Objective To assess the validity of a picture-sort approach to administering the National Cancer Institute food frequency questionnaire to older adults

Design A picture-sort interview was conducted in each respondent's home. After the picture sort, a 24-hour recall interview was administered on the same occasion. Five additional in-home recall interviews were subsequently conducted at approximately 1-month intervals.

Subjects/setting Forty-seven female and 49 male volunteers aged 66 to 100 years were recruited from among Cardiovascular Health Study participants from Maryland and North Carolina.

Main outcome measures Estimates from the picture sort and the recall for intakes of macronutrients, cholesterol, fiber, and selected vitamins and minerals exclusive of supplements.

Statistical analyses Comparison of means estimated by the two methods and correlation analyses were used. Correlations were adjusted under varied assumptions about the nature of the information contained in the six 24-hour recalls relative to respondents’ usual intakes.

Results After correction for attenuation, Pearson correlation coefficients for macronutrients ranged from .41 for protein to .74 for saturated fat and cholesterol. For vitamins and minerals, correlations ranged from .26 for beta carotene to .62 for calcium.

Applications Picture-sort estimates of mean nutrient intakes were comparable with estimates based on 24-hour recalls, and correlations with reference data were similar to those reported in the literature for conventionally administered food frequency questionnaires. This dietary assessment method may, therefore, offer a way to simplify or structure responses to improve ease of administration and increase respondents’ liking for the interview without loss of data quality. J Am Diet Assoc. 1996; 96:137-144.