The effect of missing data in the supplements to McCance and Widdowson's food tables on calculated nutrient intakes

Objective: To assess the degree of underestimation of nutrient intakes caused by missing data in McCance and Widdowson's food tables and supplements, the standard food tables used in the UK. Subjects: 1026 children aged 18 months. Setting: A research clinic in Bristol. Interventions: A 3-day dietary diary was completed for each child. These were coded and analysed using a database consisting of the unedited information in the food tables, to produce a set of 'uncorrected' nutrient intakes. 'Guesstimated' values for nutrient content were then added to our nutrient database in place of the missing values in the food tables, and the daily nutrient intakes were recalculated. Main outcome measures: Calculated daily nutrient intakes before and after substituting missing data in the food tables for guesstimated values. Results: Of the 1027 foods used in the analysis, 540 had missing data for one or more nutrient content. For several of the nutrients examined adding guesstimated values altered the calculated nutrient intake of more than 90% of the subjects. However, for nutrients such as the B vitamins and the major minerals the mean percentage underestimate was very small. We calculated the underestimation of nutrient intake to be largest for vitamins E and D (13.8% and 14.7%, respectively). The effect of missing data on calculated nutrient intakes was proportionately greater at the bottom end of the nutrient intake distribution. Conclusion: Missing data has a fairly small effect on calculated mean daily intakes. However, it can result in some individuals being misranked within a nutrient intake distribution. The availability of a standard set of 'guesstimates' to use in place of missing data would reduce this problem, and would improve comparability between dietary surveys.