Objectives To evaluate the validity of food behavior items, using a biologi
cal measure (serum carotenoids) as the criterion for validity of fruit and
vegetable intake, and the results from multiple 24-hour recalls to test con
vergent validity with nutrient intake.
Design Participants responded to 39 food-behavior questions and later compl
eted three 1-day dietary recalls. Serum carotenoid levels were determined f
or a 59% randomly selected subsample.
Subjects/setting A convenience sample of 100 English-speaking, low-income w
omen participating in a Food Stamp Nutrition Education Program in 8 Califor
nia counties.
Statistical analyses Spearman correlation coefficients were calculated betw
een responses to the food behavior items and (a) serum carotenoid levels an
d (b) mean nutrient intakes from the 24-hour recalls. Cronbach's coefficien
t a was determined for items within broad food behavior topics.
Results Responses to 10 food behavior items were significantly correlated w
ith serum carotenoid levels (correlations greater than 0.45 were found for
choosing low-fat foods and a self-evaluation of overall dietary quality). A
n additional 12 items showed hypothesized associations with the 24-hour rec
all data (with a maximum correlation 0.50 for number of eggs per week and d
ietary cholesterol). Cronbach's coefficient a ranged from 0.28 (for 5 fat a
nd cholesterol items) to 0.79 (for 9 fruit and vegetable items).
Applications/Conclusions Nutrition professionals can use these methods to v
alidate items for food behavior checklists for specific populations. The it
ems described here may be useful when designing instruments to administer t
o low-income women in a community setting.