Computerization of a dietary history interview in a running cohort; evaluation within the Amsterdam Growth and Health Longitudinal Study

Objective: In nutritional research, a growing interest in the use of computer-assisted cross-check dietary history interview methods exists in order to in prove cost-effectiveness. The introduction of such a method in an ongoing longitudinal study was evaluated with special emphasis on the effect on interviewer bias. Design: A study for the interviewer bias within and the agreement between a previously used paper-based face-to-face crosscheck dietary history interview method and a newly developed interviewer-administered computer-assisted version of this interview method. Subjects: The interviewer bias of 436 face-to-face interviews is compared with that of 352 computer-assisted interviews. A subset of 82 subjects underwent a face-to-face interview at the mean age of 27 and 32y and a computer-assisted interview at their mean age of 36y. Energy, three macronutrients (protein, fat and carbohydrate), two micronutrients (calcium and iron) and alcohol intakes obtained by these three measurements are compared to analyse the agreement between the two interview methods. Results: ANOVA showed no interviewer bias for all seven analysed nutrients within the data from the computer-assisted interview, while for the face-to-face interview method, several nutrients varied significantly among the interviewers. Five different measures, used to analyse the agreement (differences, Pearson's correlation, ICC, square weighted kappa and Bland Altman plots), showed no relevant differences between the two cross-check dietary history interview methods. Conclusions: It is concluded that the computer-assisted interview caused a reduction of interviewer bias and is of similar quality to the face-to-face interview method. Computerization of a paper-based interview can be implemented in a running cohort if a change in method is unavoidable.