SEPARATING FACT FROM ARTIFACT IN CHANGES IN NUTRIENT INTAKE OVER TIME

Objective To determine whether the differences between the procedures and nutrient databases used in the 1977-1978 Nationwide Food Consumpti on Survey (1977-78 NFCS) and those used in the 1987-88 NFCS could nota bly affect the estimated mean nutrient intakes. Design This was a spli t-sample field experiment. Seventy-two area segments were randomly sel ected; 10 housing units were selected in each segment and randomly ass igned to one of two experimental groups. Setting The study took place in the Philadelphia, Pa, metropolitan area. Subjects The subjects were 697 women aged 20 to 49 years. Treatment Group A was interviewed usin g 1987-88 NFCS 24 hour recall procedures and their nutrient intakes we re calculated using the 1987-88 NFCS food codes, weight conversion fac tors, and nutrient database. Group B was interviewed using 1977-78 NFC S 24-hour recall procedures and nutrient intakes were calculated four ways using various combinations of 1987-88 and 1977-78 NFCS food codes , weight conversions, and nutrient database. Main outcome measure Mean intakes of food energy and 14 nutrients were calculated. Statistical analysis We used two-sample, multivariate t tests; univariate t tests; univariate repeated measures analysis of variance; and univariate pai red t tests. Results Changes in the nutrient database caused by improv ed analytic techniques and increased number of foods sampled were grea t enough to warrant revising the 1977-78 NFCS estimated intakes for ir on, magnesium, and vitamins B-6 and B-12. Changes in intake of fat, vi tamin A, and thiamin, however, were caused by real changes in foods so no revisions were necessary for these nutrients. Other nutrients were unaffected. Conclusion When considering changes in nutrient intake ov er time, it is important to correct the earlier food composition estim ates so that they accurately reflect the composition of foods at the t ime they were consumed. This requires preserving the real differences in the composition of foods while correcting for artifactual differenc es attributable to improvements in the quality of nutrient data.