To assess the reliability and relative validity of a diet index score derived from a Short Food Survey (SFS).

The thirty-eight-item SFS was designed to assess recent dietary intake of 4–11-year-olds to enable calculation of the Dietary Guideline Index for Children and Adolescents. Reliability was assessed based on two online administrations of the SFS, one week apart. Relative validity was assessed by comparing intakes derived from the SFS with those from the mean of three 24 h recalls. Intra-class correlations, Bland–Altman plots and estimated biases were assessed. Cohen's κ coefficients were used to determine the level of agreement between the two methods.

Adelaide, Australia.

Sixty-three parents reported on their children's intake (mean age 7·1 (sd 2·1) years).

The intra-class correlation for reliability ranged from 0·43 for dairy foods to 0·94 for beverages, and was 0·92 for total diet index score (all P < 0·01). The intra-class correlation for validity ranged from 0·04 for meat and alternatives to 0·41–0·44 for fruit, beverages and extra foods, and was 0·44 for the total diet index score. The SFS overestimated the mean diet index score by 16 %, and the bias was consistent across levels of compliance. The percentage agreement into tertiles of index scores was 84% between the administrations of the two SFS, but only 43 % when comparing the SFS with the mean of the recalls.

The SFS can provide a consistent estimate of overall compliance to dietary guidelines for children aged 4–11 years, but overestimated the total diet index score by 16 % across all levels of compliance.

To compare diet measures from a food-frequency questionnaire (FFQ) with measures from 24-hour dietary recalls (24HDRs).

The participants answered an FFQ after completing four, repeated 24HDRs during a year.

Norway, nationwide.

Of 500 women randomly selected from The Norwegian Women and Cancer Study (the Norwegian arm of the European Prospective Investigation into Cancer and Nutrition), 286 agreed to participate and 238 completed the study.

On the group level, the FFQ overestimated absolute intake in seven and underestimated intake in six of 21 food groups. Intakes of energy, fat, added sugar and alcohol were lower in the FFQ than in the 24HDRs, whereas intake of fibre was higher. Spearman's rank correlation coefficient ranged from 0.13 (desserts) to 0.82 (coffee) for foods, and from 0.25 (β-carotene) to 0.67 (alcohol) for nutrients. Three per cent of the observations on nutrient intake fell in the opposite quintile when classified according to the FFQ as compared with the 24HDR. The median calibration coefficient, calculated by regression of the 24HDR data on the FFQ data, was 0.57 for foods and 0.38 for nutrients.

The FFQ's ability to rank subjects was good for foods eaten frequently and fairly good for macronutrients in terms of energy percentages. Weaker ranking abilities were seen for foods eaten infrequently and for some micronutrients. The results underline the necessity of performing measurement error corrections.