This was a cross-sectional study using self-administered questionnaires. The participants were students majoring in nutrition at a university in Shizuoka, Japan, and were, therefore, familiar with dietary records. The students were recruited through a bulletin board on campus. Out of a total population of approximately 100 students, the questionnaire was distributed directly to 23 participants who agreed to be surveyed, and 23 copies were collected by mail or in person (recovery rate=100%). As an incentive, participants in this survey were given a gift certificate worth JP ¥2000 (US $14).

The survey period was from July to August 2021. Participants were initially asked to respond on a scale regarding the variety of vegetables consumed in the past month. Following this, participants were asked to keep dietary records for 7 consecutive days starting the day after their responses for each scale. According to a previous study that examined the number of days of dietary records required to estimate habitual vegetable variety, we estimated that 7 consecutive days of dietary records would capture approximately 70% of the maximum theoretical number of different vegetables consumed [17]. It is reasonable to assume that an understanding of dietary habits could be achieved with 7 consecutive days of dietary records; therefore, the same number of consecutive days of dietary record collection was chosen for this study. Finally, participants were asked to provide their responses on the same scale after 1 month following their response to the first scale. After each questionnaire was collected, 2 students majoring in nutrition and 1 researcher certified as a registered dietitian checked for missing questionnaire responses and incomplete dietary records (eg, missing vegetable condiments, seasonings, and cooking oil). That missing information was then obtained from participants in person or via email.

Of the 4 scales included in this study, item-total correlation analysis was performed for each of the 3 scales that are scoring systems. Dietary records for 7 consecutive days were used as a reference standard in the validity analysis. In designing a validation study for the FFQ, administering the questionnaire before the participants create dietary records results in the questionnaire responses being related to the diet before the period during which dietary records were created. On the other hand, the intensive effort involved in creating dietary records may lead to artificially improved accuracy if the questionnaire is administered after dietary record creation [25]. To avoid optimistic estimates of the correlation in this study, the first response given by participants was used to examine the validity of each scale. Reproducibility was assessed by test-retest reliability. Since each scale to determine vegetable variety was intended to be ranked, correlation coefficients were calculated. Furthermore, because the recorded variable distributions were determined not to be normally distributed by the histograms, the Spearman correlation coefficient was used to examine the validity and reproducibility. SPSS Statistics (version 27; IBM Corp) was used for analysis. The level of significance was set at P<.05 (2-sided test).

The purpose and methods of this study were explained to participants before this study. Informed consent was obtained in writing from all participants. Parental or guardian written consent was obtained from students aged 18 or 19 years. This study was approved by the Research Ethics Committee of the University of Shizuoka (3-12).

All 23 respondents were included in the analysis. None of the participants reported restricting their vegetable intake for medical reasons.

The participants were 91% (n=21) women, with a mean age of 20.0 (SD 1.4) years (Table 1).

For scale B, only 1 item (“raw vegetables used in salad: lettuce, cabbage, etc”) had a correlation coefficient greater than 0.5 and was significant at the 1% level (Table 2). Correlation coefficients of 0.4 to 0.5 and significant at the 5% level were found for 2 items (“tomatoes, tomato ketchup, boiled tomato, and stewed tomato” and “vegetables used in cooking: cabbage and Chinese cabbage”).

For scale C, 4 items (“broccoli,” “lettuce,” “burdock,” and “lotus root”) had a correlation coefficient greater than 0.5 and were significant at the 1% level. Correlation coefficients of 0.4 to 0.5 and significant at the 5% level were found for 2 items (“carrots” and “pumpkins”).

For scale D, 6 items (“broccoli,” “spinach,” “Welsh onions [green],” “lettuce,” “burdock,” and “ginger”) had a correlation coefficient greater than 0.5 and were significant at the 1% level. Correlation coefficients of 0.4 to 0.5 and significant at the 5% level were found for 5 items (“carrots,” “pumpkins,” “komatsuna,” “Chinese chive,” and “Welsh onions [branching cultivation]”).

Since none of the items were significantly inversely correlated with the 3 scales, none of the items were removed. The validity and reproducibility of scales selected items that significantly correlated with total scores were also examined. The item selection scales were referred to as short scales B, C, and D.

For each scale, the correlation coefficients were as follows: scale A, ρ=0.28 (P=.20); scale B, ρ=0.22 (P=.31); scale C, ρ=0.51 (P=.01); and scale D, ρ=0.44 (P=.03; Table 3). This indicated that the 2 scales assessed by vegetable items (ie, scales C and D) were significantly correlated with vegetable variety based on the dietary records for 7 consecutive days. Furthermore, the correlation coefficient was ρ=0.08 (P=.73) for short scale B, ρ=0.42 (P=.04) for short scale C, and ρ=0.33 (P=.12) for short scale D, indicating that the correlation decreased with the selection of items for all 3 scales.

