and "match, overestimation, or underestimation" for amount eaten), Spearman correlation coefficients, and intraclass correlation coefficients (ICC).

RESULTS
The total match rate for items on tray was substantial (phases 1 to 5: 83%, 84%, 92%, 93%, and 89%), with items more frequently intruded than omitted. For amounts eaten, the total match rates were moderate, but generally improved throughout the study (phases 1 to 5: 65%, 64%, 83%, 83%, and 76%), with overestimations more frequent than underestimations. There was good correspondence between methods in the estimates of amount eaten in a quantitative, cup equivalent amount (fruit ICC=0.61; vegetables ICC=0.64). Significant differences (α=.05) were not observed between second- and third-grade students, respectively, in the match rate for fruits (86% and 89%) or vegetable (89% and 86%) items on tray or fruit (69% and 73%) and vegetables (74% and 76%) amount eaten. Excellent correlations were observed between amount eaten and preference for fruit (r=0.91) and vegetables (r=0.93).

CONCLUSIONS
The questionnaire offers a feasible, valid instrument for assessing F/V selection and consumption among elementary students in schools participating in the National School Lunch Program. Additional research is recommended to test the instrument's sensitivity and to reproduce these findings using an alternative reference method, such as direct observations.