Unit:
Κατεύθυνση ΒιοστατιστικήLibrary of the School of Health Sciences
Author:
Stratou Maria-Anna
Supervisors info:
Κατσουγιάννη Κλέα, Καθηγήτρια, Ιατρική Σχολή ΕΚΠΑ
Τριχοπούλου Αντωνία, Ομότιμη Καθηγήτρια, Ιατρική Σχολή ΕΚΠΑ
Δημακοπούλου Κωνσταντίνα, Ακαδημαϊκή Υπότροφος, Ιατρική Σχολή ΕΚΠΑ
Original Title:
Εκτίμηση της συνήθους διατροφικής πρόσληψης των συμμετεχόντων στο πρόγραμμα ΥΔΡΙΑ με βάση τα ερωτηματολόγια 24ωρων ανακλήσεων και προκαταρκτική εφαρμογή στην έρευνα συσχετίσεων με εκβάσεις στον τομέα της υγείας
Translated title:
Estimation of the usual dietary intake of individuals based on 24h recall questionnaires and preliminary application in the investigation of associations with health outcomes in the Hydria survey
Summary:
Introduction: The investigation of the assessment of usual dietary intake and its association with health outcomes is a common object of studies. This is difficult enough because food consumption data vary widely as some food items are consumed daily by almost everyone, while others are episodically consumed, so that dietary data have excess zeros.
Objective: The purpose of this thesis was the comparison of the estimated distributions of usual dietary intake of four food groups of the HYDRIA survey - vegetables, dairy products, legumes and fish/shellfish - as derived from a naive approach and a statistical method of the National Cancer Institute (NCI). The NCI method aims to correct the measurement error of usual intake. In addition, a naive approach and the extension of the NCI method were applied to associate these food groups with diastolic, systolic blood pressure and hypertension.
Methods: We used data from the HYDRIA survey (3951 participants). This data was provided by the Hellenic Health Foundation. Dietary intake in the HYDRIA survey was assessed by two 24-hour dietary recalls participant and by a non-quantitative food frequency questionnaire (Food Propensity Questionnaire). In order to estimate usual dietary intake, we applied a naive approach (where we used the average of two 24-hour recalls) and the NCI method. The difference between these two methods is that the NCI method fits a two-part statistical model for estimating both probability and amount of consumption. The two parts of the model are linked by allowing the two person-specific effects to be correlated and including the same covariates in both parts of the model.
In order to estimate diet-blood pressure/hypertension associations, we used the average of two 24-hour recalls in naive approach and an extension of the NCI method, proposed by Kipnis et al., in order to predict individual usual dietary intake obtained from two 24-hour recalls. One feature of the extended method was that additional covariates potentially related to usual intakes may be used to increase the precision of the estimates of usual intake and of diet-health outcome associations.
Results - Conclusions: Regarding to the estimation of the distribution of usual dietary intake, the only thing that does not differ between the two methods is the average, so we can rely on the naive method for the average population dietary intake. However, the naive approach tends to underestimate the usual intake on low percentiles and to overestimate it on high percentiles. The great difference between the two methods is in the variance of the estimated distributions, which is minimized by the NCI method.
Types of correlations of usual dietary intake with blood pressure and hypertension are the same with both methods, while small differences are presented in statistical significance. Vegetable intake has beneficial effects on diastolic blood pressure (p-values<0.001) and hypertension (p-values=0.023 (NCI) 0.011 naive)). The inverse correlation of vegetable intake with systolic pressure is statistically significant (p-value=0.045) only in naive approach. Dairy consumption beneficially affects hypertension, while its inverse correlation with systolic pressure is statistically significant only in naive method (p-value= 0.032). Legumes and are not associated with blood pressure and hypertension. Also, fish/shellfish are not associated with blood pressure, but we noted a marginally significant association with an increased risk for hypertension. It may be appropriate to use more than two 24-hour recalls in order to investigate the associations of legumes and fish/shellfish because they are episodically consumed.
Main subject category:
Health Sciences
Keywords:
Dietary survey, 24-hour dietary recall, Dietary measurement error, Estimation of usual dietary intake, NCI Method, Blood pressure, Hypertension