For each scale, the correlation coefficient of the first and second responses for scale A was ρ=0.24 (P=.27; Figure 1). Scale B was ρ=0.45 (P=.03), scale C was ρ=0.73 (P<.001), and scale D was ρ=0.74 (P<.001), confirming a correlation between the 2 responses on the vegetable subgroup scale and a strong correlation for the 2 scales using vegetable items. Furthermore, short scale B was ρ=0.40 (P=.06), short scale C was ρ=0.85 (P<.001), and short scale D was ρ=0.79 (P<.001). Although the item selection reduced the reproducibility of the vegetable subgroup scale, the 2 scales for vegetable items were still reproducible.

In this study, the validity and reproducibility of several scales to determine vegetable variety was examined and compared. Regarding validity, scales C and D significantly correlated with vegetable variety based on dietary records for 7 consecutive days. Regarding reproducibility, a significant correlation was observed for scale B, and significant strong correlations were observed for scales C and D.

Item-total correlations were conducted to examine the internal consistency of the scales. In this analysis, none of the items in all 3 scales negatively correlated with the total score. Furthermore, although items with significant correlations were selected for each scale, correlations with vegetable variety based on dietary records for 7 consecutive days were lower than the full version of the scale. Therefore, all items removed for the short scale may be necessary for determining vegetable variety, even though their correlations are weak.

For the association with vegetable variety based on dietary records for 7 consecutive days, significant correlations (ρ>0.4) were found for scales C and D but not for scales A and B. Furthermore, for the assessment of reproducibility, significant correlations were found for scale B, and significantly strong correlations (ρ>0.7) were found for scales C and D. For validation studies of FFQs, correlation coefficients with validity lower than 0.4 are considered seriously attenuate associations [25]. Compared to the scales for a single question (scale A) and vegetable subgroup (scale B), the scales for vegetable items (scales C and D) demonstrated acceptable validity and reproducibility regarding vegetable variety and, therefore, may be useful for determining the rank of vegetable variety.

Low-intake vegetables used as condiments (eg, chopped Welsh onion as a garnish), which would have little nutritional significance, were removed from the dietary records used as a reference standard in this study with a quantitative criterion for single use, based on a previous study using the 1-day food record [24]. However, other studies that have used 24-hour recalls have not specified the minimum amount of consumption [12]. Since vegetable variety varies depending on the establishment of a minimum amount consumed using either dietary records or 24-hour recalls used as a reference standard, the necessity of establishing a minimum amount and the appropriateness of the cutoff value should be examined in the future. In this study, the scales examined did not establish a minimum amount of vegetable consumption. Nonquantitative FFQs such as the BDHQ referenced in this study, are confirmed to have reasonable validity for vegetable intake using the frequency measurements only [18]. Therefore, there appears to be little need to establish a minimum amount consumed, at least for a scale aimed at determining vegetable variety.

This study has some limitations, including the use of a single survey. It has been suggested that data collected on the eating patterns of individuals should include all seasons of the year as well as days in all parts of the week [26]. Indeed, a previous study reported significant seasonal differences in vegetable intake [27]. Since vegetable variety was assessed using data collected on 7 consecutive days as a reference standard, seasonality may have affected the vegetable variety measured in this study. The BDHQ and DHQ used as a reference for scales B and C have reasonable validity for vegetable intake conducted over 4 seasons [18]. Furthermore, although this study was conducted in July 2021, the Ranking of Vegetable Consumers in Japan based on the NHNS used as a reference for scale D was assessed in November 2012 [20]. Therefore, data from this study suggested that a minimal effect of seasonal differences on the ability of each scale to determine vegetable variety. A second limitation of this study is the small sample size. Since this study included only students majoring in nutrition at 1 Japanese university (mostly women, younger in age, normal or low BMI, and living alone) and interested in this survey, it is not clear to what extent the results can be generalized. The limited number of participants in this study also made it difficult to adjust for confounding factors. Future research should assess the validity of scales in different regions and populations. Although there were no missing days for the dietary records in this study, which likely occurred since the participants were students majoring in nutrition, it is suggested that the usability of the records decreases as the records progress to the seventh day [28]. A third limitation of this study is that, because biochemical indicators such as plasma carotenoid concentrations were not assessed, the effect of underreporting on the association of vegetable variety is not clear. Finally, in cases where participants ate commercial products, ate at a restaurant, or recorded only the name of the dish, the number of vegetables consumed was estimated using information from the internet. Therefore, although the results may not reflect the actual consumption of the participants in this study, generalization could be maintained because the foods that appeared most frequently on the internet were extracted.

Correlation analysis of vegetable variety based on dietary records for 7 consecutive days and test-retest reliability suggests that the 2 scales for vegetable items have acceptable validity and reproducibility compared to the scales that used a single question or vegetable subgroup and, therefore, may determine the variety of vegetables consumed.