Errors in recalling foods eaten

The purpose of this study was to assess errors in diet recalls attributable to the actual recall of foods eaten. These errors include: not recalling foods eaten, recalling foods not eaten or modifying the description of foods eaten. It is difficult to assess the extent of these errors in free-living subjects whose actual intakes are unknown. Controlled feeding studies, in which subjects are fed experimental diets, provide an opportunity for comparison between foods consumed and foods recalled. We examined diet recalls of 18 females and 25 males participating in the Dietary Effects on Lipoproteins and Thrombogenic Activity (DELTA) Study (Protocol 1), a multicenter, controlled feeding study. Three, 24-h recalls were collected by computer-assisted telephone interview on random days using Minnesota Nutrition Data Systems. Foods recalled were compared to actual foods eaten and were coded as correct (food item eaten correctly recalled), missed (eaten but not recalled), added (recalled but not eaten) or substituted (modified description of foods actually eaten). 76% of foods eaten were correctly recalled, 10% were missed, 9% were substituted and 4% were added. Females correctly recalled foods significantly more than males (p=0.0013). Errors in recall occurred more often in some food groups than others. Foods that had a greater number of substitutions included juices, cereals and breads. Most commonly missed foods included butter/margarine and mayonnaise. These results demonstrate that different recall errors occur depending on the foods consumed which could have significant implications for assessments of food group and nutrient intakes.