Research indicates that the length of time needed to describe dietary diversity is approximately 2 weeks. This is the first study conducted to develop a dietary variety measurement tool that is sensitive to the effect of time on dietary variety without the burden of gathering data for 2 weeks.

To determine whether 3 days of 24-hour dietary recall logs collected during a 15-day period would predict food variety as well as 15 consecutive days. The study also determined which set of 3 days (consecutive vs interval days) within a 15-day period would better predict 15-day food variety.

Prospective survey of the dietary practices of children.

Seventy-two children aged 9 to 12 years attending fourth and fifth grades in a public elementary school in a Midwestern town in the fall of 2005.

Predicted 15-day cumulative dietary variety score from 3 consecutive days and 3 interval days of dietary data.

Two prediction models were obtained from multiple linear regression analyses in which natural log-transformed (log_e) 15-day variety scores were regressed on log_e 3-day variety scores (consecutive and interval days). The ability of each model to predict the 15-day cumulative variety score was assessed by comparisons of mean bias, mean-squared error, coefficient of determination (R²), and Pearson product-moment correlation coefficients.

Three days of dietary data accurately estimated dietary variety over time for this sample of 9- to 12-year-old children using the predictive equation generated in this study. Three interval days predicted 15-day food variety more precisely than 3 consecutive days.

The predictive equation is accurate in estimating food variety over time for this population and, if validated in independent samples, could be applied to similar populations.