Validity of dietary data in young populations and implications of measurement errors
|Other Titles:||Validität von Ernährungsdaten in jungen Studienpopulationen und Implikationen von Messfehlern||Authors:||Börnhorst, Claudia||Supervisor:||Pigeot, Iris||1. Expert:||Pigeot, Iris||2. Expert:||Huybrechts, Inge||Abstract:||
Abstract: A high number of relationships between dietary intakes and health outcomes has been suggested and investigated during the last decades (Ezzati&Riboli, 2012; Vargas&Thompson, 2012; Alinia et al., 2009; Howarth et al., 2005; Kushi, 1992). Diet is of special interest as it is a modifiable risk factor. However, to date little is known with certainty on the complex relations between diet and specific diseases as respective research requires accurate, quantitative information on dietary intakes. Hence, the description of dietary intakes is one of the main tasks of dietary monitoring surveys and epidemiological studies. Assessment of dietary intakes is challenging due to changes in diet during life as well as due to the day-to-day variation that characterizes dietary intakes in general. In addition, the estimation of long-term consumption frequencies and amounts is difficult for most people as it relies on long-term memory and the capability of correct averaging. Strictly spoken, dietary intakes cannot be measured without error and will presumably never be. Researchers investigating associations between diet and specific diseases or distributions of dietary intakes need to account for various measurement errors to avoid drawing erroneous conclusions. The nature and magnitude of measurement errors in dietary data depend on the study population under investigation as well as on the assessment instrument. To date, there are only few recommendations available for measuring dietary intakes among children. As young children do not have the cognitive ability to report their dietary intakes themselves, usually parents are asked to proxy-report their child's intakes. This means that additional problems emerge from meals that are not under parents' control like e.g. school meals leading to unintentional misreporting. Little is known about the validity of proxy-reported dietary data, potential determinants of misreporting and additional sources of measurement errors in young populations yet. Therefore, this thesis aimed to investigate the extent and effects of measurement errors when assessing and modeling dietary data in young children. Special emphasis was put on differential measurement errors resulting from misreporting where different methods to counteract attenuation or distortion of risk estimates were encountered and evaluated. In summary, differences in the determinants of misreporting were found for proxy-reported dietary data compared to those previously reported for self-reported data where the problem of misreporting seemed to be even more severe in case of proxy-reports. Misreporting strongly affected effect estimates of associations between diet and overweight/obesity. Results strongly depended on the chosen statistical model where even reversed signs were observed when accounting or not-accounting for reporting errors. These findings suggest that studies on diet-disease associations based on proxy-reported dietary data are problematic as there is still no formal way to handle differential measurement errors caused by misreporting. In the absence of objective validation data, the true effects remain unknown. A large number of associations reported so far in epidemiological studies may be biased due to the application of models relying on the assumption of non-differential measurement errors only.
|Keywords:||dietary intake, children, differential error, misreporting, proxy-report||Issue Date:||13-Dec-2012||Type:||Dissertation||URN:||urn:nbn:de:gbv:46-00102912-11||Institution:||Universität Bremen||Faculty:||FB3 Mathematik/Informatik|
|Appears in Collections:||Dissertationen|
checked on Jan 25, 2021
checked on Jan 25, 2021
Items in Media are protected by copyright, with all rights reserved, unless otherwise indicated.