View Record

TitleSources of error and bias in the assessment of dietary intake : 24-hour recalls and food records
AuthorMeltzer, Shelly Tracy
SubjectNutrition and Dietetics
Date2017-12-15T10:52:24Z
Date2017-12-15T10:52:24Z
Date1997
TypeThesis
TypeMasters
TypeMSc (Med)
AbstractThe first aim of this dissertation was to validate the 24-hour recall method of dietary assessment and to identify possible sources of error and bias in a cross-cultural sample of South African women. The second aim of this dissertation was to determine the source and extent of inter-researcher variability associated with the interpretation of food intake records. The first study was designed to consider whether reporting error is dependent on individual subject characteristics such as ethnicity, body mass, body fatness, age and education, or whether it is due to the dietary assessment tool (i.e., the 24-hour recall). In this study 118 women (25-55 years old) representing different ethnic and language groups (51 Xhosa-speaking, 31 Afrikaans-speaking, 36 English-speaking), different job types ( 25 unemployed, 25 general assistants, 52 medics and para-medics and 16 administrative personnel) and different levels of education (5 with 6-7 years of schooling, 35 with 8-10 years of formal schooling, 43 with 11-12 years of formal schooling and 35 with post high school diploma or degree) consumed a meal that was based on what they reported to habitually eat. All food and beverages consumed were covertly weighed and this was compared to a 24- hour dietary recall performed on the following day. Results of this study showed that the overall variance in reporting error was low. The error was, however, nutrient specific and was related to certain subject characteristics. Under-reporting was greater for subjects with a higher percentage body fat mass and a greater body mass index. Subjects with a lower level of education were more likely to under-report absolute carbohydrate (g) intake, whereas subjects with a higher level of education tended to under-report dietary fat. Subjects with a greater knowledge of food and nutrition were more likely to under-estimate protein intake and overestimate carbohydrate intake. "Seasonality" (fluctuations in food purchases due to income) affected body mass, socio-economic status and dietary reporting error. Under-reporting of fat intake (g) was greatest in subjects that experienced the most fluctuations in income. In the second study of this dissertation, three post graduate students in dietetics independently assessed and analyzed ten, 3-day food records. The specific areas of variability that were studied related to (i) the selecting of food names/codes on the computer dietary analysis program, (ii) the keying of data from the written dietary records (clerical errors) to the data storage file and, (iii) the conversion of food consumption quantities to gram weights (judgment errors). There were no systematic differences in the variability amongst dietitians in interpreting and analyzing dietary data in this study. The mean coefficient of variation for added sugar was 14.8 ± 12.6 (g) and for fat 7.1± 5.2 (g). In one subject, the range of difference in fat (kJ) intake assessed by the dietitians was as high as 1313 kJ. Conclusions: 1) Adiposity is a universal predictor for under-reporting of dietary intake. 2) Inaccuracies in dietary reporting are nutrient specific and may be influenced by social desirability bias (through education or knowledge of food), language, familiarity with certain foods and "seasonality" in food purchases. 3) These same factors influence the choices of food and beverages that subjects make in a "test meal" that is used to validate dietary intake 4) The variations in interpretation of the quality and quantity of reported food intake by the measurer is a source of experimental error that cannot be ignored and which may account, at least partially, for the difference between true and reported intake.
PublisherUniversity of Cape Town
PublisherFaculty of Health Sciences
PublisherDivision of Human Nutrition
Identifierhttp://hdl.handle.net/11427/26735